Вы находитесь на странице: 1из 282

STP 1468

Elemental Analysis of Fuels and


Lubricants: Recent Advances
and Future Prospects

R. A. Kishore Nadkarni, editor

ASTM Stock Number: STP 1468

ASTM International
100 Barr Harbor Drive
PO Box C700
West Conshohocken, PA 19428-2959
INTERNATIONAL

Printed in the U.S.A.

Library of Congress Cataloging-in-Publication Data


Elemental analysis of fuels and lubricants: recent advances and future prospects/R.A.
Kishore Nadkarni, editor.
p. c m . ~ S T P ; 1468)
Includes bibliographical references and index.
ISBN 0-8031-3494-0 (alk. paper)
1. Fuel--Analysis. 2. Lubrication and lubricants--Analysis. I. Nadkami, R.A. II.
Series: ASTM special technical publication; 1468.
TP321.E46 2005
665.5'38---dc22
2005022779

Copyright 9 2005 ASTM International, West Conshohocken, PA. All rights reserved. This material
may not be reproduced or copied, in whole or in part, in any printed, mechanical, electronic, film,
or other distribution and storage media, without the written consent of the publisher.

Photocopy Rights
Authorization to photocopy items for internal, personal, or educational classroom use, or
the internal, personal, or educational classroom use of specific clients, is granted by ASTM
International provided that the appropriate fee is paid to the Copyright Clearance Center, 222
Rosewood Drive, Danvers, MA 01923; Tel: 978-750-8400; online: http://www.copyright.com/.

Peer Review Policy


Each paper published in this volume was evaluated by two peer reviewers and at least one editor. The authors addressed all of the reviewers' comments to the satisfaction of both the technical
editor(s) and the ASTM Committee on Publications.
To make technical information available as quickly as possible, the peer-reviewed papers in this
publication were prepared camera-ready as submitted by the authors.
The quality of the papers in this publication reflects not only the obvious efforts of the authors
and the technical editor(s), but also the work of the peer reviewers. In keeping with long-standing
publication practices, ASTM maintains the anonymity of the peer reviewers. The ASTM Committee
on Publications acknowledges with appreciation their dedication and contribution of time and effort
on behalf of ASTM.

Printed in Baltimore,MD
September2005

Foreword
This publication, Elemental Analysis of Fuels and Lubricants: Recent Advances and Future
Prospects, contains selected papers presented at the symposium of the same name held in
Tampa, Florida, on 6-8 December 2004. The symposium was sponsored by Committee D02
on Petroleum Products and Lubricants. The symposium chairman and editor was R. A.
Kishore Nadkarni.

Contents
Overview

vii

Zen and the Art (or is it Science) of a Perfect Analysis--R. A. K. NADKARNI

A T O M I C EMISSION SPECTROSCOPY

Analysis of Gasoline and Diesel Fuel Samples by Inductively Coupled Plasma


Atomic Emission Spectrometry (ICP-AES), Using Pneumatic Nebulizer
and Standard Spray Chamber---c. c. ONYESO

17

Elemental Analysis of Lubricating Grease by Inductively Coupled Plasma


Atomic Emission Spectrometry (ICP-AES)--B. s. FOX

24

The Use of Microwave Digestion and ICP to Determine Elements in Petroleum


Samples--J. D. HWANG,M. HORTON,AND D. LEONG

33

Advances in ICP-MS Technologies for Characterization and Ultra-Trace


Speciation as a Tool for the Petroleum Industry--J. PASZEK,K. J. MASON,
A, S. MENNITO, AND F. C. MCELROY

42

Direct Trace and Ultra-Trace Metals Determination in Crude Oil and


Fractions by Inductively Coupled Plasma Mass Spectrometry-S. DREYFUS. C. PECHEYRAN, C. MAONIER, A. PRINZHOFER. C. P. LIENEMANN. AND
O. F. X. DONARD

51

Fuel Analysis by Filter Furnace Electrothermal Atomic Absorption


Spectrometry--P. TII~rARELLL M. PRIOLA. S. RICCHIUTO, D. A. KATSKOV, AND
59

P. NGOBENI

Rotrode Filter Spectroscopy: A Recently Improved Method to Detect and


Analyze Large Wear and Contaminant Particles in Fluids--M. LUKAS.
R. J. YURKO. AND D. P. ANDERSON

71

S U L F U R D E T E R M I N A T I O N AND X - R A Y FLUORESCENCE

Trace Levels of Sulfur in the Fuels of the Future: Analytical Perspective-R. A. K. NADKARNI

85

vi

CONTENTS

Analysis of Fuels, Lubricants, and Greases Using X-Ray Fluorescence


Spectrometry--J. WOLSKA, B. VREBOS, AND P. BROUWER

98

Determination of Sulfur Content in Crude Oil Using On-Line X-Ray


Transmission Technology--s. FESS

108

Low-Level Sulfur in Fuel Determination Using Monochromatic W D X R F ASTM D 7039-04--z. w. CHEN, F. WEI, I. RADLEY,AND B. BEUMER

116

Latest Improvements on Using Polarized X-Ray Excitation EDXRF for the


Analysis of Low Sulfur Content in Automotive Fuel--D. WISSMANN

128

Rapid Determination of Sulfur in Liquid Hydrocarbons for At-Line Process


Applications Using Combustion/Oxidation and UV-Fluorescence
Detection--s. TARKANICAND J. CRNKO

137

Pyro-Electrochemicai On-Line Ultra Low Sulfur Analyzer--J. R. RHODES

152

DP-SCD and LTMGC for Determination of Low Sulfur Levels in


Hydrocarbons--R. L. GRAS,J. C. LUONG. R. V. MUSTACICH,AND R. L. SHEARER

164

MERCURY DETERMINATION

Sampling and Analysis of Mercury in Crude Oil--s. M. WILHELM.


D. A. KIRCHGESSNER. L. LIANG, AND P. H. KARiHER

181

Determination of Total Mercury in Crude Oil by Combustion Cold Vapor


Atomic Absorption Spectrometry (CVAAS)--B. s. Fox. K. J. MASON.AND
196

F. C. MCELROY

Mercury Measurements in Fossil Fuels, Particularly Petrochemicals-P. B. STOCKWELL, W. T. CORNS, AND D. W. BRYCE

207

OTHER HETEROATOMS

Recent Advances in Gas Chromatographic/Atomic Emission Hetero-Atom


Selective Detection for Characterization of Petroleum Streams and
eroducts--F, p. DISANZO AND J. W. DIEHL

221

Improvements in the Determination of Fluorine in Fuel and Lubricants by


Oxidative Combustion and Ion-Selective Electrode Detection--L J. NASH

232

Phosphorus Additive Chemistry and its Effects on the Phosphorus Volatility of


Engine Oils--T. w. SELBY, R. J. BOSCH. AND D. C. FEE

239

Analysis of the Volatiles Generated During the Selby-Noack Test by 31p NMR
Spectroscopy--R. J. BOSCH, D. C. FEE, AND T. W. SELBY

255

Index

275

Overview
In spite of being a mature science, elemental analysis continues to play a vital role in
product manufacturing and quality characterization in many sectors of all industries. Research divisions in both industry and academia continue devising new ways of lowering the
elemental detection limits so that even the minutest amounts of elements in products could
be determined in as accurate and precise a fashion as possible.
The ASTM International D02 Committee on Petroleum Products and Lubricants through
its Subcommittee 3 on Elemental Analysis has played a large and crucial role in the last
several decades in standardizing numerous elemental analysis methods used in the oil industry. Currently there are about 75 standard test methods under the jurisdiction of SC 3,
and additionally at least 6 more are under active development and moving towards standard
designations. I have no doubt that this activity will continue in the future. These standards
comprise virtually all known modem techniques for elemental analysis of petroleum products
and lubricants.
The first ASTM D02 symposium on this subject was held in New Orleans in December
1989 at which 20 papers were presented. Of these, 13 were published as a book, Modern
Instrumental Methods of Elemental Analysis of Petroleum Products and Lubricants, ASTM
STP 1109. The current and second "quindecennial" (i.e., every 15 years) was held in Tampa,
Florida in December 2004. This was attended by over 120 people. Thirty papers were presented on diverse subjects from 64 authors from nine different countries: Brazil, France,
Germany, Italy, the Netherlands, South Africa, Switzerland, U.K., and U.S. Of these, 12
papers were from the oil industry, 15 from the instrument manufacturers, l0 from national
research organizations, and 4 from the universities.
The objective of this symposium and this book is to acquaint the readers with the latest
advances in the field of elemental analysis and to focus on what avenues of future research
to explore in this area. The subjects included are various elemental analysis techniques such
as atomic absorption spectrometry, inductively coupled plasma emission and mass spectrometry, isotope dilution mass spectrometry, X-ray fluorescence, ion chromatography, gas chromatography-atomic emission detection, other hyphenated techniques, hetero-atom microanalysis, sample preparation, reference materials, and other subjects related to matrices such
as petroleum products, lubricating oils and additives, crude oils, used oils, catalysts, etc.
Of the 30 papers presented at the symposium, 23 papers were published in the Journal of
ASTM International (JAI), and are included in this ASTM publication. As far as possible,
the papers have been arranged by analytical techniques used, although in some cases there
is some overlap: ICP-AES, XRF, sulfur, mercury, other hetero-atoms.
The first article is from the plenary lecture given at the symposium by the symposium
chairman Kishore Nadkarni. It covers total quality management practices advocated for obtaining a "perfect" analysis. Proper staff training, sampling, calibration and quality control
practices, adherence to test method details, participation in proficiency testing, accreditation
from national bodies, benchmarking, etc., are some of the critically important approaches
that need to be taken to achieve the ideal state of analytical Zen perfection.

Atomic Spectroscopy
Among the seven atomic spectroscopy papers in this book, five concern various aspects
of ICP-AES, a technique widely used for the determination of metals in petroleum products
vii

viii

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

and lubricants. Onyeso (Ethyl Corporation) presents an ICPAES method for the determination of additive elements and wear metals, principally manganese, in gasoline and diesel
fuels, with simple dissolution in kerosene and using yttrium internal standard. Accessories
such as direct injection nebulizer, ultrasonic nebulizer, chilled spray chamber, etc., were not
necessary for this analysis.
Fox (ExxonMobil Research and Engineering) presents an ICPAES method for the determination of additive elements and wear metals in lubricating greases. Since such samples
cannot be directly nebulized in the ICP plasma, alternate sample dissolution techniques were
employed: dry sulfated ashing, microwave assisted dry ashing, microwave assisted acid digestion with both open and closed vessels. This method is being developed into an ASTM
standard test method and is expected to be published by YE05. Hwang and Leong
(ChevronTexaco) also discuss the use of microwave acid digestion for sample preparation
before ICPAES measurements.
Elemental speciation using mass spectrometry in conjunction with ICPAES is a latest
advance in atomic spectroscopy, which is becoming popular in analytical research labs.
Mason et al. (ExxonMobil Research and Engineering) show how linking ICP-MS to various
liquid chromatographic techniques has enabled determination of ppm levels of metals in
hydrocarbons to ppb level measurements in refinery effluent streams. Hyphenated ICP-MS
techniques were used to provide speciation information on nickel and vanadium in crude
oils and assist in development of bioremediation options for selenium removal in wastewater
treatment plants. Similar ICP-MS technique without sample demineralization was used by
Lienemann, et al. (lnstitut Francais du Petrole) to determine the trace and ultra-trace amounts
of metals in crude oils and fractions.
Lukas et al. (Spectro Inc.) describe an improvement made in rotating disc electrode atomic
emission technology by incorporating a filter device in the rotrode, which enables to detect
particles greater than 10 i~m size.
Tittarelli et al. (SSC, Milan) employed a transverse heated filter atomizer with atomic
absorption spectrometry to determine a number of trace elements in automotive and jet fuels.
Sub-ppm detection limits were obtained. The use of filter furnace reduces the risk of elemental loss during drying and pyrolysis steps, and decreases the interferences due to molecular absorption and light scattering.

X-Ray Spectroscopy
Similar to atomic emission spectroscopy, equally widely used technique for elemental
analysis in the oil industry is X-ray fluorescence (XRF). There are four papers in this book
using this technique, three of which deal with the determination of sulfur in gasoline and
diesel.
Wolska et al. (Panalytical BV) compared performance of three XRF technologies: high
power and low power WDXRFs and a bench top EDXRE There are large differences in the
sensitivities and hence varying lower limits of detection or qualification and sample throughput, for these technologies.

Sulfur Analysis
One of the most important analyses done today on petroleum products, particularly gasoline, reformulated gasoline, and diesel, is for low levels of sulfur. Government regulations
on sulfur emissions from automobiles and other combustion sources have steadily increased;
hence, the increasing interest in devising precise and accurate methods for trace and ultra-

OVERVIEW

ix

trace amounts of sulfur in fuels of the future as evident from seven papers on this subject
published in this book.
Nadkarni (Symposium Chairman) reviewed the alternate methods available for sulfur determination in fuels. Out of about 20 ASTM standard test methods available, only about five
(D 2622 WDXRF, D 3120 microcoulometry, D 5453 UV-fluorescence, D 6920 pyroelectrochemical, D 7039 MWDXRF) are appropriate for ultratrace amounts of sulfur in
gasoline or diesel. However, in their actual industrial use only D 2622 and D 5453
predominate.
Chen et al. (XOS Inc.) describe a newly developed technology instrument based on monochromatic WDXRF for low sulfur analysis of fuels. The instrument has a significant advantage over existing WDXRF instruments in terms of increased sensitivity and improved
signal to noise ratio. This technique has been recently given the ASTM designation D 7039.
Another new instrument recently developed for sulfur by XRF determination is described
by Wissmann (Spectro, Inc.). This method uses polarized EDXRF, considerably reducing
background scatter, and achieving detection limit comparable to that of WDXRF. Recent
developments in detector technology and in closed coupled static geometry have resulted in
further improvement of sensitivity for this application. This method is also in the developmental stage for ASTM method designation.
Shearer et al. (Ionic Instruments and Dow Chemicals) describe a novel technique developed tbr low levels of sulfur in hydrocarbon matrices using a low thermal mass temperature
programmable and dual plasma chemiluminiscence detector. The method with appropriate
modification can measure individual sulfur species similar to ASTM method D 5623.

On-line Sulfur Analysis


Increasingly refineries, plants, and pipeline operators are focusing on obtaining quick turnaround for sulfur analysis rather than wait ['or time-delayed laboratory analysis. A large
number of such installations are being operated in the industry around the world. Three
papers in this book discuss applications of such on-line technology for sulfur determination
in fuels. In an on-line application of X-ray transmission technology, Fess (Spectro, Inc.)
describes the basis of this technology and its application to classification and blending of
crude oils that contain between 0.1 and 3.3 m % sulfur. Commercial instruments based on
this technology are being used in the field.
In a second on-line application paper, Tarkanic and Crnko (Antek/PAC) describe an online instrument based on ASTM Test Method D 5453, UV-Fluorescence Detection. The latter
is a widely used method in the oil industry for low and ultra-low levels of sulfur. The online instrument appears to be very stable and fast (< 1 min per analysis) over extensive
periods of field operations.
In a third on-line application paper tbr sulfur analysis, Rhodes (Rhodes Consulting), ASTM
Test Method D 6920 is applied for on-line application. This method uses pyro-combustion
followed by electrochemical detection.

Mercury Determination
Although adverse effects of mercury emissions on environment and humans has been
known lbr decades, in recent years there has been concern regarding the mercury content of
crude oils, and its emission through petroleum refining process. There are three articles in
this book discussing this issue.
Wilhelm et al. (Mercury Technology Services/EPA et al.) provide a review of the presence
of mercury in various parts of the world, its speciation, and alternate methods of determining

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

low ppm and sub-ppm levels. Fox et al. (ExxonMobil Research and Engineering) describe
a method for the determination of ppb levels of mercury in crude oils and distillation cuts
using combustion cold vapor atomic absorption spectrometry technique. Stockwell et al. (PS
Analytical Ltd.) describe the technique of atomic fluorescence spectrometry for the determination of mercury both before and after mercury removal from petrochemicals. The technique has been used for on-line measurements in installations operating around the clock for
at least 2 years.

Other Heteroatoms
DiSanzo and Diehl (ExxonMobil Research and Engineering) used GC-AED for the determination of elements such as carbon, nitrogen, sulfur, oxygen, and phosphorus in fuels
and petroleum fractions. A simplified version of comprehensive GC x GC is coupled with
atomic emission detector to reduce the hydrocarbon matrix interference using simple and
rugged modulation along with rugged wide bore capillary columns. The technique together
with other spectroscopic techniques such as GC-MS can provide information on many selected elements and compounds that may be present in fuels as additives or contaminants.
In a pair of papers, Selby et al. (Savant, hzc. and Astaris LLC) describe using phosphorus
as an indicator of volatility of engine oils. Phosphorus is volatilized during Noack volatility
test (ASTM D 5800). The volatile material is trapped and analyzed for total phosphorus
using ICP-AES, and for phosphorus species using 3~p NMR spectroscopy.
An oxidative combustion followed by ion selective electrode detection method is proposed
by Nash (Antek/PAC) for the determination of fluorine in fuels and lubricants. An ASTM
method based on this technique is in development stage.

Unpublished Symposium Papers


Some papers were presented at the Symposium; however, they were not submitted for
publication by the authors. Nevertheless, they represent interesting approaches to some specific elemental analysis issues in the petrochemical industry. It would be useful if the authors
eventually publish these articles for the benefit of others in the industry. These presentations
include the following:
I. Kelly et al. (NIST) describe an isotope dilution thermal ionization mass spectrometry
method for the determination of sulfur in fossil fuels. The method is being used in
NIST for certification of a number of liquid fuels at low sulfur concentration levels.
2. Kelly et al. (NIST) also describe a "designer" calibration standard method for sulfur
determination in fossil fuels for users to prepare NIST traceable working standards with
known concentrations and uncertainties.
3. Manahan and Chassaniol ( Cosa Instruments and Dionex) describe an oxidative combustion followed by ion chromatographic conductometric method for the determination
of a number of nonmetallic elements such as sulfur and halogens in liquid and gaseous
hydrocarbons. A standard based on this technique is under development in ASTM for
designation as a standard method.
4. Long et al. (NIST) describe another method for mercury determination in crude oils
using isotope dilution-cold vapor-inductively coupled plasma-mass spectrometry technique. The method has very high sensitivity, very low blank and high accuracy. The
technique is being used to determine mercury in a large number of crude oil samples
from Department of Energy strategic petroleum reserve in the mercury concentration
range of 0.02-10 ng/g.

OVERVIEW

xi

5. Finally, Mason et al. (ExxonMobil Research and Engineering) describe the approaches
used for assay of fresh and spent reformer catalysts to determine the precious metals
(platinum and rhenium) in them. Methods such as WDXRF, ICPAES, and classical wet
chemistry methods are used for such analysis. Precise and accurate methods are critical
for these analyses, since small errors in analysis can have a large impact in commercial
transactions of these catalysts between the catalyst vendors and the oil companies.
Hopefully, the papers included here will provide the readers with the current state-of-theart and future research trends in the field of elemental analysis in the oil industry. Most
modern techniques used in the field are represented here.

Acknowledgment
I want to thank various ASTM staff members (particularly David Bradley, Dorothy Fitzpatrick, Crystal Kemp, Hannah Sparks, and Roberta Storer) for their prompt response and
cooperation that made the symposium and subsequent efficient publication in JAI and of this
volume possible. My thanks are also due to the reviewers who did a very good job of
providing technical reviews of all original paper submissions. Their invaluable assistance in
reviewing the papers made the final publication a much better quality product.

R. A. Kishore Nadkarni
Chairman, D02.SC 3 and
Symposium Chairman

Journal of ASTM International, March 2005, Vol. 2, No. 3


Journal of ASTM International, March 2005, Vol. 2, No. 3
Paper ID JAI12964
Available online at www.astm.org

R. A. Kishore Nadkarni 1

Zen and the Art (or is it Science) of a Perfect Analysis


ABSTRACT: An analytical laboratory in any industry plays a crucial role in product quality
management and ultimate customer satisfaction. Some factors need to be considered for an aspiring
laboratory to become a perfect performer. These range from sampling, calibration, contaminationcontrol,
and use of valid test methods to statistical quality assurance. Some approaches may be utilized to achieve
a perfect analysis including: staff training, participation in proficiency testing, use of standard reference
materials in the analytical sequence, internal and external audits, agency accreditation, continuous
improvement program, benchmarking, etc. Laboratories managed in this way show demonstrated
superiority in data precision and accuracy over the labs which do not practice such quality management.
Well-managed industrial laboratories can have insignificant laboratory sigma compared with
manufacturingvariability in the plant production. For a flawless perfect analysis, determinationto excel,
mental discipline to stay the course, willingness to overcome inertia and resistance, and focus on
producing a perfect analysis at all levels of laboratory staff are essential.
KEYWORDS: analysis, quality management,perfect analysis
As Robert Pirsig wrote in his landmark iconic autobiographical novel "Zen and the Art of
Motorcycle Maintenance," the art of motorcycle maintenance is primarily a mental phenomenon
[1]. One may have the tools, but unless there is mental preparation to achieve high goals, the
tools alone will not help. A similar mindset is needed to achieve excellence in a laboratory to
make it into a perfect laboratory that produces flawless performance. Tools may be available, but
if there is no organizational passion and will to excel, the laboratory will not become a perfect
laboratory.
The culture of excellence must be pervasive throughout the laboratory organization from the
laboratory manager to the laboratory technician. Higher management especially needs to show
through visible actions that only the best will do. Perfection cannot be achieved through shortterm stop-gap measures. A long-term improvement plan must be in place and followed upon to
be effective.
What is a perfect laboratory? It is a laboratory which delivers the product (i.e., accurate and
precise data) on time; if necessary, continuously improves on itself; makes the analysis "Right
the First Time," thus eliminating repeat analysis and giving erroneous information to the
customers; communicates with its customers and sometimes educates them when necessary. This
laboratory cares about the success of its customers' business.

Customer Services
A laboratory is a microcosm of its parent organization. The product delivered from a
laboratory is quality data. Hence, the primary objective of a laboratory should be to be the best in
quality. A laboratory needs to deliver a consistent product on time which meets or exceeds
customers' expectations, and which increases customers' confidence in the laboratory's
Manuscript received 7 September 2004; accepted for publication19 October 2004; published March 2005.
i Millennium Analytics,Inc.,East Brunswick, NJ 08816.

Copyright9 2005 by A S T M International,I00 Ban"Harbor Drive,PO Box C700, West Conshohockcn,PA 19428-2959.

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

reliability and dedication to quality. A customer needs to know that the laboratory cares. "Total
Care" is the sum of impressions formed during contact with the customers. A perception by the
customer that the laboratory is a caring organization can convert a customer from being forced to
be a customer to becoming a customer by choice. Substantial or continuing violations of a
customer's justified expectations will cause the customer to feel that the laboratory, organization
simply does not care.
Pillars to Build a Perfect Laboratory

There are at least twelve components which help to produce flawless laboratory performance.
In the approximate order in which an analysis is performed, these include but may not be
conEmed to the following (See Fig. 1 on page 3):
1. Training
2. Representative Sampling and Contamination Control
3. Calibrations
4. Technical Details of Test Methods
5. Statistical Quality Assurance
6. Use of Certified Reference Materials
7. Documentation
8. Internal and External Audits
9. Proficiency Testing
10. External Agency Accreditation
11. Benchmarking
12. Ethics

Training
As Mark Twain reportedly said, "Training is everything. Cauliflower is nothing but a
cabbage with a college education." The point is that without adequate training, staff cannot
produce the best results. Training courses should include periodic refresher and new technology
courses to improve the technical ability of staff members. These courses will benefit the staff
members by helping them to think through the analysis rather than mechanically doing the test,
to identify ways to improve the methodology, to obtain better precision and accuracy, and to
improve the turnaround time.
Some of the areas in which a laboratory staff member must be fully trained include safety,
data security, laboratory instrumentation, test methods used, calibration protocols, statistical
quality assurance, use of certified reference materials, long term analytical needs and goals, and
ethical behavior. In today's culture, it is still up to the supervisors and management to filter the
Zen attitude down to the working level people. The very fundamental first step toward obtaining
a perfect analysis is through staff training. Dr. Derek Bok, former president of Harvard
University once commented, "If you think education is expensive, try ignorance."

Sampling
Obviously the first critical step in any analytical sequence is the integrity and validity of a
sample being analyzed. More often than not this sampling step does not involve actual laboratory

NADKARNI ON PERFECT ANALYSIS

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

staff; usually the chain of custody for a sample starts with the receipt of the sample in the
laboratory. Once the laboratory acquires the sample, however, it is the laboratory's responsibility
to have a system for unique identification of each sample, sample handling, storage and retention
procedures, as well as safe disposal procedures. Identification of the population from which the
sample is to be obtained, selection and withdrawal o f valid gross samples of this population, and
reduction of each gross sample to a laboratory sample suitable for the analytical technique to be
used are some of the key steps to be considered in obtaining a representative sample for analysis
[2]. Equally important is documented chain of custody procedures to authenticate and maintain
the sample integrity. Several ASTM standards deal with sampling aspects for the analysis of
petroleum products and lubricants:
*
9
9
o
9

D 4057:
D 4177:
D 4840:
D 5842:
D 5854:

Manual Sampling of Petroleum and Petroleum Products


Automatic Sampling of Petroleum and Petroleum Products
Sampling Chain of Custody Procedures
Sampling and Handling of Fuels for Volatility Measurements
Mixing and Handling of Liquid Samples of Petroleum and Petroleum Products

Additionally, some ASTM standards give instructions for specific sampling requirements for
specific analytical tests. Attention must be paid to these caveats to obtain reliable test results.
Contamination Control

Gross contamination of the sample in any analysis and in particular for trace analysis is a
serious problem which, if unchecked, will completely negate the validity of the analytical results.
The problem can become particularly insidious as one is working in the range of ppm and subppm levels of analytes. Contamination from particulates in the air, impurities in reagents, trace
elements from the sample containers as well as glass- or plastic-ware used during analyses are all
potential sources of contamination [3]. An accompanying "blank" sample used throughout the
analysis sequence may or may not accurately measure the extent of contamination, since such
contamination from air or glassware, etc. may not be uniformly present when in contact with the
blank and a real sample. The point is that both a blank determination and a rigorous protocol for
contamination control in the laboratory are essential for obtaining perfect results, particularly in
the area of trace analysis. An excellent source book for discussion on contamination control is
given in [4].
Calibration or Verification

Virtually all analytical test methods require some form of calibration or verification before
actual samples are analyzed. Different test methods require different calibration intervals. Thus,
a decision about appropriate calibration frequency must be made on a ease by case basis. There
is a tendency among many laboratories to do the bare minimum calibrations similar to their
approach toward quality control requirements. This is not the way to achieve superior
performance. Moreover, if an instrument is out-of-calibration, under no circumstances can data
from that instrument be reported to the customers.
Appropriate calibration standards must be utilized during analysis. A wide variety of such
standards are available from commercial sources, NIST, etc. Many laboratories have capabilities

NADKARNI ON PERFECT ANALYSIS

of preparing reliable in-house standards. Calibration standards identical to the samples being
analyzed would be ideal, but failing that, at least some type of standards must be used to validate
the analytical sequence. In physical measurements this is usually achievable, but it is often
difficult or sometimes almost impossible in chemical measurements. Even the effects of small
deviations from matrix match and analyte concentration level may need to be considered and
evaluated on the basis of theoretical or experimental evidence. Sometimes the use of standard
additions technique to calibrate the measurement system is a possibility. But because an
artificially added analyte may not necessarily respond in the same manner as a naturally
occurring analyte, this approach may not be always valid, particularly in speciation work9
An important aspect of calibration is the decision on calibration intervals, i.e., the maximum
period between successive recalibrations. Two basic and opposing considerations are involved:
the risk of being out of tolerance at any time of use and the cost in time and effort. The former
should be the major concern because of the dilemma of what to do with the data obtained during
the interval between the last known in and the first known out o f calibration. However, an overly
conservative approach could be prohibitively expensive. A realistic schedule should reduce the
risk of the former without undue cost and disruption to work schedules. The factors that need to
be considered in a realistic schedule include:
9

9
9
9
9
9
9
9

Accuracy requirement for the measured data


Level of risk involved
Experience of the laboratory in use o f the equipment or methodology
Experience of the measurement community
Manufacturer's recommendations
External requirements for acceptability o f data
Cost of calibration and quality control

Quality control measurements can help a great deal in deciding what calibration frequency
intervals should be used.

Test Method Details


A laboratory must have fully documented test methods that are used for analyses, and the
staff members must be competent in the details in each test method that they will be using for
analyses. Experience has shown that a major source o f analytical error is deviation from the
prescribed standard test method, whether intentional or inadvertent. Most of the time the details
given in a standard test method are there for a purpose. A laboratory wishing to deviate from the
standard test method must document the deviation and show that the modified version produces
statistically equivalent or (preferably) better results in terms o f precision and accuracy.
In most laboratories, staff members periodically and sometimes frequently change; hence, it
is important to have a system in place for periodic checking that the laboratory practice is indeed
in conformance with the test method requirements9

Statistical Quality Control and Assurance


One cannot control what one cannot measure, and one cannot improve what one cannot
control. Every measurement system is beset with variation and noise, and the only way to control

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

and reduce variation is by identifying its cause, establishing its extent, and interpreting what it is
indicating. Variability arises because no two things are absolutely alike either in nature or in
laboratory measurements. All one can try to do is to minimize it as much as possible. The
primary step in measuring variance is the use of statistical quality control (SQC) or assurance
(SQA) charts. Quality assurance must be viewed as an integral part of a complete analytical
sequence and not as an added burden with additional costs.
Calibration("--) Sample Analysis<---) SQA
The only way to prove data integrity and reliability is through SQA. The frequency of quality
control (QC) analysis will depend on the type of analysis, the instrument involved, stability of
the measurement process, importance of decisions based on the test results, and magnitude of
changes between batches. The same considerations described in the calibration section above
need to be involved in making the decision on SQA frequency. But at least one QC standard
must be analyzed with each set of samples to be analyzed. Preferably one QC standard each
should be analyzed bracketing a series of sample analyses. Ideally, a few QC standards should be
interspersed with samples to continuously monitor the data quality. Indeed there are some ASTM
test methods where such protocol is mandatory, e.g., D 4951 and D 5185 ICP-AES methods for
metals analysis. A general rule of thumb used in many laboratories is one QC sample per 5-10
actual samples. Although there are statistical protocols for reducing the QC frequency, this is
overridden if a customer (and many do) requires QC accompanying their sample analyses. If a
QC analysis indicates poor performance, immediate remedial action must be taken before
continuing the analysis or reporting the data to the customers.
Control Charts--A picture is worth a thousand words, and this is definitely true of the QC
data as well. A chart based on statistical control limits must be plotted and interpreted on time. A
control chart by itself can only indicate that there is an upset and not the cause of the upset.
Laboratory staff must investigate, identify, and eliminate the cause(s) of the upset and bring the
process back to a state of statistical control. If a control chart is merely plotted and filed away
without taking action when required it is simply a waste of time and effort.
There are a number of statistical run rules which are used to indicate out-of-control
situations. Some laboratories use only one rule of a data point beyond 3 sigmas. This in itself
may not be adequate, since it will ignore several short term and more importantly long term
indications of system instability.
Based on the QC data, monthly or quarterly laboratory capability should be calculated to
compare the laboratory precision with that given in the standard test method where available.
One way of estimating lab capability is the the equation below:
Test Performance Index = Laboratory Standard Deviation / Test Method Repeatability
Further discussion on the numerical criteria to be used for evaluating TPIs can be found in
ASTM standard D 6792. Some laboratories prefer to compare their standard deviation against the
test method reproducibility or to use an inverse calculation from that given above. In either case
an individual best laboratory will consistently produce a superior precision compared to the
industry average precision. Such periodic performance feedback should be a key feature of any
continuous improvement program.
Such a continuous improvement (CI) program can be initiated at any level and may include
obtaining a better instrument or changing to a better alternate test method. Since the lab worker

NADKARNI ON PERFECT ANALYSIS

staff is usually busy in day-to-day analyses, the initiative for such CI program usually needs to
come from supervisory or managerial staff. Such a program can continue indefinitely looking at
various ways of improving diverse areas of laboratory operations. A project ends, but a process
continues!

Use of Certified Reference Materials


Where available, reference materials need to be used for calibration or quality control. The
National Institute of Standards and Technology (NIST) and many commercial sources supply
such materials. Many high performing technology laboratories are capable of preparing their
own reference materials using the same approach used by NIST in certifying their Standard
Reference Materials (SRM). The calibration reference materials (RM) should be traceable to
national standards, and the traceability must be preserved in the laboratory's documentation
system.
It is not a good idea to use the same material for both a calibration standard and quality
control. The latter should be similar in matrix to the type of samples being analyzed, although
this is not always possible due to the lack of availability of suitable certified materials. At least
for QC it is easy enough to obtain a reasonably large quantity of plant product, analyze it by the
tests of interest multiple times, and calculate the average value and the sigma limits to initiate the
control chart.

Documentation
Laboratory analytical reports must contain all information necessary for a customer to
understand or interpret the results. This may include the unique laboratory sample ID, equivalent
customer sample identification, test methods used and any modifications done to them during
analysis, actual analytical results, supporting QC results where available, signature of responsible
authorized laboratory staff, etc. Records of calibration or QC records must be maintained for an
adequate period of time.
Errors in Data Reporting--Reporting blunders occur more frequently than one might think.
We have often seen the laboratories taking part in the ASTM interlaboratory erosscheck
programs reporting the data without paying adequate attention to the report formatting. A perfect
analysis may have been done, but if the results of this perfect analysis are reported incorrectly,
nothing is gained. We have repeatedly found that the numbers are often transposed, or wrong
decimal places are used. When we check back with these laboratories, almost always it turns out
that there was indeed an error in reporting. Some of the glaringly obvious examples that we have
come across include:
9 The total metal content is <100 ppm, yet the ash (D 482) is reported as 3 m%.
9 The ash (D 482) and sulfated ash (D 874) results are widely different by an order of
magnitude.
9 The kinematic viscosity (D 445) of a fuel sample is reported as 6300 cst @ 100~ that
would make this a practically solid sample.
9 One laboratory reported 1000~ as the flash point (D 93) of a fuel sample!
9 The pour point (D 97) of a fuel oil sample was reported as 54~ that would make it a
solid oil sample at room temperature.

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

9
9

The sulfur content (D 2622) of a reformulated gasoline sample is reported as 94 %! This


should have been 94 ppm.
The water content (D 6304) of a lubricating oil sample is reported as 9 %. This should
have been 9 ppm.

All of these errors could have been prevented easily by reviewing the data before reporting. One
would not send a defective product from a plant to a customer. Why then send the defective data
out? It takes only a few moments to go over the final report form to check for such errors; a
laboratory's credibility would be much better for having done so.

Internal and External Audits


Every laboratory striving for perfection needs to do periodic audits to check that its quality
systems are still working properly. Audits o f test methods should be conducted (perhaps
annually) to confirm adherence to the documented test methods. The performance of the entire
test should be observed and checked against the official specified test method. Many errors in
results are derived from taking liberties with the test method's detailed requirements.
It needs to be recognized that the ISO 9000 audits alone by outside registrars do not
adequately cover the technical aspects of a laboratory performance. It is imperative that internal
audits be done by the laboratory staff since they are far more familiar with the required technical
details of each test that the laboratory performs. Through such thorough reviews, a laboratory
can eliminate defects, improve on its strengths, and become a perfect laboratory satisfying
customer expectations. Findings and recommendations of internal or external audits must be
promptly reviewed and acted upon, otherwise there is little point in spending efforts in doing
audits.

Proficiency Testing
A laboratory needs to participate in relevant crosschecks organized within the company
circuit, with its customers if requested, or with the industry organizations. A completely unbiased
assessment of a laboratory's capability for precision and accuracy can only be judged by a blind
crosscheck. If the laboratory results are equivalent to those of other laboratories in the industry, it
assures the laboratory itself and its customers that the laboratory is producing reliable results. On
the other hand, if the results are less than satisfactory, it also helps the laboratory to correct its
operation, which may be deficient in some aspect. Participating in a crosscheck without
following up on deficiencies if detected is a waste of resources.
One such program widely used in the oil industry is described by Nadkami and Bover [5].
Participation in proficiency testing should not be considered as a substitute for in-house quality
control and vice versa. These are two independent and necessary activities for a well-managed
laboratory.

Laboratory Accreditation
Laboratory accreditation in the U.S. is handled by more than 150 accrediting private and
quasi-government bodies. Thus, there is a bewildering array to choose from as to which is the
best accreditation for a laboratory. This will depend to some extent on the industry and the
customers with which a laboratory does business. The most well known program is ISO 9000

NADKARNI ON PERFECT ANALYSIS

registration. Often when a chemical plant or a refinery gets ISO 9000 registered, the testing
laboratory associated with that site is also registered. It must be recognized however that having
an ISO 9000 registration does not necessarily guarantee the best laboratory data unless
supplemented by additional internal quality protocols [6,7].
Two other meaningful accreditation programs for the testing laboratories are by the
American Association for Laboratory Accreditation (A2LA) [8] and ISO Guide 17025 [9]. There
is a similar National Measurement Accreditation Service (NAMAS) in the United Kingdom and
parts of Europe, which is administered by the U.K. National Physical Laboratory [10]. An
advantage of these three accreditation schemes is that they involve actual test method
performance, laboratory quality system, and other issues vital to a credible laboratory's
performance. Given the widespread utilization of such accreditation schemes, many customers
will forgo their own quality audits of contractor laboratories if the latter have such external
certifications.

Benchmarking
Benchmarking is the search for those best practices that will lead to a company's superior
performance, and it incorporates these practices in one's own organization. To become the best,
one must learn from the best. This learning should not be limited only to the organizations in
one's industries; companies outside a specific industry may also have superior standards which
can be utilized beneficially. Ideas should be shamelessly stolen from the best, provided they are
not company proprietary materials.
Japanese people call benchmarking "Dontotsu," literally meaning "striving for the best of the
best." Xerox Corporation was a pioneer in industrial benchmarking. An excellent text on this
subject has been published by Xerox [1 t]. Some Baldrige National Quality Award winning
companies incorporate benchmarking as an essential tool in their quality management process.
Perfect laboratories also need to look at their counterparts to fred out how to improve themselves
to the level o f the best laboratories.
Many laboratories calculate the so-called Laboratory Productivity Index (LPI), defined as the
total tests/total laboratory workers per month or per quarter, etc. However, experience shows that
these indices may vary widely from laboratory to laboratory. It is not certain that LPI is a reliable
measure for interlaboratory comparison, since:
a. Different tests need different amounts of time to complete.
b. The same analysis may be done automatically by analyzers in some laboratories and
manually in others.
c. The test requirements can be different depending on the types of products analyzed.
d. The clientele and the test slate can be different for research and production laboratories.
LPI may be useful to track laboratory productivity in one site but needs caution when comparing
one laboratory against another.

Ethics
One would think that it is a simple fact that the results obtained in the laboratory should be
reported honestly; however, several incidents reported in last few years make one wonder

10

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

whether ethics should be a mandatory training course in colleges and laboratories for analytical
chemists!
In April 1996, the U.S. Department of Justice t'med a major additives company $4.75
million for providing additive packages to the government falsely certifying that they met
the military product specifications and passed specified testing.
A Linden, NJ commercial laboratory was fined $1 million in 2001 for doctoring the
laboratory reports that showed that they met EPA's Clean Air Act standards for cleaner
burning fuels. The company was also put on probation for three years, and its former
president was convicted of fraud.
Some of the causes of the erroneous and false data can be ascribed to sloppiness in analysis,
lack of training, use of improper techniques, inappropriate data manipulation, taking shortcuts in
analysis, failure to follow prescribed steps in the test methods, improper calibration of
instruments, etc. In a practice called "dry labbing," a worker merely copies the data from a
previous sample and submits it as a new sample analysis with a new identification number.
Commercial, business, and regulatory decisions based on such inaccurate data can often lead to
wrong decisions for all affected parties. Nancy Wentworth, Director of quality staff at EPA's
Office of Environmental Information, sums up advice to laboratories as: Get the right data; get
the data right; and keep the data right [12].

Laboratory QualityImprovementthrough Quality Management


Using all or many steps outlined above, is it possible to achieve ultimate laboratory
perfection? Is such a thing as a "perfect laboratory" or "perfect analysis" possible at all? The
answer is in the affirmative if efforts are spent on the underlying foundation of good laboratory
practices. These are:
9
9
9
9

The tests must be conducted exactly as written without any deviations.


The calibrations, if part of a test, must be carried out properly; if necessary, each time the
test is carried out.
Quality control must be an integral part of the analytical sequence. The QC data must be
plotted, the trends analyzed, and corrective actions taken when so indicated.
At least annually, all tests performed in a laboratory must be audited by personnel other
than those performing these tests or managing the laboratory to check that all details of
the test method are properly followed. Any deviations found must be corrected promptly.

Performance of a group of laboratories which are well-controlled in terms of the above


requirements has been shown to be superior to that of other industry laboratories based on the inhouse repeatability and their performance in ASTM interlaboratory crosscheck programs. Their
repeatability in the laboratory is even superior to that advocated in the ASTM test methods
themselves. Such precise data have implications beyond simple good analysis. A properly
calculated product specification is usually set based on both the manufacturing and analytical
testing variance.
Total sigma = (Manufacturing sigma2 + Analytical sigma2) ~

NADKARNI ON PERFECT ANALYSIS

11

If a testing laboratory is being managed to its top performance levels, often the analytical
sigma becomes virtually insignificant compared to the manufacturing variability resulting in the
width of a product specification being dependent almost solely on the manufacturing variance.
Thus, a superior analytical precision helps in both manufacturing a reliable product, eliminating
quality complaints particularly in regards to the data, and keeping the customers happy and loyal.

Lessons from Baldrige Award Winners


Since the inception of the Malcolm Baldrige National Quality Award in 1988, a number of
U.S. organizations has been recognized by the U.S. Department of Commerce as the best in total
quality management in the country. There are a number of lessons to be learned from these
winners which are equally valid in the laboratory environment to convert it into a perfect
laboratory producing perfect analysis. The basic strategies among the Baldrige winners have
been summarized as [13]:
I. Leadership and management commitment
2. Total customer delight
3. Long term efforts
4. Teamwork with employees, suppliers, and customers
5. Employee involvement and satisfaction
6. Continuing training for all
7. Statistical measurement of progress
8. Continuous quality improvement
9. Benchmarking against the best in class
10. Total and open communications with employees, suppliers, and customers

Concluding Remarks
A paradigm shift in the mindset is necessary at all levels of laboratory staff to produce a
superior product. Think of a ham and egg breakfast; the chicken is involved, but the pig is
committed. You have to be the latter. Nor are the results an instant pudding. Again, think of a
bamboo farmer; the sapling is watered for 4 years, and suddenly at the end the tree shoots up to
60 feet in 90 days! A constancy of purpose, clear vision, a prepared mind, commitment to long
term pursuit of goals, and dedication to all technical aspects of pillars of good analysis are what
will differentiate an ideal laboratory from an also-ran laboratory. The choice is yours to make.
As W. Edwards Deming wrote, "Charles Darwin's law of survival of the fittest, and that the unfit
do not survive holds in free enterprise as well as in natural selection. It is a creel and unrelenting
law. Actually, the problem will solve itself. The only survivors will be the companies with
constancy of purpose for quality, productivity, and service" [14].
Similar to Robert Pirsig's hypothesis of Zen and the art of motorcycle maintenance [1], a
mental exercise is at the heart of managing a "perfect laboratory". The mind needs to be prepared
to perform a flawless analysis and whatever is required to make it so; otherwise a perfect
analysis is simply not possible. Determination to excel, mental discipline to stay the course,
willingness to overcome inertia and resistance of others, and focus on producing perfect analysis
at all levels of laboratory staff are essential. Improvement in laboratory quality performance is a
process, not a project. A project ends, but a process continues.

12

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Inherently we all know the difference between a good laboratory and a not-so-good
laboratory. The difference is like the difference between a star in a play and the understudy. The
not-so-good laboratory (and the understudy) do their jobs and produce results, but a good
laboratory (and the star) are needed to produce exceptionally perfect results.
The technical quality management of an analytical laboratory can be a hard pill to swallow,
but it need not be. It can help prevent excessive waste in time and money, improve productivity
by eliminating duplication and waste, and most importantly lead to increased customer
satisfaction and increased profits. Unless strong and sustained measures are taken, improvement
in data quality cannot be achieved or maintained. No pain, no gain!
Only a prepared mind can meet the challenge. The bottom line is that one needs to have a
dream, a passion, and a vision to want their laboratory to be the best there is - the flawless
perfect laboratory. Unless one nurtures and pursues that vision, as Yogi Berra said, "You can't
be there if you don't know where you are going".
Additional information on this subject may be found in Refs [15-18].
References

[1]
[2]
[3]
[4]
[5]

[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[ 15]

Pirsig, R. M., "Zen and the Art of Motorcycle Maintenance," Williams Morrow and Co.,
Inc., New York, 1974.
Kratochvil, B. and Taylor, J. K., "Sampling for Chemical Analysis," Analytical Chemistry,
Vol. 53, No. 8, 1981, 924A-938A.
Mitchell, J. W., "Ultrapurity in Trace Analysis," Analytical Chemistry, Vol. 45, No. 6,
1973,492A-500A.
Zief, M. and Mitchell, J. W., "Contamination Control in Trace Element Analysis," John
Wiley & Sons, New York, 1976.
Nadkarni, R. A. and Bover W. J., "Bias Management and Continuous Improvements
through Committee D02's Proficiency Testing," ASTM Standardization News, Vol. 32, No.
6, 2004, pp. 36--39.
Nadkami, R. A., "ISO 9000 Quality Management Standards for Chemical and Process
Industries," Analytical Chemistry, Vol. 65, 1993, 387A-395A.
Nadkami, R. A., "What Every Chemical Laboratory Should Know About the ISO9000
Quality Standards," Analytical Chemistry, Vol. 66, 1994, 10(3-15G.
Locke, J. W., "Quality System for Testing Laboratories," A2LA Reprint, 1992.
de Leemput, Van, "Testing Their Calibre - A Standard to Ensure Laboratories are up to
Standard," ISO Bulletin, 16-18, June 2000.
Broderick, B. E., "Laboratory Accreditation: The Operation of an Established Scheme,"
Mikrochimica Acta, III, 17-21, 1991.
Camp, R. C., "Benchmarking," ASQC Quality Press, Milwaukee, WI, 1989.
Hogue, C., "Ferreting out Erroneous Data," Chemical and Engineering News, pp. 5 and
49-50, April 2002.
Nadkami. R. A., "A Not-So-Secret Recipe for Successful Total Quality Management,"
Quality Progress, 91-96, November 1995.
Deming, W. Edwards, "Out of Crisis," MIT Press, Cambridge, MA, 1982.
"Quality Assurance for the Chemical and Process Industries: A Manual of Good Practices,"
American Society for Quality, Milwaukee, WI, 1987.

NADKARNI ON PERFECT ANALYSIS

13

[16] Nadkarni, R. A., "The Quest for Quality in the Laboratory," Analytical Chemistry, Vol. 63,
No. 13, 1991, 675A-682A.
[17] Nadkarni, R. A., "Quality Management: A Must for the Labs Too," Today's Chemist at
Work, 34-38, December 1996.
[18] ASTM D 6792-04: Standard Guide for Quality System in Petroleum Products and
Petroleum Testing Laboratories, ASTM International, West Conshohocken, PA.

ATOMIC EMISSION SPECTROSCOPY

Journal of ASTM International, November/December 2005, Vol. 2, No. 10


Paper ID JAI12965
Available online at www.astm.org

Chris C. Onyeso, Ph.D.t

Analysis of Gasoline and Diesel Fuel Samples by Inductively


Coupled Plasma Atomic Emission Spectrometry (ICP-AES),
Using Pneumatic Nebulizer and Standard Spray Chamber
ABSTRACT: A method has been developed for accurate quantitative determination of additive elements

and wear metals in gasoline and diesel fuel in the concentration range of 0 to 50 ppm using inductively
coupled plasma atomic emission spectrometry (ICP-AES). This method requires an ICP capable of
detecting metals at the 0.02 mg/kg level, and capable of linear calibration over the range of 0.05 to 20.0
mg/kg with a correlation coefficient of 0.9999 or better. No additional ICP-AES accessory or lengthy
sample preparation is required. The PerkinElmer Optima 4300 has been shownto meet this requirement.
KEYWORDS: gasoline, diesel fuel, additive elements, ICP, wear metals, and analysis.
Introduction

Trace metals in fuels, except in the case of additives, are usually undesirable and normally
occur in very low concentrations, requiring sensitive techniques for their determination. Also,
there is a need to determine the concentration of additive elements such as manganese (Mn) in
gasoline. The industry method currently available for the determination of manganese (Mn) in
gasoline is ASTM D 3831 [1], which requires bromine reduction of the gasoline followed by
measurement with Atomic Absorption Spectrometer (AA). This method has inherent problems
with high olefin content fuels and cannot accurately measure low concentrations of Mn in
gasoline.

Use of lCP-AES
ASTM test methods D4951 [2] and D5185 [3] are commonly used for the determination o f
additive elements and wear metals in fresh and used lubricating oils respectively. The problem
with the use of ICP for gasoline analysis is that gasoline is highly volatile and extinguishes the
ICP plasma [4,5]. A few examples of ICP analysis o f gasoline are found in the literature, but
these methods involve the use of expensive ICP accessories such as direct injection nebulizer [6],
ultrasonic nebulizer with micro-porous membrane desolvator [7], chilled spray chamber [8,9], a
thermostated condenser between the spray chamber and the plasma torch [10], or lengthy sample
preparation with possibility o f contamination, such as emulsification with surfaetants [ 11-14].
In this paper, an accurate method for determining the concentration of manganese and other
metals in gasoline and diesel fuel by ICP without the use of chilled spray chamber, direct
injection nebulizer, ultrasonic nebulizer with micro-porous membrane desolvator, thermostated
condenser, or emulsification is discussed.

Manuscript received 15 November 2004; accepted for publication 13 May 2005; published November2005.
Presented at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advancesand Future
Prospects on 6-8 December 2004 in Tampa, FL.
I Ethyl Corporation, Richmond,VA.
Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

17

18

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Analysis
Calibration standards are prepared from certified Conostan| standards, using kerosene
containing Yttrium internal standard as the solvent. The kerosene containing internal standard is
also used as the solvent for the gasoline and diesel fuel samples. The calibration standards and
the gasoline samples are aspirated into the ICP instrument. Each element present in the sample
emits light at discrete wavelengths. The intensity o f light at each element's wavelength is
compared against a calibration curve generated from standards containing metals at k n o w n
concentrations to provide quantitative determination. This method also outlines the preparation
of the instrument, samples, and standards used in the analysis. The method was originally
developed for the determination of manganese (Mn) in gasoline and diesel fuel, and later
extended to include additive elements and wear metals.

Instrument
9
9
9
9
9
9
9
9
9
9

PerkinElmer Optima 4300.


The appropriate wavelength (nm) for each element (analyte) was chosen.
Internal Standard is Yttrium (Y).
Quartz Torch (no slot) for organic solvents.
Quartz Torch (single slot) for aqueous solution.
Spray chamber - cyclonic spray chamber for organic solvent.
Scott spray chamber for aqueous solution.
Nebulizer - GemCone nebulizer for organic solvent.
Crossflow nebulizer for aqueous solution.
Injector Tube - Quartz injector tube (1.2 m m i.d.) for organic solvent - Sapphire injector
tube (2.0 mm i.d.).

Spectrometer
Resolution was set at High with three replicates and integration time o f five seconds.

Plasma Parameter
9
9

Plasma View: radial.


Peristatic Pump set at a flow rate o f 2.50 ml/min.

Calibration
9
9
9

The manganese calibration standards concentrations are in mg/kg (ppm).


The instrument was calibrated using the following standards: blank (0 ppm), 5 (5.2333
ppm), 10 (10.7836 ppm), 15 (15.6748 ppm), and 20 (20.5006 ppm).
A good calibration curve with a correlation coefficient o f 0.99996 was obtained (see Fig.
1 below).

Sample Analysis
After calibration, gasoline samples o f known Mn content (as added) were analyzed. Results
are shown in Table 1.

ONYESO ON PNEUMATIC NEBULIZER

19

FIG. 1--Calibration curve.


TABLE 1--Determination of Mn in gasoline samples by ICP.
Gasoline Sample
Given Cone.
Measured Cone.
Difference

(mg/kg)
1) Citgo
0
2) Motiva
0
3) Philips
10
4) Philips
2
5) Philips
6
6) Gasoline from unknown source
18
7) Gasoline from unknown source . . . . . . 8.3

(mg/kg)
0
0
10.3
1.8
6.4
19
9.2

0
0
0.3
-0.2
0.4
1.0
0.9

Sample Preparation
Samples were prepared by diluting 1.0 g of gasoline or diesel fuel with 99.0 g of kerosene
containing Yttrium (Y) internal standard.
Example: Gasoline sample wt. = 1.0601 g
Kerosene wt. = 98.9727 g
Gasoline is very volatile; the weighing of the One,gram and the addition of kerosene must be
done very quickly to avoid loss of weighed gasoline by evaporation. Weighing the kerosene first
and then adding in the gasoline has been found to reduce gasoline loss.
Based on accurate results obtained on the gasoline samples with known Mn content (the
gasoline samples were spiked with a known amount of Mn), many gasoline and diesel fuel
samples were analyzed for Mn content and the ICP results (Alton) compared with those obtained
by AA at a contract Lab (Tables 2 and 3). The target values are the spiked values.

20

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

......
TABLE 2--Determination of Mn in diesel fuel.
Sample Type
Contract Lab. D3831
Alton R & D
Target Value
(AA) mg/kg
ICP mg/kg
mg/kg
Diesel 1: Run 1
2.7
2.3
Run2
1.9
3.1
Average
2.3
2.7
3.0
Diesel 2: Run 1
2.5
3.0
Run 2
3.2
3.1
Average
2.9
3.1
4.5
Diesel 3: Run 1
4.3
4.3
Run 2
4.4
4.3
Average
4.4
4.3
6.0
Diesel 4: Run 1
0.0
0.0
0.0
0.0
Average
0.0
0.0
0.0
TABLE 3--Determination of Mn in gasoline samples.
Sample
Contract Lab
Alton R & D
Target Value
D3831 mg/kg
ICP mg/kg
mg/kg
Gasoline 1: Run 1
0.0
0.0
Run 2
0.0
0.0
Average
0.0
0.0
0.0
Gasoline 2: Run 1
5.6
7.0
4.5
6.0
Average
5.1
6.5
8.0
Gasoline 3: Run 1
14.7
16.2
Run 2
12.4
14.7
Average
13.6
15.5
18
Gasoline 4: Run 1
20.7
22.9
Run 2
21.0
19.2
Average
20.8
21.1
25
Gasoline 5: Run 1
36.2
31.0
Run2
33.3
31.8
Average
34.8
31~.4.
40.0
In six out of the nine diesel and gasoline samples analyzed, the ICP results were closer to the
target manganese values than the AA results. The differences seen between AA and ICP results
are not unreasonable when one considers that the data sets were obtained using two different
analytical methods in two different laboratories.

Precision
The manganese method has a repeatability of 0.24 mg/kg. At 12.0 mg/kg Mn level, the
precision of the method is +/- 1.16 mg/kg. Diesel fuel was diluted 100-fold to keep the method
consistent; detection limits and precision could be better for diesel at lower dilutions since diesel
is not as volatile as gasoline.

ONYESO ON PNEUMATIC NEBULIZER

21

Expansion of the Method to Include Additive Elements and Wear Metals


Having successfully used this method to determine Mn in gasoline and diesel fuel, it was
desirable to extend the method to include determination of additive and wear metals. The method
was calibrated with a Conostan| standard containing the desired elements, and a good
calibration curve was obtained as shown below (Table 4).
TABLE 4--Multi-element calibration and limits of detection.
Analvte
Corr. Coef.
Limit of Detection mg/kg
Mn (293.305)
0.999997
0.9
A1 (308.215)
0.999817
2.1
Ca (315.887)
0.999962
1.8
Cu (324.752)
0.999938
1.3
Fe (238.204)
0.999937
1.2
Mg (285.213)
0.999987
0.7
Pb (217.000)
0.999972
3.1
V (290.880)
0.999975
0.6
Zn (206.200)
0.999617
3.2
With this calibration, gasoline, furnace oil, and diesel fuel samples were analyzed in
duplicate. Table 5 shows elements detected in these samples. The minimum detection limits for
the analytes (elements) were determined at 100-fold dilution and are shown in Table 4 above.
TABLE 5--Elements detected in samples.
Sample
Mn (mg/kg)
Run 1
Run 2
FBG-10513 (gasoline)
17
17
R05905 (gasoline)
22
21
EC194995-2 (gasoline)
12
11
EC194995-1 (gasoline)
21
20
Furnace Oil
0
0
Low Sulfur Diesel
0
0
R06241201 Diesel
20
21

Avg
17
21.5
11.5
20.5
0
0
20.5

Zero results indicate concentrations below the minimum detection limit. The gasoline, diesel
fuel, and furnace oil samples listed above were analyzed for the calibrated elements listed in
Table 4. A1, Ca, Cu, Fe, Mg, Pb, V, and Zn were below the method detection limit in Furnace
Oil, Low Sulfur Diesel without additive, diesel fuel, and four gasoline samples. Only Mn was
detected in those gasoline and diesel fuel samples with additives. The above results have shown
that this method can be used to determine additive and wear metals in gasoline, diesel fuel, and
furnace oil in concentrations that are within the detection limit of the instrument.

Aqueous Analysis
Another way that gasoline can be analyzed by conventional ICP without the use of additional
accessories or lengthy sample preparation is to analyze it in aqueous solution. For gasoline and

22

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

diesel fuel samples containing solid impurities or particulates that are insoluble in kerosene,
aqueous analysis is a very viable option. Determination of metals after total destruction o f
organic matter is a very reliable procedure for petroleum and petrochemical samples.
The gasoline samples were digested with nitric acid in a microwave digestion oven. The
resulting acidic solution was diluted with de-ionized/distilled water down to about 5-10 % HNO3
and analyzed by aqueous ICP method (Table 6). Sample dilution in the aqueous analysis is about
1 to 50, unlike organic analysis, which is 1 to 100.
TABLE 6--Aqueous ICP analysis of gasoline ......
Gasoline Samples (results in mg/kg)*
Elements Gasoline 1 Gasoline 2 Gasoline 3 Gasoline 4 Gasoline 5
Fe
3
7
4
25
23
Mo
0
0
1
0
0
Zn
0
0
0
0
0
Pb
3
0
4
15
0
B
0
0
0
4
1
Si
11
22
29
31
9
*Sn, Cr, P, Ni, Ba, and Mn were below the detectionlimit in these six gasoline samples.

Gasoline 6
7
4
23
3
1
12

The differences between the results shown for Fe, Pb, Ca, and Si in Tables 5 and 6 are due to
the fact that the gasoline samples are different, the dilution factors are different, and the matrices
are different (kerosene versus water).
Conclusions

An ICP method for the accurate determination of additive elements and wear metals in
gasoline and diesel fuel has been developed. The gasoline and diesel sampleswere dissolved in
kerosene with yttrium internal standard and analyzed by ICP without any additional accessory.
Samples were bracketed with matrix matched standards, and an internal standard was also used.
Based on good calibration curves (correlation coefficients of 0.9999 and higher), lack of spectral
interference, and accurate results obtained on gasoline samples with known amount of Mn, the
results for the samples with an unknown amount of metals must be correct. For the organic ICP
method, the standard deviations for four measurements of each of the twelve (12) gasoline
samples ranged from 0.00 to 0.21. This study has demonstrated the ease of gasoline and diesel
fuel analyses both in organic and aqueous matrices by ICP without the use of additional
expensive accessories, such as direct injection nebulizer, ultrasonic nebulizer with micro-porous
membrane desolvator, chilled spray chamber, a thermostated condenser between the spray
chamber, and the plasma torch or lengthy sample preparation.

Acknowledgments
Thanks go to Ken Garelick and Penny Hurt for their assistance in the preparation and analysis
of the samples and to Afton Chemical Company.

ONYESO ON PNEUMATIC NEBULIZER

23

References

[1]
[2]
[3]

[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]

ASTM Standard D 3831, "Standard Test Method for Manganese in Gasoline by Atomic
Absorption Spectrometer," ASTM International, West Conshohocken, PA.
ASTM Standard D 4951-02, "Standard Method for Determination of Additive Elements in
Lubricating Oils by ICP-AES," ASTM International, West Conshohocken, PA.
ASTM Standard D 5185-02, "Standard Test Method for Determination of Additive
Elements, Wear Metals and Contaminants in Used Lubricating Oils and Determination of
Selected Elements in Base Oils by ICP-AES," ASTM International, West Conshohocken,
PA.
Botto, R. I., Spectrochim. Acta, Part B, Vol. 42, 1987, p. 181.
Kreuning, G. and Maessen, F. J. M. J., Spectrochim. Acta, Part B, Vol. 44, 1989, p. 367.
Botto, R. I., Canadian Journal of Analytical Sciences and Spectroscopy, Vol. 47, No. 1,
2002, pp. 1-13.
Botto, R. I., J AnaL At. Spectrom., Vol. 8, No. 1, 1993, pp. 51-57.
Ryan, A., Chemistry in New Zealand, Vol. 61, No. 6, 1997, pp. 9-11.
Hausler, D. W. and Taylor, L. T., AnaL Chem., Vol. 53, 1981, p. 1223.
Maessen, F. J. M. J., Seeverens, P. J. H., and Kreuning, G., Spectrochim. Acta, Part B, Vol.
39, 1984, p. 1171.
Saint Pierre, T. D., Dias, L. F., Pozebon, D., Aucelio, R. Q., Curtius, A. J., Welz, B.,
Spectrochim. Acta, Part B, 2002, 57B(12).
Brenner, L B., Zander, A., Kim, S., Shkolnik, J. J., AnaL At. Spectrom., Vol. 11, No. 2,
1996, pp. 91-97.
A1-Swaidan, H. M., Atomic Spectroscopy, Vol. 14, No. 6, 1993, pp. 170-173.
Reimer, R. A. and Miyazaki, A., Journal of Analytical Sciences, Vol. 9, No. 1, 1993, pp.
157-159.

Journal of ASTM International, September2005, Vol. 2, No. 8


Paper ID JAI12966
Available online at www.astm.org
B r i a n S. F o x 1

Elemental Analysis of Lubricating Grease by Inductively


Coupled Plasma Atomic Emission Spectrometry (ICP-AES)
ABSTRACT: Lubricating grease is a mixture with complex properties that poses far more difficult
analytical challenges than do its individual components. At many petroleum testing laboratories,
Inductively Coupled Plasma Atomic Emission Spectroscopy, or ICP-AES, is used for the analysis of
additive elementsand wear metals in oil. However,it may also be used to measure the concentrationof
both additive metals and contaminantsin lubricatinggrease. However,unlike some of the other matrices
that are routinely analyzed, grease cannot be simply diluted for direct sample introductioninto the
instrument without some form of acid digestion. This paper will discuss how ICP-AES is used as a
grease analysistool. Several samplepreparationschemes will be covered, includingclassicalsulfatedash
on a hot plate, microwaveassisted dry ash, and microwaveassisted acid digestion. Comparisondata from
different digestiontechniqueswill illustratepotentialproblemsthat may be encounteredby the analystin
each of these methods. The advantages of closed-vesselmicrowavedigestion for the analysisof elements
that are often volatilizedtrader normal digestionconditionswill be discussed.
KEYWORDS: lubricatinggrease, additives,ICP-AES

Introduction
In 1885, Grant McCargo formulated one of the earliest petroleum grease mixtures b y
blending oil and soap together. Today, lubricating grease finds use in almost all bearings around
the world. Lubricating grease is composed o f approximately 90 % additized oil and soap or
other thickening agent [1]. In the absence of an oiling system, grease provides a film of
additized oil to protect the metal surfaces that are in contact with one another. The thickener in
the grease helps to keep it located in the bearing assembly, where it can do its job, Some of the
other additives in lubricating grease mirror the functions that they perform in typical lubricating
oil. In lubricating oil, the additive and wear metal concentrations are determined by several
atomic spectroscopy methods, among them being ASTM D 4951 or ASTM D 5185. However,
these tests have no counterparts for testing grease samples.
The focus of most standardized grease testing has been upon performance and appearance,
rather than determining the concentration of chemical components. In part, this is likely due to
the relative volume of grease demand when compared to other lubricants. But the primary
reason may be that grease is a rather difficult matrix with which to work from the perspective o f
the analytical laboratory. Lubricating oil may be diluted in a solvent and introduced directly into
an Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-AES) for rapid elemental
measurements. Grease, on the other hand, does not dissolve in any common laboratory solvent
that is compatible with ICP-AES.

Manuscriptreceived 18 November2004; accepted for publication7 March 2005; publishedSeptember2005.


Presented at ASTM Symposiumon ElementalAnalysisof Fuels and Lubricants:Recent Advancesand Future
Prospects on 6-8 December2004 in Tampa, FL.
i AdvancedResearch Associate, ExxonMobilResearch and Engineering,Paulsboro, NJ 08066.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohockon, PA 19428-2959.

24

FOX ON ANALYSIS OF LUBRICATING GREASE

25

In addition to ICP-AES, an X-ray fluorescence spectroscopic technique is sometimes used to


quantify elemental concentrations in grease samples. However, it suffers from two major
drawbacks. The first is the inability to detect two important lubricating grease components,
lithium and boron. Lithium compounds are used as a thickener in the vast majority o f lubricating
grease types on the market. Boron is sometimes used with lithium as a complexing agent. The
second drawback emerges in the case o f the analysis o f used grease obtained from a bearing. A
grease sample that is displaced b y fresh grease and obtained from outside a bearing housing m a y
contain dirt from the environment along with wear metals. The grease that is of primary interest
is that which is sampled from the load zone o f a disassembled bearing [2]. Obtained from this
source, there is often insufficient material available to fill the X-ray sample cup for a proper
analysis.
Despite the difficulty in working with grease samples, there are well over a dozen elements
o f interest to the lubricating grease supplier in the form o f additives, contaminants, and wear
metals. Table 1 shows some o f these elements in lubricating grease and the role that each may
play. Determining their concentrations can be an important aspect o f grease manufacture. In
addition, a reliable analysis technique can also assist in the process o f troubleshooting problems
with new and used grease in the field. For example, some o f the thickeners used in the various
types o f lubricating grease m a y be incompatible with one another [3]. Quantifying the signature
elements o f these thickeners could determine the level o f accidental mixing o f two incompatible
grease types.
TABLE 1--Potential
Element

Aluminum, A1
Antimony, Sb
Boron, B
Barium, Ba
Bismuth, Bi
Cadmium, Cd
Calcium, Ca
Chromium, Cr
Copper, Cu
Iron, Fe
Lead, Pb
Lithium, Li
Magnesium, Mg
Manganese, Mn
Molybdenum, Mo
Nickel, Ni
Phosphorous, P
Sulfur, S
Silicon, Si
Tin, Sn
Zinc, Zn

elementsfound in lubricatin~ ~rrease.

Potential Sources

Clay or other thickener


Extreme pressure additive
Complexing additive for thickener
Rust inhibitor
Extreme pressure additive
Wear metal, contaminant
Thickener and extreme pressure additive
Wear metal, contaminant
Wear metal, contaminant
Wear metal, clay thickener
Extreme pressure additive
Thickener
Clay thickener minor element
Clay thickener minor element
Molybdenum disulfide / extreme pressure additive
Wear metal
Additives such as zinc dithiophosphate(ZDTP)
Additives such as ZDTP
Contaminant,
clay thickener trace element
Contaminant
Additives such as ZDTP

Concentration Range

<20 to 3000 #g/g


<20 to 1000 #g/g
<20 to 1000 #g/g
<20 to 1000 #g/g
<10 to 1000 #g/g
Trace to 1000 #g/g
<20 to 3000 #g/g
Trace to 1000 #g/g
Trace to 1000 #g/g
<10 to 5000 #g/g
<20 to 1000/zg/g
<20 to 3000 #g/g
Trace to 100 #g/g
Trace to 100 #g/g
<20 to 3000 #g/g
Trace
<20 to 1000 #g/g
<20 to 1000 #g/g
<10 to 1000 #g/g
Trace to 100/zg/g
<20 to 1000/~/~

This paper examines various techniques that are available to prepare lubricating grease
samples for analysis b y ICP-AES. Each o f the techniques has specific advantages and m a y be
applied to different grease sample types, depending upon composition and expected levels o f
analyte concentration. Special attention will be given to the use o f high-pressure, microwave
assisted acid digestion. This technique provides the optimal solution for complete digestion o f

26

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

the organic matrix, while minimizing loss of additive elements that may be volatilized during
other preparation schemes.
Instrumentation

The measurements were performed using a Thermo Elemental ]RIS Inductively Coupled
Plasma Atomic Emission Spectrometer (ICP-AES). A 2 kW crystal-controlled radio frequency
(RE) generator operating at 27.12 MHz powers the plasma source. An Echelle optical system
with a 381-ram focal length diffracts the light from the plasma source before it is focused onto
the Charge Injected Device (CID) camera detector [4].
A fully de-mountable ABC TJA IRIS Radial torch was equipped with a Twister cyclonic
spray chamber for the non-hydrofluoric (HF) acid analyses. Both are available from Glass
Expansion Inc, Pocasset, Massachusetts, USA. When analyzing samples prepared with a dilute
HF acid solution, an alumina center tube and HF resistant spray chamber (Glass Expansion) were
substituted. The nebulizer used was a Teflon Mira Mist that is available from Burgener Research
Inc., Mississauga, Ontario, Canada.
The microwave digestion oven used for sample preparation is an MDS 2000. It is available
from CEM Corporation, North Carolina, USA. The microwave digestion vessels are high
pressure, Teflon lined, 45-mL Model 4782 bombs that are available from Parr Instrument
Company, Moline, Illinois, USA. For microwave assisted dry-ashing of lubricating grease
samples, a CEM Microwave Ashing System MAS 7000 was used. The open vessel microwave
digestion apparatus that was used for this work is a Prolabo Microdigest 401.
Reagents and Materials

All water used is 18 Mfl de-ionized water polished by a Simplicity ion exchange system
available from Millipore Corporation, USA. Trace Metal Grade sulfuric, nitric, and perchloric
acids that were used in the procedures are available from Fisher Scientific, Pittsburgh,
Pennsylvania, USA. Commercially prepared, multi-element, aqueous ICP-AES standards and
the blank acid matrix were obtained from VHG Labs, Inc., Manchester, New Hampshire, USA.
The multi-element, oil standards used for validation are available from Conostan Division,
Conoco Specialty Products, Ponca City, Oklahoma, USA.
For sulfated ash digestions, either 250-ml Vycor beakers or dishes may be used (Fisher
Scientific). The microwave digestion vessels used were high pressure, Teflon lined, 45-mL
Model 4782 bombs that are available from Parr Instrument Company, Moline, Illinois, USA.
The polypropylene volumetric flasks used to dilute digested samples to volume are available
from Fisher Scientific.
Procedure

The following descriptions of the digestion methods are simplified versions. Depending
upon sample difficulty, other steps may beadded to obtain a clear solution that is relatively free
of the organic matrix and un-dissolved material.

Sulfated Ash Digestion


The sample size depends upon expected analyte concentration. Approximately 1-2 g of

FOX ON ANALYSIS OF LUBRICATING GREASE

27

material are weighed into a Vycor container of suitable size. The sample is then charred on a hot
plate until it is reduced to about 0.5 g. A heat lamp is used to assist the process. Then 1-2 mL of
sulfuric acid are added to the residue and heated until the fumes cease to evolve. The charred
sample is placed in a muffle furnace at 530~ until the black color is gone. This typically takes
about 2 h, but it was originally felt that a sample may be left overnight to obtain complete
decomposition. Approximately 5 mL of hydrochloric acid are added, and the sample is gently
heated to dissolve remaining solids. The solution is brought to volume in a 50-mL volumetric
flask with water.

Microwave Assisted Dry Ashing


Approximately 1 g of sample is weighed into a crucible. Vycor or platinum crucibles have
both been used for this work. The MAS 7000 furnace insert allows the use of metal vessels in
the microwave unit. The crucible is placed into the microwave furnace. The temperature is
ramped up to 525~ over a two-hour period. It then holds the temperature for one hour. After
the sample cools, approximately 5 mL of a 1:1 nitric acid-water solution are added to the
crucible and heated gently to dissolve the ash. It is critical that nitric acid only be added after the
organic matrix has been ashed and cooled. The sample is quantitatively transferred to a 25-mL
volumetric and brought to volume with water.

Open- Vessel Microwave Digestion


Approximately 1 g of sample is weighed into the digestion vessel, and 10 mL of nitric acid
are added. The microwave unit is set to 30 W to cause a gentle reflux of the nitric acid. After
about one hour or when the sample is nearly decomposed, the drop-wise addition of perchloric
acid will typically complete the digestion. Perchloric acid is an extremely strong oxidizer that
may react violently with organic material. Experienced personnel only should use it with utmost
caution. The sample is heated until about 5 mL of solution remain in the vessel. The contents are
washed into a 50-mL volumetric flask and brought up to volume with water.

Closed- Vessel Microwave Digestion


From a safety perspective, it is critical that no more than 0.1 g of sample be used for this
technique. The sample is weighed into the Teflon digestion vessel. Approximately 4 mL of
nitric acid are added. The vessel is capped and placed into the microwave oven. Four vessels
are simultaneously processed. The microwave is set at 125 W for 15 min. The oven then ramps
up to 190 W for another 15 rain. Care must be taken not to keep internal temperature and
pressure within the capability of the vessels. Excessive heat and pressure will cause the
digestion bombs to deform and potentially leak. After the cycle is finished, the vessels are
placed into an ice bath for at least one hour to cool. The dissolved sample is washed into a 25-ml
volumetric flask and brought to volume with water.

ICP-AES Instrument Conditions


The instrument operating conditions are shown in Table 2. Optimization of conditions is
important to attain a robust plasma that is minimally affected by differences in acid concentration
between standards and samples. Calibration was done using a series of external standards with

28

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

yttrium (Y) as the internal standard. The internal standard was continuously added to standards
and samples via a "tee" fitting and mixing loop.
TABLE 2--ICP-AES operating conditions.
Forward Power
1150 W
Coolant Gas, Argon
16 1 min "l
Auxiliary Gas, Argon
0 1 min "1
Nebulizer Gas, Argon
0.65 1 min -l
Sample uptake rate
1.1 ml min 1
Discussion

Lubricating grease does not possess the required qualities to make it a suitable Standard
Reference Material, or SRM. The oil in the grease tends to separate from the thickener after a
few months on the shelf. Most of the following work uses samples with expected concentrations,
based upon known quantities of well-characterized additive. Commercial certified standards
were also used to validate the digestion procedure and to determine detection limits.
Table 3 shows the elements in lubricating grease, which was prepared with an organic
thickener as opposed to a metallic soap. All of the results are expressed in weight percent. Four
repeat analyses were made using each digestion scheme. The expanded uncertainty of each
measurement is calculated by multiplying the value for of one experimentally derived standard
deviation of the data times the critical value. The critical value of 3.18 is determined from a
student's double-sided t-distribution [5] with 3 degrees of freedom and a confidence level of
95 %. The average of the sulfated ash digestion indicates that nearly 12 % of the phosphorous is
lost from the sample. The loss is likely due to volatilization of phosphorous as opposed to
sample spatter, based upon the fact that the zinc result is closer to the expected value. The dry
ash technique shows less average phosphorous loss. However, it is still slightly more than that of
Parr Bomb or closed vessel, high-pressure microwave digestion scheme.
........ TABLE 3--Additive concentration results in lubricating grease with organic thickener.
Expected
2-h Sulfated
Open-Vessel
Parr Bomb
Element
Concentration
Ash
Dry Ash
Microwave
Microwave
Sb, Mass% <0.001
<0.001
<0.001
n/a
<0.002
B, Mass%
<0.001
<0.001
<0.001
n/a
<0.001
Li, Mass% <0.001
<0.001
<0.001
n/a
<0.001
P, Mass%
0.19
0.168 + 0.013 0.181 + 0.009 n/a
0.186 + 0.005
Zn, Mass% 0.212
0.222 + 0.007 0.217 + 0.006 n/a
0.212 + 0.003
The microwave digestion procedure is limited to a maximum sample size of about 0.1 g.
This is due to the pressure generated by the evolution of carbon dioxide and other gaseous
digestion by-products from the acid solution. Despite the small sample size, microwave
digestion provides additive concentration results that exhibit little or no increase in variability
from those procedures for which sample size is not a constraint. However, this may not be the
case for trace analyte concentrations where some loss of precision will inevitably occur.
The grease in Table 4 represents a more typical lithium soap formulation. Phosphorous loss
is almost negligible in this grease sample. However, boron suffers a nearly 25 % loss from the

FOX ON ANALYSIS OF LUBRICATING GREASE

29

expected value when the sample is prepared with the sulfated ash technique. A nearly complete
loss of boron has also been observed when digesting a commercial multi-element oil standard by
the sulfated ash technique. Open vessel microwave digestion results were consistent with
expected values. However, this technique has limited utility due to safety concerns about using
perchloric acid to complete the digestion. Another problem with the perchloric acid reagent was
the presence of boron. Despite the use of trace metal grade acid, an unacceptable level of boron
was detected. This forced the use of a blank correction that could lead to expanded uncertainty.
Subsequent to the generation of these data, a new source of perchloric acid was found that
appears to have suitably low levels of boron contamination.
The four digestion schemes were applied to a second lubricating grease sample with lithium
soap thickener. These results are shown in Table 5. The sulfated ash digestion resulted in loss of
antimony, boron, and phosphorous, while lithium and zinc remained consistent with expected
values.
TABLE 4--Additive concentration results in lithium thickener lubricating grease.
Expected
2-h Sulfated
Open-Vessel
Parr Bomb
Element
Concentration
Ash
Dry Ash
Microwave
Microwave
Sb, Mass% <0.001
<01001
<0.001
<0.001
<0.002
B, Mass%
0.03
0.023 +0.009 0.027 +0.006 0 . 0 3 1
0.029+0.002
Li, Mass%
0.19
0.189+0.003 0.185+0.004 0.191
0.184+0.006
P, Mass%
0.19
0.194+0.008 0.192+0.011 0.189
0.193+0.006
Zn, Mass% 0.212
0.223 + 0.006 0.205 + 0.006 0.216
0.221 + 0.003
*Result is reportedto two significantfiguresdue to loss of precision fromacid blank correction.
.... TABLE 5--Additive concentration results in lithium thickener lubricating; grease.
Expected
2-h Sulfated
Open-Vessel
Parr Bomb
Element
Concentration
Ash
Dry Ash
Microwave
Microwave
Sb, Mass%
0.026
0.02i + 0.008 0.022 + 0.007
0.025
0.025 + 0.002
B, Mass%
0.03
0.025 + 0.011 0.027 + 0.005
0.031
0.034 + 0.002
Li, Mass%
0.18
0.190 + 0.003 0.186 + 0.004
0.190
0.192 + 0.006
P, Mass%
0.195
0.166 + 0.016 0.192 + 0.011
0.192
0.191 + 0.009
Zn, Mass%
0.197
0.190 + 0.004 0.193 _+0.006
0.204
0.206 + 0.004
*Result is reportedto two significantfiguresdue to loss of precision fromacidblank correction.
The dry ash did not appear to result in nearly as much analyte loss due to volatilization.
However, this procedure has shown a higher propensity to spatter or ignite, resulting in ejection
of analyte from the container. There is some evidence for this in the fact that the dry ash
technique sometimes shows the lowest results for non-volatile grease components. It also can be
more difficult to dissolve the ash from a dry combustion, especially in used grease samples. This
may be the result of the formation of recalcitrant oxides that require severe heating of the acid to
complete the dissolution of the ash.
In the absence of a certified reference material that matches the sample, a generally accepted
validation technique involves subjecting an analogous certified material to the digestion
procedure and extrapolating the results to samples. This validation exercise was done on a
commercial, multi-element oil standard. The results are shown in Table 6. As expected, an
average of nearly 65 % of the boron was lost due to volatilization during the sulfated ash

30

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

procedure. However, there was little or no significant loss of the phosphorous over the four
repetitions. The Parr bomb microwave digestion consistently gave good recovery for all
analytes.
TABLE 6---Additive elementsfrom commerciallyprepared metal in oil standard.
Element
Certified Concentration
2-h Sulfated Ash
Parr Bomb Microwave
B, Mass%
0.030
0.011 + 0.012
0.031 + 0.001
Ca, Mass%
0.100
0.098 + 0.004
0.101 + 0.003
Mg, Mass%
0.100
0.098 + 0.006
0.098 + 0.006
P, Mass%
0.100
0.102 _+0.008
0.099 + 0.004
Zn, Mass%
0.100
0.105 _+0.006
0.102 + 0.003
The satisfactory phosphorous recovery shown in Table 6 may lead a researcher to believe
that phosphorous volatility is not an issue. However, Table 7 shows the effect that the sulfated
ash scheme has on an additive material. A zinc dithiodialkylphosphate (ZDDP) additive was
tested by the ASTM D 5185 method. The results obtained were consistent with the expected
values. The sulfated ash procedure yielded an average loss of nearly 25 % of the phosphorous
with much higher variability between measurements than the microwave digestion.
TABLE 7--Comparison of ZDDP additive elements by ICP-AES.
Expected
2-h
Parr Bomb
Element
Concentration
ASTM D 5185
Sulfated Ash
Microwave
B, Mass%
<0.001
<0.001
<0.001
<0.002
Ca, Mass%
<0.001
<0.001
<0.001
<0.002
Mg, Mass%
<0.001
<0.001
<0.001
<0.002
P, Mass%
9.5
9.4 + 0.4
7.3 + 1.5
9.4 + 0.5
Zn, Mass%
10.6
10.4 + 0.5
10.4 + 0.5
10.3 + 0.6
In this series of analyses, the average loss of phosphorous was more severe when ashing the
pure ZDDP additive than that observed while ashing the additized grease samples. It is possible
that the presence of the metallic soap in the grease may contribute to minimizing the loss of
volatile components in the short term. As the sulfated ash conditions become even more severe,
a greater portion of volatile analyte is lost. When the samples are allowed to heat overnight at
530~ it was observed that nearly all of the antimony and boron are lost. More than half of the
phosphorous was also lost.
The advantages of retaining volatile analytes, while ensuring complete digestion of a grease
sample when using the high-pressure Parr Bomb, start to become apparent. One of the major
disadvantages of the closed vessel microwave digestion scheme is the sample size limit of
approximately 0.1 g. Since 4 mL of acid are used to digest the sample, the analyst is either faced
with an extremely large dilution factor or must deal with the effects of a strong acid
concentration in the final solution.
Because these digestion schemes may result in widely varying acid concentrations in the
final solution, the ICP-AES conditions require careful optimization for this work. Researchers
have seen that increasing acid concentration often causes a depression in signal intensity for
some lines when using pneumatic sample introduction systems [6]. The effect may be especially
prominent under non-robust plasma conditions. The ICP-AES conditions were optimized using

FOX ON ANALYSIS OF LUBRICATING GREASE

31

the Mg II 280.270-nm/Mg I 285.213-um as described by Mermet [7]. Varying the nitric acid
concentration from less than 4 % v/v to over 18 % v/v caused less than a 2 % change in signal
intensity for the lines of the yytrium internal standard.
The cleanliness of the laboratory becomes an even more germane issue as the sample size
decreases. Trace contamination becomes magnified when dilution factors are approximately 250
or more. The sample size constraint of the high-pressure microwave digestion scheme tends to
raise the minimum detection limits of the analytes in a sample. Table 8 shows the result of
digesting a certified Conostan standard. The elements in the table represent several of those that
may be found at relatively low concentrations in lubricating grease, either due to contamination
or wear. For the standard with a certified concentration of 100 #g/g, the recovery is good for all
listed elements with the relative difference within 4 %. Uncertainty and recovery began to
deteriorate for the standard with a certified concentration of 10 #g/g. Replacing the pneumatic
nebulizer with an ultrasonic nebulizer (USN) would likely result in generally lower minimum
detection limits for analytes in the samples.
TABLE 8 Elemental analysis of certi~ed standards after closed-vessel microwave digestion.
Element
Certified,
Found,
RPD, %
Certified,
Found,
RPD, %
A1
Cr
Cu
Fe
Mn
Mo
Na
Ni
Pb
Si
Sn
Zn

100
100
100
100
100
100
100
100
100
100
100
100

103 + 8
102 + 3
103 + 3
104 + 5
100 + 4
100 + 2
97 + 10
102+6
101 + 8
98 + 7
99 + 9
103 + 6

3
2
3
4
0
0
3
2
1
2
1
3

10
10
10
10
10
10
10
10
10
10
10
10

13 + 2
l0 + 1
9+ 2
13 + 2
10 + 1
11 + 1
9+3
10+2
8+ 2
16 + 3
7+ 2
14 + 2

26
0
10
26
0
10
10
0
22
46
35
33

Certain clay-based greases may have silica as a trace element. In addition, silica is often
found as dirt in used grease samples. Table 8 indicates good recovery for silicon despite using
only nitric acid for the digestion of the multi-element oil standard. This level of recovery is not
likely to be seen when analyzing samples that have significant amounts of inorganic silicon. We
have found that the addition of about 100-200 #1 of 48 % hydrofluoric acid (I-IF) to the nitric
acid in the Parr bomb prior to digestion yields clear solutions when analyzing either clay-based
or used grease samples. Feng et al. [8] have found that addition of a small quantity of HF
enhances aluminum and silicon recoveries and may be analyzed without passivation by boric
acid. They also found that potential calcium precipitation in the presence of trace fluoride did
not negatively impact their calcium recovery. Since both boron and calcium are important
additives, these findings are of interest. To date, we have not done much work to validate mixed
acid, closed vessel microwave digestions with lubricating grease, but the application to accurate
silica quantification in both clay-based and used grease is promising.

32

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Conclusion

Modem lubricating grease is evolving with new additives and formulations to increase its life
in bearings, gearboxes, and other applications. Extended service life means greater attention to
additive decay and contamination. The development of elemental analysis techniques for
lubricating grease are somewhat hampered by its difficulty as a sample matrix. The lack of
standard reference materials and the tendency of grease to de-oil challenge the ability of the
analyst to validate the measurement system.
The fact that lithium thickener still dominates the grease market means that any atomic
spectroscopy technique to be considered should include the capability of determining the
concentration of this important analyte. The complex nature of the grease sample matrix makes
it difficult to get into a form that is compatible with ICP-AES. However, the elemental
composition may be effectively measured if a suitable digestion technique can be developed.
Four digestion techniques were examined. Each of them has advantages that should be
considered when analyzing a given sample. However, the closed-vessel microwave digestion
seems to provide more advantages than classical hot plate digestion or dry ashing. It tends to
yield clean solutions, ready for ICP-AES analysis, in less time than the other schemes, while
minimizing the chance of contamination by the laboratory environment. Despite its sample size
constraints, closed-vessel microwave digestion offers the benefit of good recovery and precision
for analytes that may be volatilized under other severe decomposition techniques that are open to
the atmosphere.
References

[1] Marino, J. H., "From Mystery to Profit: Grease Lubrication," Lubrication and Fluid Power,
Vol. 2, Issue 2, May 2001, pp. 11-14.
[2] Herguth, B., "Grease Analysis Monitoring Grease Serviceability and Bearing Condition,"
Practicing OilAnalysis, March-April 2002, pp. 18-25.
[3] Dresel, W. H., "Performance of Mixtures of Lubricating Greases with Different Thickeners,"
Paper presented October 29, 2002 at National Lubricating Grease Institute (NLGI) 69th
Annual Meeting, Preprint #0217.
[4] Iris Plasma Spectrometer Standard Vertical Hardware Guide, Thermo Electron Corporation
1994.
[5] Anderson, R. L., Practical Statistics for Analytical Chemists, Van Nostrand Reinhold
Company Inc., 1987, p. 299.
[6] Carre, M., Lebas, M., Marichy, M., Mermet, M., Poussel, E., and Mermet, J. M., "Influence
of the Sample Introduction System on Acid Effects in Inductively Coupled Plasma Atomic
Emission Spectrometry," Spectro Chimica Acta, 50B, 1994, pp. 271-283.
[7] Mermet, J. M., "Use of Magnesium as a Test Element for Inductively Coupled Plasma
Atomic Emission Spectrometry Diagnostics," Analytica Chimica Acta, 250, 1991, pp. 85-94.
[8] Feng, X., Wu, S., Wharmby, A., and Wittmeier, A., "Microwave Digestion of Plant and
Grain Standard Reference Materials in Nitric and Hydrofluoric Acids for Multi-Elemental
Determination by Inductively Coupled Plasma Mass Spectrometry," s Anal. At. Spectrom.,
14, 1999, pp. 939-946.

Journal of ASTM International, November/December2005, Vol. 2, No. 10


Paper ID JAI12967
Available online at www.astm.org

J. David Hwang, Ph.D., 1Meaghan Horton, 2 and David Leong3

The Use of Microwave Digestion and ICP to Determine


Elements in Petroleum Samples
ABSTRACT: ICP-AES analyses are prone to errors caused by changes in the power level, nebulization
rate, plasma temperature, and sample matrix. This is especially true for ICP-AES and ICP-MS organic

applications in the petroleum industry where extremely complicated sample matrices are common. As a
result, accurate analyses of petroleum lubricants, lubricant additives, fuels, crude oils, residues, and
related materials often require bracketing with matrix matched standards, the use of internal standards,
and a flexible ICP-AES/ICP-MS system. Many new innovative analytical instruments were developed to
better improve elemental analysis in the last few decades. However, sample preparation remains the
formidable challenge because it is always labor-intensive, cumbersome, and does not permit fast
operation. It is encouraging to see the application of microwave radiation field as the energy source for
sample digestion to pursue more reliable, but simpler, safer, more rapid, and less expensive methods,
particularly in the petroleum industry.
KEYWORDS: microwave digestion, sample preparation, petroleum and petrochemical products,
inductively coupled plasma-atomic emission spectrometry (ICP-AES)
Introduction

Although petroleum consists predominantly of hydrocarbons, many elements play a vital role
in petroleum production and finished products. Metals such as nickel, iron, vanadium, sodium,
and non-metals like sulfur and nitrogen in crude oils have an adverse effect on the refinery and
processing operations, generally acting as catalyst poisons. Vanadium and sodium in fuel oils
produce deleterious effects on refractory linings, metal surfaces o f modem furnace, boilers, and
turbines. Data on metals and sulfur concentration in fuel oils are also important for evaluating
emissions for oil-fired power plants. OrganometaUic compounds o f calcium, zinc, and other
metals are useful additives to petroleum products, such as lubricant oils [1]. Wear metals like
iron, lead, silver, copper, and chromium in used lubricant oils are often determined as indicators
o f possible or incipient beating damage. Common analytical procedures for the analysis o f
petroleum and petrochemical products b y ICP may be divided generally into two main groups:
a) Determined b y direct introduction (e.g., aspiration) into analyzer after a simple dilution
or dissolution o f the sample by suitable organic solvent or solvent mixture.
b) The metal determination is performed after the sample "prepared" b y dry or wet ashing
digestion has been converted into an aqueous solution.

Manuscript received 25 October 2004; accepted for publication 23 May 2005; published November 2005. Presented
at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
Senior Staff Scientist, Chevron Energy Technology Company, Richmond, CA, USA.
2 Research Assistant, Chevron Energy Technology Company, Richmond, CA, USA.
3Analytical Chemist, Chevron Energy Technology Company, Richmond, CA, USA.

Copyright 9 2005 by A S T M International,I00 Ban" Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.
33

34

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Direct introduction of organic sample solution may have transport interference problems [due
to differences in physical properties (e.g., viscosity and surface tension) of the samples and
calibration standards], background problems (caused by organic nature of the sample which may
result in spectral interference), matrix effects (arising from the complexity of hydrocarbons and
other contents), and standardization (calibration problems resulting from different type of
organometallic compounds present in the sample). The analysis after digestion removes almost
all the effects of direct organic introduction because digestion usually converts all organometallic
compounds into an inorganic form [2]. The determination of metals performed after total
destruction of organic matter is still [he most reliable procedure for determining metals in
petroleum and petrochemical samples.
Sample dissolution is one of the most common operations in analytical laboratories. This is
particularly true for petroleum analytical laboratories in which quantitative trace element
analysis of organic materials and minerals are needed and performed on a daily basis. Because
most analytical techniques require that samples be introduced in liquid form, thousands of
sample dissolutions are performed every day in analytical laboratories all over the world. Despite
the importance and widespread applicability of sample dissolution, most traditional digestion
procedures are time-consuming, labor-intensive, and the reagents used (such as perchloric acid)
are dangerous and potentially harmful to laboratory analysts [3,4]. Many new innovative
analytical instruments were developed to better improve elemental analysis in the last few
decades. However, sample preparation remains the formidable challenge and the bottleneck of
laboratory fast operation. Fortunately, microwave dissolution, a promising alternative to the
furnace and hot plate, has appeared and makes it possible to speed the preparation of solid,
sludge, organic matrix, and other difficult samples by combining the rapid heating ability of
microwave energy with the use of sealed digestion vessels [5-7].
The advantages of microwave dissolution include faster digestion that results from the high
temperature and pressure attained inside the sealed containers. The use of closed vessels also
makes it possible to eliminate uncontrolled trace element losses of volatile species that are
present in a sample or that are formed during sample dissolution. It is well known that significant
amounts of elements such as arsenic, boron, chromium, mercury, antimony, selenium, and tin are
lost at relative mild temperature with some open vessel acid dissolution procedures [8,9].
Another advantage of microwave dissolution is to have better control of potential contamination
in blank as compared to open vessel procedures. This is due to less contamination from
laboratory environment, unclean containers, and smaller quantity of reagents used.
Spectrometric oil analysis was first applied in the early 1940s in an effort to detect wear in
diesel locomotive engines by the railroad industry. In the early 1960s, the US Air Force adopted
spectrometric oil analysis to monitor the wearing in turbojet aircraft engines [10]. It has been
reported that direct dilution methods may not be accurate to determine metals in oils by either
Flame Atomic Absorption (FAA) or Inductively Coupled Plasma (ICP) because of significant
matrix effects [11-13]. It has been found that viscosity modifier polymers interfere with the
formation of aerosol, thus affecting the sample delivery to the plasma torch which causes the ICP
signal suppression as much as 50 %. By using both internal standard and appropriate dilution
rate, adequate ICP results were obtained. However, the particle size of the wear metals in used
oils may create different problems for spectrometric analysis. The detection and identification of
metallic particles in the lubricant are very critical because their presence indicates engine oilwetted component wear. The inability to detect large wear particles has led to engine component
failure without prior "warning" by spectrometric methods. Several severe incidents have been

HWANG ET AL. ON MICROWAVE DIGESTION

35

reported in which large metallic wear particles produced by severe wear mechanisms have not
been quantitatively determined by spectrometric techniques employed [14]. Although "trend
analysis" by rotating-disk electrode atomic emission spectrometry (RDE-AES) may serve the
purpose to monitor the wear metals in engine, particle-size-independent techniques [15-17], and
other dissolution approaches (such as microwave digestion) would give more accurate elemental
determination in used oils. The use of high solid nebulizers such as V-groove nebulizer can
minimize the clogging and blocking problems, but the short residence time (-2-3 milli-seconds)
in ICP (DCP or flame) would not be able to fully decompose all particulates in used oils, which
would result in lower analytical results.
Conventional ashing and air-refluxing methods have been widely used in the Petroleum
industry's analytical laboratories for many years. These traditional methods are always timeconsuming, labor-intensive, and tedious. Additionally, contamination is very likely to occur from
the acids (especially sulfuric acid) and/or vessels used during the sample preparation. Moreover,
sulfur determination may not be achieved because sulfuric acid is always used in these methods.
In 1975, the microwave was finally applied to digest samples in the analytical laboratory
[18-20]. The idea that something as revolutionary and simple as the microwave could also be
applied to help with digestions was unheard of until then. Using a heating source as fast and
reliable as a microwave would digest samples at a greater speed without having to be constantly
attended to and would decrease the time it took to get their results back. A solution to the
analytical scientists' problem was finally found in simplicity. When the idea of using
microwaves started being tested, people used regular household microwaves; this soon proved to
be troublesome due to a number of problems that would often be encountered. The microwaves
used in homes were made to heat large quantities of food, which was a different case than
analytical samples, which tend to be smaller. Analytical samples could also be explosive, give
off toxic fumes, or lose valuable elements during the heating process.
Utilizing a high pressure microwave system, many innovative and rapid analytical methods
were developed in our atomic spectroscopy lab recently. The high operating temperature and
pressure attainable in this microwave digestion system ensures total decomposition of complex
petroleum crude oils and petrochemical products, with no residual organic content. Sample
preparation time had been shortened from days (or, in some cases, even more than a week) to
minutes. Our analytical methods can now also be applied to the determination of sulfur in crude
oils and polymer samples because sulfuric acid is no longer required in the microwave digestion
process. The microwave digestion system also eliminates contamination previously happening
due to the use of more traditional digestion vessels and from airborne particulates in the lab
environment.
In this work, several approaches such as direct dilution, wet-ashing, and a newly developed
microwave digestion technique were used to prepare and analyze petroleum samples for the
purposes of method and result comparison. The samples used for this study were a NIST
lubricant additive Standard Reference Material (SRM) 1848, Residue Fuel oil SRM 1634B,
1634C, and a polymer resin.

Experimental Section
Microwave Oven

A Milestone laboratory microwave oven, MLS 1200 MEGA model, with a 6 vessel carousel
rotor was used for all microwave digestion work in this study. The MLS-1200 MEGA

36

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

microwave system consists of the microwave labstation, EM-45 exhaust module, and model 240
control terminal. Power to this system is a 1200 W microwave generating magnetron with 1000
W delivered inside the working chamber. Power emission is microprocessor controlled from 10
to 1000 W in 10 W increments. The microwave power frequency is 2450 MHz. To ensure
explosion proof safety, the hardware of the unit is all stainless steel. An external coating was
applied to ensure protection against acids and organic solvents. The inside of the working
chamber is coated with a 5-layer PTFE coating and is provided with a direct connection to the
EM-45 module to exhaust acid vapors from the working chamber itself. The working chamber
was purged with nitrogen to create an oxygen free working environment to prevent any fire
incident from handling petroleum hydrocarbons.
The Microwave Digestion Rotor (MDR) technology has brought microwave digestion to
"almost" routine analytical laboratory procedures. It provides fast, reliable, and very safe
digestions. In our laboratory, >95 % samples can be completely digested in a one-step digestion
procedure. A microwave vessel consists of a PEEK sleeve, adaptor, and spring, as well as a TFM
vessel, ring, and cover. The use of TFM and PEEK components would minimize trace element
contamination and provide very safe operations.

TraceCLEANAuto System
All microwave Teflon vessels were soaked in 1:1 HC1 overnight prior to cleaning by nitric
acid using a TraceCLEAN Auto System from Milestone Inc. TraceCLEAN auto system is
specifically designed for cleaning o f vessels and labware based on the principle of acid vapor
cleaning. The bottom of the glass vapor container is filled with the nitric acid vapor for cleaning.
The bottom plate of the vessel holder sits over the acid level and is automatically lifted up and
down (controlled by a built-in control terminal). The bottom plate has several different holes to
insert vapor tubes of different lengths or blind stoppers. The labware is placed onto the vapor
tubes for cleaning.
The acid is heated by an electric temperature controlled heating plate. The acid vapors enter
through the vapor tubes into the vessels. The acid vapors then condense on the cold vessel walls
and form droplets, thus washing down the vessel surface. The acid droplets gather on the PTFE
bottom of the vessel holder and flow into the collecting channel of the glass vapor container.
After the system has cooled, the dirty acids can be pumped out through an outlet connection.

Wet Ashing
A specially made, cylinder shape, quartz vessel (12 cm height and 3 cm ID) is used for wet
ashing digestion. Common laboratory hot plate and programmable muffle furnace are also
needed to conduct wet ashing procedures.

Inductively CoupledPlasma Atomic Emission Spectrometer (ICP-AES)


All the analytical measurements were made on a Thermo Electron Iris ICP. The Iris ICP uses
Echelle optics and a unique Charge Injection Device (CID) solid-state detector to provide the
flexibility of choosing wavelengths (as a scanning sequential ICP) and also the high throughput
(as a simultaneous multi-channel ICP). The ICP system provides complete and continuous
wavelength coverage over the analytical range from 165 to 1000 nm. When the optional XUV
configuration is chosen, the range is extended to 130 nm, which allows C1 determination at

HWANG ET AL. ON MICROWAVE DIGESTION

37

134.724 nm. All operations of the ICP spectrometer are controlled by the TEVA (Thermo
Electron Validated Analysis) software, a 32-bit applications program that runs under Microsoft|
Windows | software.
TM

Reagents
All inorganic acids used for microwave digestion are o f "Ultrex | grade from LT. Baker |
Chemical Co. Hydrogen peroxide used is Suprapur | from EM Science. For wet ashing, acids
used are ACS grade from Fisher Scientific Co, and the flaming sulfuric acid used is from J.T.
Baker | Chemical Co. De-ionized water of 18 Mf//cm specific resistivity is always used; it is
prepared with either the Millipore | Milli-Q water and a Rios/Elix reverse osmosis purification
system or a Nanopure Diamond water system from Barnstead |
For direct dilution organics ICP, USP/FCC grade mineral oil and reagent ~rade O-xylene
from Fisher Scientific are used. Organometallic standards from Conostan are used for
calibration and internal standard.
Procedure

Microwave digestion conditions were established and optimized for our Milestone
microwave MLS 1200 Mega system with a 6-vessel carousel MDR. Optimal conditions for
microwave digestion depend on sample weight, composition, volume of digestion reagents, and
microwave system used. A method was developed and used to digest >95 % of the samples in
our laboratory. Approximately 0.1 g of the samples was accurately weighed into a pre-cleaned
microwave Teflon digestion vessel. Eight ml of nitric acid, and 2 ml of hydrogen peroxide were
added into the digestion vessel. The digestion vessels were placed in the 6 vessel MDR rotor and
into the closed, high pressure microwave digestion system to completely convert the samples
into an aqueous form. An accurate amount of internal standard (scandium) was added to the
samples and brought up to a final volume with deionized water. The prepared samples were then
analyzed using IrisT M ICP-AES or ElanT M DRC ICP-MS.
For the wet ashing approach, approximately 1-2.5 g of samples were accurately weighed into
a quartz digestion vessel. Two ml of fuming sulfuric were added into the container and placed on
a hot plate (around 120~ and heated underneath an infrared lamp for 2-3 h. The container was
then transferred into a programmable muffle furnace to be heated and ashed for 72 h until the
oxidation of carbon was completed. The ash was dissolved in 8 ml of aqua regia with gentle
boiling. The digested solution was finally filled up to a 25 ml mark with deionized water when
the digestion was completed prior to ICP-AES analysis.
For direct dilution organics ICP approaches, modified ASTM D 5185 and D 4951 methods
were used in our laboratory. O-xylene is the solvent of choice, and internal standard is used for
all organic ICP applications.
Results and Discussion

In Table 1, the direct dilution ICP approach gave lower results because the short residence
time (2-3 milli-seconds) of sample aerosol in the ICP plasma was not able to fully decompose
the large dry aerosol generated by the nebulization of the used oil. Contrary to direct dilution
ICP, microwave digestion "dissolved" and "converted" all analytes into aqueous solution prior to
ICP analysis. Thus, the results are much better and reliable. Different from samples #1 and #2, #

38

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

3 is a fresh oil, and microwave digestion results correlate with direct dilution (modified ASTM D
5185 method) quite well.
TABLE 1--Comparison of microwave digestion and direct dilution ICP for used~fresh oils

(ppm).
Analytical
Approach
Microwave
Digestion/ICP
Direct ICP
Microwave
Digestion/ICP
Direct ICP
Microwave
Digestion/ICP
Direct ICP

Sample #

ca

Fe

Used Oil 1

8.88 %

5100

Used Oil 1
Used Oil 2

1.67 %
2.53 %

Used Oil 2
Fresh Oil 3
Fresh Oil 3

Mg

Mo

Si

760 ........i'106

6.23 %

1229

960
5214

170
1273

211
305

1.65 %
2.55 %

85
6.55 %

1.21%
0.62 %

670
28.9

284
25.4

177
106

1.61%
0.74 %

27.3 %
18

0.62 %

25.2

21.2

109

0.74 %

14

For lubricant additive (NIST SRM 1848), as shown in Table 2, boron results are not available
because of the formation of volatile boron compounds (e.g., boric acid) when acid was added and
heated. Loss has been associated with some boron species, but not with others. Loss of boron
was not noted with a Conostan| boron standard. Sulfur results are not available because sulfuric
acid is used in the ashing process. Other element results align more closely with the alternative
preparation methods. While the precision for boron looks good for the microwave digestion
preparation, compared to the two direct dilution methods, the results are slightly lower. This
could be attributed to a safety feature of the microwave vessel that causes it to vent under
excessively high pressure. Thus, volatile boron species may be lost. Calcium results by the direct
dilution additive method compare well with wet ashing and microwave digestion results.
Calcium results are 4-5 % lower when direct dilution was applied. All four methods compare
well for magnesium and zinc.
TABLE 2--Comparison of analytical methods using ICP-AES to determine elements in
lubricant additive (%).
Element

Certified Value

Microwave
Wet Ashing
Direct Dilution
Di~esti0n
B
0.137 +/- 0.019
0.123 +/- 0.008 a N.A. b
0.142 +/- 0.004
Ca
0.359 +/- 0.011
0.359 +/- 0.010
0.358 +/- 0.004
0.348 +/- 0.009
Mg
0.821 +/- 0.038
0.835 +/- 0.014
0.847 +/- 0.034
0.858 +/- 0.016
P
0.788 +/- 0.028
0.793 +/- 0.017
0.750 +/- 0.022
0.750 +/- 0.036
S
2.327 +/- 0.0043
2.223 +/- 0.061
N.A. c
2.236 +/- 0.12
Zn
0.873 +/- 0.022
0.833+/- 0.026
0.877 +/- 0.020
0.847 +/- 0.019
aThe microwavevessel was "popped;"thus someboronloss mightoccur.
bThe boronspecieswouldreact with acid to formvolatileboric acid and lost duringthe samplepreparation.
cSulfuric acid was used in wet ashing; thus sulfur is not available.
The closed vessel, high-pressure microwave digestion gave good results for calcium,
magnesium, phosphorous, sulfur, and zinc. It has been proven that closed-vessel microwave
digestion can be used to accurately prepare samples containing "volatile" boron and silicon

HWANG ET AL. ON MICROWAVE DIGESTION

39

compounds. However, in this case, the lower Boron results may have been caused by the vessels
"releasing" under high pressure, causing the loss of volatile boron (boric acid). Had the vessel
not vented, the Boron result would have been higher and much closer to the certified value.
Sulfur information was available because sulfuric acid was not required in the microwave
digestion step. More importantly, the microwave digestion approach only took 30 min.
The direct dilution approach works well for lubricant additives and fresh oils. However, this
approach cfinnot handle used oils containing particulate or volatile compounds. Accurate analysis
for sulfur by ICP can be problematic if the samples contain volatile sulfur such as in sulfide
form. There is a "vapor enrichment" phase separation that occurs during the pneumatic
nebulization process in ICP. Hence, the use of closed vessel, high-pressure microwave digestion
can be used to accurately prepare samples containing "volatile" sulfides. ICP can be used to
determine the accurate sulfur result.
For petrochemical application, as shown in Table 3, microwave digestion ICP results of
cobalt were pretty much in line with wet ashing approach. The microwave digestion approach
took about 30 min, but it would take at least 4 days or even up to 2 weeks for the wet ashing
approach.
TABLE 3---Comparison of cobalt content in polymer resin.
Wet Ashing
Cobalt Content (%)
Sample # 1
0.97
Sample #2
0.96
Sample #3
0.97
Microwave Digestion
Sample # 1
1.03
Sample #2
1.05
.......
Sample #3
1.04
As another example, a certified fuel residue SRM 1634 C was used to "validate" the
microwave digestion with other proven methods. Thus, microwave digestion is still able to
produce better results than other traditional methods, such as wet ashing or direct dilutions. In
Table 4, the microwave data is able to achieve better results than wet ashing for both 1634B and
1634C SRMs. However, the direct dilution data are just about the same as microwave digestion
data.
The use of microwave digestion has not only helped our laboratory improve the productivity
(and turn-around time), but also provided a reliable tool to handle some difficult analytes and
solve very tough problems. It appears that the main advantages of microwave digestion over
conventional wet- or dry-ashing procedures are speed (wet ashing approach took up to two
weeks, but microwave digestion can handle the samples within hours), simplicity of microwave
digestion procedure, less sample and reagents needed, and much better contamination control.
More than 95 % of the samples can be completely digested with our single microwave digestion
method. Others need HC1 and/or HF added to completely digest "high" content metals or silica
(in dirt of used grease samples). However, it is difficult to completely digest >0.5 g of petroleum
hydrocarbon samples by microwave digestion. Thus, for certain samples, wet ashing may be the
method of choice when larger samples are required to get consistent sampling and representative
samples. The other drawback for microwave digestion is to have "poor" method detection limit
because a small quantity of samples is used resulting in higher dilution rates (-100) than the wet

40

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

ashing approach (~20). To ensure accurate and reliable analytical results, microwave vessels
need to be cleaned thoroughly prior to use. Common airborne contaminants, such as iron,
sodium, silicon, and calcium are very difficult to completely remove unless a clean room is used.
For boron and silicon analysis, deionization is not sufficient to remove these weak bases
effectively. Deionization with reverse osmosis water purification system should be used instead
to ensure that the purified water is free of borate and silica.
TABLE 4---Comparison of different methods determining trace metals in fuel residue.
Nickel (ppm)
Vanadium (ppm)
Sulfur (%)
NIST Certified Value
28 +/- 2
55.4 +/- 1.1
2.80 +/- 0.05
Microwave Digestion
1634B
26.4
55.1
2.72
1634B Dulpicate
27.0
55.5
2.73
Direct Dilution
1634B
27.9
56.9
2.96
1634B Duplicate
27.7
56.7
2.95
1634B Triplicate
27.8
56.8
2.98
NIST Certified Value
17.54 +/- 0.21
28.19 +/- 0.40
Supplemental Value
Microwave Digestion
1634C
17.1
29.5
2.11
1634C Duplicate
17.0
29.8
2.13
Direct Dilution
1634C
17.1
29.5
2.11
Wet Ashing
1634C (2.5 g)
2.95
21.61
NA
1634C Duplicate (2 g)
8.74
24.56
NA
1634C Triplicate (1 g)
15.05
27.49
NA

Summary
Microwave digestion has been proven to help speed up and improve the productivity and
quality of the sample preparation in our laboratory. It has become the method of choice for many
unknown and/or complicated matrix samples. "Obtaining a representative sample" can be a
problem when smaller quantities (<0.2 g) are used for microwave digestion. Other applications,
such as speciation, organic extraction, and sample preparation for wet chemistry are some areas
to which we can apply this powerful tool in order to further improve our laboratory practice.

Acknowledgments
The authors would like to thank Karsia Yip, and other ICP Laboratory colleagues for their
contribution and assistance.

References
[1]

Mortier, R. M. and Orszulik, S. T., "Chemistry and Technology of Lubricant," Blackie


Academic & Professional, 1997.

HWANG ET AL. ON MICROWAVE DIGESTION

[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]

41

Hwang, J. D. and Wang, W. J., "Application of ICP-AES to Analysis of Solutions,"


Applied Spectroscopy Reviews, Vol. 30, No. 4, 1995, pp. 231-350.
Bock, R., "A Handbook of Decomposition Methods in Analytical Chemistry,"
International Textbook Company, 1979.
Matusiewicz, H. and Sturgeon, R. E., "Present Status of Microwave Sample Dissolution
and Decomposition for Elemental Analysis," Prog. AnaL Spectrom., Vol. 12, 1989, pp. 2139.
Kingston, H. M. and Jassie, L. B., AnaL Chem., Vol. 58, 1986, pp. 2534-2541.
Kingston, H. M. and Jassie, L. B., "Introduction to Microwave Sample Preparation-Theory
and Practice," American Chemical Society, 1988.
Kingston, H. M. and Haswell, S. J., "Microwave-Enhanced Chemistry-Fundamentals,
Sample Preparation, and Applications," American Chemical Society, 1997.
Hoffman, J. L. and Lundell, G. E. F., Journal of Research of the National Bureau of
Standards, Vol. 22, April 1939, pp. 465--470.
Zief, M. and Mitchell, J. W., "Contamination Control in Trace Element Analysis," John
Wiley & Sons, 1976.
Eisentraut, K. J., Newman, R. W., Saba, C. S., Kauffman, R. E., and Rhine, W. E., AnaL
Chem., Vol. 56, 1984, 1087A.
Sychra, V., Lang, I., and Sebor, G., Prog. Analyt. Atom. Soectrosc., Vol. 4, 1981, pp. 341426.
Bowman, J. A. and Willis, J. B., Anal Chem., Vol. 39, 1967, p. 1210.
Bansal, J. G. and McElroy, F. C., "Accurate Elemental Analysis of Multigrade Lubricating
Oils by ICP Method, Effect of Viscosity Modifiers," SAE Technical paper 932694, 1993.
Kagler, S. H., Jantzen, E., and Fresenius Z.,AnaL Chem., Vol. 310, 1982, p. 401.
Saba, C. S. and Eisentraut, K. J., AnaL Chem., Vol. 49, 1977, p. 454.
Saba, C. S. and Eisentraut, K. J., AnaL Chem., Vol. 49, 51, 1979, p. 1927.
Brown, J. R., Saba, C. S., and Rhine, W. E., Anal. Chem., Vol. 52, 1980, p. 2365.
Abu-samra, A., Morris, J. S., and Koirtyohann, S. R., Anal. Chem., Vol. 47, 1975, p. 1475.
Barrett, P., Davidowski, L. J., Penaro, K. W., and Coopeland, T. R., Anal Chem., Vol. 50,
1978, p. 1021.
Nadkami, R. A., AnaL Chem., Vol. 56, 1984, p. 2233.

Journal of ASTM International, October 2005, Vol. 2, No. 9


Paper ID JAI12968
Available online at www.astm.org

Joseph Paszek, 1, Kelly J. Mason, 1Anthony S. Mennito, 1 and Frank C. McElroy 1

Advances in ICP-MS Technologies for Characterization and


Ultra-Trace Speciation as a Tool for the Petroleum Industry
ABSTRACT: Particular elements can, even at trace levels, impact product quality, deposit formation,
and regulatory compliance. Characterization o f trace and ultIa-trace level constituents in aqueous and/or
organic matrices has often required the need to employ more than one analytical technique. In recent
years, improvements in ICP-MS technologies have made it a preferred technique for multi-element trace
level characterization in a variety o f matrices. Coupling ICP-MS with various liquid chromatography
techniques has enabled a unique capability to speciate organo-metallics at ppb levels. This paper will
show how ICP-MS has been successfully used to measure ppb levels in refinery effluent streams. It will
demonstrate how hyphenated ICP-MS techniques assisted in development of bioremediation options for
selenium removal in wastewater treatment plants.
KEYWORDS: selenium, selenocyanate, selenite, selenate, ICP-MS, and ion chromatography

Introduction

Today's analysts no longer focus on just trace elemental analysis, but rather they face the
challenge of being able to speciate and quantify many high profile elements. ICP-MS coupled
with suitable chromatographic separation techniques offers a unique combination of
simultaneous multi-element capabilities with ppb and ppt detection limits and the ability to
perform elemental speciation. In this paper, we will present how Inductively Coupled Plasma Mass Spectrometry (ICP-MS) has been successfully used with liquid chromatography to provide
trace level speciation information.
Selenium enters the refinery as contaminates of processed crude oils. As the refiner processes
the crude oil, the selenium contaminate migrates from the crude oil into the refinery's striped
sour water (SSW) streams. Once in the SSW stream, the selenium contaminate, which is an
environmental concern, is now part of the wastewater stream. The problem for the refiner is how
to efficiently remove the selenium prior to discharging the water to an appropriate catch, such as
a bay or municipal water treatment plant.
It has been shown that the predominate state of the selenium in these SSW streams is as the
selenocyanate 2(.SeCN-) [1]. The selenocyanate in SSW is oxidized by refinery biotreaters to
selenite (SeO3 "), which can then be removed via an iron based treatment process. One of the
most effective of these iron-based processes is iron co-precipitation process. However, the
selenium can also occur as s e l e n a t e (8eO42-), which is not removed during the iron coprecipitation process [2]. In order to design effective treatment protocols to remove the selenium,
more than just the concentration of the metal must be determined. Without knowledge of the
types of selenium species present in the water, design of an effective removal treatment protocol
Manuscript received 25 October 2004; accepted for publication 11 April 2005; published October 2005. Presented at
ASTM Symposium on Elemental Analysis o f Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
i ExxonMobil Research and Engineering, 1545 Route 22 East, Annandale, NJ 08801.
* Primary author to whom correspondence should be addressed.

Copyright9 2005 by ASTMInternational,100 Ban"HarborDrive,PO Box C700, West Conshohockoa,PA 19428-2959.


42

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

43

is impossible. The need to know that the species of selenium present, as well as the fact that the
metal would be present in the untreated SSW at levels ranging from a few parts-per billion to
low parts-per-million, make the analysis very difficult.
While ICP-MS has a demonstrated capability for being able to provide accurate and precise
selenium measurements at these levels, there still remained the problem of the identification of
the selenium species present in the samples. Research efforts defined here focused on developing
a quick and reliable separation technique to separate the selenium species present in the various
SSW and wastewater samples prior to quantification by ICP-MS.
Building upon a study that was performed for the Western States' Petroleum Association
(WSPA) [1] to optimize iron-based selenium removal processes, a scheme was developed to
measure total selenium, and the individual selenium species present in the SSW and aqueous
streams. Total selenium was measured by ICP-MS after a peroxide/acid digestion pretreatment
of the samples. Speciation of the selenium was accomplished using an anionic ion
chromatographic separation followed by quantification of the selenite, selenate, and
selenocyanate using ICP-MS.
The goal of the analytical program was to develop a measurement method with a detection
limit of 10 ixg SeL "1. By the end of the program, we were able to not only demonstrate that this
goal was achievable, but that detection limits of 2 ~tg selenite'L"1 and 0.2 ~tg selenate'Ll were
possible.

Instrumentation
The instrumentation used for this work to quantify and speciate selenium in WWS is listed in
Table 1.
TABLE 1--Instrumentation usedfor Selenium Speciation Study.
Instrumentation/Software
Manufacturer/Model
Perkin Elmer/Sciex Elan 6000 ICP-MS
ICP-MS
Chromatographic Integration Program - (P/E software)
Integration Software
Dionex Model 2000i Tertiary Ion Chromatography System
IC
Analytical Column
Dionex AS-4-SC (4.0 m m x 250 mm.)
Dionex AG-4-SC (4.0 mm x 50 mm.)
Guard Column
Gilson
Peristaltic Pump
Spray Chamber
Scott
Meinhard
Nebulizer

Reagents and Materials


All eluents were purged with helium. In addition, helium was used to pressurize and blanket
eluent bottles as well as operate the IC pneumatics. All of the standards materials and eluent
constituents were obtained from Aldrich Chemical. The stock 1000-mg Se'L"1 selenite and
selenate standards were made from their corresponding sodium salts, while the selenocyanate
stock was made from the potassium salt.
The germanium internal standard was made by diluting a 1000 m g L "l aqueous commercial
standard to 10 mg/L in 300 mM NI-hOH. The reason for using ammonium hydroxide was to
avoid the precipitation of germanium oxide when the internal standard was mixed with the IC
effluent. The eluents were made by dissolving the appropriate amount of pure material in

44

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

distilled deionized water that was prepared using a Millipore Milli-Q deionization system.
Complete dissolution of the p-cyanophenol in the eleuant required the use of an ultrasonic bath.
The sodium hydroxide solutions were made fresh and blanketed under helium in the eluent
reservoirs. The 50 % (m/m) sodium hydroxide concentrate was discarded every 30 to 45 days to
minimize the influence of any dissolved carbonate that may have accumulated in the sodium
hydroxide during that time period.
The 30 % (w/w) hydrogen peroxide used in the digestion was commercially available reagent
grade material.
The concentrated nitric acid used was Ultrex grade acid available from Baker Chemicals. The
1 % (v/v) used in the digestion procedures was prepared by diluting 10 ml of the Ultrex grade to
1 L using deionized distilled water (DDW).

Experimental
An analytical scheme was developed [3] to characterize both the concentration and nature of
the selenium species present in the various streams measured (Fig. 1). Because an in-depth
discussion of all of the techniques would be beyond the scope of this paper, the reviewed work
will focus on the efforts to measure total selenium as well as selenite, selenate, and
selenocyanate. A Sciex Elan Model 6000 Inductively Coupled Plasma-Mass Spectrometer
system equipped with a water-cooled spray chamber was used for all of the selenium
measurements. The operating conditions used for the ICP-MS are presented in Table 2, while
conditions for the IC are presented in Table 3.

Scheme for Selenium Speciation

FIG. 1--Scheme for selenium speciation.

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

45

TABLE 2--General ICP-MS instrument conditions.


Parameter
Nominal Value
RF Power
1.2 KW
Nebulizer Gas flow
0.6 L/min-1
Plasma Gas Flow
15 L/min-1
Auxiliary Gas Flow
1.0 L/min"1
Sample Flow Rate
0.5 ml/min"1
Nebulizer
Meinhard

Time
(min.)
0
0.1
0.3
5.0
7.0
7.1
18.0
18.1
20.0

TABLE 3 - - I C conditions used for selenium speciation.


%El
%E2
%E3
5
5
5
50
50
0
0
5
5

0
95
0
95
0
95
0
50
0
50
40
60
40
60
0
95
0
95
E1 = Ammonium Carbonate (8 ~ )
E2 = 50 mM p-cyanophenol in 78 mM NI-I4OH
E3 = Distilled Deionized Water

.....
Comments

Inject
Load

End Run

Initially, the interface between the Ion Chromatography (IC) and ICP-MS used y-tubing
fitting where the effluent from the IC could feed one arm of the y-tubing fitting. A peristaltic
pump fed germanium (Ge) internal standard into the other arm of the y-tubing fitting. The output
of the y-tubing fitting was connected to a Meinhard nebulizer of the ICP-MS. However, with
this arrangement, observed fluctuations in the intensity of the internal standard were attributed to
incomplete mixing of the eleuant and internal standard. In the final setup, which was used for all
of the speciation measurements in this program, the effluent from the IC system was interfaced
to the ICP-MS via a short length of 0.010 mm ID PEEK tubing connected to one of the side arms
of a plastic "T" fitting. The other arm of the T was connected to the output of the peristaltic
pump supplying the Ge internal standard. The position of the arm supplying the internal standard
was at a 90 ~ angle to the flow of the IC effluent to the ICP-MS. This arrangement was used to
ensure adequate mixing of the internal standard with the stream carrying the selenium species.
Results and Discussion

The samples of striped sour water (SSW) used for this study were refinery sour water streams
that have had the bulk of the ammonia and hydrogen sulfide removed or striped. No
preservatives were added to the samples prior to analysis. All of the samples were kept
refrigerated with a minimum amount ofheadspace in the bottles to minimize the reactivity of the
selenium species present in the various samples.

46

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Because the Se in the SSW samples could be present in many ways, such as particulate, all
total Se measurements were carried out on unfiltered samples of the water. These samples were
digested prior to any total Se measurements by using a procedure that employed a combination
of hydrogen peroxide and nitric acid to convert all of the selenium to soluble selenium.
In the total selenium procedure, an aliquot of the unfiltered sample is treated with 3-4 ml of
hydrogen peroxide and heated to 95~ to digest the bulk of the organic materials present in the
sample. The volume of the solution is reduced to half its original volume, and an additional 3-4
ml of hydrogen peroxide is added to the solution. When the peroxide-treated solution has been
further reduced to approximately one-fiRh its original volume, one more addition of peroxide is
performed. The solution is then heated until all effervescence subsides and the solution become
colorless. At this point, 0.1 ml or more of the 1 % (v/v) nitric acid is added to dissolve any and
all precipitated material. Once the solution is clear and colorless, it is transferred to a 50-ml
volumetric flask and made up to volume using DDW.
The final step used to treat the digested SSW samples prior to analysis by ICP-MS was to
remove the bulk of the sodium present in these samples. Removing the sodium prior to
measurement would minimize the effects high sodium levels could have on the plasma
characteristics, thereby impacting the overall precision of the total selenium measurements.
Sodium was removed by passing the digest solutions through a Bio-Rad AG50W-X8 cation
exchange column prepared according to manufacturer's directions. The first 1-2 column bed
volumes are discarded and the rest retained and measured by ICP-MS.
All connections between columns and instruments were made using 0.007 mm ID PEEK
tubing. A schematic of the instrument arrangement, as well as representative chromatograms of
the selenium speciation, are presented in Figs. 2 and 3.

FIG. 2--Schematic of the LC/IC-ICP-MSarrangement.

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

47

50000

100 ppb Standard


SeO= ="
~000~

Se04 ;!-

~UUUt

_=
20000

O0

40C

230

3C0

~4~0
lime

~CO

.300

?CO

300

9C0

FIG. 3--Selenium Speciation Standard.


The 82Se mass was used to measure the selenium for both the total and speciated
measurements. The reasons for choosing this line, which is not one of the more abundant lines
(-8.7 %), were based on minimizing the mass overlap interference from Ar2, which would have
quenched the more abundant 78Se and 8~ signals. However, even the 82Se mass line suffered
from mass overlaps from Kr, HBr, and SO3 that, if left uncorrected, could affect the overall
accuracy of the selenium measurements. In order to correct the 82Se signal, the responses for
these masses were monitored in all of the measured samples and mathematically corrected for in
the 82Se signal. Monitoring the 78Se mass line in conjunction with the 82Se line could be used as a
means of testing the level of interference from sulfur species present in the samples being
measured.
Although we did not pursue it as part of this program, we were able to demonstrate that the
lower limit of quantification could be extended by the use of a short concentrator cartridge
mounted in place of the injection loop on the IC sample introduction section. This concentrator
cartridge was typically a short 5-cm column containing the same substrate as the analytical
column. Using this approach, we were able to load 1 ml of the SSW water sample without any
selenium breakthrough. This approach effectively gave us a 40-fold improvement in the lower
selenium detection limit.
The concentrations of the mixed standard (selenate, selenite, and selenocyanate) were
confirmed by measuring the total selenium concentration of the solution. The mixed standards
were also carried through the digestion procedure as a quality control check of the recovery.
Recoveries were generally in the range of 95-98 %. Examples of the spike information, as well
as a comparison of total Se data versus the sum of the calculated species can be found in Table 4.
Additionally, selected samples of the SSW stream were analyzed on different days to confirm
the repeatability of the measurements. Figure 4 shows repeat measurements for the three
selenium species. The duplicate measurements for three selenium species for various samples
differed by no more than 4 % of the mean.

48

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

TABLE

4--Comparison of speciated and digested total Se measurements.


Se (I.tgL"l) as

Sample ID
1

2
3
4
5

Total
(Digested)
102
10
<10
112
103
70
5
<10
75
70
71
5
<10
76.0
80
142
19
<10
161
154
500
500
500
1500
1430
Note: % Recovery = (Total Speciated/Total Di~ested)*100

SeO3

SeO4

FIG. 4

SeCN

Total

%
Recovery
109
107
95
105
105

Selenium species repeatability.

Once the ability of the IC-ICP-MS system to repeatedly separate the three selenium species
was confirmed, efforts focused on investigating process variables which would influence the
concentration o f various selenium species, specifically selenocyanate (SeCN) in the SSW. One
such process variable studied was pH. The study of the pH effects on the conversion of the SeCN
to other selenium species, such as selenate and selenite, was undertaken to define the pH
boundaries for any pre-treatment or treatment conditions for the SSW. Knowledge o f this
behavior would influence the design of any systems, e.g., iron co-precipitation, built to remedate
the Se in the SSW streams.
A series o f samples o f a single SSW stream was buffered at several different pH levels to
determine the effect pH and time had on the SeCN. The study showed that the pH o f the streams
studied had to be strongly acidic, pH less than or equal to 2, to cause any significant change in
the concentration of the SeCN (Fig. 5).
A plot o f the results for the buffered solutions shows that, even after several days, there is not
an appreciable loss of the SeCN signal in the samples whose pH was greater than 2 (Fig. 6).
However, at a pH of 2 there was an appreciable loss o f SeCN over the course of several days
with the level dropping below the quantification limit after six days.
Knowing not only the level o f the selenium present but also its form and how it reacted to
various chemical treatments was critical to the success o f the remediation efforts. The

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

49

ruggedness of the separation technique coupled with the sensitivity and precision of the ICP-MS
guided the development of effective, efficient means for treating refinery streams.
z-60.7-

SeCNin HC1Matrix
pH = 2

~o-

InitialSeCN= 1180ppb

ao-

~o"

I Day1, SeCN=590ppb 1

lo

.... ~ .... , .... , .... , .... i .... , .... ~ .... ""'lb

.... ,,,,,I

.... ""',6

.... ' .... ~' .... ,,,,,&,,,,I,,,,~

.... ,,,'

FIG. 5--SeCN degradation at p H = 2.

FIG. 6---Low p H SeCN degradation.


Conclusion

ICP-MS offers sensitivity, accuracy, and precision with multi-element capabilities, which is
not available with other analytical techniques. The unique qualities of ICP-MS, when coupled
with proper chromatographic techniques, provides the analyst with concentration and nature of
the elemental species present in their samples. Other elemental techniques, which have been
coupled with chromatographic systems, do not exhibit the same robustness as has been
demonstrated with the ICP-MS [4,5]. Exploiting this ability, we were able to use ICP-MS
coupled with an anionic chromatography separation, to determine ideal conditions for
remediation of selenium contamination in various aqueous refinery streams.

50

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

References

[1] Brown and Caldwell, Western States Petroleum Association Removal Of Selenium By IronBased Treatment Processes Final Report, California Public Health Foundation, May 1995.
[2] Cotton, F. A. and Wilkinson, G., Advanced Inorganic Chemistry, A Comprehensive Text,
Second Edition, John Wiley & Sons, Inc., 1966.
[3] Roehl, R., "Recent Advances in the Speciation of Selenium in Petroleum Refinery and
Municipal Waste Waters Using ICP-MS and ICP-MS Coupled with Liquid
Chromatography," Transcript of Oral Presentation, 1996 Winter Conference on Plasma
Spectrochemistry, Ft. Lauderdale, FL, January 8-13, 1996.
[4] Zhang, X. H. and Koropchak, J. A.,AnaL Chem., 71, 3046, 1999.
[5] Roychowdhury, S. B. and Koropchak, J. A., AnaL Chem., 62, 484, 1990.

Journal of ASTM International, October 2005, Vol. 2, No. 9


Paper ID JAI12969
Available online at www.astm.org

S. Dreyfus, 1,2 C. Pdcheyran, 1 C Magnier, 2 A. Prinzhofer, 2 C. P. Lienemann, 2 and O. F. X.


Donard 1

Direct Trace and Ultra-Trace Metals Determination in Crude


Oil and Fractions by Inductively Coupled Plasma Mass
Spectrometry
ABSTRACT: A method was developed to analyze direct trace and ultra-trace metal elements in crude oil
and its fractions (maltenes-asphaltenes) by ICP-MS after sample dilution in xylene. Efficient introduction
of organic compounds requires addition of O2 for complete combustion of the sample; carbon deposit on
cones (interface) and extraction lenses was minimized by optimization of argon to oxygen ratio in the
plasma. A PFA-100 (100 ~tl.mn1) MicroFlow Nebulizer and a Scott "double pass" chilled spray chamber
were associated for an optimal introduction of petroleum products. The Q-ICP/MS was equipped with
platinum cones to limit drifts caused by cone corrosion during an organic analysis. A standard addition
method was realized for the calibration procedure in order to control the matrix effects. Method
validation was completed by analyzing three certified reference materials from the National Institute for
Standards and Technology (NIST). NIST 1085b, NIST 1084a, and NIST 1634c accuracies were
approximately 10 %. Detection limit of ultra trace elements in xylene were at pg.g-t level. A clean PFA
filter system was developed to separate the oil fractions (maltenes and asphaltenes) by precipitation of
asphaltenes (heavy fraction) in n-heptanc. For optimal detection conditions, the samples were diluted in
xylene according to their pre-estimated element concentration range. Total mass balance shows a
recovery close to 100 % for Ni, V, Cu, Mo, Ag, Sn, Ba, and Pb. These results show that elements
analyzed are highly concentrated in the asphaltenic fraction.

KEYWORDS: inductively coupled plasma mass spectrometry, organic solvents, microflow nebulizer,
trace and ultra trace metals in crude oils, maltenes, and asphaltenes
Introduction

In recent years, intensive studies have been dedicated to trace and ultra trace metals
determination in petroleum products. The need o f the petroleum industry in this field is highly
related to exploration and exploitation activities: geochemical characterization o f source rocks
and basins, and corrective actions during oil production and refining (e.g., prevention o f catalyst
poisoning, corrosion, and pollution control). The existence o f metals in fossil fuels was first
established b y Alfred Treibs in the 1930s [1]. In the 1960s, E. W. Baker has started his studies on
the characterization o f petroporphyrins [2]. Since then, m a n y works have been published on the
geochemistry o f organometallic complexes [e.g., Ni(II), VO(II), Cu(II), Fe(iII), Ga(III), Mn(III)
porphyrins] in geological materials (e.g., bitumens and marine sediments) [3,4]. Trace metals are
incorporated into crude oils as organometallic compounds (e.g., geoporphyrins), but also as
inorganic components originating from formation and production o f water (e.g., contamination
of B a and As) and mineral matter (e.g., trace elements in clay minerals) [5]. The inorganic or
Manuscript received 8 October 2004; accepted for publication 22 March 2005; published October 2005. Presented
at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
:LCABIE/CNRS UMR 5034, 2, Avenue P. Angot, 64053 Pan Cedex 9, France.
E-mail: sebastien.dreyfus@mtiv-pau.fr.
2Institut Frangais du Pdtrole, 1 et 4, Av. de bois Prdau, 92852 Rueil Malmaison Cedex, France.

Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

51

52

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

metal organic species of crude oils (generally less than 1%) is important for two reasons. Firstly,
a few organometallic species have damaging effects during processing and upgrading of crude
oils. Secondly, trace metals in crude oils may be used as geochemical biomarkers in oil-oil or oilsource rock correlation, in the identification of source rock depositional environments, and in the
quantification of oils' thermal maturity and biodegradation levels.
The evolution of elemental signatures, such as Ni and V, in crude oils has already been used
to determine the level of reservoir biodegradation. However, new parameters are needed to
improve the general knowledge about migration, maturation, mixing, and biodegradation of the
oil from an inorganic or bioinorganic angle. Sensitive methods are then required to analyze other
elements occurring in the petroleum in very low concentration (e.g., Mo, Pb, Cu, Cd, Sn, Ba,
etc.). The instrumental limitations of the most common techniques Used so far (AAS, ICP-OES)
have been the main reason for the poor knowledge about inorganic constituents ofoil [6].
Recently, inductively coupled plasma mass spectrometry (ICP-MS) has been used to analyze
trace elements in oil and other petroleum products [7-9]. Performances of ICP-MS allow the
detection of very low concentrations in an organic matrix. A sample introduction method in the
plasma is fundamental for the use of organic aerosol. The presence of organic vapors in the
plasma causes carbon deposit on cones and extraction lenses. Such deposits may affect the
transport efficiency of ions and plug cones, and they may even blow out the plasma. In addition,
polyatomic interferences caused by carbon-argon-oxides species do not allow the quantification
of certain elements (e.g., Mg, Cr," Fe) with a Q-ICP-MS. Moreover, a significant quantity of
organic vapors introduced in the ICP can also increase matrix effects [10,11]. In order to
minimize these problems, several solutions have been found, with more or less success.
Mineralization of petroleum samples can be performed in order to remove the organic matrix
[12]. However, this method has some disadvantages: the sample preparation is time consnming,
volatile elements can be easily lost, and the blank contamination may be important. Ultrasonic
nebulization (USN) used in addition to a desolvation membrane permits an efficient desolvation
of the organic aerosol [6,13]. Although this nebulizer allows one to achieve excellent detection
limits and to reduce carbon based interferences, certain metal species (e.g. volatiles species) can
be lost during the desolvation process [13]. Furthermore, crude oil and its fractions can easily
foul the USN and its desolvation membrane. A MicroFlow Nebulizer associated to a chilled
spray chamber appears to be the best introduction tool for an optimal analysis of petroleum
products by ICP-MS. This combination allows working with very low volumes of petroleum
products and minimizes the quantity of organic solvent entering into the plasma. The major
advantages of this technique are very low detection limits, no loss of metallic species, and no
contaminations [ 14]. This methodology has never been applied to the analysis of trace elements
in crude oil and its fractions by ICP-MS.
In this study, we have developed a technique to analyze directly trace and ultra trace metal
elements in crude oils and its fractions (maltenes-asphaltenes) by ICP-MS after sample dilution
in xylene. This method was then used to produce a new set of data, using the above-mentioned
methodology, on the distribution of metal trace elements between asphaltenes and maltenes in
crude oils from a Brazilian petroleum basin.

DREYFUS ET AL ON DIRECT TRACE AND ULTRA-TRACE

53

Materials and Methods

Instrumentation
Samples were analyzed on a X7 Series CCT ICP-MS (Thermo Electron Corporation,
Winsford, UK) equipped with a PFA-100 MicroFlow Nebulizer (Elemental Scientific Inc.,
Omaha, NE) fitted to a Scott-type double pass chilled spray chamber and with Platinum Sampler
and Micro Skimmer cones (Thermo Electron Corporation, Winsford, UK). Oxygen was supplied
continuously into the nebulizer gas flow for complete combustion of the sample and reduction of
carbon deposits on the interface cones. A 1.14-mm injector orifice de-mountable torch was fitted
to the instrument. It is essential when running organic solvents to increase the forward power to
help break apart the organic bonds. The power had to be set to 1550 W. The Q-ICP/MS was
equipped with platinum cones to limit drifts caused by cone corrosion during an organic analysis.
The ICP-MS operating conditions are listed in Table 1.
TABLE 1--ICP-MS operating conditions for xylene introduction.
ICP-MS system
ICP/MS X7 CCT THERMO| Electron
RF power (W)
1550
Cones (orifice diameter, mm) Sampler (Platinum, 1.0); Skimmer (Platinum, 0.7)
Spray chamber
Scott type double pass at -7~
Nebulizer
PFA-100 MicroFlow
Injector diameter (nun)
1.14
Ar plasma (l/rain)
15.5
Ar outer (1/mn)
1.1
Ar nebulizer (l/mn)
0.5
02 nebulizer (1/mn)
0.066
51~, 58-,.~-. 60~T- 62~T- 6 3 ~
65,'~ 9 8 ~ r
100~,
107-Isotopes measured (m/z)
V,
1N1, I N I ~ 1~11, k.,U, U U , lVlO~
_rVI.O, Ag,
l~
llSSn' ]2~ ' 137Ba' 138Ba' 2~ ' 207pb' 208pb
Data acquisition parameters
Resolution (amu)
0.8
Sweeps
100
Dwell time (ms)
20
Replicates
5

Reagents
Xylene "Baker Analyzed" ACS and n-Heptane "Baker Analyzed" were purchased from J. T.
Baker (MaUinckrodt Baker Inc., Phillipsburg, NJ). Membrane filters in Regenerated Cellulose
(0.45 ~tm, O: 47 ram, type 184) were obtained from Sartorius AG (Goettingen, Germany).
Conostan| S-21 Blended Standards (100 ppm organic multi element standard in oil) was
used in this work (ConocoPhillips, Houston, TX) for the samples' calibration. Method validation
was completed with three Standard Reference Materials (SRM) from NIST (National Institute
for Standards and Technology, Gaithersburg, MD): 1085b, 1084a lubricating oils, and 1634c
(residual'fuel oil).

54

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Oil Fractionation
In order to limit filtration blank contamination, a clean PFA filter unit was developed to
separate the oil fractions by precipitation of asphaltenes (heavy fraction) in n-Heptane.
Membrane filters of size 0.45 ~tm in regenerated cellulose were used with the filter unit. One
gram of crude oil was dissolved in 50 ml of n-H~ptane (in a 250 ml PFA balloon). The solution
was then heated for 20 rain at 60~ under magnetic stirring and finally filtered under vacuum.
Asphaltenes were accumulated on the filter and stocked in 50 ml polypropylene tubes at room
temperature in the dark for further analysis. Maltenes, which are soluble in n-H~tane, were
collected in a polyethylene Erlenmeyer and evaporated under pure nitrogen flow in 50 ml
polypropylene tubes.
Results and Discussion

Sample Introduction and Measurement


The sample introduction technique used to introduce organic aerosol into the ICP-MS is
critical for success. Experience shows that the minimization of solvent quantity brought in the
plasma and the oxygen to argon ratio are the critical parameters for analysis of mixed organic
products (e.g., diluted crude oil) by ICP-MS in order to reduce polyatomic interferences and
matrix effects. The spray chamber had to be chilled in order to reduce the amount of solvent load
entering the mass spectrometer. For the same reason, a 1.14 mm injector orifice de-mountable
torch had to be fitted to the instrument. With this torch the aerosol enters faster in the plasma.
Thus, the nebulization gas flow can be decreased without any loss of sensitivity in order to
reduce the quantity of aerosol introduced in the plasma. As a result, the signal detected by the
mass spectrometer is more stable, the signal to background noise ratio is better, and thus the limit
of detections are lower (Fig. 1). Detections limits (LOD-3o) for all trace elements in xylene
obtained with this torch are below the 0.3 ng.gq level (LOD are at pg.g-i level for ultra trace
elements such as Pb, Ba, Sn, etc.). LOD were calculated from the sensitivity (slope) of the
analyte addition curve and the standard deviation of a xylene blank.
Rinse efficiency of the spray chamber is also an important factor. A good rinsing efficiency
limits chamber fouling by heavy organic compounds and thus minimizes signal drifts and
memory effects on analytes. Figure 2 shows the signal intensity of three isotopes analyzed in
crude oil diluted 100 times in xylene as a function of stabilization, acquisition, and rinsing times.
A stabilization time of 2-3 min is necessary before measuring intensities. For H3Cd isotope, the
signal is still not stabilized after 3 min, and a drift of the intensity remains during all of the
acquiring time.
Matrix effects can be important during the analysis of petroleum products. They generally
disturb intensities emitted by a detected isotope [10]. Organic matrix may modify element
ionization in the plasma and consequently cause a variation of sensitivity. For these reasons, a
standard addition method was performed for the calibration procedure to control the matrix
effects. Concentrations of metals in analyzed organic samples must be compatible with ICP-MS
potential in order to obtain reliable results. Experience shows that concentrations detected by the
mass spectrometer ideally should be in the range of 1-100 ng/g. Thus, for optimal detection
conditions, samples were diluted in xylene according to their pre-estimated' element
concentration range.

DREYFUS ET AL. ON DIRECT TRACE AND ULTRA-TRACE

55

FIG. 1--Detection limits (3a) of trace elements in xylene obtained with the 1.14 mm injector
orifice torch compared to the detection limits obtained with the standard torch (1.5 mm orifice
injector).

FIG. 2--Intensity of98Mo, t13Cd, and 2~


rinsing times.

as a function of stabilization, acquisition, and

Validation of the Methodology

Accuracy and repeatability of the analytical procedure was assessed by analyzing certified
elements in three SRM, diluted 2000 (NIST 1085b), 800 (N-IST 1084a), and 200 (NIST 1634c)
times in xylene, and by using the standard addition method. Table 2 summarizes concentrations

56

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

o f trace elements and percentage error (E, %) obtained on NIST 1085b, 1084a, and 1634c and
compared to the certified values. Accuracies are better than 20 % for all elements and thus
permit to validate this analytical technique on the analysis of trace metals in petroleum products
by ICP-MS.
TABLE 2--Concentrations and percentage error (E, %) compared to the certified value
obtained on NIST 1085b, 1084a, and 1634c.
Element

Certified Values'
NIST 1085b (~tg.~') z

This Work (~tg..g


-1)'-

E (%)

V (51 V)

297.8 4.61

241 4

-19

Ni (62 Ni)
Cu (63 Cu)
Zn (66 Zn)
Mo (98 Mo)
Ag (107 Ag)
Cd (112 Cd)
Sn (118 Sn)
Ba (138 Ba)
Pb (208 Pb)

295.9 7.4
295.6 + 8.5
296.8 + 6.8
300.6 q- 3.2
304.6 8.9
302.9 + 5.1
299.4 ~- 4.8
300.1 + 2.4
297.7 6.8
NIST 1084a (~g.g-1)2

303 a: 6
313 :~ 7
303 4
255 :~ 7
340 4
347 ~: 5
244 ~: 5
307 :~ 8
289 :L 7

2
6
2
-15
11
14
-18
2
-3

V (51 V)
Ni (62 Ni)
Cu (63 Cu)
Mo (98 Mo)
Ag (107 Ag)
Sn (118 Sn)
Pb (208 Pb)

95.9 + 9.4
99.7 4- 1.6
100 + 1.9
100.3 4- 1.4
100.4 3.8
97.2 2.6
101 1.3
NIST 1634c (ixg.g'l)2

81 + 8
108 3
117 9 4
92 5
103 9 6
105 -~ 4
95 5

-16
8
17
-8
2
8
-6

V (51V)
28.2
26.7 0.8
-5
Ni (62 Ni)
17.5 0.2
16.2 0.6
-8
1All concentrationsin ~tg.g"1,errors on concentrationswere calculated from 95 % confidenceintervals~
~l'hose SRM are initially supplied for ICP-AES analysis. The certified values of NIST 1984a and NIST 1985b are
too high for an ICP-MS analysis (respectively 100 ~tg.g"l and 300 ~tg.g'l). Both SRM need to be highly diluted
before their introductionin the ICP-MS. This dilution may induce errors in the measurements.

Analysis of Crude Oil and lts Fractions, Total Mass Balance (TMB) and Repeatability of Oil
Fractionation
Once the methodology for trace element determination in petroleum products by ICP-MS
was validated, it was applied to several crude oils (supplied by Petrobr~ts) and their fractions
originating from a Brazilian petroleum basin. In order to prove the reliability o f analytical results
obtained after oil fractionation, total mass balance (TMB) was calculated for each element
measured. Table 3 shows an example o f crude oil and its fractions analyzed by ICP-MS. TMB
results were in the range o f 90-110 % for most o f the elements analyzed. In order to confirm the
repeatability o f oil fractionation and analysis, each sample (oil + fractions) was analyzed three

DREYFUS ET AL. ON DIRECT TRACE AND ULTRA-TRACE

57

times with a different sample preparation (including oil fractionation, samples' dilution, and
addition calibration procedure) for each time. This study indicates that trace metals, such as Ni,
V, Cu, Mo, Ag, Sn, Ba, and Pb, are highly concentrated in the asphaltenic fraction.
TABLE 3--Mean concentrations (in I~g/g), standard deviations (n = 3), and Total Mass
Balance (TMB) of trace elements obtained after fractionation o/'15 crude oil sample.
Element
V
Ni
Cu
Mo
Ag
Sn
Ba
Pb

Crude oil (lg)


11.5
17.1
1.07
0.088 0.008
0.019 4- 0.003
0.029 0.004
0.45
0.97 0.075

Asphaltenes 6 % Maltenes 94 %
944-2.1
5.4
146
8.7
9.3+0.3
0.51
0.98 0.11
0.044 0.004
0.31 0.042
< LOD
0.48 4- 0.057
< LOD
4.7
0.23
15.6 0.4
0,08 0.006

TMB 100 %
93
99
98
114
98
109
111
105

Acknowledgments
Petrobr~s is acknowledged for providing crude oils. Thermo Electron is acknowledged for
the loan of instrumentation.
References

[1]
[2]
[3]
[4]

[5]

[6]

[7]

[8]

Treibs, A., "Porphyrins in Coal," Ana. Chem., Vol. 520, 1935, pp. 144-151.
Baker, E. W., "Porphyrins," Organic Geochemistry, G. Eglinton and M. T. J. Murphy,
Eds., 1969, pp. 464-497.
Baker, E. W. and Louda, J. W., "Porphyrins Geochemistry of Atlantic Jurassic-Cretaceous
Black Shales," Organic Geochemistry, Vol. 10, 1985, pp. 905-914.
Eckardt, C. B., Wolf, M., and Maxwell, J. R., "Iron Porphyrins in the Permian
Kupferschiefer of the Lower Rhine Basin, N.W. Germany," Organic Geochemistry, Vol.
14, 1989, pp. 659-666.
Filby, R. H., "Origin and Nature of Trace Elements Species in Crude Oil, Bitumens and
Kerogen: Implications for Correlation and Other Geochemical Studies," Geofluids: Origin,
Migration and Evolution of Fluids in Sedimentary Basins, Geol. Soc. Sp. Pub., J. Parnell,
Ed., Vol. 78, 1994, pp. 203-221.
Duyck, C., Miekeley, N., Porto da Silveira, C. L., and Szatmari, P., "Trace Element
Determination in Crude Oil and Its Fractions by Inductively Coupled Plasma Mass
Spectrometry Using Ultrasonic Nebulization of Toluene Solutions," Spectrochimica Acta
Part B: Atomic Spectroscopy, Vol. 57, 2002, pp. 1979-1990
A1 Swaidan, H. M., "The Determination of Lead, Nickel and Vanadium in Saudi Arabian
Crude Oil by Sequential Injection Analysis/Inductively-Coupled Plasma Mass
Spectrometry," Talenta, Vol. 43, 1996, pp. 1313-19.
Hausler, D., "Trace Element Analysis of Organic Solutions Using Inductively Coupled
Plasma-Mass Spectrometry," Spectrochimica Acta Part B, Vol. 42, 1987, pp. 63-73.

58

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

[9] Kahen, K., Strubinger, A., Chirinos, J. R., and Montaser, A., "Direct Injection High
Efficiency Nebulizer-Inductively Coupled Plasma Mass Spectrometry for Analysis of
Petroleum Samples," Spectrochimica Acta Part B, Vol. 58, 2002, pp. 397-413.
[10] Botto, R. I., "Matrix Interferences in the Analysis of Organic Solutions by Inductively
Coupled Plasma-Atomic Emission Spectrometry," Spectrochimica Acta, Vol. 42B, 1987,
pp. 181-189.
[11] Botto, R. I., "Method for Correcting for Acid and Salt Matrix Interferences in ICP-AES,"
Spectrochimica Acta, Vol. 40B, 1985, pp. 397-412.
[12] Wondimu, T., Goessler, W., and Irgolic, K. J., "Microwave Digestion of "Residual Fuel
Oil" (NIST SRM 1634b) for the Determination of Trace Elements by Inductively Coupled
Plasma-Mass Spectrometry," Frenesius, Vol. 367, 2000, pp. 35-42.
[13] Botto, R. I. and Zhu, J. J., "Use of an Ultrasonic Nebulizer with Membrane Desolvation for
Analysis of Volatile Solvents by Inductively Coupled Plasma Atomic Emission
Spectometry," Journal of Analytical Atomic Spectrometry, Vol. 9, 1994, pp. 905-912.
[14] McElroy, F., Mennito, A., Debrah, E., and Thomas, R., "Uses and Application of
Inductively Coupled Plasma Mass Spectrometry in the Petrochemiehal Industry,"
Spectroscopy, Vol. 13, No. 2, 1998, pp. 43-52.

Journal of ASTM International, July/August 2005, Vol. 2, No. 7

Paper ID JA112976
Available online at www.astm.org
Paolo TittarellL 1Marco Priola, 1 Silvia Ricchiuto, ~Dmitri A. Katskov, 2 and Prince Ngobeni 2

Fuel Analysis by Filter Furnace Electrothermal Atomic


Absorption Spectrometry
ABSTRACT: The determination of Mn, Fe, Co, and Zn was performed in gasoline, jet, and diesel fuel

samples by electrothermal atomic absorption spectrometry using the Transverse Heated Filter Atomizer
(THFA). Thermal conditions were experimentally defined for the investigated elements.
The elements were analyzed without the addition of chemical modifiers, using organometallic standards
for the calibration. Gasoline samples were analyzed directly, while jet and diesel fuel samples were
diluted 1+3 with n-beptane.
The following LODs were obtained using 40 ~tL injections: 0.08 ~tg/LMn, 0.35 lag/LFe, 0.54 ~tg/LCo,
and 0.06 I.tg/LZn. The limits of determination, calculated for diesel and jet fuel samples, were 1.2 ~g/kg
Mn, 4.9 ~tg/kgFe, 7.6/ag/kg Co, and 0.8 I~g/kgZn. The corresponding limits of quantitation for gasoline
samples were about four times higher.
KEYWORDS: electrothermal atomic absorption, transverse heated filter atomizer, gasoline, diesel fuel,

trace elements
Introduction

Electrothermal atomic absorption represents a suitable instrumental technique for the analysis
o f elements in petroleum products [1-5]. The technique shows very low detection limits, similar
to or even better than those found for other spectroscopic techniques, such as Inductively
Coupled Plasma Atomic Emission Spectrometry (ICP-AES) [6] and ICP-Mass Spectrometry
(ICP-MS) [7-12]. Some problems were evidenced in the use o f ETAAS when elements like Ni
[3] and Pb [13] are analyzed, due to the different behavior o f organo-metallic species during the
thermal treatment o f standards and samples.
Recently, the use o f transverse heated filter furnace for the analysis o f trace elements in
automotive fuels was proposed [14]. This instrumental technique showed interesting results for
volatile elements, such as Cd and Pb. The filter furnace approach was the subject o f several
papers and o f continuous developments in order to improve the design [15], to extend the range
to low volatile elements [16], to employ the furnace in transverse heated atomizers [17], and to
determine trace elements in aqueous and organic solutions [ 18,19].
The most interesting aspects o f the filter furnace concept are: the capability to inject large
volumes due to the large volume o f the zone allocated to the sampling, the limited non-specific
absorption b y molecular species due to the filtering action o f the furnace, and no need for
chemical modifiers due to the furnace design.

Manuscript received 6 October 2004; accepted for publication 29 December 2004; published July 2005. Presented
at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
i Stazione Sperimentaleper i Combustibili, San Donato Milanese MI, 20097, Italy.
2 Tshwane University of Technology, Pretoria, 0001, South Africa.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

59

60

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Furthermore, in the case of petroleum products, the high surface area of the substrate in the
furnace can be wet easily and homogeneously by the organic medium.
In particular, the injection of large volumes, up to 100 IxL, can be loaded onto the fiber of the
filter furnace and can be carried out without any particular problems, while the loading of
organic solution samples greater than 15-20 ~tL onto conventional THGA platforms can lead to
analytical problems, due to the spread of the liquid on the graphite surface [20].
The aim of this investigation was to extend the range of analyzed elements, to examine
petroleum products other than gasoline and diesel fuel (namdy jet fuel and fuel oil), to verify the
stability of some liquid fuels as far as the determination of trace dements is concerned, and to
validate the THFA procedure using standard reference materials and the comparison with an
independent technique (microwave digestion followed by ICP-MS measurements).
The behavior of manganese, iron, cobalt, and zinc was investigated. Copper was also
examined to compare the actual figures of merit with those obtained in a previous investigation
on automotive fuels using THFA [ 14].

Experimental
Instrumentation
A Perkin Elmer SIMAA 6000 atomic absorption spectrometer, equipped with Perkin Elmer
hollow cathode lamps (Co, Cu, Fe, Mn, and Zn), was employed. The instrument was set up
according to the manufacturer's instructions [21] and run in the single element mode.
Filter furnaces were assembled according to the procedure outlined in Ref. [17]. The design
of the THFA is reported in Fig. 1. In particular, the platform was removed from the standard
Transverse Heated Graphite Atomizer (THGA) tube, and a filter made of porous graphite in the
form of a spool was inserted in the tube. The filter was made from a 6 mm rod of AGKSP
(Union Carbide) graphite with 28 % porosity. The total length of the filter was 17 ram, and the
collector length was 6 mm. The collector was filled with carbon fiber (Levertex G-868), by
wrapping the fiber around the filter spool. The average fiber loading was about 15 mg. Before
use, the fiber was treated with a stream of chlorine to remove trace contaminants.

FIG. 1--Transverse heated filter furnace: a) tube, b) filter, c) collector.


The sample, injected through the dosing hole, distributes along the surface of the fiber that is
much wider than that of a conventional platform. In particular, petroleum products
homogeneously distribute onto the fiber surface, due to the wetting capability of the organic
medium. During the heating of the external tube, the sample is forced to migrate through the
porous graphite of the filter. The filtering action of the porous graphite strongly reduces the
formation of molecular species and non-specific absorption during the atomization step. While in
the case of THGA sampling vaporization and atomization occur in the same tube zone, in the

TITTARELLI ET AL. ON FUEL ANALYSIS

61

case of THFA sampling, vaporization and atomization occur in three distinct locations of the
tube, i.e., collector, filter, and analytical zone.
A Perkin-Elmer SCIEX ELAN 9000 ICP-MS was employed for element determination after
microwave digestion of fuel samples, performed using an Anton Paar Multiwave 3000
microwave digestion system.

Reagents
The multielement standard in oil Conostan $21 (Conoco, Ponca City, OK) was employed for
calibration and recovery studies. N-heptane was used as the solvent (Merck, Darmstadt,
Germany). The standard 10 mg/kg content of each element was diluted with n-heptane to obtain
the required concentration of element.
The microwave digestion was performed using nitric acid Tracepure and hydrochloric acid
Suprapure (Merck, Darmstadt, Germany) and hydrogen peroxide (J. T. Baker, Deventer,
Holland).
NIST 1634c Trace elements in fuel oil were employed to verify the accuracy of THFAETAAS and ICP-MS measurements.

Procedure
ETAAS Measurements--The following wavelengths were employed: Mn 279.5 nm, Fe 248.3
nm, Co 242.5 nm, Zn 213.9 nm, and Cu 324.8 nm.
Gasoline samples were examined directly, without dilution with solvent. Diesel fuel and jet
fuel samples were diluted with n-heptane 1+3 by mass.
The thermal program employed is reported in Table 1. The vaporization temperature (step 2)
was set for all types of petroleum products, while pyrolysis and atomization temperatures were
chosen according to the thermal properties of element under investigation.
Forty ~tL of sample were usually injected, although THFA furnace shows excellent results in
vaporizing volumes up to 100 ~tL [17,18]. However, the sample volume was limited to 40 IxL, to
achieve a suitable compromise between improvement in detection limit and steadiness in
analytical results.
TABLE 1--Instrumental conditions for THFA-ETAAS.
Step
Temperature, 0C ..... Ramp, s
Holdl s
1

110

2
400
20
20
3
*
5
5
4
*
0
*
5
2450
1
5
*Temperatureand hold time set accordingto the element. Gas stop during atomization,40 #L injection.
The determination of all elements was performed without the addition of chemical modifiers.
Three replicate readings were performed for each determination.
Determinations were also carried out in consecutive days on jet and diesel fuel samples to
evaluate the stability of samples spiked with known amounts of trace elements.

ICP-MS measurements--Microwave digestion followed by ICP-MS was performed on SRM


1634c samples. About 0.2 g were digested in microwave oven using 6 mL nitric acid, 1 mL

62

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

hydrochloric acid, and 2 mL hydrogen peroxide, following the procedure outlined in Ref. [22].
The power program is reported in Table 2. The acidic solution was diluted to 25 mL with
deionized water. Further dilution was required according to the elements under investigation.
The instrumental conditions of ICP-MS measurements are reported in Table 3. Rhodium was
used as the internal standard.
TABLE 2--Powerprogramformicrowave digestion.
Step
Power, W
Time, min
1
0-1100
30
2
1100
10
3*
0
20
* Systemcoolingwith air flow.
TABLE 3--Instrumenta ! conditions for ICP-MS measurements.
RF Power
1200 W
Argon flow rates
plasma
15 L/min
nebulizer
0.5 L/min
Scan mode
peak hopping
Sweeps/reading
20
Readings/replicate
1
Replicates
5
Dwell time
50 ms
Results and Discussion

The main difference between gasoline and diesel fuel, as far as ETAAS are concerned, is the
sample volatility. Gasoline sample can be injected directly in the electrothermal atomizer and
vaporized without formation of any residue, while diesel fuel undergoes partial pyrolysis during
the thermal treatment before the atomization occurs. The pyrolysis can lead to the formation of a
carbonaceous residue. The carbonaceous buildup affects the atomization of carbide forming
elements such as Ni and V [3]. In this work, diesel fuel samples were diluted with a suitable
solvent to reduce the formation of carbon residues.

Manganese
Pyrolysis and atomization curves of manganese are shown in Fig. 2. The pyrolysis curve was
obtained using 2300~ as the atomization temperature. The selected pyrolysis temperature for
Mn in THFA is about 1600~ Using this pyrolysis temperature, the optimum atomization
temperature corresponds to 2100~
These pyrolysis and atomization temperatures do not
correspond to the THGA suggested temperatures [21], as the furnace design is different and the
rate of the vaporization process is also different. Some loss on manganese is observed above
2100~ due to the volatility of the atomic species at high temperature.
A typical atomization peak of manganese is shown in Fig. 3. In this case, due to the very low
Mn content in diesel fuel, manganese is added to the diesel, so that 0.2 ng Mn are injected
together with the sample aliquot. The peak shape shows some tailing till the end of the
atomization step. The background is very limited and almost continuous during the step.

TITTARELLI ET AL. ON FUEL ANALYSIS

63

0.~.$ -

0.L5

0.10

0.85

0.00
1300

17OO

2100

T~l~m~

2500

('C)

FIG. 2--Pyrolysis (~) and atomization (/X) curves of manganese.

,%
/ \

0.08

--

atomicabsorption

~0.04

<i

0
0

10

Time (s)
FIG. 3--Manganese atomization peak. Diesel#el spiked with 0.2 ng Mn. Pyrolysis 1600~
atomization 2100~
The figures of merit of Mn are shown in Table 4. The limit of detection (LOD) is calculated
as three times the standard deviation of the average of six blank runs. The limit of quantitation
(LOQ) is calculated as three times LOD. The data reported in Table 4 are referred to diesel
samples that are diluted 1+3 with n-heptane. The corresponding LOQ for gasoline samples are
four times lower, as gasoline is injected without preliminary dilution.
TABLE 4---Limits of detection and quantitation for diesel fuel.
LOD, #g/L a
LOQ,/zgJkg o
Mn
0.08
1.2
Fe
0.35
4.9
Co
0.54
7.6
Zn
0.06
0.8
a 3a of n-heptanr (n=6); b 3-LOD.

64

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Iron
Pyrolysis and atomization behavior of iron are reported in Fig. 4. The optimum conditions
employed for further measurements are 1600~ and 2300~ for pyrolysis and atomization steps,
respectively.
Figure 5 shows a tailing of the atomization peak more pronounced than that observed for
manganese. Some memory effect can be expected if a cleaning step is not performed at the end
of the vaporization cycle. Background absorption is very limited.
o~.

3 ~
0.20"

~ o.t~-

.e!
~ O.lO.
~ o.115.
0.~
1300

1700
2100
T e ~ ' a l m ~ ('C)

2500

FIG. 4--Pyrolysis ( I ) and atomization (lk) curves of iron.

FIG. 5--Iron atomization peak. Diesel fuel spiked with 0.6 ng Fe. Pyrolysis 1600~
atomization 2300~
As in the case of manganese, 0.6 ng iron are added to the diesel fuel, due to the very low
content in most samples. As already reported for some elements, the content of trace elements in
modem automotive fuels is very low, often below the quantitation limit of THFA-ETAAS [3].
The limits of detection and of quantitation for iron in diesel fuel are reported in Table 4.

TITTARELLI ET AL. ON FUEL ANALYSIS

65

Cobalt
The thermal properties of cobalt in THFA are shown in Fig. 6. The pyrolysis temperature can
be set to 1600~ However, the continuous increase of the atomization curve indicates that cobalt
atomization is not quantitative even at the maximum temperature allowable to the THFA tube
(2500~
The atomization profiles (Fig. 7) evidence a remarkable tailing of peaks. A behavior similar
to that of cobalt was observed for nickel [3]. Both elements in fact vaporize at high temperature.
The continuous increase of the atomization curve and the contemporaneous tailing can be
attributed, as for nickel, to the formation of element carbides in the graphite tube. This situation
was evidenced in the analysis of vacuum gas oils in a conventional end-heated atomizer [3].
0.25.

0.'~0 -

0as.
~

0.I0,

0.05 -

0.00
1000

6
1500

'2000

:aSl)O

? m p ~ a ~ (*r

FIG. 6--Pyrolysis (ll) and atomization (Z~) curves of cobalt.

FIG. 7--Cobalt atomization peak. Diesel fuel spiked with 1.6 ng Co. Pyrolysis 1600~
atomization 2400~
In the case of cobalt, 1,6 ng cobalt are added to the diesel sample, as the element shows a
lower sensitivity in comparison to manganese and iron. Limit of detection and quantitation of
cobalt are affected by the low sensitivity of THFA for low volatile elements (Table 4).

66

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

~nc

Zinc is extremely sensitive to temperature changes of pyrolysis and atomization steps as


indicated in Fig. 8.
The highest pyrolysis temperature without element loss can be set to about 700~ The
atomization curve shows a maximum at about 1800-1900~
The sensitivity to thermal conditions is reflected in the atomization profiles that exhibit
uneven behavior (Fig. 9).
As for all elements investigated, background is very limited. Due to the high volatility of
zinc, LOD and LOQ (Table 4) are very low and comparable with those obtained with
conventional THGA tubes [21 ].
0.2

0,15

,.~ 0,1

~ O.O5

~0

1000

1500

2000

2500

Temloemtu-re(*C)
FIG. 8--Pyrolysis (R) and atomization (/x) curves of zinc.

FIG. 9--Zinc atomization peak. Diesel fuel spiked with 0.1 ng Zn. Pyrolysis 700~
atomization 1900~

Performance of THFA
The limits of detection obtained with THFA-ETAAS are listed in Table 5 with those
obtained recently using various instrumental techniques.

TITTARELLI ET AL. ON FUEL ANALYSIS

TABLE 5--Detection limits of analytical techniques.


ETV-ICP-MS
USN-ICP-MS
DIHEN-ICP-MS
ETAAS
Gasoline [9]
SRM 1634c [ 1 1 1
xylene [ 12]
gasoline [4]
Mn
0.02
0.6
Cu
0.22
0.1
0.4
Pb
0.2
1.46
1.8
Fe
0.5
3.0
Ni
0.38
0.4
0.8
0.8
USN: UltrasonicNebulizer; DIHEN:DirectInjectionHigh EfficiencyNebulizer.

67

THFA
heptane
0.08
0.27 [14]
0.37 [14]
0.35
0.6 [14]

In general, THFA-ETAAS compares well with conventional ETAAS and various ICP-MS
approaches. Detection limits of copper and nickel were obtained in a previous work focused on
automotive fuels investigation.

THFA Tubes Lifetime


The lifetime of a THFA tube was evaluated by consecutive 40 #L injections of a 40 #g/L
copper solution in n-heptane, using instrumental conditions already defined in Ref. [14]. Figure
10 shows a slight increase in peak area absorbance, possibly due to some memory effect. The
lifetime can be considered to last around 900-1000 firings. The visual inspection of THFA
components shows limited corrosion of the ends of the filter. However, the used fiber can be
removed and replaced by a new one.
0.35

0 . 3 0

~0.25
lad

0.20
200

400

600
Fi r ~

800

i0~0

1200

no .

FIG. 1O---Lifetime of THFA tubes.

Stability of Fuel Solutions


Jet and diesel fuel samples were spiked with known concentrations of manganese, copper,
and iron and stored in dark bottles (20 #g/L spike). The element content was checked over a
period of 25 days using calibration solutions prepared daily. The results obtained for manganese
are reported in Fig. 11. While manganese content in diesel fuel samples remains constant during
the timeframe investigated, the element content in jet fuels drops dramatically after about two

68

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

weeks. The same behavior holds for copper and iron. Hence, the time elapsed between jet fuel
sampling and analysis should be kept to a minimum to avoid element loss and erroneous results.
The same behavior of jet fuel can be postulated for gasoline.
25

....
0

10

15

~0

25

30

Tiaue(da~)
FIG. 11--Stability of manganese in jet fuel ( A ) and dieselfuels (~ , Vq).

Comparison with ICP-MS Measurements


Fuel oil SRM 1634c was examined to compare the accuracy of the results obtained by
THFA-ETAAS with those obtained by ICP-MS after microwave digestion (Table 6). The
average of three independent determinations is shown in the Table.
Manganese and copper contents obtained by THFA are higher than those obtained by ICPMS. Vanadium and nickel were not examined by TI-IFA due to the high atomization temperature
required and the memory effects associated with the incomplete atomization of these elements.
TABLE 6--SRM 1634c. Comparison between THFA-ETAAS and ICP-MS.
Certified, mg/kg
THFA
ICP-MS
Mn
0.35 + 0.03
0.27 0.02
Cu
0.23 0.02
0.13 0.04
Pb
0.10 0.02
0.12 0.03
Co
0.1510 0.0051
0.16 0.01
0.12 0.02
V
28.19 0.40
28.0 1.0
Ni
17.54 0.21
16.3 0.4

Conclusions
The THFA approach is suitable for liquid fuel analysis. In particular, the injection of large
sample volumes represents an advantage for ultra-trace analysis. The handling of large sample
volumes still has to be exploited in its full capability to enhance the analytical performance of the
technique. However, problems are encountered when stable elements, such as V and Ni, are
analyzed.
The validation of instrumental procedures for trace element analysis at /zg/kg levels is
hindered by the lack of automotive/aviation fuel SRMs. Furthermore, light fuels (gasoline/jet

TITTARELLI ET AL. ON FUEL ANALYSIS

69

fuel) are prone to element loss. Nevertheless, the stability of diesel fuels is quite promising for
possible SRMs development.
References

[1] ASTM Standard D 6732-02, "Standard Test Method for Determination of Copper in Jet
Fuels by Graphite Furnace Atomic Absorption Spectrometry," ASTM International, West
Conshohocken, PA.
[2] Silva, I. A., Campos, R. C., and Curtius, A. J., "Determination of Lead and Copper in
Kerosene by Electrothermal Atomic Absorption Spectrometry: Stabilization of Metals in
Organic Media by a Three-Component Solution," Journal of Analytical Atomic
Spectrometry, Vol. 8, 1993, pp. 749-754.
[3] Anselmi, A. and Tittarelli, P., "Determination of Nickel in Petroleum Heavy Ends by
Electrothermal Atomic Absorption Spectrometry," La Rivista dei Combustibili, Vol. 48,
1994, pp. 435-441 (in Italian).
[4] Campos, R. C., Santos, H. R., and Grinberg, P., "Determination of Copper, Iron, Lead,
Nickel in Gasoline by Electrothermal Atomic Absorption Spectrometry Using ThreeComponent Solutions," Spectrochimica Acta Part B, Vol. 57, 2002, pp. 15-28.
[5] Aucelio, R. Q. and Curtius, A. J., "Evaluation of Electrothermal Atomic Absorption
Spectrometry for Trace Determination of Sb, As and Se in Gasoline and Kerosene Using
Microemulsion Sample Introduction and Two Approaches for Chemical Modification,"
Journal of Analytical Atomic Spectrometry, Vol. 17, 2002, pp. 242-247.
[6] Botto, R. I., "Applications of Ultrasonic Nebulization in the Analysis of Petroleum and
Petrochemicals by Inductively Coupled Plasma Atomic Emission Spectrometry," Journal
of AnalyticalAtomic Spectrometry, Vol. 8, 1993, pp. 51-57.
[7] van Heuzen, A. A., "ICP-MS in the Petroleum Industry," Application of Inductively
Plasma Mass Spectrometry, A. R. Date and A. L. Gray, Eds., Blackie, London, 1989,
Chapter 7.
[8] A1-Swaidan, H. M., "Trace Determination of Vanadium and Nickel in Saudi Arabian
Petroleum and Petroleum Products by Microemulsion ICP-MS," Atomic Spectroscopy,
Vol. 14, 1993, pp. 170-173.
[9] Kumar, S. J. and Gangadharan, S., "Determination of Trace Elements in Naphtha by
Inductively Coupled Plasma Mass Spectrometry Using Water-in-Oil Emulsions," Journal
of Analytical Atomic Spectrometry, Vol. 14, 1999, pp. 967-971.
[10] Saint Pierre, T. D., Dias, L. F., Pozebon, D., Aucelio, R. Q., Curtius, A. J., and Welz, B.,
"Determination of Cu, Mn, Ni and Sn in Gasoline by Electrothermal Vaporization
Inductively Coupled Plasma Mass Spectrometry, and Emulsion Sample Introduction,"
Spectrochimica Acta B, Vol. 57, 2002, pp. 1991-2001.
[11] Duyck, C., Miekeley, N., Porto da Silveira, C. L., and Szatmari, P., "Trace Elements
Determination in Crude Oil and Its Fraction by Inductively Coupled Plasma Mass
Spectrometry Using Ultrasonic Nebulization of Toluene Solution," Spectrochimica Acta
PartB, Vol. 57, 2002, pp. 1979-1990.
[12] Kahen, K., Strubinger, A., Chirinos, J. R., and Montaser, A., "Direct Injection High
Efficiency Nebulizer-Inductively Coupled Plasma Mass Spectrometry for Analysis of
Petroleum Samples," Spectrochimica Acta Part B, Vol. 58, No. 200, pp. 397-413.
[13] Kowalewska, Z., Bulska, E., and Hulanicki, A., "Organic Palladium and PalladiumMagnesium Chemical Modifiers in Direct Determination of Lead in Fractions from

70

[14]

[15]
[16]
[17]
[18]
[19]

[20]
[21]
[22]

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Distillation of Crude Oil by Electrothermal Atomic Absorption Analysis," Spectrochimica


Acta Part B, Vot. 54, 1999, pp. 835-843.
Anselmi, A., Tittarelli, P., and Katskov, D. A., "Determination of Trace Elements in
Automomotive Fuels by Filter Furnace Atomic Absortion Spectrometry," Spectrochimica
Acta Part B, Vol. 57, 2002, pp. 403-411.
Katskov, D. A., Marais, P. J. J. G., and Tittarelli, P., "Design, Operation and Analytical
Characteristics of the Filter Furnace, a New Atomizer for Electrothermal Atomic
Absorption Spectrometry," Spectrochimica Acta Part B, Vol. 51, 1996, pp. 1169-1189.
Katskov, D. A., Marais, P. J. J. G., Katkovnik, V. J., and Tittarelli, P., "Adaptation of the
Filter Furnace Atomizer for Atomic Absorption Determination of Less Volatile Metals,"
Spectrochimica Acta Part B, Vol. 52, 1997, pp. 1377-1394.
Katskov, D. A., Marais, P. J. J. G., and Ngobeni, P., "Transverse Heated Filter Atomizer
for Electrothermal Atomic Absorption Spectrometry," Spectrochimica Acta Part B, Vol.
53, 1998, pp. 671-682.
Marais, P. J. J. G., Panichev, N. A., and Katskov, D. A., "Performance of the Transverse
Heated Filter Atomizer for the Atomic Absorption Determination of Mercury," Journal of
Analytical Atomic Spectrometry, Vol. 15, 2000, pp. 1595-1598.
Mbileni, C., Ngobeni, P., Katskov, D. A., and Panichev, N., "Determination of Lead and
Cadmium in Organic Solutions by Electrothermal Atomic Absorption Spectrometry with a
Transverse Heated Filter Atomizer," Journal of Analytical Atomic Spectrometry, Vol. 17,
2002, pp. 236-241.
Tserovsky, E. and Arpadjan, S., "Behaviour of Various Organic Solvents and Analytes in
Electrothermal Atomic Absorption Spectrometry," Journal of Analytical Atomic
Spectrometry, Vol. 6, 1991, pp. 487-491.
The THGA Graphite Furnace, Techniques and Recommended Conditions, PerkinElmer,
13berlingen, 1999.
Bettinelli, M., Spezia, S., Baroni, U., and Bizzarri, G., "Determination of Trace Elements
in Fuel Oils by Inductively Coupled Plasma Mass Spectrometry after Acid Mineralization
of the Sample in a Microwave Oven," Journal of Analytical Atomic Spectrometry, Vol. 10,
1995, pp. 555-560.

Journal of ASTM International, March 2005, Vol. 2, No. 3


Paper ID JAI12970
Available online at www.astm.org

Malte Lukas, 1 Robert J. Yurko, 2 and Daniel P. Anderson 3

Rotrode Filter Spectroscopy: A Recently Improved Method to


Detect and Analyze Large Wear and Contaminant Particles in
Fluids
ABSTRACT: Routine spectrochemical analysis of used oil samples with rotating disc electrode (RDE) or
inductively coupled plasma (ICP) emission spectrometers has become standard practice in machine condition
monitoring programs to determine elemental content. Unfortunately, both of these techniques become less
responsive to particles as they increase in size. To overcome this limitation, several complementary methods
and analytical techniques to detect large particles in used oil have been applied, including acid digestion,
ferrography, and X-ray analysis. Unfortunately, these techniques are not practical or economical to implement
in a high sample throughput machine condition-monitoring program. One technique that has demonstrated
success as a complementary method for detecting large particles is rotrode filter spectroscopy (RFS). RFS
extends the capability of rotating disc electrode (RDE) spectrometers to detect particles greater than 10 lam in
size,

KEYWORDS: Rotrode Filter Spectroscopy (RFS), rotating disc electrode spectrometer, RDE, used oil

analysis
Introduction

Spectrometric oil analysis has been applied for more than 40 years as a routine and costeffective condition monitoring technique [1]. It is used to determine the elemental concentration
in parts per million o f wear metals, contaminants, and additives in a used oil sample.
Commercial oil analysis laboratories report on as many as 20 different elements. With the
knowledge o f the wear metal and contaminant limits for the machine or engine being monitored,
a determination may be made as to whether or not that equipment is operating properly.
A fact that has recently become more widely understood is that spectrometric oil analysis
detection efficiency decreases as the wear particle size increases [2]. This partial limitation is
not a problem with reciprocating engines that generate mostly small particles, but it is sometimes
a problem in fatigue failures o f rolling element bearings, such as those in military gas turbines,
which generate large particles at the outset o f failure without generating many small particles. In
addition, some wear modes such as spalling, severe sliding wear, and cutting wear generate large
particles that may go undetected b y spectroscopy [3]. This is true to a greater degree for certain
types o f spectrometers than others. It is also more important for certain types o f machines and
engines.
Atomic absorption spectroscopy (AAS) and inductively coupled plasma (ICP) spectrometers
suffer the most from particle detection inadequacies.
Rotating disc electrode (RDE)
spectrometers are responsive to somewhat larger particles, but the upper limit is still
approximately 10 ~tm [4]. Ferrography and X-ray analyzers address this limitation and are
Manuscript received 6 October 2004; accepted for publication 28 November 2004; published March 2005.
Chief Technology Advisor, Spectro Inc., Littleton, MA 01460.
2 V.P. R&D and Engineering, Spectro Inc., Littleton, MA 01460.
s V.P. Sales and Marketing, Spectro Inc., Littleton, MA 01460.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

71

72

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

capable of identifying large wear particles as a complementary technique. However, these


instruments are not always practical or cost-effective in a used oil analysis laboratory
environment that competes on the price of samples and may process hundreds or even more than
1000 samples per day.
The RFS method makes use of the fact that carbon disc electrodes used in RDE
spectrometers are porous and can be used as a filter to trap particles in oil samples. Early
versions of the RFS fixture used standard disc electrodes with some success and acceptable
sensitivity [5]. However, standard disc electrode geometry caused the process to be slow, and
lack of manufacturing control of disc porosity resulted in poor repeatability. Due to its ease of
use and inherent advantages, the RFS concept was recently revisited with the intent of further
improving performance. The updated RFS technique uses custom ring-shaped graphite
electrodes and a semi-automatic process to improve sensitivity and repeatability and to reduce
sample preparation time.
This paper describes the theory of operation of the RFS technique, the evolution of the
concept based on experience, and how it is applied in commercial used oil analysis laboratories.

Theory of RFS
A microscopic inspection of the carbon disc electrodes used in rotating disc electrode (R.DE)
spectrometers reveals that they are porous. Rotrode filter spectroscopy (RFS) makes use of this
fact, and a funnel-based fixture (Fig. 1) is used to clamp the discs so that used oil samples can be
drawn through the outer circumference of the disc when a vacuum is applied. The filtering
process through the disc captures the particles in the oil. The oil is then washed away with
solvent, and the disc is allowed to dry. The particles are left on the outer circumference of the
disc electrode so that they are vaporized and detected when sparked on the RDE spectrometer.

FIG. 1--RFS disc electrode filtering oil through its circumference.

LUKAS ET AL. ON ROTRODE FILTER

73

The RFS technique is used as a comparative method due to the unavailability of oil standards
with known gravimetric concentration of particles for each element measured by the
spectrometer. In practice, a used oil sample is first analyzed using the standard RDE technique,
which provides an analysis of dissolved and small wear particles. A second analysis of the same
sample using the RFS technique detects large particles. The two analyses provide an indication
of the wear particle size distribution in the sample. A sudden presence of large wear particles
will not be seen by conventional analysis alone. Their presence, however, will be readily evident
through the RFS analysis.

The RFS Fixture


The original RFS fixture to prepare electrodes for the analysis of large particles was first
developed in the early 1990s (Fig. 2). The fixture used standard disc electrodes, which were
clamped in the funnels to filter and capture large particles. It also had five stations so that five
samples could be processed simultaneously. A vacuum pump was used to pull the sample and
filter it through the electrode. A manual rinse with heptane was used to wash away any
remaining lubricant. Sample preparation time of an electrode for RFS analysis varied with the
viscosity and contamination of the oil sample. Preparation time could be as short as 4-5 min for
relatively clean, used lubricating oil samples, such as from turbines, electric motor bearings, and
hydraulic systems. Engine oil samples with high soot levels require the longest filtration times,
sometimes half an hour or more.

FIG. 2--Early RFSfixture.

74

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

It has always been a problem that standard disc electrodes are not manufactured with
consistent porosities, which affects the capture efficiency of the particles. This variability hurts
repeatability and thus makes data trends, on which used oil analysis maintenance
recommendations are based, more difficult to establish. Furthermore, long sample preparation
times for highly sooted samples from diesel engines were not practical for some laboratories.
These factors led to a new system that addressed these concerns and offered additional
improvements. The updated and improved system is known as automated rotrode filter
spectroscopy (A-RFS) (Fig. 3). It greatly improved repeatability through the use of controlled
electrodes, automation, and a vacuum/pressure pump to reduce sample preparation times.

FIG. 3--A-RFS system.


The A-RFS system is a semi-automated sample preparation instrument with a five-station
fixture of funnels. It works on the same principle as the original RFS system. However, it uses a
special electrode manufactured to more consistent porosity specifications, it has an automated
cleaning cycle, and its filtration times are reduced by the application of a vacuum/pressure pump
to both pull and push the used oi ! sample through the electrode.
Conventional off-the-shelf electrodes were used as a filter in the original RFS method.
Experimental data over time confirmed that the porosity of these electrodes was not controlled in
the manufacturing process, and thus it had a detrimental effect on repeatability when new
batches of electrodes were purchased. The new A-RFS sample preparation process begins with a
specially designed RFS electrode. The RFS electrode is custom manufactured to consistent
porosity specifications. Thinner wails improve repeatability and reduce filtration times. The
dimensions of the standard disc electrode used for conventional RDE analysis and the new RFS
electrode are shown in (Fig. 4).

LUKAS ET AL. ON ROTRODE FILTER

75

FIG. 4---Conventional electrode and RFS electrode.


In routine operation, the A-RFS disc electrodes are mounted on an electrode clamp assembly
for installation in one of the five preparation stations. The electrode clamp assembly is made of
two separate pieces so that a new electrode can be mounted on it directly from its container, thus
limiting the possibility of contamination through handling (Fig. 5a.). The clamps are numbered
so they can be easily identified and associated with their matching oil sample.
The electrode clamp assembly with a fresh carbon disc is then installed in one of the RFS
sample funnels. An oil sample is poured into the funnel, and the funnel is clamped shut (Fig.
5b). With the start of the automated process, a vacuum/pressure pump creates a pressure drop of
approximately 5 arm across the disc electrode, causing the sample to flow through the disc
electrode. The filtration process captures and concentrates large wear particles on the surface of
the electrode. A sensor determines when the sample has been filtered through the electrode and
automatically starts the electrode cleaning cycle followed by a drying process. Heptane is used
as a solvent to wash the remaining oil from the RFS electrode and to re-suspend any particles
remaining in the funnel so that they can be filtered and captured.
The RFS preparation process is complete when all of the oil has passed through the disc
electrode, residual oil has been washed away, and the electrode is dried. The operator then
removes the electrode clamp assembly and installs the shaft complete with the electrode in the
spectrometer for analysis (Fig. 5c).

FIG. 5--Preparation process of the A-RFS electrode.

76

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

RFS Data Interpretation


Unfortunately, calibration standards prepared with actual wear metal particles at various sizes
and concentrations are not available. Also, even if they were available, shelf-life, particle
settling, and the actual distribution of particle sizes could be problematic. Thus, in order to
provide the basis for benchmarking RFS calibration and to assure that data are consistent,
spectrometers are calibrated with certified commercial or military metallo-organic calibration
standards. This insures the spectrometer has the same response over time and permits trending
of wear metals and contaminant particles. It is important to note that RFS data is therefore not in
true parts per million (ppm), but in scalar thaits that the spectrometer can reproduce over time.
RFS is a powerful analytical tool because it provides additional information on actual wear
and contaminant particles in an oil sample. In routine operation, laboratories perform two
analyses on a used oil sample. The first is conventional oil analysis that provides quantitative
and qualitative analysis of dissolved and small particles in the sample. The second is RFS
analysis that provides an indication of large particles and their elemental composition. Taken
together, trends of the two analyses can be used as a trend to provide a much clearer picture as to
the mechanical health of oil wetted systems.
A typical data trend from an aircraft engine is shown in Tables 1 and 2. The first set of
analyses (Table 1) is with conventional oil analysis for dissolved and small particles in the oil
samples over a period of about 2 months. Conventional oil analysis shows no trends or
significant concentrations of wear metals and contaminants. The second set of data (Table 2) is
for the same samples but with RFS analysis for large wear particles. Note that the concentrations
of several elements such as iron, magnesium, sodium, and silicon are significantly higher than
conventional analysis. In addition, other wear metals such as aluminum, chromium, nickel, lead,
and zinc are now detected. The RFS analysis also identifies a major trend change at 299 h
denoted by a big jump in the iron, magnesium, and silicon concentrations. This signifies that
there is a big increase in large particles. Conventional analytical data (Table 1) showed no trend
or major concentrations for any of the elements. The aircraft was taken out of service at 312 h,
the engine was inspected, and a beating was replaced. Continued service could have resulted in
complete failure.
TABLE 1--Aircraft engine oil analysis trend, conventional spectroscopy analysis.
Date
12/10/03
12/12/03
12/15/03
12/16/03
12/18/03
01/05/04
01/08/04
01/15/04

H
262
268
271
278
293
299
307
312

Fe
0
1
0
1
1
1
0
0

Ag
0
0
0
0
0
0
0
0

A1
0
0
0
0
0
0
0
0

Cr Cu
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Mg
0
0
0
0
0
0
0
0

Na
0
0
0
0
0
0
0
0

Ni Pb
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Si
0
0
0
0
0
0
0
0

Sn Ti
4
1
4
1
5
1
4
1
4
0
3
1
3
0
3
0

B
0
0
0
0
0
0
0
0

Mo
0
0
0
0
0
0
0
0

Zn
0
0
0
0
0
0
0
0

V
0
0
0
0
0
0
0
0

LUKAS ET AL. ON ROTRODE FILTER

77

TABLE 2--Aircraft engine oil analysis trend, RFS analysis.


Date
12/10/03
12/12/03
12/15/03
12/16/03
12/18/03
01/05/04
01/08/04
01/15/04

H
262
268
271
278
293
299
307
312

Fe
24
15
10
18
14
52
58
106

Ag
0
1
0
0
0
0
0
1

A1
2
6
3
6
0
3
7
4

Cr
1
1
1
1
1
3
3
8

Cu
1
1
0
2
0
1
3
2

Mg
18
9
10
7
4
30
47
19

Na
14
9
8
3
1
18
30
14

Ni Pb Si
2
2
43
1 2
19
0
1 22
1 2
18
1 2
12
4
0
80
3
1
113
9
3
45

Sn Ti
0
2
0
1
0
1
0
1
0
0
0
3
0
4
0
2

B
1
1
0
0
2
1
2
2

Mo Zn
1
7
1
2
0
8
1
6
1
1
0
8
1
11
1
12

V
0
0
0
0
0
0
0
1

It is common knowledge that an analysis is only as good as the oil sample. This is even more
important with RFS analysis since RFS analyzes large wear and contaminant particles. It is thus
important to always take the oil sample from the same place and in the same manner to improve
analytical performance. Variations in RFS analysis also will be observed due to particle settling
in the sump if samples are taken at varying times after a system is shut down.
RFS analyses, as shown by this example, are capable of identifying and quantifying large
particles that would normally be missed by conventional spectrometric oil analysis.

A-RFS Field Experience


RFS has shown itself to be a valuable additional tool to military oil analysis programs.
However, it started in industrial applications and continues to provide rapid and valuable
information for in-house and commercial laboratories. Experimental data from military
application and a commercial case history will be reviewed in this section.

Experimental Data with Metal Powders


In order to verify that the RFS technique can detect particles greater than 10 gm, the U.S.
Department of Defense Joint Oil Analysis Program Technical Support Center (JOAP-TSC)
prepared samples using M50 alloy particles. M50 alloy is often present in aero-derivative
turbines. Some failure modes of J52 engines appear to be sudden catastrophic failures where the
root cause began with the lack of lubrication of the 489 beating area. Initially, analysis of the
samples by RDE spectroscopy did not indicate the failure mode [6]. The composition and
specifications for M50 alloy are as shown below (Table 3).

TABLE 3--Composition of M50 alloy."


Fe
Cr
89.8
3.75-4.5
in % concentration.

Ni
0.3 max

Si
0.2-0.6

Mo
3.9-4.75

V
0.8-1.25

78

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

The M50 alloy was sieved into known particle ranges and mixed at several concentrations in
MIL-L-23699 oil in groupings of 25-38 ~tm and 45-63 Ixm. The samples were analyzed using
conventional spectroscopy and also RFS [7].
Conventional atomic emission spectroscopy (Table 4) did not detect any of the M50 alloy
mixed into the new MIL-L-23699 oil. This was to be expected and confirms the inability to
detect large particles in an oil sample.
The same samples also were analyzed with the new A-RFS technique (Table 5). The data
verify the capabilities of RFS in detecting large particles and the presence of the M50 alloy, a
frequently used alloy for roller beatings in aviation gas turbines. The response for particles in
the 25-38 ~tm range is good. The data analyst can readily detect the presence of large particles,
and the alloy can also be identified by its metallurgy. As the particle size increases to the 45-63
~tm range, the excitation potential of the spectrometer is insufficient, and only a portion of the
particles is vaporized.
TABLE 4---Conventional spectrometric analysis of M50 alloy particles in MIL-L-23699 oil.
M50 Particle Size / Vanadium
Concentration
25-38 Ixm / 3 ppm
25-38 ~tm / 12 ppm
25-38 ~tm / 20 ppm
45-63 Ixm / 3 ppm
45-63 ~tm / 7 ppm
45-63 ~tm / 15 ppm

Fe

Cr

Ni

Si

Mo

0.0
0.9
1.5
0.0
0.1
0.0

0.0
0.1
0.1
0.1
0.1
0.0

0.0
0.0
0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.2
0.0
0.1

0.0
0.0
0.0
0.0
0.0
0.0

0.0
0.0
0.0
0.0
0.0
0.0

TABLE 5--RFS analysis of M50 alloy particles in MIL-L-23699 oil.


M50 Particle Size / Vanadium
Concentration
25-38 ~tm / 3 ppm
25-38 lam / 12 ppm
25-38 ~tm / 20 ppm
45-63 ktm / 3 ppm
45-63 ~tm / 7 ppm
45-63 ~tm / 15 ppm

Fe

Cr

Ni

Si

Mo

463
1454
3083
26.9
21.1
25.8

17.5
59.6
157
0.9
0.6
0.9

25.1
86.6
219
1.2
0.8
1.0

15.0
17.5
42.9
1.4
2.0
2.1

21.2
81.6
227
0.9
0.7
0.4

3.2
12.1
37.1
0.0
0.1
0.3

Microscopic inspection of the RFS electrode as well as an automatic particle counter and
shape classifier have been used to verify that the effective RFS capture range of particles from a
used oil sample is from 2 ~tm to more than 100 ~tm. However, the upper particle size limit that
can be completely volatilized by the spectrometer's arc is closer to 50 ~tm. This does not
preclude the RFS analysis from detecting larger particles, because even partial volatilization of a
particle above 50 ~tm will add signal and concentration to the analysis. The purpose of RFS
analysis is to detect and trend the presence of large particles in a used oil sample that are missed
by routine spectrometric analysis, and not to quantify the exact size or concentration of these

LUKAS ET AL. ON ROTRODE FILTER

79

particles. The mere presence of large particles is most significant, and the RFS analysis directs
the data analyst to additional analytical tests.
It also may be observed that the RFS technique is very sensitive. Just a few ppm of particles
in the oil give a big response by the spectrometer. This is due in part to the RFS filtration
process concentrating particles on the disc surface from a relatively large oil volume for
subsequent vaporization and detection.

Commercial Case History


The following is a typical case history [8] from a Northeastern power company. The item
being sampled is a condensate vacuum pump. In this case, the commercial laboratory provided
oil analysis services for lubricant physical properties and metals including RFS for larger
particles. Although the laboratory provides data for 20 wear metals, contaminants, and additives,
a trend was observed only for iron and silicon (Fig. 6).

FIG. 6---Spectrometric analysis trends for normal (fine) and RFS (coarse) analyses.
Normal spectrometric analysis, "fine," did not show significant wear. In most industrial
systems such as this pump, concentrations of wear metals will fluctuate in accordance with oil
added and can vary by as much as 10 ppm. The wear as indicated by normal spectrometric
analysis therefore did not exceed expected limits, and did not result in any maintenance
recommendations. On the other hand, the RFS analyses as indicated by "Coarse" in (later in Fig.
8) showed a clear trend resulting in maintenance recommendations.
The silicon readings by themselves were not cause for alarm, but they were when combined
with RFS data for iron. The RFS trend for iron jumped enough for' the laboratory to issue a
severe alarm after the January 9, 2003 sample. The maintenance response was to "feed and
bleed," i.e., drain some lubricant and top it off. The next two RFS analyses showed a
corresponding reduction in the iron analyses. However, the May 22, 2003 sample indicated that

80

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

there was still a problem, and a severe alarm was again generated. This time the maintenance
personnel opened up the pump and found that a bearing had a loose fit and was fretting on the
bearing housing, bearing, and shaft (Fig. 7). The RFS data also prompted the laboratory to
prepare a ferrogram of this sample. Ferrographic analysis verified that large particles were
present and that cutting wear (Fig. 8) was taking place.

FIG. 7--Bearing housing, outer, and inner race fretting due to a loose fit.

FIG. 8--Ferrogram showing cutting wear.

LUKAS ET AL. ON ROTRODE FILTER

81

In this example, normal spectrometric analysis did not detect the large wear particles
emanating from the bearing-fretting problem. The RFS analysis, however, did provide early
waming to repair, clean up, and solve the problem before it could lead to a major failure with
possible secondary damage. The next three samples show a return to a normal trend after the
lubricant was cleaned and replaced.
Conclusion

RFS is an important enhancement to standard RDE analysis. It provides new information on


the condition of the machine being monitored by expanding the particle detection capability of
the RDE technique. Sometimes it is also used to screen samples to determine if more time
consuming analysis, such as ferrography, is required.
Rotrode Filter Spectroscopy (RFS) is a recently improved technique that has been shown in
actual field applications to provide important additional information about large wear particles,
information that may be missed with conventional techniques. It is applied with existing
instruments to provide two analyses on the same used oil sample. The first, using the standard
RDE atomic emission technique, detects and quantifies dissolved and small wear metals and
contaminants. The second analysis, by RFS, qualifies and semi-quantifies larger particles. This
paper has shown how the early version of the RFS system has been improved and updated to
provide better sensitivity, repeatability, faster sample throughput, and improved particle capture
capability.
RFS analysis is not used by itself, but as a second test that expands the particle detection
capability of spectroscopy to provide a comprehensive tool to direct the laboratory to identify
and solve a maintenance problem. Frequently, the RFS analysis is also used as a screening tool
to determine if additional tests, which are too time-consuming or expensive to perform on all
samples, should be performed. Supplementary techniques such as analytical ferrography or
particle shape characterization and particle counting are then applied to verify a severe wear
condition.
References

[1] Sawyer, J. W., "Sawyer's Turbomachinery Maintenance Handbook," Turbomachinery


International Publications, CT, 1980, pp. 5-8.
[2] Lukas, M. and Giering, L. P., "The Effects o f Metal Particle Size in the Analysis for Wear
Metals Using the Rotating Disc Atomic Emission Technique," International Symposium on
OilAnalysis, Erding, Germany, July 1978.
[3] Anderson, D. P., "Wear Particle Atlas (Revised)," Prepared for Advanced Technology
Office, Naval Air Engineering Center, Lakehurst, NJ, June 28, 1982.
[4] Rhine, W. E., Saba, C. S., and Kaufinan, R. E., "Metal Particle Detection Capabilities of
Rotating-Disc Emission Spectrometers," Lubrication Engineering, Vol. 42, No. 12, 1986, p.
755.
[5] Anderson, D. P. and Lukas, M. "Rotrode Filter Spectroscopy: A Method for MultiElemental Analysis of Particles in Used Oil Samples," Proceedings of the International
Condition Monitoring Conference, Pensacola, FL, November 16-20, 1992.
[6] Humphrey, G. R. "Filter Debris Analysis by Energy Dispersive X-Ray Fluorescence
Applied to J52 Engines," Joint Oil Analysis Program International Condition Monitoring
Conference, Pensacola, FL, April 18-22, 2004.

82

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

[7] Yurko, R. J. "Applying Rotrode Filter Spectroscopy (RFS) and LaserNet Fines (LNF)
Technology to Detect Large Wear Particles of M50 Alloy from J52 Engine Beating
Failures," Joint Oil Analysis Program International Condition Monitoring Conference,
Pensacola, FL, April 18-22, 2004.
[8] Case History Compliments of National Tribology Services, 5 Lakeland Park Drive,
Peabody, MA 01960, U.S.A.

SULFUR DETERMINATION
AND X-RAY FLUORESCENCE

Journal of ASTM International, March 2005, Vol. 2, No. 3


Paper ID JAI12978
Available online at www.astm.org

R. A. Kishore Nadkarni 1

Trace Levels of Sulfur in the Fuels of the Future: Analytical


Perspective
ABSTRACT: Sulfur levels in both gasoline and diesel fuels are being restricted because of
environmental pollution concerns throughout the world. The U.S. Environmental Protection Agency has
mandated the levels of sulfur which may be allowed in most road and off-road vehicles. Although there
are many standard test methods available for determining sulfur in petroleum products, presently only
about four can qualify for their usefitlness at the expected low levels in the future fuels. A large
interlaboratory study undertaken by ASTM to estimate the capabilities of these methods is described.
Also, data from the ASTM interlaboratory crosscheck programs is reviewed to identify the trends that the
industry laboratories are showing. Finally, a brief overview of some new methods being developed in the
industry for such analyses is presented.
KEYWORDS: sulfur analysis, gasoline analysis, diesel analysis, intedaboratory crosscheck programs,
ultralow sulfur in diesels
Origin and Occurrence of Sulfur in Petroleum Products
Sulfur and its compounds are and will be present in virtually all petroleum products and
lubricants from crude oils to ultra low sulfur fuels o f the future. The sulfur levels in these
products vary from the highest amounts in the crude oils and some lubricants to trace levels in
the currently proposed fuels.
The inorganic elements present in crude oil originate from its marine animal and vegetative
origins deposited with sediment in coastal waters in pre-historic times. Over the millennia,
sulfur, oxygen, nitrogen, and other volatile compounds evolved out o f this primordial mixture
due to bacterial action, leaving behind a mixture o f hydrocarbons containing some o f these
elements and metals and other compounds. Apart from carbon, hydrogen, and oxygen, sulfur is
one o f the main heteroatom present in the crude oils. Sulfur up to 6 m % has been reported in
some crude oils. Sulfur content is one o f the more important criteria in evaluating the crude oil
value and its processing. As the sulfur level increases, so do the complexity and the expense o f
the refining operations. Low boiling point sulfur containing compounds such as hydrogen sulfide
will be evolved during the distillation process, while other sulfur compounds will remain in the
heavier distillation residue.
Sulfur content is always determined as a part o f the "crude assay" o f the crude oils.
Enormous amounts o f such data exist in each oil companies' crude assay database for the crude
oils o f their interest. Some examples o f the sulfur content o f the major oil fields o f the world are
given in Table 1.

Manuscript received 14 September 2004; accepted for publication 2 November 2004; published March 2005.
i Millennium Analytics, Inc., East Brunswick, NJ 08816.
Copyright9 2005byASTMInternational,100Ban"HarborDrive,POBoxC700,WestConshohocken,PA 19428-2959.
85

86

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

TABLE 1--Typical sulfur content of crude oils from some giant oil fields [1].
Country
Abu Dhabi
Algeria
Canada
Colombia
Egypt
Indonesia
[ran
[raq
Kuwait
Libya
Nigeria
Saudi Arabia

Sulfur, m%
0.62-0.77
0.02-0.31
0.2-3.67
0.25-1.11
0.84-2.06
0.07-0.18
0.76-3.68
1.36--2.1
1.82-2.58
0.13-1.04
0.10-0.26
1.25-3.91

Country/State
United States
Alaska (N. Slope)
California
Louisiana
Oklahoma
Texas
Wyoming

Sulfur, m%
1.07
0.2-1.2
0.1-0.8
0.2-1.18
0.1-2.4
0.1-3.44

GasoBne and Diesel


Sulfur oxides formed during combustion of gasoline in the internal combustion engine are
undesirable because o f damage both to the engine and to the environment. Also, o f major
concern are the poisoning effects of sulfur on Nox reducing catalysts and adsorbers employed in
mobile sources (vehicles) [8]. Sulfur acids increase rusting and corrosion of engine parts, piston
ring, and cylinder walls. In the atmosphere, sulfur oxides will convert to sulfuric acid by reacting
with moisture, with potential harm to vegetation and animal-human life.
In diesel fuels, sulfur compounds can be present as mercaptans, sulfides, disulfides, and
heterocyclic compounds, such as thiophenes. Their combustion by-products can cause wear in
diesel engines and increase the amount o f deposits in the combustion chamber and on the
pistons. Some typical sulfur concentrations reported in diesel fuels are given in Table 2.
TABLE 2--Typical sulfur content of diesel fuels [2].
Diesel Fuel
Diesel Fuel Oil Grade 1 D
Diesel Fuel Oil Grade 2 D
Diesel Fuel Oil Grade 4 D
Kerosene
Premium Diesel
Railroad Diesel
Marine Distillate Diesel

Sulfur, m%
0.50
0.50
2.0
0.12
0.30
0.50
1.2

Fuel Oils
Depending upon the source of the crude oil, the refining process, and the fuel grade, varying
a m o u n t s o f sulfur may be present in different types o f fuel oils. Combustion of the sulfur
containing fuel oils produces sulfur oxides, which pollute the atmosphere and cause corrosion
problems in boiler equipment. They may form sodium and vanadium complexes, and such
deposits on external surfaces o f superheater tubes, economizers, and air heaters cause equipment
corrosion and loss o f thermal efficiency.

NADKARNI ON TRACE LEVELS OF SULFUR

87

Lubricants
In case of the lubricant additives and lubricating oils including automatic transmission fluids,
gear oils, turbine oils, etc., additional amounts of sulfur in the form of metal sulfonates are added
to enhance the performance characteristics of the oils in the internal combustion engines. Two
principal sources of sulfur in engine oil are zinc dialkyl dithio phosphates and base stock. Some
typical examples of sulfur levels in such lubricants are given in Table 3. See [4] for a discussion
of the metal compounds used for this purpose and the effect they have on engine performance.
Additional information on the presence and significance of sulfur in petroleum products can be
found in [2-4].
TABLE 3--Sulfur levels in lubricants.
LubricantProduct
Automatic TransmissionFluids
Gear Oils
Lubricating Oils
AutomotiveLubricantAdditives

Sulfur, m%
0.1 - 0.3
1.5 - 2.0
0.3 - 0.7
0.5 - 3.5

Government Regulations
Because of the deleterious effects of sulfur emissions from motor vehicles, a number of
government regulatory agencies, principally in North America and Europe, has been vigorously
controlling and gradually reducing allowable sulfur contents of the fuels used for automotive,
marine, aviation, off-road, utilities, and home heating. The allowed sulfur levels are steadily
being tightened from a current level of about 330 mg/kg sulfur in gasoline and about 500 mg/kg
in diesel to 30 and 15 mg/kg, respectively, in gasoline and diesel by 2006, and near zero by 2010.
European diesel specification EN 590 mandates sulfur content of diesels from 350 mg/kg
maximum in 2000 to 10 mg/kg maximum by January 1, 2005 in European Union countries.
One method of sulfur removal from refinery streams is by hydrodesulfin'ization (HDS) in the
refineries. This step also directly impacts the characteristics of low sulfur diesel fuels, such as
density, aromatics content, cetane number, and cloud point. The magnitude of these changes will
depend upon the type and setup of refinery HDS units. However, in the end some refractory
compounds in fuel, e.g., 4,6-dimethyl dibenzothiophene, are very resistant to desulfurization,
owing to the inaccessibility of the organically bound sulfur atom. Lower pressure HDS units
which can work satisfactorily at 350 mg/kg sulfur levels, may have difficulty achieving reduction
to 50 or 10 mg/kg sulfur level.
In May 1999 the U.S. Environmental Protection Agency (EPA) released its proposed "Tier
2" emission standards for vehicles and gasoline sulfur standards for refineries [5-7]. They
mandated the ASTM D 2622 X-ray fluorescence (XRF) method for sulfur determination, and in
later announcements D 6428 became the mandatory method for sulfur determination in diesel
fuels. Over the years through discussions and litigations, these rules have been modified to the
point where alternate test methods can be used instead of only the mandated methods. The
alternate methods allowed would be: (a) those developed by the voluntary consensus based
standards writing bodies, such as ASTM, or (b) industry methods for which data will have to be
provided to EPA to establish their correctness under Performance Based Measurement System
(PBMS). Discussions between the industry and EPA are ongoing to further define the statistical
basis for judging the available methods' acceptance and other related issues.

88

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Analytical Perspectives
Several test methods are available for determining sulfur in diverse petroleum product and
lubricant matrixes from sub-mg/kg to m% concentration levels. One of the earliest methods
issued by the D02 committee was for the sulfur determination by an oxygen bomb method (D
129) first issued in 1922. This was the 15th standard ever issued by the D02 Committee. The
techniques currently used for the sulfur determination include classical wet chemistry (D 129, D
1266, D 1552, and D 2784), X-ray fluorescence (D 2622, D 4294, D 4927, D 6334, D 6443, D
6445, and D 7039), atomic spectroscopy (D 4951 and D 5185), and various thermal combustion
microelemental methods using various detection techniques such as microcoulometry (D 3120
and D 3246), UV-fluorescence (D 5453 and D 6667), electrochemistry (D 4045 and D 6920),
GC-FPD (D 7041), etc. A list of such available test methods is given in Table 4 along with other
pertinent details as to the applicability of the method.
Although nearly 20 test methods for sulfur are listed in Table 4, appropriate selection of a
method for analysis would depend on the matrix, sulfur concentration level, and the desired
precision of the analysis. Thus, out of the 20 methods above, only 4 typically qualify to be
considered as potentially useful methods for the determination of low levels of sulfur in the ultra
low sulfur fuels of the future: D 2622, D 3120, D 5453, andD 6920.
Several of these ASTM test methods have also been separately issued by other international
standards writing bodies such as IP, ISO, DIN, JIS, and AFNOR. These equivalent test methods
are listed in Table 5 [9]. In an ever-expanding global market, it is useful in commerce to
recognize the analytical methodology used in other industrialized nations.

ASTM Research Report D02-1456


In a very thorough research study done at SouthWest Research Institute (SWRI), San
Antonio, TX, three sulfur test methods (D 2622, D 4294, and D 5453) were studied for fitness of
use for the analysis of fuels at sulfur levels below 500 mg/kg [10]. All three were found to be
equivalent for measurements in the 150-500 mg/kg range. But D 2622 and D 5453 were found to
be of equivalent fitness for use down to 20 mg/kg, and D 5453 even to as low level as 1 mg/kg
sulfur. It was also found that the latter method gives the best precision at such low sulfur levels.
Later work in ASTM crosschecks confirms these conclusions.

European Sulfur Cross Check


In a very large crosscheck conducted in Europe, 69 laboratories from 9 countries analyzed
fuels with sulfur ranges between 5 and 500 mg/kg in 8 gasoline and 7 diesel samples [11]. ISO
equivalent methods ofASTM D 2622, D 4294, D 5453, D 3120, and D 1266 were employed. At
these levels all five test methods were found to produce equivalent test results, but the precision
of different methods varied considerably. The best reproducibility was found for methods D
2622 and D 5453. Based on this study, it was proposed that D 1266 and D 4294 methods be
dropped from consideration for analyzing fuels at the 30 mg/kg sulfur levels in the 2006 and
onward fuels. Only the D 5453 method was considered suitable for the determination of sulfur at
<10 mg/kg. These conclusions are consistent with the work done at SWRI mentioned above.

NADKARNI ON TRACE LEVELS OF SULFUR

T A B L E 4--Methods for sulfur


Technique
ASTM
Method
Bomb combustion and BaSO4
D 129
precipitation
D 1266
D1552
D 2622
D2784
D 3120

determination in petroleum products and lubricants.

Lamp combustion and gravimetry


or titrimetry
High temperature combustion and
iodate titration or IR detection
Wavelength Dispersive XRF
Oxy-hydrogen lamp combustion
and gravimetryorturbidometry
Oxidative pyrolysis&
microcoulometry

Matrix

No longer widely used in the


industry

Light liquid hydrocarbons

Chlorine and nitrogen may


interfere; precision is not very
good
Chlorine and nitrogen may
interfere
Not widely used in the industry

LPG fuels

D 4294

D 4927
D 4951

Wavelength Dispersive XRF


ICP-AES

Additives and lube oils


Additives and lube oils

D 5185

ICP-AES

D5453

Combustion U-v'-fluorescence
detection

Additives, lube oils, and


used oils
Liquid hydrocarbons at
low sulfur levels

D 6334

Wavelength Dispersive XRF

D 4045

D 6443
D 6445
D 6920

Wavelength Dispersive XRF


Energy Dispersive XRF
Oxidative combustion and
electrochemical detection
D 7039
Monochromatic wavelength
dispersive XRF
D 7041
Combustion gas chromatography
, with flame [0hotometricdetection

Comments

Lube oils, adpacks,


greases, other petroleum
products
Gasoline, kerosene,
naphtha
Petroleum cokes, other
products
All liquid petroleum
products and oils
LPG fuels

Oxidative pyrolysis &


microcoulometry
Hydrogenolysis and rateometric
colorimetry
Energy Dispersive XRF

D 3246

89

Naphtha, kerosene,
alcohol, jet fuel, etc.
All liquid petroleum
products and oils

Gasoline and oxygenate


blends
Lube oils and additives
Gasoline
Gasolines, diesels, and
oxygenates
Gasoline and diesels
Gasoline and diesels

No longer widely used in the


industry
Nitrogen and chlorine may
interfere
Expensive iustrument, yet one of
the most widely used methods
Halogens may interfere

Widely used in industry because of


low cost instrumentation; poor
precision at low sulfur levels
Widely used in additives industry
Widely used in additives industry
for metals analysis
Widely used in additives and oil
industry for metal analysis
Widely used in the industry for
very low levels of sulfur; excellent
precision at ultra low sulfur levels
Standards must match oxygenated
samples

Precision poorer than other popular


methods D 2622 or D 5453
New method; little field experience
On-line sulfur method widely used
in the field

T A B L E 5--International standards for the determination of sulfur.


Analysis
ASTM
IP
DIN
AFNOR
JIS
ISO
Bomb Method
D 129
61
51-577
T60-I09
. . . . . .
Lamp Method
D 1266
107
...
M07-031
. . . . . .
HT Method
D 1552
...
M07-025
WDXRF
D 2622
...
51-400T6
K 2541
14596
M07-022
K 2276
16591
OxidativeMicroeoulometry
D 3120
...
OxidativeMicroeoulometry
D 3246
373
...
M07-052
...
EDXRF
D 4294
336
M07-053
...
8754
WDXRF
D 4927
407
51-391T2
. . . . . .
UV-Fluorescenee
D 5453
......
M07-059
. . . . . .
(IP: Institute of Petroleum, U.K.; DIN: Deutsche Institut fur Normung (Germany); AFNOR: Association Francaise
de Normalisation (France);/IS: Japan Industrial Standards; ISO: International Organization for Standardization)

90

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Reliability of EDXRF Method D 4294

This method is widely used in the industry for low levels o f sulfur determination because o f
its comparatively low cost instrumentation and ease o f operation in spite o f the fact that the
scope o f the method limits it to >150 mg/kg sulfur levels. Indeed, the method produces quite
satisfactory results at these and higher levels o f sulfur. See Table 6 for comparison between four
widely used methods which all produced equivalent results at m % levels o f sulfur in diverse
products.
T A B L E 6--Methods for sulfur determination at higher levels.
ASTM
D 2622
D 4294
D 4951
D 5453
Sample
WDXRF
EDXRF
ICPAES
UV-FL.
CO 0403
1.008 0.019(14)
1.042 0.038(76)
...
GO 0404
...
1.856 0.069(6)
1.831 ~ 0.148(7)
...
GO 0312
_.
1.832 0.060(8)
1.758 0.038(4)
...
GG 0403
0.2089 0.0055(12)
0.2057 0.0098(22)
...
RFG 0404
0.0416 0.0016(58)
0.0438 0.0022(35)
...
0.0411
MG 0404
0.0271 0.011(30)
0.0276 0.0040(51)
...
0.0258
RFG 0312
0.0238 0.0013(63)
0.0258 0.0029(38)
...
0.0230
All values in m% and expressed as average value standard deviation (number of results).
A S T M interlaboratory crosscheck programs, however, have repeatedly demonstrated that at
low levels o f sulfur this method gives biased high results compared with other reliable test
methods, and it has very poor reproducibility sometimes approaching 100 % or more o f the m e a n
sulfur value. The results o f such inaccuracy or imprecision could result in a false positive result
for sulfur, indicating that the product is exceeding the regulatory sulfur limits and that the
customer-supplier will have on-going quality disputes as to the true levels of sulfur in the
shipped products i f the two parties are using two different test methods. Some recent data
obtained b y three most widely used methods from A S T M crosschecks on low level sulfur in
motor gasoline, diesel fuel, and reformulated gasoline are summarized in Table 7, which clearly
indicate that the E D X R F method is not useful for determining low levels o f sulfur in the fuels o f
the future.
Because o f such poor quality o f data, the U.S. E P A did not include D 4294 as one o f the
preferred methods for determining sulfur in fuels o f the future. With active help from instrument
vendors, A S T M is looking into the possibility o f improving the accuracy and the precision o f
n e w technology E D X R F methods. Several such initiatives are underway.
T A B L E 7--Comparison of three widely used sulfur
ASTM Sample
D 2622
D 4294
WDXRF
EDXILF
Mogas 0208
13.4 4.8 (36)
31 21 (39)
Diesel 0202
32.6 6 (80)
44 43 (137)
RFG 0204
23.6 4.1 (64)
36 16 (27)
RFG 0205
55.4 + 5.4 (67)
77 26 (30)
RFG 0206
25.6 4.9 (72)
32 16 (25)
RFG 0211
9.5 3.4 (63)
24 12 (27)
RFG 0212
32.6 4.7 (66)
44 14 (31)
RFG 0303
19.1 4.2 (67)
27 17 (34)
RFG 0304
19.2 3.9 (66)
42.1 26.6 (34)
RFG 0307
16.8
31

test methods.
D 5453
UV-F1.
11.7 1.9 (22)
32.0 4.7 (45)
22.9 3.2 (34)
54.3 6.8 (33)
22.4 2.6 (34)
8.2 + 1.8 (37)
31.6 3.9 (32)
16.3 3 (40)
16.9 2.3 (39)
15.0

All values are expressed as robust mean robust standard deviation in mg/kg (# labs).

NADKARNI ON TRACE LEVELS OF SULFUR

91

A Benchmark ASTM Interlaboratory Study


In 2000, the U.S. EPA published requirements for tier 2 fuels and also mandated the test
methods that can be used for sulfur determination in these fuels. ASTM D 6428 was chosen as
the mandatory method for diesel fuels with ASTM D 2622, D 3120, and D 5453 methods
allowed as alternates if their accuracy and precision were shown to be equivalent to those of D
6428 method. The petroleum industry widely uses the latter three test methods for sulfur but had
virtually no experience with the mandated D 6428 method, which is under the jurisdiction of
ASTM D 16 Committee on Aromatic Hydrocarbons and Related Chemicals. The method was
written for the sulfur determination of liquid aromatic hydrocarbons, their derivatives, and
related chemicals. The precision quoted in this method was for aromatic hydrocarbons and not
for petroleum products. Moreover, it was inadequate (10 analyses by one operator on two
samples) and did not conform to the statistical protocols used in the D02 Committee test methods
for calculating the precision and bias estimates.
Consequently in the fall of 2001, D02.SC 3 undertook the responsibility to conduct a major
Interlaboratory Study (ILS) to evaluate the capabilities of these four test methods - D 2622, D
3120, D 5453, and D 6428 - for the determination of low levels of sulfur in gasoline and diesel
motor fuels of the future. In its scope and eventual extent this was one of the most exhaustive
ILS ever undertaken by ASTM D02 Committee. There were a total of 16 samples each of
gasolines and diesels. These also included biodiesels, oxygenated fuels, reformulated gasolines,
and NIST gasoline and diesel SRMs. They contained varying levels of sulfur from 1-100 mg/kg.
A total of 70 laboratories including those from the U.S. and abroad volunteered for this ILS,
eventually, however, only 56 laboratories returned their results: 38 oil industry laboratories, 14
instrument vendor laboratories, and 4 commercial analytical service laboratories. Multiple
vendor instruments were used in all methods. Several laboratories also used multiple test
methods for the analysis. Some laboratories determined the required density determination in
these samples. Thus, overall there were about 6000 data points used in the statistical evaluation
of the data. The full report of this ILS is available as RR-D02-1547.
The statistical evaluation of this data indicates the following important points:
9

Although a similar European study [10] used somewhat different protocols and statistics,
there was no dramatic difference in the precisions obtained in the European and this
ASTM ILS.
9 The EPA mandated D 6428 test method had the poorest precision compared to three other
methods used: D 2622, D 3120, and D 5453.
9 At the time of this study there were only 2 vendors available for D 6428 instruments. The
repeatability but not the reproducibility of these 2 instruments was found to be
significantly different.
9 The test method D 6428 would not be precise enough for regulatory compliance at the
proposed sulfur limits. For example, the reproducibility at 30 mg/kg sulfur in gasoline
would be 23, and that for diesel would be 16 at 15 mg/kg sulfur in diesel
9 The preferred methods for regulatory compliance would be D 2622 and D 5453 in terms
of their accuracy and precision.
Table 8 summarizes the precisions obtained from this ILS. As a follow-up to this landmark
study, all three existing methods were revised to include the precisions obtained, and the D 6428
method of the D 16 committee was revised with the new precision data and issued separately as

92

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

D 6920 under the jurisdiction o f D02 committee. Eventually, E P A abandoned the requirement o f
using mandatory D 6428 method in light o f the p o o r precision shown in this ILS.
T A B L E 8--Precis!ons for sulfur
ASTM Test
Technique
Method
D 2622
D 3120
D 5453
D 6920

determination in ultra-low sulfur fuels of thefuture.

WDXRF
Microcoulometry
UV Fluorescence
Eleetroehemistr/.

Range

3 mg/kg-5.3 m%
3-100 mg/kg
1-800 mg/kg
1-100 m~Jk~

30 mg/kg
Gasoline
r
R
3.8 7.4
3.85 17.7
1.95 11.7
4.7
22.8

15 mg/kg
Diesel
r
R
3.4 6.7
1.33 8.07
1.20 6.26
3.1
15.6

ASTM Interlaboratory Crosscheck Programs


Since 1993, A S T M D02 C o m m i t t e e ' s Coordinating Subcommittee 92 has been conducting
crosschecks on a scheduled basis. This is probably the largest industry proficiency testing
program in the world. A t present over 2000 laboratories w o r l d w i d e (of these 45 % are non-U.S.
based laboratories) participate in 18 programs spanning diverse petroleum products and
lubricants [12,13]. S o m e o f these programs include the determination o f sulfur from m g / k g to
percent levels depending upon the product being analyzed. The most c o m m o n l y used methods
for such analysis include D 2622, D 4294, and D 5453. Other test methods such as D 1266, D
3120, D4045, and D 6344 are used to a m u c h lesser extent. Overall, there appears to be good
agreement between the m e a n results obtained by alternative methods for a particular product.
H o w e v e r , there is considerable difference between the precisions obtained with the individual
methods.
S o m e examples o f data from m o r e recent cross checks are shown in Table 9. In most
examples, only the data at sulfur level b e l o w 50 m g / k g are shown to illustrate the capability o f
the alternate techniques.
TABLE
ASTM Sample
Mogas 0208
Mogas 0312

9---Analysis for low levels of sulfur (nfuel products.


D 2622
....
13.4 ~: 4.8 (36)
6.3 q- 3.1 (24)

D 4294
31 :~ 21 (39)
27 + 31 (39)

....... D 5453
11.7 -~ 1.9 (22)
3.74 q- 0.85 (37)

Diesel 0202
Diesel 0402

32.6 + 6 (80)
9.3 q-4.3 (67)

44 ~-43 (137)
26 a: 23 (91)

32.0 ~: 4.7 (45)


8.22 -~ 1.86 (61)

Avgas 0307
Avgas 0403

63.6 5.4 (62)


9.6 q- 2.8 (60)

75 :~ 24 (117)
26 -~ 20 (94)

59.7 9 6.0 (43)


9.0 1.5 (56)

RFG 0204
23.6 4.1 (64)
36 16 (27)
22.9 d: 3.2 (34)
RFG 0205
55.4 5.4 (67)
77 26 (30)
54.3 q- 6.8 (33)
RFG 0206
25.6 4- 4.9 (72)
32 16 (25)
22.4 2.6 (34)
RFG 0211
9.5 :~ 3.4 (63)
24 ~: 12 (27)
8.2 1.8 (37)
RFG 0212
32.6 4.7 (66)
44 + 14 (31)
31.6 :~ 3.9 (32)
RFG 0303
19.1 4.2 (67)
27 17 (34)
16.3 3 (40)
RFG 0304
19.2 :t: 3.9 (66)
42.1 + 26.6 (34)
16.9 2.3 (39)
RFG 0307
16.8 3.5 (58)
31 + 18 (28)
15.0 2.7 (40)
RFG 0308
22.9 4.8 (60)
31 ~- 22 (31)
20.7 + 2.9 (42)
RFG 0310
38.4 4.0 (65)
66 23 (36)
36.8 4.4 (46)
RFG 0403
10.8 q- 4.2 (60)
28 :~ 19 (24).
9.31 :~ 1.42 (54)
All values are in robust mean robust standard deviation in mg/kg sulfur (number of labs).

NADKARNI ON TRACE LEVELS OF SULFUR

93

This data excerpt clearly shows that the agreement between D 2622 and D 5453 is quite
good, and both produce comparable precision. The EDXRF method D 4294, however, produces
biased high results with reproducibility often approaching or exceeding 100 %, making it an
unsuitable method for commerce or regulatory compliance at such low levels of sulfur.
ASTM ULSl) ILCP Program
Since the beginning of January 2004, the ASTM D02 Committee added a new
Interlaboratory Cross Check Program for proficiency testing. The intent of this new program was
to evaluate the industry laboratories' capability for determining ultra low levels of sulfur in
diesel fuel. This is a monthly program, and only four test methods (D 2622, D 3120, D 5453, and
D 6920) are allowed to be used. If other test methods have been approved by the U.S. EPA under
its Performance Based Measurement System (PBMS) protocol, they may also be used. Generally
the level of sulfur in these samples is between 1-15 mg/kg. It is also recommended that a
certified reference material sample be analyzed along with the crosscheck sample and that data
be returned to ASTM. A suggested standard for this work was NIST SP,,M 2723a with a certified
sulfur value of 11 + 1.1 mg/kg. This program has proved quite successful in the short time that it
has been offered since EPA is interested in using this database for evaluating individual industry
laboratories' capability for such analyses. A summary of results from first six months' of
crosschecks is given in Table 10. Based on these data it appears that:
9
9
9

9
9

Most laboratories use D 2622 and D 5453 test methods. Very few laboratories use the D
6920 method mandated by EPA at one time.
Generally, the results by D 2622, D 3120, and D 5453 methods are comparable.
All laboratories that used NIST SRM obtained correct results for it. However, only 20 %
of the laboratories used this SRM. An equal number of laboratories did not use any CRM,
and the rest used other commercial or in-house prepared CRMs.
The reproducibility obtained in the crosschecks is similar to that given in the standard
methods.
In some samples (0402 and 0403) the sulfur levels were below the pooled limits of
quantitation given in the test methods; yet, the labs were able to produce reasonably good
results even though the uncertainty around the mean value was quite high. However,
since the mean values were well below the regulatory compliance limits, the high
uncertainty should not be a limiting factor for determining sulfur at these low levels.
The sample 0406 had the lowest sulfur levels (sub-mg/kg) in this series of crosschecks.
As a result, the uncertainty around the mean value was very large, sometimes
approaching or exceeding 100 % of the mean value. Also, the agreement between the
mean values by different methods was not very good. Obviously the capability limits of
the currently available test methods become evident for sub-mg/kg sulfur determination.
It would also be interesting to learn which test methods the refineries and blenders of
these ultra low sulfur diesels are using to certify such low sulfur levels when it is clear
from these crosschecks that the currently known test methods are not capable of such
precision or detection.
Discussions concerning all of these methods suggest that when the operator and
equipment are prepared for the 10 mg/kg and below range, better precision can be
obtained. This would be generally true for most instrumentation and test methods that

94

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

ideally the calibration range and the analyte concentration in the sample should closely
match each other to produce the best results.
TABLE l(Y--Results of ultra-low sulfur in diesel ASTM cross check programs.
ASTM Sample #
D 2622
D 3120
D 5453
D 6920
January 0401
15.2 2.3 (80)
i3.35 3.46 ('i4)
15.1 + 1.5 (99)
13.8; 15.2 (2)
February0402
2.35 + 1.76 (86)
1.53 + 0.52 (14)
1.58 + 0.47(104)
0.3; 0.9; 1.1 (3)
March 0403
1.86 + 1.52 (87)
1.53 + 0.40 (10)
1.41 + 0.44(118)
0.96 + 0.54 (8)
April 0404
15.41 + 2.13(96)
14.84 + 1.07 (14)
14.9+ 1.45(100)
16.20 2.96 (8)
May 0405
10.48 1.79(95)
9.89 + 0.67 (16)
9.83 0.98(113)
0.3 - 9.8 (6)
June 0 4 0 6
0.48+0.57(70)
0.90+0.63(14)
0.66
<0.1-1.2(6)
July 0407
4.01 1.47 (89)
3.57 0.50 (12)
3.49 0.59(124)
2.91 1.18 (8)
All values are expressed as robust mean + robust standard deviationin mg/kg (numberof results).
For intercompany commerce as well as regulatory compliance, reproducibility of the data is a
key parameter to deciding the validity of certificates of analysis sent from the suppliers to the
customers. Hence, the reproducibility obtained for the four methods used in these crosschecks
are compared with those given in the standards in Table 11. In the 21 data sets in this table, the
test method reproducibility is better in 3 cases, about the same in 9 cases, and worse in 9 cases
compared with the reproducibility obtained in these crosschecks. Thus, there is no clear
indication that the field precisions have improved or worsened since the standards were issued.
TABLE 11--Reproducibilities of the crosschecks versus standard test methods.
ASTM
Sulfur,
D2622
D 3120
D 5453
ULSD #
mgJkg
C.C. STM C.C. STM C.C, STM
0401
~ 15
6.38 3.64 9.98 7 . 2 2 4 . 0 9 6.29
0402
-2
4.89 4.75 1.49 1.13 1.31 1.35
0403
- 1.5
4.22 4 . 5 3 1.10 1.13 1.22 1.29
0404
N 15
5.92 8.27 2.57 8 . 0 9 3 . 9 4 6.23
0405
~ 10
4.96 7 . 2 4 1.88 5.43 2 . 7 2 4.37
0406
~0.5
1.59 3 . 7 9 1.77 0.80 1.11 1.01
0407
~4
6.01 5 . 4 0 1.77 2.18 1.64 0.88
(C.C.: Crosscheck; STM: StandardTest Method)

Reference Materials for Sulfur Determination


One o f the key aspects in proving the applicability of an established or a newly developed
method or analytical instnamentation is how well it performs in its accuracy and precision of
measurements. Although the precision can be locally determined in a laboratory b y multiple
analyses of a stable homogenous product over a period of time, well-characterized and preferably
certified materials need to be available to determine the accuracy o f an analytical system.
Although there are a number of commercial vendors who provide so-called "Certified Reference
Materials" (CRM), their certified values are generally based merely on gravimetric blending of
an analyte in an appropriate solvent, or they have been analyzed by one standard method. The
National Institute of Standards and Technology (NIST) has pioneered in producing extremely
well-characterized Standard Reference Materials (SRM) with infallible analytical values
assigned to them. Generally, the NIST certification involves determination of an analyte by two
independent primary analytical methods and collaboration by other secondary methods [14].

NADKARNI ON TRACE LEVELS OF SULFUR

95

M a n y sulfur S R M s have been produced b y N I S T o v e r the years that h a v e proven to be highly


useful in determining the bias and accuracy o f existing and new test methods. Very recently,
N I S T has been preparing some ultra-low sulfur fuel standards to help the industry in coping with
the analyses o f such "clean" fuels. A partial list o f such available N I S T S R M s for sulfur in
petroleum products is given in Table 12. There is also w o r k underway toward certifying other
low level sulfur SRMs.
TABLE
Product
Crude Oils
Diesel Fuels

Gasolines - RFG

Gasolines - High Octane


Kerosines
Lubricating Oil (X 5)
Lubricating Oil Additive
Petroleum Coke
Residual Fuel Oils

Used Oils

12--Sulfur SRMsfrom NIST.


NIST SRM #
2721
2722
1624d
2723a
2724b
2770
2294
2295
2296
2297
2299
2298
1616b
1617a
1819a
1848
2718
2719
1619b
1620c
1621e
1622e
1623c
1634c
2717a
1083
1084a
10854

S Concentration
1.5832 0.0044 m%
0.21037 0.00084 m%
0.3882 0.0020 m%
11.0 1.1 mg/kg
426.5 + 5.7 mg/kg
41.57 0.39 mg/kg
40.9 1.0 mg/kg
308 2 mg/kg
40.0 0.4 mg/kg
303.7 1.5 mg/kg
13.6 1.5 mg/kg
4.7 1.3 mg/kg
8.41 0.12 mg/kg
0.17307 0.00034 m%
0.04235 - 0.6135 m%
2.327 0.0043 m%
4.7032 0.0079 m%
0.8877 0.0010 m%
0.6960 0.0077 m%
4.561 0.015 m%
0.9480 0.0057 m%
2.1468 0.0041 m%
0.3806 0.0024 m%
(2 m%)
2.9957 0.0032 m%
(980 mg/kg)
(1700 mg/kg)
(4500 m~/k~)

Similarly, Laboratory o f G o v e r n m e n t Chemist (LGC) in U.K. has produced several fuel


standards o f low sulfur levels:
Product
Gas Oil
Gas Oil
Diesel
Diesel
Diesel

LGC CRM #
3000
3001
3023
3021
3022

Sulfur Concentration
94.4 2.4 m g / k g
448 11
52.4 1.1
10.67 0.27
~ 32 In Preparation

96

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

On-Line Sulfur Determination


For many operations measuring on-line is a preferred choice rather than waiting for the offline analysis in a laboratory. At present, the U. S. EPA has not usually allowed on-line methods
to be used for fuel products certification. Many on-line methods are extensions of wellestablished test methods such as D 2622, D 4045, D 4294, D 5453, etc., while others were
developed for on-line analysis only. Some of the known on-line methods based on vendor
supplied information are given below. This is not intended as a complete list of such available
technology.
9

9
9

A technology based on combustion-UV Fluorescence method similar to ASTM D 5453


has been widely used in the field. An ASTM standard based on this application is being
developed.
A method based on combustion followed by gas chromatographic separation of sulfur
species and flame photometric detection of S02 has been widely used in the field. This
method was issued as ASTM standard D 7041 earlier this year.
A hydrogenolysis and rateometric colorimetry technique based on ASTM D 4045 is
being developed.
An EDXRF method similar to ASTM D 4294 is available, but it may not have enough
sensitivity to be used for levels in the ultra low sulfur diesel fuels.

There are also other vendors who are working on other technologies to market new on-line
sulfur analyzers for an expected rise in demand for instrumentation in the refineries and blend
terminals.

Emerging New Technology Test Methods for Sulfur


Given the increasing importance of precisely measuring lower and lower quantities of sulfur
in fuels of the future, it is not surprising that a number of organizations are working to develop
better methods for such analyses. Some of these methods, which have been presented for ASTM
review, include:
9

Monochromatic WDXRF: Traditional wavelength dispersive XRF instrument uses


polychromatic excitation. This new technology uses monochromatic focused excitation.
Preliminary work has shown a reproducibility of about 2 mg/kg at a level of 10 mg/kg of
sulfur in gasolines and diesel fuels. Earlier this year this method was issued as ASTM D
7039.
9 Polarization EDXRF: Polarized excitation geometry used with energy dispersive XRF
employing low power tubes and X-ray end window tube with palladium or rhodium
anode and beryllium window is claimed to lower the sulfur detection limit to 1 mg/kg.
The precision of the method has not yet been established.
9 EDXRF with proportional Counter: A low background proportional counter is key in a
new energy dispersive XRF based method to measure lower levels of sulfur. This low
background proportional counter suppresses the noise generated when incident X-rays are
absorbed near the wall with resulting incomplete charge collection. An electrode shield
close to the wall detects incomplete charge collection, and associated electronic detection

NADKARNI ON TRACE LEVELS OF SULFUR

97

circuitry rejects those events, resulting in lower spectral background and consequent
lower detection limit compared to conventional proportional counters.
There is a number of papers being presented in this DO2 Symposium which also deals
with various aspects of new technology for low level sulfur determination in liquid fuels.
Additional information on this subject may be found in Refs [15-18].
References

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]

[15]
[16]

[171
[18]

Magee, E. M., Hall, H. J., and Varga, G. M., "Potential Pollutants in Fossil Fuels," EPA
R2-73-249, National Technical Information Service Publication PB 225-039, 1973.
Dyroff, G. V., "Manual on Significance of Tests for Petroleum Products," ASTM MNL 1,
5th ed., 1989.
Boldt, K. and Hall, B. R., "Significance of Tests for Petroleum Products," ASTM STP 7c,
1976.
Nadkami, R. A., "Modem Instrumental Methods of Elemental Analysis of Petroleum
Products and Lubricants," ASTM STP 1109, ASTM International, West Conshohocken,
PA, 1991.
U.S. Environmental Protection Agency - Federal register 62 (133); 40 CFR Part 80 Part II,
page 37337, July 11, 1997.
U.S. Environmental Protection Agency- Federal Register 64 (1992); 40 CFR Parts 80, 85,
and 86, page 26055, May 13, 1999.
U.S. Environmental Protection Agency - Federal Register 65(28); 40 CFR Part 80, Page
6752 - 74, February 10, 2000; 66(12); 40 CFR Part 80, page 5002-5141, January 18, 2001.
U.S. Environmental Protection Agency - Tier 2/Sulfur Regulatory Impact Analysis (EPA
420-R-99-023, December 22, 1999, Chapter IV: Technology Feasibility, page 311).
Nadkami, R. A., "Guide to ASTM Test Methods for the Analysis of Petroleum Products
and Lubricants," ASTM MNL 44, ASTM International, West Conshohocken, PA, 2000.
Kohl, K, "Fitness for Use of Sulfur Methods," ASTM RR-D02-1456, 1999.
Tittarelli, P., "Round Robin Exercise for Sulfur Test Methods for EN 228 and EN 590 Fuel
Specifications, CEN/TC 19/WG 27, April 2000.
Nadkami, R. A., "Does Your Lab Measure Up? ASTM Program Helps Build
Proficiency," Lubes'N'Greases, April 2000, pp. 28-34.
Nadkarni, R. A. and Bover, W. J., "Bias Management and Continuous Quality
Improvements through Committee D02's Proficiency Testing," ASTM Standardization
News, Vol. 32, No. 6, June 2004, pp. 36-39.
May, W., Parris, R., Beck, C., Fassett, J., Greenberg, R., Guenther, F., et al., "Definition of
Terms and Modes Used at NIST for Value-Assignment of Reference Materials for
Chemical Measurements," NIST Special Publication 260-136, 2000.
Nadkarni, R. A., "Determination of Trace Amounts of Sulfur in Petroleum Products,"
World Refining, Vol. 10, No. 5, June 2000.
Nadkami, R. A., "Detecting Sulfur in Diesel Fuel," Today's Refinery, August 2000.
Nadkarni, R. A., "Determination of Sulfur in Petroleum Products and Lubricants: A
Critical Review of Test Performance," Amer. Lab., Vol. 32, No. 22, November 2000, pp.
16-25.
Nadkami, R. A., "The Challenge of Sulfur Analysis in the Fuels of the Future," ASTM
Standardization News, Vol. 32, No. 6, June 2004, pp. 32-35.

Journal of ASTM International, January 2006, Vol. 3, No. 1


Paper ID JA112972
Available online at www.astm.org
J. Wo]ska, 1 B. Vrebos, 1 a n d P. B r o u w e r 1

Analysis of Fuels, Lubricants, and Greases Using X-ray


Fluorescence Spectrometry
ABSTRACT: X-ray spectrometry (XRF) is a versatile instrumental method for elemental analysis in a

wide variety of materials. The performance of three different XRF systems will be compared: a high
power wavelength dispersive x-ray spectrometer (WDXRF), a low-power WDXRF, and a bench-top
energy dispersive instrument (EDXRF).
The data obtained show quite clearly the large difference in sensitivities between instruments. This
translates into varying lower limits of detection (LLD) or quantification and sample throughput. For
quantitative analysis well above the LLD, the accuracy of the results obtained on the different instruments
is similar, or - for practical purposes of quality control in many applications - the differences in accuracy
between the different instruments can be considered insignificant.
Most of the applications and methods developed by a variety of laboratories and regulatory organizations
are applicable to specimens of a specified nature and matrix. Some of the more versatile procedures
include a method to account for differences in matrix between standard reference materials and
unknowns. One of the effects that is usually neglected in this case is the wedge effect, which is caused by
the fact that the volume that is excited by the beam and the volume from which characteristic radiation
can be observed are not the same and vary (for a given spectrometer system) depending upon the density
and the quantity of the sample. This effect will be illustrated and discussed.
KEYWORDS: Elemental analysis, x-ray spectrometry, analysis of fuels, lubricants and greases, XRF,

WDXRF, EDXRF
Introduction

As regulations regarding the environmental impact o f petroleum products like fuels and
lubricants become more and more strict, the need for accurate and fast chemical analysis
increases. Moreover, accurate and fast chemical analysis also strongly supports the production
and ultimately the quality o f the end product. The demand to reduce the concentrations o f
possibly harmful elements like sulfur imposes new challenges for detection technology.
X-ray spectrometry (XRF) is a versatile, fast, and accurate instrumental method for elemental
analysis in a wide variety o f materials. The performance o f three different XRF systems will be
compared: a high power wavelength dispersive x-ray spectrometer (WDXRF), a low-power
W D X R F , and a bench-top energy dispersive instrument (EDXRF).
Because liquids comprise the majority o f samples in the petroleum industry, specific issues
regarding the analysis o f liquids with X R F will be discussed. Such issues are related to the low
mass attenuation coefficient and therefore the relatively large penetration depth for these
specimens, for example.
Manuscript received 8 October 2004; accepted for publication 13 June 2005; published January 2006. Presented at
ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
i Panalytical BV, Lelyweg 1, 7602 EA, Almelo, the Netherlands.

Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

98

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

99

Principles of Operation
A quick overview of the principles of operation of both WD-XRF and ED-XRF instruments
will be given in this section. The main difference between these two techniques is in the
detection system: wavelength-dispersive systems (Fig. 1, top) use a single crystal
(monochromator) to select a particular wavelength of interest from the radiation emanating from
the specimen. As a result of this, the task of the detector and its associated counting electronics is
limited to counting pulses; in principle, all pulses detected correspond to the same wavelength.
Some fairly straightforward and simple pulse height analyzers can take into account the presence
of pulses corresponding to higher order reflections on the crystal and the escape peak.
Energy-dispersive systems, on the other hand, lack these crystals (Fig. 1, bottom). As a
result, all wavelengths in the beam emanating from the specimen are entering the detector.
Hence, the detector and the counting electronics also have the task of 'sorting' the incoming
photons according to their energy.
Recently, several new implementations have been developed, combining the above principles
with polarization and/or monochromatizing of the incident beam using curved crystals. In Fig. 1,
top right, a WDXRF instrument with a monochromator in the incident beam is represented, while
in Fig. 1, bottom right, the diagram of an EDXRF instrument using a polarizer is shown. These
instruments will not be discussed further in this paper.

FIG. 1--Schematic diagrams of different XRF techniques.

100

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Lower Limit of Detection

The lower limit of detection has been a key issue with the industry. Regulatory bodies, such
as the EPA in the USA and the EC have pushed the allowable levels of several compounds (such
as sulfur) down significantly over the last two decades. Instruments and methods have to follow
these directives, enabling the industry to provide proof that their products are meeting those
requirements. Actually, both of the new developments mentioned earlier are driven by the fact
that the instrumental lower limit of detection can be improved upon by lowering the intensity of
the background radiation. The lower limit of detection (LLD) is dependent on the intensity of the
background radiation as given by:

where LLD is the instrumental detection limit [ppm], S is the sensitivity [1/s/ppm], B is the
intensity of the background [l/s], and T is the measuring time [s]. The parameter n determines
the level of confidence and is usually taken equal to 3. Equation 1 clearly shows the effect of the
counting time, the sensitivity, and the intensity of the background radiation on the value of the
lower limit of detection. It is customary to work with mg/kg, although mg/L can also be used.
The performance of the different X-ray fluorescence systems is largely determined by the
power of the x-ray tube (read: sensitivity) and the specifications of the detector system (read:
intensity of the background radiation). In general, high-end wavelength dispersive systems offer
the best performance because they use high power excitation sources and they exhibit low
background distortions due to the presence of the monochromator. In many instances of quality
control, however, the mechanically much simpler energy-dispersive bench top instntments offer
viable alternatives.
TABLE 1--Typical LLDs in lubricating oils for different type o f XRF spectrometers. All
. numbers . are in ppm
. and are. calculated
. for 100. s.
.
System
Highpower
WDXRF
Lowpower
WDXRF
Highpower
EDXRF
Bench-

Mg
0.9

A1
0.4

Si
0.3

P
0.1

S
0.3

C1
0.8

K
0.2

Ca
0.2

Ti
0.4

Cr
0.3

Mn
0.3

Fe
0.2

Ni
0.1

Cu
0.1

Zn
0.1

Cd
0.4

Sn ""Ba'
0.6 0.7

Pb
0.3

15

13

12

0.5

0.3

0.9

1.3

1.9

1.1

0.6

0.5

3.5

4.1

3.4

12

10

10

12

10

1.1

1.6

0.8

0.7

05

0.2

0.2

0.2

0.8

2.3

0.4

100

34

20

15

10

10

30

16

23

10

top
EDXRF

The data in Table 1 show quite clearly the large difference in the lower limits of detection
(LLD) between instruments. This translates also in significant differences of sample throughput.
For quantitative analysis well above the LLD, the accuracy of the results obtained on the
different instruments is similar, or - for practical purposes of quality control in many
applications - the differences in accuracy between the different instruments can be considered
insignificant. The high-power WDXRF and EDXRF instruments are the more expensive ones
(and have comparable purchase prices), and the low-power WDXRF instrument is about a third
o f these, while the bench-top EDXRF instruments cost about 10-20 % of the high-end ones.

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

101

FIG. 2--Calibration graphs f o r sulfur obtained on two different spectrometers. The upper
curve (squares) is achieved on a high-power WDXRF instrument, while the lower curve is
achieved on a benchtop E D X R F spectrometer.
Matrix Effects

X-ray fluorescence analyzes the material in a condensed form (solid or liquid), compared to
techniques that work in the gaseous or plasma state. As a result o f this, the intensity of the signal
of a particular analyte is usually not directly proportional to its concentration, but is also subject
(to a lesser but not negligible degree) to the concentration o f the other elements present ('the
matrix'). The matrix elements influence the intensity in two distinct ways: absorption (generally
causing a reduction o f intensity) and enhancement (an increase in intensity). The measured
intensity is obviously the net result of these matrix effects. As a result, the calibration curves can
be convex or concave (see Fig. 3). The physics of XRF, however, are well understood, and
correction algorithms have been developed to deal with these effects in a mathematical way.
Also, compensation methods (in which the effect is made to cancel out) have been developed. A
popular one is the use o f an added internal standard. In this method, a fixed concentration o f a
particular element is added to the specimen. The element is chosen so that it is affected by the
matrix effect (absorption and enhancement) in a way similar to the analyte's radiation. The
intensity ratio is then independent of the matrix effects. An example of such method is ISO
14596, which uses Zr as an internal standard for the determination of S in oil, or ASTM 5059,
which uses Bi as an internal standard for determination of Pb in gasoline.

102

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 3--Influence o f absorption and enhancement effect on a calibration curve.

Sample Thickness
The matrix in oils consists primarily of low atomic number elements, which have a low
absorption for the x-ray energies o f interest. This causes the primary beam to penetrate much
deeper in the specimen than usually anticipated. Also, the characteristic radiation of the analyte
elements experience low absolute absorption. Between specimens, however, the relative
differences can still be significant, leading to matrix effects. The low absolute absorption causes
the photons to travel a long way - and just like the photons of the primary beam, geometry
becomes an issue as indicated in Fig. 4. The critical depth, d, is given by:
4.61 .
d =
sin V~
(2)
PP
where d = critical depth [cm];/d = mass attenuation coefficient [cm2/g];/9 = density [g/cm3];
and ~2 = take off angle.

FIG. 4--Critical depth.

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

103

This thickness corresponds to the thickness required to absorb 99 % of radiation in a material


with a given density and at a specified energy.
The data in Table 2 give the critical depths (or infinite thickness as they are also referred to)
for lead, iron, silica, lithium tetraborate, and water for photons corresponding to Mg Ka, CrKa,
and Sn Kct. Mg K a radiation (with an energy of 1.25 keV) is readily absorbed within a very
small distance - less than 20 microns in all materials listed. Some photons corresponding to Cr
Kcx (energy of 5.41 keV) will travel a distance of roughly one millimeter in water. This means
that it takes a layer of water 1 mm thick to absorb 99.9 % of the fluorescent photons. The small
remaining fraction will eventually be absorbed deeper in the material. For Sn Kc~, the situation
becomes quite different, as we are now talking about path lengths of centimeters.
Please note that the distances travelled in oil are not much different from those given in
water; even though there is some minor influence by the C/H ratio and the content of oxygen in
the oil considered.
TABLE 2--Critical depth for different materials for
distances in microns, unless indicated otherwise.
Material
Mg Kcr
Cr Kcz
Lead
0.7
4.5
Iron
1
35
SiO2
8
110
Li2B407
13
900
H20
16
1000

Mg Ka, Cr Kv~ and Sn Ka. All


Sn K a
55
290
0.9 cm
4.6 cm
5.3 cm

In XRF, the specimens are commonly assumed to be infinitely thick for all photon energies
considered (including the incident beam). This is certainly justified in industrial applications
when dealing with alloys. So, many quantitative methods are implicitly derived under this
assumption. However, as the matrix of the specimens consists of more and more low atomic
number elements, this assumption is no longer justified, and applying the methods developed for
.analysis of alloys will lead to significant bias. This is caused primarily by the fact that the
incident beam is not fully absorbed within the specimen. If the fraction absorbed is constant
between specimens and standards, a workable solution is achieved in many cases. The
assumption that the complete incident beam is absorbed and causes fluorescence in the specimen
is then merely replaced by the fact that a constant fraction of the beam is absorbed; as shown
above in Fig. 4. Several methods have been derived with the explicit assumption that the
specimens are not infinitely thick, but these are beyond the scope of this paper. An extensive
review can be found in the work by van Grieken and Markowicz [1].

The Wedge Effect


Another effect compounding the situation is the wedge effect. The algorithms developed for
specimens of less-than-infinite thickness are usually derived with the assumption that the
thickness is small compared to the surface area. This implies that the surface of any layer within
the specimen is constant. This assumption is no longer justified when the thickness (or rather the
path length) is comparable to the diameter of the specimen; see Fig. 5.
Consider the sample as being built up by many very thin layers of the same composition,
parallel to the measured surface of the specimen. The elements in each of these layers are excited

104

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

by the photons of the primary beam and emit their characteristic radiation. The attenuation of the
primary beam and of the secondary beam by other layers on the way from the x-ray source and
toward the detector can be taken into account. As long as the penetration depth of the radiation is
small compared to the surface area, the area of each layer (and thus the amount of material
contributing to the measurement in each layer) can be considered constant.

FIG. 5--Schematic illustration of the wedge effect.


As the penetration depth becomes larger and larger, eventually the volumes that are not
excited or where fluorescent radiation generated is not detected become significant. This effect is
called the shadow effect or the wedge effect (see Fig. 5) and finds its origin in geometrical
aspects of the instrumentation. The fraction of each layer that contributes to the measured signal
decreases as the layer is positioned further from the specimen surface. This decrease is additional
to the decrease caused by absorption by intermediate layers.
The intensities ofSn Kct and Cu Kc~ of butanol samples containing 500 ppm Sn and 500 ppm
Cu each were measured, but with different volume (thickness). All samples had the same
diameter of 37 mm. The penetration depth from Sn Kct in butanol is about 5 cm, versus only
about one millimeter for Cu Kot (see Table 2, and noting that butanol consists of C, H, and O; the
value for H20 is used for reference purposes only; the theoretical calculations are performed with
the correct matrix type).
The data in Fig. 6 and Table 3 show that the measured intensity reaches its maximum value
for a thickness of 20 mm. A sample with a thickness of 8 mm already produces 85 % of the
maximum reachable intensity. Also, from Fig. 7 it can be concluded that one way to avoid the
wedge effect is to analyze small amounts of sample: if the thickness of the specimen is small
(e.g., below 1 mm, although this depends on the energy of the analyte's radiation), the effect
becomes negligible. When this approach is taken, however, it must be clear that always the same
mass of specimens must be used for analysis, or a finite thickness correction must be used.
For Cu-Kct radiation the effect is much less pronounced than for Sn-Kc~ radiation, because
the penetration depth for Cu-Kot radiation is much lower than that of Sn-Ko; as illustrated in Fig.
7.
Methods based on an internal standard (such as ISO 14596, which uses Zr as an internal
standard for the determination of S in oil) will implicitly correct for tiffs wedge effect and for the
matrix effect because the effect is the same for both the internal standard and the analyte lines.

105

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

60

.4,- ~

SO

12O

40

W
9

30

f
I

W
C

o.

2o

10

15

20

25

T h i c k n e s s [ram]

F I G . 6--Measured intensity (solid line) and calculated intensities of Sn Ko~ Dashed line:
calculated without wedge correction; dotted line: calculated with wedge correction of samples
with different thicknesses. Note that the wedge effect is significant for Sn Ka in butanol (refer to
the penetration depth of Sn K a in [-120 in Table 2).
TABLE
Thickness
[mini
4.1
8.2
12.2
16.3
20.4

3--Penetration depth.

,,

Intensities [kcps]
Calculated with wedge
effect
20
30.6
35.4
37
37

. . . . Calculated without
wedge effect
20
34.3
44.3
51.4
56.4

Measured
20
31.8
36
37.3
37.4

41.00

J,

40.00.

<<-"
Q.
CJ

........

J_

. . . . . . . . . .

39.00.

,,~ 38,00 9

36.00 -

35.00
5

10

Thickness

15

20

25

[ram]

7--Measured intensity (solid line) and calculated intensities of Cu Ket (dotted line with
wedge correction, dashed line without wedge correction) of samples with different thickness.
FIG.

106

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

In case of greases where introducing the internal standard into the sample itself is often not
possible, as it is in liquids, the alternative may be analyzing the thin grease layer distributed in
the sample cup. The thickness of each sample has to be the same, and instead of internal
standard, a plate or foil from Nb in this case can be used. This practical approach is illustrated in
Fig. 8.
But for methods without internal standard, care must be taken to matrix-match the standards
and the unknowns, and to work with constant weight in the sample cups. The alternative
procedure is to account for the wedge effect and to use a full fundamental parameter approach to
the analysis.

FIG. 8--Preparation of a grease sample.


Conclusion

XRF is a well-established and versatile instrumental technique for elemental analysis in a


broad range of mostly industrial applications. Where high-power WDXRF yields the best results
in terms of LLD, the more economical bench-top, on the other hand, often gives quite acceptable
results, especially in production control.
The wedge effect is significant for high energetic radiation in low-Z matrices. The effect can
be modelled adequately by taking it into account in the Fundamental Parameter equations.
When analyzing heavy elements in low-Z matrices, two scenarios can be used: either
measure only a small amount of sample to reduce the wedge effect, or take the wedge effect into
account if larger amounts are measured. In both cases, finite thickness correction is required.
Measuring only small amounts will not affect the sensitivity drastically.
A number of XRF methods currently exist in ASTM literature and are widely used. These
include D 2622, D 6334, D 4294, D 5059, and D 7039 for fuels; and D 4927 and D 6443 for
lubricants.

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

107

Reference

[1] "Quantification in XRF Analysis of Intermediate-Thickness Samples," Handbook of X-Ray


Spectrometry, 2nd edition, R. van Grieken and A. Markowicz, Eds., Marcel Dekker, New
York, Basel, 2002.

Journal of ASTM International, September 2005, Vol. 2, No. 8

Paper ID JAI12974
Available online at www.astm.org
Scott F e s s 1

Determination of Sulfur Content in Crude Oil Using On-Line


X-ray Transmission Technology
ABSTRACT: This paper describes the function and use of X-ray Transmission (XRT) analytical
instrumentationas applied to the classificationand blending of crude oil that contains0.1-3.3 % sulfur by
weight. In the XRT method, sulfur absorbs X-rays transmitted between an X-ray source and detector.
The techniquecompensates for density changes in the oil and minimizesthe effects of variations in the
C/H ratio of the oil. When used on-line the instrument obtains the sulfur measurementin the highpressure flow conditions required to pump crude oil, typically 800 psig in pipelines. Automatic
compensationfor variationsin oil densityand temperature are achievedusing integrateddensitometerand
temperature sensor readings taken by the instrument as the sulfur is being measured. In this paper
calibration,precision,and long-termstabilityresults are shown. Experimentaldata were obtainedusing a
commerciallyavailableon-line XRT analyzer.
KEYWORDS: sulfur, fuels, crude oil, refinery,pipeline

Introduction
As sulfur is tightly regulated in petroleum fuels, such as gasoline and diesel, the sulfur
content of the oils must be closely monitored throughout the entire refining process, beginning
with crude oil. Determining the sulfur content in crude oil is of importance for the classification
and blending of the crude. Although government regulations control the allowable sulfur content
in fuel oils, the sulfur content in crude oil is not directly regulated. Thus, the quality and price of
crude oil depends on the sulfur content. Sweet crude contains a maximum of 0.50 % (wt/wt)
total sulfur, whereas sour crude may have a maximum o f 2.99 % total sulfur. Some crude oils
may contain as much as 6 % total sulfur. On-line analysis is highly desirable for the
classification and blending of crude oil, requiring a fast and robust method of analysis.
This paper examines the use of the XRT (X-ray Transmission) technique as applied to the
analysis of sulfur in crude oil. XRT may also be referred to as XRA (X-ray Absorption).

Analytical Techniques and Instrumentation


X-ray Fluorescence (XRF) is a common technique for sulfur determination in hydrocarbon
oils. At-line and laboratory XRF analysis is covered in the ASTM standard test methods for
sulfur in petroleum products: ASTM Standard Method D 4294 (for Energy Dispersive XRF,
EDXRF) [1] and ASTM Standard Method D 2622 (for Wavelength Dispersive XRF, WDXRF)
[2]. Polarized EDXRF [3] is also used at-line for the determination of very low sulfur content
(<10 ppm sulfur) in diesel and gasoline fuels.
Manuscript received 8 October 2004; accepted for publication 27 January 2005; published September 2005.
Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future
Prospects on 6-8 December2004 in Tampa, FL.
i SPECTROAnalyticalInstruments,Inc., 1515N. Highway 281, MarbleFalls, TX 78654.
Copyright9 2005byASTMInternational,100Ban"HarborDrive,POBoxC700,WestConshohocken,PA19428-2959.
108

FESS ON ON-LINE X-RAY TRANSMISSION

109

EDXRF is well suited for the on-line measurement of sulfur content for less dense oils that
are pumped at low pressures, typically at a maximum of 30 psig. Refineries, pipelines, and
blending operations, however, require on-line analysis of thick, dense, viscous crude oil pumped
at high pressure. These environments require a robust analysis technique with instrumentation
capable of accurately measuring crude oil pumped at a pressure of 800 psig. XRF is not
desirable for such an analysis.
XRF measures the fluorescent X-ray emission of sulfur, which occurs at an energy of 2.308
keV (K-alpha emission line). Absorption effects are therefore much greater than in the XRT
method. Matrix effects that occur in XRF can be minimized using XRT. Also, because the
sulfur fluorescent X-rays are low energy, the path length that they travel before being completely
absorbed is a tiny fraction of the path length of X-rays with energy 21 keV used in the XRT
technique. Therefore, coatings that may build up on the flow cell window will greatly affect the
XRF measurement. Furthermore, the thin windows necessary to efficiently transmit the soft
fliaorescent X-rays cannot sustain pressure differentials as high as 800 psig.
For these reasons, XRT is a more desirable technique for the on-line analysis of sulfur in
crude oil. ~,a XRT, X-rays are emitted from an X-ray source and pass through the volume ofoil
to a detector on the opposite side of the sample flow cell. X-rays are emitted from an X-ray
source at a high enough energy to pass through the volume of oil to a detector on the opposite
side of the sample flow cell. The presence of sulfur in the oil absorbs these source X-rays, and
so calibration is obtained by relating concentration to X-ray count rate. This calibration is an
inverse relationship: as the sulfur concentration increases, the X-ray intensity reaching the
detector decreases. In this way the sulfur is indirectly measured by measuring the amount of Xray radiation transmitted through the oil.
The transmission (or absorption) of X-rays is given by the following equation [4]:
I/Io = exp-dt[ ~tm(1-Cs)+lasCs]
where:
I = final X-ray intensity, after absorber [counts]
Io = initial X-ray intensity, before absorber [counts]
d = density of absorbing material [g/cm3]
t = thickness of absorber [cm]
#m = mass absorption coefficient for the matrix [cm2/g]
/~ = mass absorption coefficient for sulfur [cruZ/g]
Cs = weight fraction of sulfur [% wt/wt]

(1)

Table 1 shows the mass absorption coefficients for the hydrogen, carbon, and oxygen
components of organic oil, as well as ~t for sulfur. Because the sulfur ~t is on the order of ten
times greater than those for hydrogen, oxygen, and carbon, the transmission of X-rays are
primarily proportional to the sulfur content.
TABLE 1--Selected mass absorption coefficients at X-ray energy 21 ke V [5].
Element
Mass Absorption Coefficient, ~t [g/cm2]
H
0.37
C
0.41
O
0.79
S
5.82

1 10

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Possible interferences include the presence of chlorine in the oil and coating of the flow cell
beryllium windows. Chlorine may interfere with the sulfur measurement if the chlorine is
present in a concentration ratio to the sulfur of greater than 1:10 and the chlorine concentration
varies independently of the sulfur. Flow cell window coating has negligible effect on the sulfur
measurement in XRT. This is because a coating build-up of a few hundred microns is a
negligible part of the entire X-ray path length.
The SPECTRO Model 682T-HP on-line XR.T analyzer was used to obtain the data presented
in this paper [6]. All measurements and data were obtained with the sample in a static condition,
completely filling the sampling pipes (flow cell and densitometer paths). X-rays and X-ray
transmission are not affected by sample flow.

Description of the Instrument


Figure 1 shows a drawing of the analyzer. The essential components are: sample flow cell;
X-ray source; X-ray detector; and densitometer. Figure 2 shows the sample flow and data output
of the analyzer.

Detector

CabinetDoor in the
Open position

SampleFlow Cell
FIG.

l--Component diagram.

X-raySource

FESS ON ON-LINE X-RAY TRANSMISSION

11 1

RETURN TO PROCESS

I" 600# ANSI RF FLANGE--~

A/02 = Density
CaUbratlon S e l e c t
D/J2 = Remote S t o r t / S t o p
I)/[ll = Common At~rm
D/It

I]/Q2 = Resu~UDd~ted(PuLse)

MANUALSAMPLEINPUT/VENT

POVER
110/220VAC
3/1,5 A, 60/50Hz

SPECTR~]
XRA/XRTANALYZER
INPUT/OUTPUT

PURGESUPPLY
9 ) 5 kg/r
PSI)
) 0.22~ /hP r $CFH)
Depends on Area
Ct~s$1FIc~tlon

~
1" 60D# ANSI RF FLANGE

PRESSURE RELIEF VALVE

PROVIDED--,-BY SPECTRD

PROVIDED
BY CUSTOMER

MANUAL DRAIN

I I E~
S V { ~

DRAIN

INLET FRDMPROCESS
10 - 60"C (50 - 140"F), 56 kg/c~e(BOOPSI) MAX

FIG. 2--Sample flow and data output.

Sample Flow Cell


The oil sample enters the bottom o f the flow cell tube and exits at the top. The inner
diameter of the flow cell tube is 4.90 cm. The X-ray measurement is taken as the sample flows
through the tube. Beryllium windows allow the X-rays to enter and exit the analysis volume, the
X-rays traversing a total path length of 8.64 cm.

X-ray Source
The X-ray source is a 30 kV side-window X-ray tube with a rhodium anode target. X-ray
tube voltage and emission current are set in order to yield source X-rays with energy above 25
keV and approximately 30 000 cps at the detector. For these tests the following X-ray tube
parameters were used: tube voltage of 27 kV and emission current of 75 uA. To obtain a more
monochromatic X-ray source in the range 20-30 keV, a palladium filter is mounted over the Xray tube. This gives source X-rays primarily at 21 keV, the K-line emission of palladium.

X-ray Detector
The detector is a gas-filled proportional counter with an argon fill gas. The bias voltage is set
such that the detector operates in the normal proportional counter mode, detecting all X-ray
energies between 3 keV and the tube voltage setting.

112

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Densitometer
Oil from the process stream simultaneously flows through the X-ray sampling flow cell and a
densitometer. The densitometer measures the oil density and temperature and inputs these values
into the data processing software for corrections to the calibration and final analysis results.

Data Output
Total X-ray counts from the detector are simply accumulated and output using a
programmable logic controller (PLC). The total counts are expressed as count rate (total counts
divided by analysis count time) and sent to the software for use in calibration and analysis
algorithms.
Performance

This section describes typical results obtained for calibration, precision, sensitivity, and longterm stability. A comparison o f repeatability is also made with ASTM D 4294. All
measurements were obtained using an analysis count time of 100 s, and results are quoted as %
wt/wt.

Calibration
Calibration was performed b y analyzing a suite o f certified crude oil standards. [7] The assay
values for these samples are given by the certification shown in Res 7. A broad range
calibration was built for the range o f sulfur from 0.1-3.3 %, with the density increasing with
sulfur content. To refine the calibration over the specific range used in blending operations, a
second calibration was built over the range o f 0.1-1.5 % sulfur, with independently varying
densities.
,,, TABLE 2--Broad range correlation results.
Element: S SEE = 0.1624
Density [g/cm3j Assay Value [%J MeasuredValue [%J
Sample
0.763563
0.10
0.089
1
0.836063
0.58
0.599
2
0.883313
1.01
1.098
3
0.862688
1.97
1.758
4
0.958500
2.48
2.689
5
1.009310
3.28
3.187
6
TABLE 3--Blending range correlation results.
Element: S
SEE = 0.051
Sample
Density [g/era3] Assay Value [%] MeasuredValue [%]
11
0.902630
0.10
0.092
7
0.798000
0.30
0.370
12
0.900380
0.30
0.273
9
0.848630
0.50
0.461
10
0.882190
1.00
1.000
13
0.898880
1.00
1.033
8
0.801000
1.50
1.440
14
0.897188
1.50
1.538

FESS ON ON-LINE X-RAY T R A N S M I S S I O N

113

TABLE 4---Precision-based on broad range calibration.


Element: S
Sample
Assay Value*
Mean Value
1a Std.
%
[%]
[%]
Dev.
Relative
1
0.10
0.104
0.005
4.6
6
3.28
3.225
0.003
0.1
* As givenby the certification. See Ref. [7].
Broad Range Correlation Plot
3.500
3.000
==
2.500
2.000
1.500

1.000
0.500 t
0.000

0.00

1.00

2.00
3.00
Assay Value

4.00

FIG. 3--Broad range correlation plot.


Blending Range Correlation Plot
1.800 ,

=e 1.200 ]
0.900 1

1
0.00 0.30 0.60 0.90 1.20 1.50 1.80
Assay V a l u e

FIG. 4--Blending range correlation plot.

Precision and Sensitivity


The precision results for selected samples shown in Table 5 were obtained by making 100
repeat analyses of each sample using a count time of 100 s per analysis. A detection limit was
obtained using the empirical method. The Minimum Detection Limit (MDL) here is defined as
the 3tr observed standard deviation of 100 analyses of a blank sample containing no sulfur.
Using heavy mineral oil as the blank sample, the MDL for sulfur was observed to be 0.02 %.

114

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

TABLE 5--Precision-based on blending range calibration.


Element: S
Sample
Assay Value* [%]
Mean Value [%]
lo Std. Dev.
11
0.10
0.093
0.006
7
0.30
0.344
0.005
10
1.00
0.991
0.006
13
1.00
1.022
0.005
14
1.50
1.536
0.004
* As givenby the certification. See Ref.7.

% Relative
7.0
1.4
0.6
0.5
0.3

Long-term Stability
Long-term stability data were gathered over 3090 continuous measurements of a static
sample over a 92-h time period. The results are shown in Table 6.

Comparison to ASTM D 4294


Section 15.1.1 (Repeatability) of ASTM D 4294 expresses the r-value for the measurement
of sulfur in oils using EDXRF to be:
r = 0.02894(X+ 0.1691)
(2)
where Xis the sulfur concentration in mass %.
Table 7 shows a comparison of the precision results of the XRT method (derived from Table
5) and the r-value (repeatability) given in D 4294. The table shows the 95 % confidence value
(2.77 o) for the precision results using XRT as compared to the maximum 95 % confidence rvalue listed in D 4294.
These results indicate that the XRT on-line method is equivalent to the repeatability
performance quoted in ASTM D 4294 (for EDXRF) for sulfur concentrations above 0.10 %.
TABLE 6---3090 repeat analyses at 100 s per analysis.
Element: S
Sample
Assay Value*
Mean Value
la Std.
%
[%]
[%]
Dev.
Relative
14
1.50
1.538
0.005
0.3
* As givenby the certification. See Ref. 7.

Sample
11
10
7
13
14

TABLE 7--XRT results compared to ASTM D 4294.


XRT
XRT 2.770
r-value from D
MeanValue(%Sulfur)
Std. Dev.
4294
0.093
0.017
0.008
0.991
0.017
0.034
0.344
0.014
0.015
1.022
0.014
0.034
1.536
0.011
0.049

FESS ON ON-LINE X-RAY TRANSMISSION

1 15

Conclusions

9
9
9
9

The X-ray Transmission technique is a viable method for the determination of sulfur
content in crude oil.
Precision for sulfur above 0.1% is equivalent to the r-value (repeatability performance)
quoted in ASTM D 4294.
The XRT method is better suited than XRF for the process control of sulfur in crude oil.
Future work is required to determine the response time of the instrument to step changes
in sulfur concentration and to test flowing conditions of the sample.

References

[1] ASTM D 4294, Standard Test Method for Sulfur in Petroleum Products by EnergyDispersive X-ray Fluorescence Spectroscopy, ASTM Annual Book of Standards, ASTM
International, West Conshohocken, PA, 19428.
[2] ASTM D2622, Standard Test Method for Sulfur in Petroleum Products by X-ray
Spectrometry, ASTM Annual Book of Standards, ASTM International, West Conshohocken,
PA.
[3] SPECTRO Analytical Instruments GmbH & Co. KG, Boschstr. 10, 47533 Kleve, Germany.
[4] Liebhafsky, H. A., Pfeiffer, H. G., Winslow, E. H., and Zemany, P. D., X-ray Absorption and
Emission in Analytical Chemistry, 2nd ed., John Wiley & Sons, Inc., 1966, pp. 130-133.
[5] Robinson, J. W., CRCHandbook of Spectroscopy, Vol. 1, CRC Press, Inc., 1974, p. 208.
[6] SPECTRO Analytical Instruments, Inc., Marble Falls, TX.
[7] Analytical Services, Inc., The Woodlands, TX. Standards certified and assayed using ASTM
D 2622.

Journal of ASTM International, September 2005, Vol. 2, No. 8


Paper ID JAI12971
Available online at www.astm.org

Z. IV. Chen, 1 Fuzhong Wei,11an Radley, i and Berry Beumer 1

Low-Level Sulfur in Fuel Determination Using


Monochromatic WD XRFmASTM D 7039-04
ABSTRACT: A monochromatic wavelength dispersive X-ray fluorescence (MWDXRF) technique has
been successfully developed for measuring low-level sulfur in fuel. In this technique, two doubly curved
crystal (DCC) optics were used to provide monochromatic excitation and fixed channel wavelength
dispersive analysis. Using highly efficient DCC optics, compact bench top MWDXRF analyzers for
sulfur analysis have been successfully produced. A new ASTM standard test method for low-level sulfur
determination in diesel and gasoline, D7039-04, has been developed based on this technique. Data for
ultra-low-sulfur fuel were collected and analyzed using this new method. In this paper, the results of
repeatability, reproducibility, and bias are presented and discussed. The repeatability for 10-ppm sulfur
fuel was shown to be around 1 ppm. The pooled limit of quantification (PLOQ) for ultra-low-sulfur
diesel was found to be less than 1.5 ppm in this study. The reproducibility of 15-ppm sulfur diesel fuel
was determined to be better than 3 ppm (95 % confident level). The limit of detection for a single
analyzer was found to be 0.36 pprrL The effect of matrix composition was investigated. The correction
due to matrix for low-level sulfur diesel and gasoline is generally not significant.
KEYWORDS: sulfur, diesel, X-rays, D7039, fluorescence, optics
Introduction

To deliver low-sulfur fuel to meet the new EPA regulation, it is important to monitor the
sulfur level during the fuel-refining process and transport in a pipeline terminal. X-ray
fluorescence (XRF) analysis has been demonstrated to be a reliable and accurate method for online measurement o f sulfur in petroleum processes. However, conventional XRF using energy
dispersive spectrometry (EDS), without the use o f X-ray optics, is limited in detection o f lowlevel sulfur in oil due to a relatively poor signal-to-background ratio (S/B). Another serious
issue for any XRF/EDS system is that the strong scattering o f the primary beam from the sample
can adversely inundate the energy dispersive detector resulting in large detector dead times
during measurement. This limits the intensity of the primary beam due to the counting
limitations o f the energy detector.
Alternatively, XRF with a wavelength dispersive
spectrometer (WDS) gives improved sensitivity and precision over EDS systems. However, high
performance WDS systems require high power X-ray tubes with water-cooling and extensive
maintenances. These systems are not suitable for distribution terminal quality control and online applications.
XRF analysis using monochromatic excitation can provide a much better S/B than the
polychromatic excitation for EDS and WDS XRF techniques described above by eliminating the
scattering o f bremsstrahlung (continuum X-rays) from the X-ray tube. Laboratory
monochromatic XRF methods have been hindered b y the lack o f efficient X-ray focusing
Manuscript received 9 November 2004; accepted for publication 27 January 2005; published September 2005.
Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future
Prospects on 6-8 December 2004 in Tampa, FL.
I XOS, 15 Tech Valley Drive, East Greenbush, NY 12061.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

116

CHEN ET AL. ON MONOCHROMATIC WDXRF

117

monochromators until recently. With innovative point-to-point focusing doubly curved crystal
(DCC) optic devices [1,2], intense monochromatic focused beams can be achieved using low
power compact X-ray sources. The basic configuration of a compact monochromatic WDS XRF
sulfur analyzer with single channel analysis based on DCCs is shown in Fig 1. It consists of an
X-ray tube, a point-focusing DCC for X-ray excitation, a sample mount, a focusing DCC for
collection of characteristic X-rays from the sample, and an X-ray counter. The X-ray tube has
power of 30-75 W and its spot size is in the range of several hundreds of microns. In this
system, the first-point focusing DCC captures a narrow bandwidth of X-rays from the source,
typically a characteristic line of the X-ray tube, and focuses an intense monochromatic beam to a
small spot onto a sample. The target material of the X-ray tube is chosen to have proper
characteristic line for the excitation of sulfur atoms. During sample excitation, sulfur K~x
secondary characteristic fluorescence X-rays are emitted from the excitation volume of the
sample. The second DCC, the collection crystal, collects the sulfur Kct X-rays and focuses them
at the detector. The X-ray detector can be a proportional counter or a solid-state detector. The
intensity of the detected X-rays has linear relationship with the sulfur elemental concentration in
the specimen. With proper calibration, the sulfur concentration in a sample can be directly
determined.
Sam )le

--

,,

<5 tor

Source
Count

DCC
Optic2
DCC
Optic1

FIG. l--Principle of Monochromatic WavelengthDispersiveXP~Ffor sulfur analysis.


The most important advantage of the monochromatic wavelength dispersive X-ray
fluorescence (MWDXRF) approach is that the S/B is improved by using the monochromatic
excitation of the X-ray source characteristic line. The primary beam is highly monochromatic,
and the bremsstrahlung photons from the X-ray tube with energies corresponding to the sulfur
fluorescence peak can only reach the detector by an unlikely two-step successive scattering
process. Therefore the S/B ratio is drastically improved compared to polychromatic excitation.
The other advantages of this unique MWDXRF sulfur analysis system are that is has no moving
parts, is very compact, and requires very little maintenance.
An ASTM standard, D 7039-04, has been recently designated for this MWDXRF technique.
In this paper, data for sulfur analysis in diesel and gasoline using D7039-04 are presented and
discussed.

1 18

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Materials and Methods

Samples used in this study included commercially available sulfur in diesel standards,
gravimetric samples prepared in-house, NIST SRMs, and refinery diesel and gasoline products.
The commercial sulfur diesel standards were gravimetricaUy based and were obtained from
AccuStandard in New Haven, CT, and Analytical Service Inc. in Woodlands, TX. The in-house
samples were made by mixing n-butyl sulfide (CsH18S) with decalin (C10H18), tetralin (Cl0Hl2),
or decane (CloH22). The NIST SRMs included 1616a, 2723a and 2299.
Bench top MWDX_RF analyzers were used in this study. Measurements were performed in
two different laboratories. Five different bench top instruments were used in the XOS R&D
laboratory and one instrument was used for measurements acquired in a refinery lab.
For laboratory measurements, samples were measured directly with minimal sample
preparation. Prior to analysis, the samples were prepared by placing --4 mL of liquid sample into
a Chemplex XRF disposable plastic liquid sample cell fitted with a Chemplex cardboardmounted mylar window (3.6-pza thickness). The analytical volume of the MWDXRF method is
small due to the focusing nature of the incoming X-ray beam. The full size of the excitation
beam is about 1 m m x 1 mm on the Mylar window and the sulfur X-rays 1/e escape depth for
fuel is ~0.1 ram. Therefore, only a small amount of the specimen near the center of the film is
analyzed during measurement. The sample cell was vented immediately after sample
preparation if gasoline or another volatile sample was used. The Mylar film surface was kept
dust free, flat, and wrinkle free near the center area of the sample cell to avoid errors during
measurement.
Results and Discussion

Calibration and Linearity


The bench top analyzer was first calibrated using commercial gravimetrically based diesel
standards. Two sets of commercial diesel standards and a set of in-house prepared standards
were used to generate three different calibration curves for comparison. The first set of standards,
made from No. 2 diesel (Vendor A), included sulfur concentration levels of 0 ppm (blank), 5
ppm, 10 ppm, 100 ppm, 500 ppm, 1000 ppm, and 3000 ppm. The second set, made from a
synthetic diesel fuel matrix (Vendor B), included sulfur levels of close to 0 ppm (blank), 20 ppm,
50 ppm, 100 ppm, 200 ppm and 500 ppm. The in-house standards were made by mixing different
ratios of decalin and n-butyldisulfide (CsHjsS) gravimetrieally with resulting sulfur
concentration levels of close to 0 ppm (blank), 10 ppm, 48.6 ppm, 87.6 ppm, 320 ppm, and 946
ppm. The measured sulfur fluorescent X-ray intensity, in counts, for each standard is listed in
Table I.
Based on the repeatability statement in the ASTM D 7039-04 method, the precision of the
MWDXRF method is a function of the concentration and it can be best fit by a square root
function. Therefore, a weighted linear least square fit is needed to obtain a proper calibration
curve over a wide range of concentration levels. Since X-ray counting is governed by Poisson
statistics, and the uncertainty due to Poisson statistics is comparable to the precision statement of
the ASTM D 7039-04 method, a satisfactory fitting result can be achieved by using the Poisson
fluctuation as a weighted error bar. The linear regression function can be written as: Y =
A+B*X, where Y is the X-ray counts obtained in 300 s, X is the concentration, A is the
calibration intercept, and B is the slope of the calibration curve. Weighted linear regression

CHEN ET AL. ON MONOCHROMATIC WDXRF

119

results are compared with non-weighted fitting results (Table 2). The regression output includes
the intercept, the slope, and the linear correlation probability R.
TABLE 1 Sulfur X-ray intensity with corresponding concentration for three sets of
calibration standards.
No. 2 Diesel Set (VendorA)
Concen~afio~
Measured
pprn
Counts
0.2
301
5
943
10
1 611
100
13 965
500
68 507
1000
136 784
3 000
407 712

SyntheticDiesel Set (Vendor B)


Concengafio~
Measured
ppm
Cotm~
0.5
291
20
3 042
50
7 186
100
14 020
200
27 772
500
68 464
.
.
.
.
.
.
.

In-houseSet, Decalin-based
C o n c e n ~ a t i o ~ Measured
ppm
Counts
0
244
10.0
1 582
48.6
6 905
87.6
1 2334
320
4 4114
946
1 29686
.
.
.
.
.

TABLE 2--Comparison of linear regressio n using non-weighted and weighted data.


CalibrationSet
Intercept, A, Counts
Slope, B, Counts/ppm
Linear Correlation
Probability,R2
Weighted
NonWeighted
NonWeighted
Nonweighted
weighted
____weighted
No. 2 Diesel Standard
271 13
418 :~132 136.2a: 0.2
135.8
1.0000
1.0000
Synthetic Diesel Standard
234 19
342 :~57
137.1 0.5 136.4 0.3
0.9994
1.0000
Deealin-basedStandard
242 + 8
273 29
136.9 0.2
136.8
1.0000
1.0000
It is clearly shown that non-weighted linear fit gives high uncertainty of the intercept. This
can adversely affect the accuracy o f the low ppm measurement. Particularly for the two
commercial calibration standard sets, the biases for the low ppm measurements can be as much
as 1 ppm if non-weighted fitting is used. The linearity of the calibration curve is exceptional for
all three sets of standard. There is no significant difference among the three calibration curves.
The in-house standard has the smallest fitting error. The final calibration for this study was set at
A = 242 and B = 136.9. The calibration line with the data points is shown in Fig. 2. The
relationship between the X-ray counts and the sulfur concentration is given: W (in ppm) =
((counts/T)-B/300)/(A/300), where T is the measurement time in seconds.

160k-

8k

~'120k-

4k
Ok

"~ 8Ok.
@
~ 4Ok- S

i=242+136.9~
w

R=I,O000

Ok
0 2o0 46o soo eoo lobo
Sulfur Concentration W (ppm)
FIG. 2 - - T h e calibration curve for a bench top sulfur analyzer." the open circles are the

calibration data points and the solid line is the linear fit with weighted error bars.

120

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Repeatability and Precision


Repeatability measurements were performed at two sites: the XOS R&D laboratory and a
refinery laboratory.

XOS Results
All the measurements were done with the bench top analyzer with the calibration curve
shown in Fig. 2. Three bottles of diesel samples with sulfur levels of 5 ppm, 10 ppm, and 300
ppm were measured periodically over 20 days. For each measurement, a specimen was taken
from a sample and placed into the bench top analyzer and a reading was obtained. Two
successive measurements were performed on each day for each sample. The two successive
measurements were done within 15 min. of each other. Forty measurements were obtained in
total for each concentration level and the results are shown in Fig. 3. The difference between
each pair of measurements, total 60 pairs, is shown in Fig. 4.
Single site repeatability can be determined from the data shown in Figs. 3 and 4. The results
are listed in (Table 3). The definition of repeat measurement standard deviation is based on
ASTM D 6300. There is no significant difference between the paired repeat standard deviation
and the standard deviation of the 40 total measurements. This indicates that there is no
observable measurement drift over 20 days. The standard deviation of the 40 measurements for
the 5-ppm and 10-ppm samples are very close to the Poisson standard deviation based on the Xray photons\counts. Conversely, at high concentrations, the Poisson fluctuation is small, and
other instrumental factors become important. Table 3 shows that at 300 ppm the difference
between the measured standard deviation and the Poisson standard deviation is significant. EPA
recently issues the precision criteria for selecting an ASTM testing method for measuring sulfur
in diesel fuel. With at least 20 tests over 20 days, the maximum standard deviations for the
selected method are 0.7 ppm and 7 ppm for a 5-15-ppm sulfur diesel fuel and 200-500-ppm
sulfur diesel fuel, respectively. The precision obtained above for both low-level and high-level
sulfur is well within the EPA precision criteria.

Refinery Site Results


Data were also collected from a different bench top sulfur analyzer at a refinery laboratory
over a period of one month. Again, all the measurements were carded out with one calibration
curve. For the repeatability evaluation, two diesel products and a gasoline product were used.
The gasoline and one of the diesel samples contain ~ 50 ppm sulfur. The other diesel sample has
10 ppm sulfur. For each sample, two successive readings were obtained within 10 rain. from
two different specimens by the sample operator. A total of 26, 36, and 32 pairs of results were
obtained over a month for the 50-ppm diesel, 10-ppm diesel and the gasoline samples
respectively. The difference between each pair was calculated and the results are shown as
histograms for the three samples (see Fig. 5). The repeat measurement standard deviation and
the site repeatability were found and the results are listed in Table 4.

CHEN ET AL. ON MONOCHROMATIC WDXRF

121

310
e9

9 ~

306
# 9

~149

t
302

:=
298

Ca
0

11.5

10.5

.*t

,t*,.= **e*tt*t

***~

9.5

"l

4,

9120

9/25

9/30

10/5

10/10

10115

Date

FIG. 3--Twenty pairs of repeats over 20 days for diesel samples with sulfur level at 5 ppm,

l Oppm, and 300 ppm.

FIG. 4---The Histograms for the difference (A) between each pair at 5 ppm, l Oppm, and 300

ppm sulfur level.


TABLE 3--Repeatability of a MWDXP,F bench top analyzer for single site.
Diesel C, ppm cr of repeats, ppm Repeatability, ppm a of whole set, ppm Poisson c, ppm
5
0.24
0.71
0.22
0.23
10
0.25
0.74
0.34
0.30
300
1.5
5.3
2.9
1.5

122

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 5--Histograms for the difference between the two values of each pair.
TABLE 4--The repeat measurement ,standard deviation and the repeatability at a refinery lab.
Sulfur level, ppm RepeattL ppm Repeatability,r, ppm
50, diesel
0.88
2.7
50, gasoline
0.87
2.6
10, diesel
0.46
1.3
The precision o f the analyzer was also evaluated. Two gasoline samples with ~50 ppm and
~10 ppm sulfur and one diesel sample with ~10 ppm sulfur were used for the study. For each
measurement, a specimen was obtained from the sample and placed into the analyzer. Each
measurement was taken at least 7-h apart. A total of 26-28 measurements were carried out for
each sample over the period o f one month. The results for the three samples are shown as
histogram plots in Fig. 6. The site precision was determined and the results are listed in Table 5.
Again there is no significant difference between the repeatability and the site precision,
indicating no observable drift for the instrument. The measurement results were also in good
agreement with the XOS data.

FIG. 6--Histogram plot for extensive repeat measurements on 10- and 50-ppm sulfur
gasoline, and 1O-ppm sulfur diesel.

123

CHEN ET AL. ON MONOCHROMATIC WDXRF


TABLE
Sulfur level, ppm
50, gasoline
10, gasoline
10, diesel

5--Siteprecision for low-sulfur gasoline and diesel.


Mean, ppm
45.8
7.0
8.4

......

o, pprn
0.96
0.24
0.40

Precision, ppm
2.5
0.6
1.0

Repeatability and Reproducibility for Ultra-low-sulfur Diesel Cross Check Samples


An in-house round robin study was conducted using the ultra-low-sulfur diesel crosscheck
samples from the A S T M D2 study group. Five bench top instruments were used with a different
operator for each instntment. Calibration standards from different commercial sources were used
for the calibration o f these instruments. There were a total o f 8 diesel samples from the
crosscheck program. NIST SRM2723a was added to the 8-sample group so a total o f 9 samples
were used in the study. Two successive repeat measurements were made for each sample on each
analyzer. The measurement results are listed in Table 6.
TABLE &-Measured sulfur concentrations (ppm) of the ultra-low-sulfur diesel cross check
samples and SRM 2 723a with five analyzers.
Analyzer
A
B
C
D
E

ULSD04
03
1.26
1.33
1.70
1.55
1.24
1.40
2.00
1.95
1.40
1.70

ULSD04
04
14.50
13.92
14.33
13.66
13.77
13.33
13.85
14.58
14.60
15.50

ULSD04
05
9,94
9,74
10.94
11.13
10.07
10.55
10.80
10.85
9.60
10.30

Sample
ULSD04 ULSD04
06
07
0.39
3.26
0,47
4.11
0.63
4.11
0.68
4.14
0.11
4.09
0.23
4.55
0.80
3.38
0.63
3.68
1.00
3.90
1.10
4.10

ULSD04
08
14,30
14.42
15.16
15.54
16.30
16.18
16.03
16.88
14.00
13.40

ULSD04
09
0.31
0.17
0.69
0.48
0.43
0.23
0.65
0.65
0.60
0.60

ULSD04
010
15.51
15.77
15.42
15.32
15.28
15.52
15.25
13.93
15.80
16.30

SRM
2723a
10.00
9,80
10.50
10.40
10.70
10.40
10.80
9.80
10.10
10.90

Following the guidelines o f A S T M D 6300, repeatability (r) and reproducibility (R) were
determined from the data shown in Table 6. The results are given in Table 7 along with the mean
o f each sample and the standard deviations o f r and R. The repeatability results show significant
improvement over that o f the A S T M D 7039-04. The reproducibility results are similar to that o f
inter-laboratory study (ILS) in 2002 [3]. Further interlaboratory round robin studies will be
needed to confirm the reproducibility.
TABLE 7--Standard deviations, repeatability, and reproducibility for ultra-low-sulfur
diesel fuel, in ppm.
Sample
ID
Mean
at
OR
r
R

U L S D 0 ULSD0
403
404
1.55
14.20
0.12
0.48
0.29
0.64
0.44
1.75
1.05
2.33

ULSD0
405
10.39
0.28
0.56
1.03
2.05

ULSD0 ULSD0
406
407
0.60
3.93
0.08
0.33
0.33
0.40
0.29
1.19
1.21
1.45

ULSD0
408
15.22
0.35
1.21
1.29
4.40

U LSD 0 U LS D 0 SRM
409
4010
2723a
0.48
15.41
10.34
0.10
0.46
0.42
0.20
0.63
0.40
0.38
1.68
1.53
0.72
2.28
1.45

124

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Accuracy and Bias


The accuracy and bias of the analyzer was evaluated using three NIST SRMs and a
commercial gravimetric standard. One specimen was taken from each sample and 10 readings
were recorded. The mean of the 10 data points was obtained and was compared with the
reference value. The comparison is shown in Table 8. Biases for these reference samples were
found to be non-significant except for the gasoline sample.
TABLE 8 - - Accuracy and bias: comparison of the measured concentrations and the
referenced values.
Sample
Reference Value, ppm Measured Value, ppm Bias,ppm Biassignificance
SRM 1616a,kerosene
146.2
149 4-2.50
2.8
n
SRM 2723a, diesel
11.0 4- 1.1
10.8 4-0.6
-0.1
n
SRM 2299, gasoline
13.6 4-1.5
11.6 4-0.6
-2.0
y
Commercialstandard, diesel
300 4- 3.0
303.2 4- 3.0
3.2
n

Matrix Effects
The absorption o f both the excitation beam and the characteristic sulfur X-rays o f the sample
is matrix dependent. In general, the matrix effect is minor and there is no need for correction.
However, if the sample measured has a high oxygen content or very different hydrogen-tocarbon H/C ratio compared to the calibrant, a correction will be necessary. Due to the
monochromatic excitation, the matrix correction is simple if the major composition o f the matrix
is known. The primary fluorescence intensity for an element in the matrix can be simply found
from the first principle for monochromatic excitation [4]:

Ii(Ei) = ~(Ei)

Qi(Ei)I(E~

(1)

Wi

Co (/4Eo) + c/4Ei))
where rl(E) is detector efficiency, Q(Ei) is the fundamental parameter for element I and I(E) is
the excitation beam intensity, Wi is the concentration o f the element by weight, Ix(E) is the
average mass absorption coefficient at energy E for the matrix, and Go and G are geometrical
factors related to the beam and sample configuration. The term ~M = (~t(Eo)+Gl.t(~)) is the only
parameter related to the matrix.
The current sulfur analyzer has a Cr Kct excitation beam and the fluorescence element is
sulfur. The G factor for the beam configuration in the analyzer design is .--0.87. For diesel and
gasoline, the matrix consists o f C, H and O. Assuming their weight percentage as Wc, WH, and
Wo respectively, the matrix effect term can be written based on the mass absorption coefficients
for C, H, and O at Eo = 5.4keV and E4 = 2.3 keV [5] and is the following:
/.t M = 1 9 8 W c + 0.35W,~ + 4 4 0 W o

182Rc.,,0 =

Wo)

(Rc. u + 1)

o.350-

wo) +

440W o

(2)

(Rc, n + 1)

where Rc~ is the C/H mass ratio in the matrix. The matrix correction factor, CM, then can be
defined as CM = ~M (specimen)/~m (calibrant), and the correct sulfur X-ray counts of the
specimen is Ic=Cu*Ir, where Ir is the recorded sulfur net counts from the analyzer.

CHEN ET AL. ON MONOCHROMATIC WDXRF

125

To validate Eq 2, measurements were carded out for eight gravimetric diesel standards with
known C/H ratios to determine the C factor experimentally. The experimental C factor can then
be compared with the C factor calculated from Eq 2. Relatively high sulfur concentration
samples were used since the relative precision o f the method is better for high concentration. The
study results are listed in (Table 9). Measurements on gasoline samples with added ethanol were
also performed to validate the effect o f oxygen. The results are also given in Table 9. Please keep
in mind, decalin is the matrix for the calibration material and the correction factor for decalin
samples is 1.

Sample
Diesel,
Kerosene, SRM1616a

TABLE 9--Effects of C, H, and 0


C/H ratio,
Wo
Sulfur Concentration
RC,H
W, ppm
6.18
0
300

composition.
Expected
Counts, I~
41312

Recorded
Counts, Ir
41720

CM= IJ
Ir
0.991

5.51

146

20229

20708

0.977

5.42

951.5

130502

135731

0.961

6.03

950.55

130372

133464

0.977

Decalin
Mixture of Decalin
and Tetralin

6.67

946.1

129763

129694

1.000

8.09

948.5

130091

128036

1.016

Tetralin

9.93

951.5

130502

125231

1.042

Gasoline
Gasoline + 3.31%
ethanol
Gasoline + 6.71%
ethanol
Gasoline + 8.30 %
ethanol

6.56

205

28312

28032

1.005

6.68

0.0115

195.4

27384

26768

1.018

6.81

0.0233

188.4

26294

25331

1.033

6.85

0.0288

185.2

25966

24706

1.046

Decane
Mixture of Decane
and Decalin

The experimental values o f CM are plot vs. RC,Hat different oxygen levels and compared with
the theoretical value from Eq 2 (See Fig. 7). The experimental results follow the theoretical
curves quite well. Therefore the theoretical curve can be used as a general guideline for practical
use. The C/H mass ratio is in the range o f 5.3-6.7 for diesel products and 5.7-7.3 for gasoline.
I f the synthetic diesel calibration set (commercial standards with C/H mass ratio o f 6.18) is used,
the biases due to the matrix effect will be no more than 2 % in the extreme case. The biases will
be insignificant for low-sulfitr diesel and gasoline, and there is no need for matrix correction.
For gasoline with oxygenates, the matrix correction will also be insignificant, provided the
oxygen content is less than 2 %, which corresponds to no more than ~ 1 1 % o f MTBE or ~6 % o f
ethanol.

Limits of Detection
The minimum limits o f detection o f an analyzer can be determined b y measuring the standard
deviation o f the background. Decalin was used for background measurement. Twenty repeat
measurements were carried out on this sample and the measured X-ray intensities are given in

126

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Fig. 8. Each measurement was done with a 300s counting time. The average X-ray counts in
300s for the background is B = 241 counts with standard deviation o f era = 16.9 counts. From
the calibration curve in Fig. 2, the background standard deviation can be written in terms o f p p m :
era = 16.9/A = 0.12 ppm. Therefore the minimum detection limit is given as:
L O D = 3or B = 0 . 3 6 p p m

Z.9%

1.10.

Tg

1.1[~ 9

1,00/

0.95.

J J

~I~

,
oe*e.os, oo eo
Mixture of decane, deealin and tetrstin

J.

9 Gasoline+ ethanol,1.2%Oxygen

o
13

9 Gasoline,no ethanol
O Diesel
Kerosene

0.90

C/H Mum

Ratio

FIG. 7--Matrix effects: measured correction factor and the theoretical correction curves f o r
C/H mass ratio and total oxygen wt. %
300-

Mean = 241, u.=16.9

9
9

~
U

240.

W
am

200too

10

111

20

observalion

FIG. 8 - - M e a s u r e d background X-ray counts.

CHEN ET AL. ON MONOCHROMATIC WDXRF

127

Conclusion

Compact bench top analyzers have been successfully produced for low-sulfur determination
in fuel. A new ASTM method, D 7039-04, was the basis for the study in this paper. The bench
top analyzer provides a wide dynamic range for sulfur measurements. The linearity of the
technique is excellent and it holds from 0-3000 ppm. During the calibration, linear regression
with a weighted error bar is necessary for obtaining good accuracy for low ppm sulfur
measurements. The repeatability for 10-ppm sulfur in fuel was shown to be around 1 ppm. The
PLOQ for ultra-low-sulfur diesel was found to be less than 1.5 ppm in this study. The precision
of an analyzer over one month without further calibration remains the same as that of repeat
measurements taken over very short intervals, such as minutes. The reproducibility of 15-ppm
sulfur diesel fuel was determined to be better than 3 ppm (95 % confident level). The limit of
detection for a single analyzer was found to be 0.36 ppm. The correction factor due to matrix
composition was measured and compared to the theoretical curve. The correction factor
predicted by the theoretical curved is accurate within the measurement error. The correction due
to the matrix for low-sulfur diesel and gasoline is generally not significant.

Acknowledgments
We would like to acknowledge the assistance of David Aloisi and Dr. Tomasin Miller with
measurements.
References

[1] Chen, Z. and Gibson, W. M., "Doubly Curved Crystal (DCC) X-ray Optics and
Applications," Powder Diffraction, Vol. 17, No. 2, 2002, p. 99.
[2] Chen, Z. W. and Wittry D. B., "Microprobe X-ray Fluorescence with the Use of Pointfocusing Difffactors," Applied Physics Letters, Vol. 71, No. 13, 1997, p. 1884.
[3] Lubeck, A. J., ASTM Research Report RR:D02-1552
[4] Jenkins, R., Gould, R.W. and Gedcke, D., Quantitative X-ray Spectrometry, Marcel Dekk~r,
New York, 1995.
[5] Goldstein, J. I., Newbury, D. E., Echlin, P., Joy, D. C., Roming, A. D., Jr. Lyman, C. E.,
Fiori, C., and Lifshin, E., Scanning Electron Microscopy and X-ray Microanalysis, Plenum
Press, New York, 1992.

Journal of ASTM International, October 2005, Vol. 2, No. 9


Paper ID JAI12975
Available online at www.astm.org
Dirk Wissmann I

Latest Improvements on Using Polarized X-Ray Excitation


EDXRF for the Analysis of Low Sulfur Content in Automotive
Fuel
ABSTRACT: Polarized X-Ray EDXRF has been used for the analysis of low levels of a broad range of
elements in light matrices for a long time. By using the polarization of the excitation radiation, the
background scattered by the sample is drasticallyreduced. For the analysis of low sulfur in automotive
fuels, this meant a detectionlimit comparableto the WDXRFtechnique.
Recent developments in detector technology and in closed coupled static geometry result in further
improvementof sensitivityfor this application.
This paper shows the developmentof this technique for the analysis of low sulfur content in petroleum
products. It shows the results of a roundrobin test, performedby CEN TC 19 WG 27 in 2001.
The technicalbasis of these data is compared to the recent technicaldevelopments,and future options are
shown. The influence by sample preparation and sample presentation, which might be special for this
analyticaltechnique,is investigatedand discussed.
KEYWORDS: EDXRF,polarization,sulfur, automotivefuels
Introduction
XRF basically operates always in the way that the excitation radiation is transferred into the
sample, where fluorescence radiation is produced. An energy dispersive detection system allows
the separation of this fluorescence radiation by energy and the counting of the intensity collected
by energy.
If an X-ray tube is used as a source for the excitation, the radiation can be modified by using
collimators, filters, secondary targets, crystals, and focusing optics.
The applied detection systems can be characterized by the energy resolution, the maximum
throughput, the peak-to-background ratio, and other parameters.
Excitation Geometry
When the radiation of the X-ray tube is directly transferred to the sample (maybe optimized
by collimators and filters), this is called direct excitation. Figure 1 shows such a setup.
If a direct excitation geometry is used for the analysis of automotive fuels, a Ti X-ray tube
can be used. Typically those tubes use a power o f maximum 30 W. Figure 2 shows an example
spectrum for a sample with a sulfur content of 100 mg/kg. The peak-to-background ratio
achieved in the example from Fig. 2 is 2.5:1.
Manuscriptreceived 8 October 2004; accepted for publication 11 April 2005; publishedOctober2005. Presentedat
ASTM Symposiumon ElementalAnalysisof Fuels and Lubricants:Recent Advancesand FutureProspects on 6-8
December2004 in Tampa,FL.
Spectro AnalyticalInstruments,160 AuthorityDr., Fitchburg,MA 01420.
Copyright9 2005byASTMInternational,I00BarrHarborDrive,POBoxC700,WestConshohocken,PA19428~
128
Copyright*2005 by ASTM International

WISSMANN ON POLARIZED X-RAY EXCITATION EDXRF

129

FIG. 1--Direct excitation geometry.

FIG. 2--Spectrum of a sample with a sulfur content of l O0 mg/kg using direct excitation; the
Ti escape line appears at the displayed energy in combination with a Si based detector.
Polarization can be implemented in this geometry by using a crystal, which reflects and
polarizes the excitation radiation. The reflection by a crystal creates a monochromatic excitation
radiation with an energy that can be close to the absorption edge of the dement of interest. With
this reflection the radiation is also polarized, and by using a Cartesian geometry the background,
which is created by scattered radiation, can be reduced [3]. Figure 3 shows the excitation setup.
If the X-ray tube power is increased to 400 W and a Pd target tube is used in the geometry
shown in Fig. 3 together with a HOPG (highly oriented pyrolithic graphite) crystal, Pd L-lines
can be used for the excitation. This will lead to an improvement in the sensitivity, which can be
seen in Fig. 4, where a spectrum, measured with a sample with sulfur content of 100 mg/kg, was
analyzed. The peak-to-background ratio is improved to 3.7:1; the intensity of the S fluorescence
line is increased by a factor o f - 3 in comparison to the example of direct excitation from above.

130

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 3--Polarization geometry.

FIG. 4---Spectrum, taken in polarization geometry (blue), in comparison to direct excitation


(red).
One of the limitations of such a setup certainly is the use of a higher power in the X-ray tube
in comparison to a direct excitation setup. This is caused by losses due to longer paths in the
excitation as well as by losses in the reflection using the crystal.
Using a fine focus X-Ray tube helps compensate these effects. The excitation radiation of
such a tube can be transmitted optimally to the sample by X-ray lenses or X-ray mirrors. As a
mirror for X-rays, again a crystal can be used, which in this case is concave-shaped. The
reflected radiation is polarized [2,4,5]. Figure 5 shows this in a linear plane schematic.
By reflecting and focusing the X-ray beam, a higher flux of excitation radiation is transmitted
to the sample. This increases the sensitivity, and lower power X-ray (less than 50 W) tubes can
be utilized. A spectrum with a sulfur in oil sample at a sulfur concentration of 100 mg/kg is
displayed in Fig. 6. The peak-to-background ratio is improved to 5.5:1, and the intensity of the
sulfur fluorescence line is increased by a factor of -2 in comparison to the example of
polarization excitation and by a factor o f - 6 in comparison to the example of direct excitation
from above.
Additional improvements in this geometry in the future can be seen in using differently
shaped mirrors and shorter coupling of tube, sample, and detector.

WISSMANN ON POLARIZED X-RAY EXCITATION EDXRF

131

FIG. 5--Polarization geometry using curved crystal with fine focus X-ray tube.

FIG. 6--Spectra comparison focusing polarization geometry (black), polarization geometry


(blue), direct excitation (red).
Detection

The detection of the fluorescence X-rays should be done by using a detection system with a
spectral resolution enabling the easy separation between the excitation radiation (in this case
energy of Pd-L-lines) and the fluorescence radiation of sulfur.
Detection systems such as Si(Li) detectors and silicon drift detectors were found to be
suitable for this application. They typically work in a range of 150-175 eV (at Mn K-alpha
radiation). With this spectral resolution it is even possible to analyze low contents of chlorine as
well using the same excitation. Figure 7 shows spectra of a blank sample (displayed in red), a
second sample containing sulfur 100 mg/kg (displayed in blue), and a third sample containing
100 mg/kg chlorine (displayed in black). The fluorescence lines can be clearly separated, and
even a concentration of 100 mg/kg has no significant effect on the background of the sulfur
fluorescence line.
The major differences in the choice of a detection system for this application between Si(Li)
and Silicon drift detector is the Peak-to-Background (P/B) ratio, which typically is better using a
Si(Li) detection system. Future developments in the SDD technology show that P/B ratios are
improving.

132

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 7--Spectra of a blank sample (red), a second sample containing 100 mg/kg sulfur
(blue), and a third sample containing 100 mg/kg chlorine (black).

Sample Preparation and Influence of Film


Liquid samples in all cases are analyzed by preparing a sample cell for liquid samples. This
sample cell is closed at the analytical end by a film, which should be as transparent as possible
for the excitation and fluorescence radiation and should be robust, not destroyed by the sample
itself. Various types of film are available with various thicknesses. Also, impurities, which might
be inside the film, must be considered, if they are at higher level or in varying concentrations.
Table 1 shows the transmission of sulfur fluorescence radiation relative to the most
transparent film for sulfur.
TABLE 1--Transparency for Sulfur K radiation and impurities for different types offilm.
X-Ray Film
Thickness Transparency
Impurities
Mylar
1.5 ~tm
100 %
Si, (P)
Mylar
2.5 ~tm
94 %
P, Ca
Mylar
3.5 ~tm
87 %
P, Ca
Mylar
6.0 ~tm
75 %
P, Ca
Polypropylene
4.0 ~tm
96 %
...
Polypropylene
6.0 ~tm
93 %
...
Polycarbonate
4.0 ~tm
86 %
...
Polycarbonate
5.0 ~tm
81%
...
Hostaphan
3.5 ~tm
83 %
Si, (P)
In all cases of XRF analysis, a part of the excitation radiation will be scattered by the sample,
including the X-ray film used. This effect certainly is smaller when using polarized excitation
radiation but still can be seen because the degree of polarization typically is only 90 %. Figure 8
shows the Cartesian geometry with polarized excitation radiation.

WISSMANN ON POLARIZED X-RAY EXCITATION EDXRF

133

The influence by the scattering from the X-ray film depends on its material and thickness.
The scattering of the excitation radiation will influence the intensity of the Pd-L-lines in the
measured spectrum if a Pd- X-ray tube is used for the excitation. This again will have an impact
on the background in the spectral region of the S fluorescence radiation, and by this it influences
the achievable detection limit. Table 2 shows the background intensity in that spectral region for
various X-ray films relative to the highest intensity measured using a blank sample.

FIG. 8--Principle setup of a polarization excitation system in Cartesian geometry.

TABLE 2--Background intensity at Sulfur K radiation for different types offilm, determined
from the measurement of a blank sample.
X-Ray Film
Hostaphan
Mylar
Mylar
Mylar
Polycarbonate
Polycarbonate
..... Polypropylene

..... Ttlickness
3.5 xm
3.5 lm
2.5 lm
1.5 xm
5.0 lm
4.0 lm
4.0 lm

Backgroun~l ,
100 %
91%
83 %
64 %
64 %
39%
40 %

In addition, the stability of the X-ray film is very important for a safe operation of the
analyzer. For the analysis of sulfur in automotive fuel, two films are used, one for closing the
sample cup itself and the second as a safety window in the holder for the sample cup. The use of
the safety window with an additional film has become a standard technique, as this gives
additional safety in the case of a leaking sample cup. Figure 9 shows the sample cell with the Xray film together with the sample cup and its safety window.
The most common and most appropriate technique is to use Polycarbonate film for the
sample cup and for the safety window.

134

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 9--Image of sample cup with X-ray film and sample cup with safety window.

Analysis Results
In 2001 CEN TC19 WG27 [1] performed a round robin test for the analysis of low sulfur
levels in fuel, where 102 laboratories participated in the test. Out of these, five laboratories have
used EDXRF instruments (from one manufacturer) using polarized X-rays for excitation. Tables
3 and 4 show a comparison of the achieved analysis results for diesel and petrol samples in
comparison to those results achieved with WDXRF instruments with and without internal
standard.
In the evaluation of the round robin test the reproducibility r and the repeatability R were
calculated for various analytical techniques. WDXRF was differentiated between low power
(LP) and high power (HP) instruments with both methods with and without using internal
standard. Tables 5 and 6 show the comparison for diesel and petrol samples.
TABLE 3--Analytical results of diesel samples using polarization EDXRF versus WDXR
(high power (> 3 kW) systems only).
Ident.
QCG
OG
PG
RG
SG
TG
VG
WG
ZG

WD-XRF
No IS
31.48
4.47
52.57
21.39
10.33
40.96
36.48
8.11
8.44

WD-XRF
IS
32.78
4.79
54.13
22.15
10.74
42.21
37.51
8.46
8.80

EDXRF
Polarized
32.44
5.41
53.75
22.49
11.15
42.48
37.73
9.16
8.73

WlSSMANN ON POLARIZED X-RAY EXCITATION EDXRF

135

TABLE 4---Analytical results of petrol samples using polarization EDXRF versus }VDXRF
(high power (>3 kW) systems only).
Ident.
QCP
AP
BP
CP
DP
EP
FP
GP

WD-XRF
No IS
20.65
15.34
38.49
1.23
22.30
6.44
56.32
44.16

WD-XRF
IS
21.22
16.27
40.17
1.50
23.87
6.94
58.47
45.82

EDXRF
Polarized
21.11
15.37
40.00
0,98
23.11
6.81
59.18
45.81

TABLE 5--Reproducibility r and the repeatability R for diesel samples at 10 and 30 mg/kg
sulfur using different analytical techniques.

WD-XRF IS HP
WD-XRF IS LP
ED-XRF non polarized
E D - X R polarized
WD-XRF HP
WD-XRF LP
UV Fluorescence
Microcoulometry

r
10
1.85
2.46
7.38
1.59
1.92
2.22
1.15
1.69

r
30
2.59
3.28
7.95
1.59
2.16
2.71
2.21
2.36

R
10
3.33
3.97
10.4
2.14
2.99
5.19
2.24
3.29

R
30
4.67
5.29
11.2
2.14
4.97
6.34
4,48
4.60

TABLE 6---Reproducibility r and the repeatability R for petrol samples at 10 and 30 mg&g
sulfur using different analytical techniques.
r
r
R
R
WD-XRF IS HP
WD-XRF IS LP
E D - X R non polarized
ED-XRF polarized
WD-XRF HP
WD-XRF LP
UV Fluorescence
Microcoulometry

10
2.12
3.77
6.29
2.00
1.92
3.24
0.98
1.32

30
3.08
3.77
7.87
3.00
2.57
4.32
2.24
2.63

10
3.54
5.96
10.05
2.28
3.45
5.93
2.71
3.58

30
4.95
5.96
12.56
3.43
5.74
7.90
6.21
7.17

In one laboratory the long-term stability over a period of two weeks was tested with a diesel
sample containing sulfur close to 10 mg/kg. The measurements were performed in duplicates on
nine different days. An average concentration of 9.63 mg/kg was analyzed. The standard
deviation is calculated as 0.55 mg/kg. The results are displayed in Fig. 10.

136

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. l O---Long-term repeatabilityfor a diesel sample with a sulfur content close to 10 mg/kg.
Summary and Outlook
In the last few years EDXRF using polarized and focused excitation has become a very
powerful analytical technique for the analysis of low S content in automotive fuel. The technique
shows the potential to achieve repeatability and reproducibility values, which are comparable to
reference methods for this application, In preparation for a new ASTM and EI norm, a new
round robin with a larger number of participants is under way.
The future for this technique will involve improved crystal optics for improved sensitivity
and new detection systems with improved peak-to-background ratios helping to achieve even
lower detection limits.
References
[1] CEN TC 19 WG 27 2002, Test Methods for the Determination of Sulfur Content, Round
Robin report 2001.
[2] Grigorieva, I. G. and Antonov, A. A., "HOPG as Powerful X-Ray Optics," X-Ray
Spectrometry, Vol. 32, No. 1, January/February 2003, pp. 64-68.
[3] Heckel J., "Bragg and Barkla Polarization in EDXRF," EDXRS Conference, Lisboa, 1996.
[4] Heekel J., "Bragg Polarization Optics in EDXRF," 49 th Annual Denver X-ray Conference,
2000.
[5] Ryon, R.W. and Heckel, J., "Polarized Beam X-ray Fluorescence Analysis," Handbook of Xray Spectrometry, 2nd ed., R. E. Van Grieken and A. A. Markowicz, Eds., Marcel Dekker
Inc., New York, 2002, pp. 603-630.

Journal of ASTM International, October 2005, Vol. 2, No. 9


Paper ID JA112980
Available online at www.astm.org

Steve Tarkanic I and John Cl'nko 1

Rapid Determination of Sulfur in Liquid Hydrocarbons for AtLine Process Applications Using Combustion/Oxidation and
UV-Fluorescence Detection
ABSTRACT: With the reductionin sulfurlevelsin gasolineand dieselproducts,itisnecessaryto have a
method availablewhich can preciselyand accuratelymeasure the very low levelsof sulfurin thesefuels.
A S T M testmethod D 5453, Sulfurby Combustion and UltravioletFluorescence(UVF), is a laboratorybased technologythathas been widely used in the industryfor such purposes for more than a decade.An
on-line/at-processversionof thistechniqueis describedwhich can complete a fastanalysisin lessthan
one minute.Thus, near real-timedata can be obtainedwhen refineriesand terminalsneed to make rapid
determinationsof sulfurlevelsfor process controland pipelinedistributiondecisionsinvolvingthe new
generation"clean"hydrocarbon fuels.Particularlydiscussedisthe advantage gained for decisionmaking
involvingdetectionof transportboundariesand relatedcustodytransferissues.

KEYWORDS: rapid sulfur analysis,gasoline analysis,diesel analysis,ultra low sulfur in diesel,


combustion/oxidationUV-fluoresccnccdetection,faston-line/atprocess sulfurdetermination

Contemporary Challenges for Sulfur Determination


Sulfur levels in gasoline and diesel fuels have been decreasing steadily over the last several
years, primarily in response to government regulations both in the U.S. and in Western Europe.
For example, U.S. Environmental Protection Agency's (EPA) Tier 2 emission standards mandate
sulfur level reduction from current 330 and 550 mg/kg sulfur in gasoline and diesel, respectively,
to 30 and 15 mg/kg by 2006 and near zero by 2010 and later years [I-3]. Regulations in
California are even more restrictive. Similarly, E N 590 European diesel specification mandates
sulfur content of diesels from 350 mg/kg m a x i m u m in 2000 to 10 mg/kg m a x i m u m by January l,
2005 in European Union countries.
These mandated sulfur reductions, coupled with other regulatory driven specification
changes, such as fuel volatility and oxygen content, have spurred large capital and operational
cost increases for a wide range of refinery and downstream (distribution)business models. In
fact, a major time compression in the entire Well-to-Wheel dynamic is evolving as the oil
industry invests in projects ranging from installation of sulfur reduction/removal processes to
significant "remodeling" of final product blending, storage, and transportation schemes. Strong
product demand, industry capacity limitations, crude oil price fluctuations, and the need to
realize a timely return on investment has oil industry administrative, operational, and engineering
personnel realizing the need for all types of instantaneous or near real-time data. The benefits of
very rapid and accurate product quality data can now, more than ever, bring value to the entire
petroleum and petrochemical industry.

Manuscriptreceived2 December 2004; acceptedforpublication21 March 2005; publishedOctober2005.


PresentedatA S T M Symposium on ElementalAnalysisof Fuels and Lubricants:Recent Advances and Future
Prospectson 6-8 December 2004 in Tampa, FL.
iAntek Instntments/PAC,300 Bammel WestfieldRoad, Houston,TX 77090.
Copyright9

2005 by ASTM International, 100 BaITHarbor Drive, PO Box C700, West Conshohoekcn, PA 19428-2959.
137

138

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

As the new "clean" fuels enter the marketplace, both refinery process control and
maintenance of product quality during distribution are crucial. Availability of near real time
sulfur data, particularly in downstream applications, can have significant value when time critical
decisions must be made for a custody transfer and for detecting transport boundaries of similar
fuel streams flowing through a pipeline system.
One of the parameters historically used for determining product transition in a pipeline is API
gravity. However, there is not necessarily a correlation between the gravity of a fuel and its
sulfur content. When an ultra-low sulfur fuel and "other" similar fuel streams with higher sulfur
levels are being transported through the same pipeline, in order to minimize contamination and
loss, it is critical to be able to identify a change of sulfur levels in the streams with as little delay
as possible. Ideally the change in sulfur concentration is reported at the boundary between the
materials. This interface between different product streams is called trans-mix region, and
pipeline companies strive to keep this region to a minimum. Depending on factors such as
pipeline flow rate and speed of analysis, the amount of cross-contamination and waste which can
occur as the two streams are directed to separate destinations can vary. Often, in an effort to
minimize the chance for contamination occurrence, larger volumes of material near the interface
region of the two streams are diverted or cut to special trans-mix holding vessels.
Test methods that might be used for sulfur determination applications in fuels include
techniques such as lead acetate paper tape, oxidative combustion followed by gas
chromatographic separation for flame photometric detection [4], and energy dispersive X-ray
fluorescence with coaxial proportional counter detectors [5]. The first of these two methods was
recently issued as a new standard D 7041 by ASTM. However, these and other commercially
available analyzers, such conventional on-line/at process sulfur by combustion and UVF
analyzers typically requirr analytical cycle times of 4-10 rain. This delay in reporting the sulfur
concentration levels limits the real time detection capability of changes in the sulfur
concentration of fast moving or rapidly changing transport or process streams.
A frequent, very fast, and accurate determination of sulfur contamination contained in the
hydrocarbon materials being transported in a pipeline or involved in a controlled blending
situation can facilitate more precise material management and significant reductions in waste.
This paper will describe such a rapid technique for on-line/at process determination of ultra-low
levels of sulfur in hydrocarbon engine fuel streams based on the D 5453 oxidative combustion
UV-fluorescence technique. The described technology can continuously determine and report
ultra low levels of sulfur at a rate of once per minute.

Analytical Laboratory Techniques for Sulfur Determination


There are about 20 laboratory based ASTM standard test methods available for the
determination of sulfur in various petroleum products and lubricant samples [6]. These utilize
diverse analytical techniques and have applicability range spanning from m% to low mg/kg
levels. However, at the very low end of sulfur analysis there are only three or four test methods
which can adequately determine sulfur in such fuels. These lab-based standard test methods
include ASTM D 2622 - wavelength dispersive X-ray fluorescence, D 3120 - oxidative
microcoulometry, D 5453 - combustion UV-fluorescence, and D 6920 - oxidative combustion
electrochemical detection methods. Without a doubt, the most widely used two methods out of
these in oil industry laboratories are D 2622 and D 5453. Studies have shown that at truly ultralow levels of sulfur only D 5453 can deliver accurate and precise results. This conclusion has

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

139

been corroborated in a Southwest Research Institute study [7] and an extensive cross check
undertaken by a CEC work group in Europe [8].
ASTM D 5453 Test Method

The D 5453 method "Standard Test Method for Determination of Total Sulfur in Light
Hydrocarbons, Motor Fuels and Motor Oils by Ultraviolet Fluorescence" has been widely used
in the industry laboratories for over 15 years with a great deal of success. It is one of the
preferred methods for such analysis approved by the California Air Resources Board (CARB)
and recognized by the U.S. EPA. In a major study conducted by ASTM, it produced the best
repeatability, reproducibility, and pooled limit of quantitation compared with D 2622, D 3120, or
D 6920 test methods [9].
In this standard test method, a sample is combusted in an oxygen-rich environment at about
1050~ All sample components are completely oxidized, with various sulfur species in the
sample being converted to sulfur dioxide (SO2). Hydrocarbons are converted to water vapor and
carbon dioxide. The water vapor produced during the combustion is removed with a membrane
dryer, and the gases are exposed to short wavelength ultraviolet light. The sulfur dioxide in the
gas absorbs the energy from the ultraviolet light and is converted to excited SO2. The
fluorescence emitted from the excited SO2 as it returns to a stable state is detected by a
photornultiplier tube. The resulting signal is a measure of the sulfur concentration in the original
sample.
R-S + 02 --" 1050~ ~ SO2 + Combustion Products
S02 + Xv -,. S0"2 ---- S02 + Xv'
The complete vigorous combustion (oxidation) of the entire sample and the elimination of
the resulting water byproduct virtually eliminates interferences. The method has been used for
the analysis of a wide variety of liquid petroleum products such as naphtha, distillates, oils,
ethanol, fatty acid methyl ester, gasoline, oxygen enriched gasoline, reformulated gasoline,
diesel, biodiesel, and jet fuel for sulfur levels as low as 250 gg/kg. A method based on the same
technique and used for the analysis of gaseous hydrocarbons and liquefied petroleum gases has
been designated ASTM D 6667.
The Lab to On-Line/At Process Environment Transition

On-line/at process instrumentation based on the aforementioned ASTM D 5453 technology


has been widely used in oil industry and petrochemical processes for over 10 years. Applications
have ranged from feedstock analysis and catalyst protection to quality control of finished fuels
and other chemical products. Common problems encountered whenever a technology is moved
from the laboratory into the process environment include development of effective sample
collection, treatment, and handling mechanisms. Once the sample is effectively isolated and
prepared, it must promptly and properly be placed into the analyzer. The apparatus needed to
execute the analytical technology must be appropriately packaged and made safe for operation in
a variety of environments. Overall instrument ruggedness and ease o f maintenance and
downtime frequency issues are often constraints. Data are typically displayed locally and
transmitted over large distances so they can be used for process product quality control purposes.
Figure 1 shows a very basic block diagram applicable for most on-line analyzers. Sample

140

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

collection and conditioning, sample injection, packaging of the technology apparatus, and data
handling are depicted.

Sample
Collectio~n
end
,
Conditioning

~[ Sample
-1 Introduclion

Combuslion ~
and Sulfur
Detection

Quantifcation
and
Reporting

FIG. 1--Basic block diagram of a typical on-line/at process analyzer.


On-Line/At Process Sulfur Determination by Combustion and UVF

As the laboratory based sulfur by combustion and UVF technology was placed in the at
process application environment, the associated design constraints and operational concerns of
any high temperature analysis technique were realized. Application specific solutions were
developed for concerns ranging from specialized flammable sample handling of multiple streams
to various explosion proof apparatus enclosure requirements. Figures 2 and 3 illustrate a typical
at process sulfur by combustion and UVF analyzer.

FIG.
2---Conventional
sulfur
by
combustion and UVF on-line/at process
instrument: Note the man-machine-interface
(computer screen), metal safety enclosure,
and purge system.

FIG. 3----An internal view of a typical


on-line/at process sulfur by combustion
and UVF analyzer: Note separation oJ
furnace section (top) and electronics and
detector enclosure (bottom).

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

141

Operations of Typical On-Line~AtProcess Sulfur by Combustion and UVF Analyzer


As with most on-line analyzers, a non-static (flowing) material of interest is supplied from
and returned to its source through a flow path loop designed for the sample conditioning system.
Velocity or flow rate through this commonly called "fast" loop is set so that material presented to
the sample conditioning panel is representative of the process being monitored and can be
correlated to the time of sample aliquot acquisition (injection) and the corresponding result.
Once the fast loop presents the material of interest, the sample conditioning panel filters a
portion of this fast loop flow to an acceptable particle size. Pneumatic inputs are then used to
select either the filtered process stream, a calibration standard, or a validation material. The
selected sample from this sample conditioning system is then flowed through a sample injection
valve. At the point of injection, a fixed volume of the sample is inserted into the cartier gas flow
and introduced into the analyzer through a small volume injection tube. Once the sample enters
the high temperature combustion zone (fumace), the sample is analyzed and sulfur content is
measured in the same manner described earlier for the laboratory based D 5453 test method.
Manual controls or another man-machine interface allows the user to operate the analyzer,
manipulate analysis parameters, and input the concentration values for calibration materials. The
detector output is integrated to produce a value that is proportional to the total amount of sulfur
that flowed through the detection chamber during the analysis. This proportional value is
compared to the value produced by a standard to determine its concentration. This value can be
corrected for the density of the standard or sample being analyzed. The resultant concentration
can then be locally displayed, written to a register for use on a serial bus, and/or used to set a
representative value on a 4-20 milliamp (mA) current loop.

Design Challenges for a High Speed Sulfur Analyzer for Hydrocarbons


Significant advances in sample introduction, combustion technique, and the handling of
combustion byproducts have been incorporated into the basic on-line Sulfur by UV Fluorescence
analyzer. The implementation of these new techniques has yielded a continuous rapid sulfur
measurement capability that is based upon the reliable and well-established D 5453 standard test
method. Figure 4 describes a typical high speed sulfur analyzer.
F'dbdon
and

Cultbn~n
Slmdnnl

VdidJon C m ~ n r
91andml Ou

UV Light

10Q0.1U60~C

Cmnbus~n

Gae

Density

High
Vdlqe

FIG. 4---Block diagram of typical high speed sulfur by combustion and UVF in an on-line~at
process application.

A Summary of Enabling Techniquesfor a High Speed Sulfur Determination


A central issue for enabling rapid response for any at process instrumentation is to ensure
that the analyzer is promptly supplied with a sample that corresponds to the process being
monitored. Two issues emerge: elimination of sample discrimination and lag time. Materials

142

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

obtained for analysis must represent a true random example of the substance of interest and must
be presented in a prompt reproducible timeframe that can be correlated to the system being
monitored and the frequency of result.
A properly designed sample conditioning system (SCS) is critical for reliable high speed
sulfur measurements. The main function of the SCS is to prepare a representative process
sample for introduction into the analyzer with as little delay as possible and to select what is sent
to the analyzer for analysis or calibration. A correctly designed SCS will remove particulates and
water, regulate the sample pressure, and control the sample flow to ensure sample introduction
under constant conditions.
A fast action injection valve constructed of materials compatible with the material being
analyzed is essential. This injection valve precisely inserts the sample material into a carrier gas
stream that transports the sample into the analyzer. A suitable injection valve will also be
compatible with system pressures.
All gas flow rates must be of sufficient volume to ensure that the sample and its combustion
products are passed through the system in a timeframe that allows the analyzer to promptly
return to baseline conditions prior to the next injection. Favorable gas flow rates will eliminate
or minimize sample valve "wetting," ensure sufficient sample combustion, provide sharp
Gaussian peak characteristics, and enhance detector response.
A specially designed combustion tube allows the sample to be injected at much higher carrier
flow rates than can be achieved with the standard combustion tube design. The sample inlet and
furnace mixing zone promote complete sample oxidation. A proper combustion tube is typically
constructed of quartz.
A controlled zone of heating is applied to the combustion product's flow path and to the
sulfur detector reaction chamber. Water vapor produced by the high temperature oxidation of
each hydrocarbon sample injection is maintained in the vaporous state throughout detection and
measurement of SO2 fluorescence. Appropriate heating prevents any water vapor condensation,
thereby eliminating SO2 solubility and any problems with water buildup in the analytical flow
path.
Materials and Methods

Most work discussed here regarding standard Antek on-line/at process analyzers, laboratory
based equipment, and the High Speed Sulfur (HSS) on-line analyzer was carried out in the PAC
Houston R&D Laboratory and Production facility. Other data were collected from various
industry and public sources and are so identified in the associated text and graphics. Various
solvents, sulfur sources, and standard reference materials were used in association with both
historical and new data. Solvents and chemicals, such as iso-octane, toluene, and dibutyl sulfide,
were of typically reagent grade purity and were obtained from readily available commercial
sources, such as Aldrich chemicals. When appropriate, these materials were analyzed before
use.

In all cases the various materials analyzed by both the standard and HSS on-line analyzers
used injection valves and typical gas flows and pressures found in end-use applications. In some
instances, sample flow rates to the injection valve were slowed in order to preserve limited
supplies of costly calibration materials or limited volumes of sample.
Validation criteria for the HSS instruments included meeting or surpassing the performance
established by the standard Antek on-line analyzers (6000/6200 series). Where possible,
comparisons were also made to D 5453 lab instrumentation results. Due to the large number of

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

143

data collected in some experiments (large number of individual injections), some results are
documented graphically.
The on-line/at process industry often describes precision and linearity as a percentage of the
full scale concentration. Full scale, for example, is the maximum concentration that is expected
to be analyzed or the concentration that is equivalent to 20 mA on the 4-20 mA current loop.
When used in this work, the term "% RSD of full scale" is the standard deviation for the results
of a sam_pie analysis compared to the full scale value of the range that the analyzer is set for.
Standard deviation as a percent of full scale indicates the maximum deviation that can be
expected on a sample of any concentration that is within the analysis range. Because % RSD is
dependent on the concentration being analyzed, it would have to be determined for each
concentration in the range. Percent of full scale is a convenient way to define the precision of an
analyzer that allows the maximum anticipated RSD to be inferred for any concentration in the
range. See Calculation 1:
Calculation 1 - (% Full Scale) X (Full Scale Value) / (A Specific Concentration) =

Maximum Expected % RSD


Results and Discussion

Minimizing or Eliminating Instrument Lag


In on-line/at process analysis, rapid indication of changes in sulfur concentration depends
upon proper "fast loop" / sample conditioning design and the ability of the analyzer to produce a
near immediate shift in detector response when a material with a different sulfur concentration is
first presented to the analyzer. A delay or lag in the shift to a new level of detector response
caused by residua from previous samples or by poor detector response can bias results. A good
benchmark for demonstrating that an analyzer system does not suffer from lag or memory effects
is to determine the system's capability to report very different sulfur concentrations within two
injections following a significant sulfur concentration range. Figure 5 shows a typical response
to sudden large sulfur concentration changes for the standard 6000/6200 on-line analyzers.
Rapid Response of Sulfur in Diesel When Switching from 1800 to 400 ppm
2000
GasOil 2 ~
1500 C~

E 1000"
500'

,wwwo-r
9

GasOil 1

"111 IIIIIIIIIIIIIIIIIIIIII

16

IIIIIIIIIIIIIIII1||

26

36

46

45

Injections - 1 minute Analysis

FIG. 5--Rapid response of sulfur in dieselfuel when switchingfrom 1800ppm to 400 ppm.

144

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Figure 5 is an example of the analyzer response for sulfur in gas oil/diesel samples using
standard Model 6000/6200 analyzers when switching from -1800 mg/kg to -400 mg/kg sulfur
levels. The circled data points indicate the first injection after the sample was switched from
high to low concentration. Injection rates were at five-minute intervals.
Successful execution of a continuous high speed sulfur determination at injection rates of
once per minute reduces the time necessary to recover from memory effect. The memory effect
when changing from a high concentration sample to a low concentration sample is shown in Fig.
5 above. Data exemplified in Fig. 6 demonstrate two important characteristics: the HSS analyzer
exhibits no significant memory effects; and the recovery from the minor memory effect of the
analyzer is dependent upon the number of injections made after a concentration change, not on
the amount of time between the injections.
i

Response Time for Sulfur in Gasoline at 110, 225 and 15 ppm

,,,

/11_
,, ......

,,,

10

,,

jl

,J,

,,

+.t

,,

20
30
40
Minutes (1 analysis each minute)
i

ill

ii ill ii ill

50
i

60
i

FIG. 6---Response time of high speed analyzer.


Figure 6 demonstrates analyzer response for sulfur in gasoline samples using the HSS
analyzer when switching from 225 mg/kg to 15 mg/kg sulfur levels. The arrow points to data
from the first injection after the sample was switched from high to low concentration. Injection
rates were at one-minute intervals.
Table 1 lists some of the HSS analyzer results illustrated in Fig. 6. These data were collected
just before, during, and after a change in sulfur contamination from the 225 mg/kg level to the
lower 15 mg/kg level. An evaluation of the data points at injection times 17-23 finds a mean
concentration of 221.9 mg/kg of S. Average (mean) sulfur values for data points at times 26--31
are found to be 14.3 mg/kg. In this case where the initial concentration was about 12 times
greater than the lower sulfur concentration, the second result after concentration change (Result #
25) was found to be accurate within the analyzer's precision limit for the lower concentration.
See Calculation 2.

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

Calculation 2 - V25 - M = D
M = Mean of samples 26-31 (mg/kg)
D = difference in M & V25 (mg/kg)
samples 25-30.

145

V25 - value of injection 25 (mg/kg)


D is less than the standard deviation of

Therefore, if an injection is made every minute, it would take only three minutes for the
analyzer to begin yielding results within precision for the lower concentration and only one
minute to indicate accurately a significant step change in sulfur concentration at over 95 %
change.
TABLE 1--Data summary of HSS response to changes in sulfur concentration.
Sample 1
mg/kg
Sample 2
mg/kg
Sample Sample
1
2
Injection #
Inj ecti0n #
17
224.2
23
22.2
Average
221.9
14.3
18
222.5
24
17.2
Standard
2.13
0.75
Deviation
25
14.7
19
221.4
%RSD
0.96
5.20
26
15.3
20
221.7
27
14.7
21
223.4
28
13.2
22
218.1
29
13.8
30
14.7
31
14.3

Precision of Results
Users also need to know that the analyzer responsible for monitoring a fast changing material
stream can shift to various concentrations and quickly generate data with good precision. Figure
6 also shows the excellent repeatability e r a 110 mg/kg gasoline sample repeatedly injected about
30 times in a high speed sulfur analyzer. It resulted in a standard deviation o f 1.26 mg/kg or a
relative standard deviation (RSD) of 1.14 %. Then a sample o f - 2 2 5 mg/kg sulfur was injected
10 times, resulting in a standard deviation o f 1.91 with a RSD of 0.86 %. This was followed
by about 30 injections of a -15 mg/kg sulfur sample, resulting in a standard deviation o f 0.79
mg/kg or a RSD of 5.7 %. The ~225 mg/kg sulfur material was then analyzed for about five
cycles.
High speed sulfur analyzer data can now be compared to historical online analyzer
performance. Figure 7 shows repeatability of a 50.44 mg/kg gasoline sample repeatedly injected
about 70 times in the Standard 6000 process analyzer. It resulted in a standard deviation of
mg/kg or a relative standard deviation (RSD) of 0.93 %. Then a sample of 29.9 mg/kg sulfur was
injected about 20 times, resulting in a standard deviation of~- 0.23 with a RSD of 0.75 %. This
was followed by about 10 injections of a 4.95 mg/kg sulfur sample, resulting in a standard
deviation of 4- 0.14 mg/kg or a RSD of 2.87 %. See Table 2. At this stage, the sample was spiked
with ~103 mg/kg of sulfur for four cycles, and then returned to the lower range of 50 mg/kg
sulfur. Through all these cycles in different ranges of sulfur concentrations both the HSS and
standard combustion UVF analyzers maintained excellent response to drastic changes in sulfur
concentration.

146

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Repeatibility of Sulfur in Gasoline at 50, 30 and 5 ppm


110 .
9O

m
m

--- 7 0

50

="=

_.,__A .....

=. . . . .

,-.__

.==

-..~

. . . . . . . . . .

30 10""
0
0

20

40

60
Injections

80

100

120

FIG. 7--Repeatability data at different sulfur concentrations in gasoline for model 6000.
TABLE 2--Precision summary for Fig. 7 gasoline.
mg/kg S # of Injections Standard Deviation mg/kg % RSD
50.44
~70
0.47
0.93
29.2
~20
0,23
0,75
4.95
-10
0.14
2.87

Linear Response
Demonstration of linear response enables the use of single point calibration techniques,
which can offer considerable cost savings regarding the number of calibration materials that
must be maintained and the time it takes to calibrate the on-line analyzer. A 2000 ppmw standard
was made with 7.88 g of pure dibutyl sulfide diluted in 1 L of toluene. This standard was further
diluted sequentially 1:1 nine times until the expected sulfur concentration in the sample was
down to 3.91 mg/kg. Each of these 10 samples were injected multiple times in a 6000 series
analyzer that was set for a 2500 ppmw full scale range and calibrated with the 2000 ppmw
sample (see Figs. 8 and 9). Data from the results of the 16.13 ppmw sample were lost due to a
data acquisition problem. The data, shown in Table 3, indicate a % RSD of better than 1 % in
the sulfur range 62-2000 mg/kg. At lower sulfur levels of 4-31 mg/kg, the %RSD is 11-2 %.
To demonstrate the linear response of the High Speed Sulfur Analyzer, a 750.1 mg/kg
reference diesel sample was diluted eight times sequentially on a 1"1 basis with a near-zero
sulfur light hydrocarbon of similar density. Each of the nine samples was injected multiple times
in a HSS series analyzer that was set for a 1000 ppmw full scale range and calibrated with the
neat reference material. The data are summarized in Table 4, and the correlation is plotted in Fig.
10. The r2 correlation coefficient was 0.999. The data in Table 4 indicate a % RSD of better than
10.1% in the 4-28 mg/kg sulfur range. At the higher sulfur levels, between 50--750 mg/kg, the
%RSD is between 4-0.6%. Figure 10 illustrates the correlated linearity data.

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

147

2000

lOOq

500
250
125
0

,62.5

~31.25

~--7.81
\3.9

Injections
FIG. 8--Graph of standard model linearity data (Table 3).

2000
1800
1600

Correlation = 0.999

1400
"8

1200

> 1000
800

ooo

0 ~

'~.

200

400

600

800

1000

1200

1400

1600

CalculatedValue

FIG. 9--Standard model correlated linearity data (Table 3).

1800

20oo

148

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TABLE 3--Linearity of sulfur measurements.


Average Sulfur
Number of
Standard
Concentration
Measurements
Deviation,
Found, mg/kg
,,
mg/k~
1993.57
43
14.81
998.67
41
8.41
489.6992
162
8.12
248.20
84
1.86
123.68
195
1.31
62.58
40
0.42
34.01
15
0.69
7.05
10
0.25
3.72
12
0.39

Expected Sulfur
Concentration,
,
mg/kg
2000.00
1000.00
500.00
250.00
125.00
62.50
31.25
7.81
3.91

Relative
Standard
Deviation
0.74
0.84
1.66
0.75
1.06
0.66
2.01
3.54
10.56

TABLE 4 Linearity of sulfur measurements for the..high speed sulfur


Average Sulfur
Number of
Standard Deviation,
Concentration Found,
Measurements
mg/kg
mg/kg . . . . . .
752.1
7
4.3
383.0
9
4.2
197.4
10
3.4
100.47
12
2.3
48.91
12
2.0
28.0
11
2.3
15.3
12
1.5
8.2
10
0.7
4.8
10
0.4

(HSS) analyzer.
% Relative
Standard
Deviation
0.57
1.10
1.72
2.31
4.01
8.13'
10.07
8.40
9.20

High Speed Linearity


700 I
Correlation = 0.999
600

.'"

"a 500
:>
"o

400

O
N

:~

300

e,.

<

200
100

O~

100

200

300
400
Calculated Value

500

600

700

FIG. 10---HSS correlated linearity data (Table 4).

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

149

Stability of Results
The user also wants to know that the analyzer is stable over time and does not drift. Figure
11 shows the long range stability of measurements for the HSS analyzer. In this case, a gas oil
(diesel) sample containing about 130 mg/kg of sulfur was measured repeatedly over a period of
24 h. An average value of 131.4 mg/kg with a 0.662 RSD was found. Figure 11 illustrates the
long-term stability of the HSS

High Speed Sulfur Analyzer 24-hour Repeatability


Sample: Gazole/Gasoil @ 131.4 mg/Kg
16o1
~4o ! ....................................................................

I
I

120

8o!
60,

4o!

Calibrated Average = 131.4 ppm, Standard Deviation = 0.869, %RSD = 0.662%

20,
0

0 "1 2
Hours (1 minute Analysis)

FIG. 11--Long-term stability of HSS.

Accuracy of Results
Finally, Table 5 compares the results obtained on six ASTM Ultra-Low Sulfur in Diesel
(ULSD) cross check samples. Data from a D 5453 laboratory analysis and data using a process
analyzer 6200 and 6200 HSS are compared with ASTM D 5453 consensus values based on about
100 measurements from the participating industry laboratories. Both the standard Model 6200
process analyzer and the 6200 High Speed Sulfur analyzer results are in excellent agreement
with the ASTM consensus values as well as the D 5453 results obtained in our laboratory.
ASTM Standardization of the New Method

Work is underway to standardize the test method described here as a new ASTM standard.
An Inter-Laboratory Study (ILS) is proposed to include at least nine each of gasoline and diesel
Tier 2 class materials containing sulfur levels between <10 and 80 mg/kg in gasoline and
between <10 and 80 mg/kg in diesels. Some blind National Institute of Standards and
Technology (NIST) Standard Reference Materials (SRM) and solvent blanks are also planned for
the cross check. The ILS will follow a protocol of an initial analysis of a set of samples and
solvent blanks, and after an elapsed time of at least 12 h but no more than 72 h, a second set of
measurements on the same set of samples and blanks will be completed. We expect this ILS to
be completed by the end of 2005.

150

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

ASTM
ULSD #
0401
0402
0403

TABLE 5--Sulfur results of ASTM


D 5453
ASTM
High
Lab
Consensus
speed
Results
Values*
sulfur
(n = 2)
analyzer
14.5
15.08 1.27
14.90 q(99)
0.43
(10)
1.6
1.58 q-0.07
2.21 :e
(104)
0.25
(10)
1.6
1.41 0.07
1.30 :L
(118)
0.20

ULSD cross check samples (mg/kg).


% RSD
HSS

% Full
Scale

Standard
Online
Analyzer
6200
15.63 ~0.16
(12)
1.46 + 0.12
(10)

% RSD

% Full
Scale

2.87

0.43

1.01

0.31

11.07

0.25

7.98

0.23

15.5

0.20

1.51 0.08
(10)

5.16

0.16

3.58

0.52

15.2 -4-0.27
(7)

1.75

0.53

4.01

0.40

9.66 0.32
(7)

3.33

0.64

(lO)
0404

16.3

14.92 q-0.18
(104)

0405

7.9

9.83 0.13
(113)

0406

0.6

14.52 q0.52
(10)
9.98
0.40
(10)
0.62
0.17 (10)

0.66 0.28
27.12
0.17
0.64 0.08
12.8
0.17
(101)
(16)
All values are in mg/kg sulfur.
* Mean value :~ standarddeviation(numberof results).
The 6200 was set for a full scale concentrationof 60 mg/kg and calibrated with a 60 mg/kg standard that that was
produced from thiophenein a matrix of tolueneand iso-octane.
The 6200 HSS analyzer was for a full scale concentration of 100 mg/kg and calibrated with a 13 mg/kg diesel
reference material.

Conclusions
Rapid determination of sulfur levels in fuel streams is an important parameter of low sulfur
fuels produced in refineries today. It can be even more critical to have such capability in
blending, fuel transport, or pipeline installations. The instrument described here can deliver
updated sulfur data every minute, which will allow a more accurate delineation 0fboundary lines
in the trans-mix region of diverse fuel streams. Timely information in such situations is critical
for making cut-point decisions which impact revenues and potential cost savings. The analyzer
described in this paper is capable o f accurately ( ~ 1 % RSD) and very rapidly (~1 min)
determining very low levels of sulfur (1 mg/kg) in gasoline and diesel streams with a minimal
recovery time between successive sampling of different material streams, thus enabling a
refinery or a pipeline terminal operation to achieve significant savings of time and costs.
The authors wish to thank Dr. R.A. Kishore Nadkami for his editing assistance and Franek
Olstowski for some data generation regarding the production of this paper.

References
[1] U.S. Environmental Protection Agency - Federal Register 62(133); 40 CFR Part 80 Part II,
Page 37337 (July 11, 1997).
[2] U.S. Environmental Protection Agency - Federal Register 64(1992); 40 CFR Parts 80, 85,
and 86, Page 26055 (May 13, 1999).
[3] U.S. Environmental Protection Agency - Federal Register 65(28); 40 CFR Part 80, Pages

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

151

6752-74 (February I0, 2000); and 66(12); 40 CFR Part 80, pp. 5002-5141(January 18,
2001).
[4] Johnson, W., "Measuring Low Sulfur Levels," Hydrocarbon Engineering, Vol. 8, No. 3,
2003, pp. 49-52.
[5] Hietala, M., "Sensitive ASTM D 4294 Sulfur Analysis," Hydrocarbon Engineering, Vol. 8,
No. 3, 2003, pp. 53-54.
[6] Nadkami, R. A., "Guide to ASTM Test Methods for the Analysis of Petroleurn Products and
Lubricants," ASTM MNL 44, ASTM International, West Conshohocken, PA, 2000.
[7] Kohl, K., "Fitness for Use of Sulfur Methods," ASTM RR-DO2-1456, ASTM International,
West Conshohocken, PA, 1999.
[8] Tittarelli, P., "Round Robin Exercise for Sulfur Test Methods for EN 228 and EN 590 Fuel
Specifications," CEC/TC 19/WG 27, April 2000.
[9] Nadkami, R. A. and Bly, K. J., "An Interlaboratory Study for the Evaluation of Test Method
Capabilities for the Determination of Low Levels of Sulfur in Gasoline and Diesel," ASTM
RR-DO2-1547, ASTM International, West Conshohocken, PA, 2003.

Journal of ASTM International, September2005, Vol. 2, No. 8


Paper ID JAI12983
Available online at www.astm.org

John R. Rhodes I

Pyro-Electrochemical On-Line Ultra Low Sulfur Analyzer


ABSTRACT: Current and upcoming regulations worldwide require maximum concentrationsof sulfur in
fuels to be in the range 10--50mg/kg. This paper describes a new on-line analyzer for ultra low sulfur, the
first to employ "Pyro-electrochemicar' technology, which is significantly simpler than other pyrolysis
methods such as Pyro-microcoulometry and Pyro-UV fluorescence.
The technology is based on ASTM D 6920. The instrument is described with special reference to the
modifications needed to adapt the technique to on-line analysis. Experimental results are given for long
term stability, sensitivity, precision, speed and linearity of response, temperature stability, and sensitivity
to nitrogen.
KEYWORDS: ultra low sulfur, fuels, on-line process analyzers, thermal oxidation, pyrolysis,
electrochemical sensors, nitrogen converters

Introduction
The maximum sulfur content o f fuels has been reduced steadily over the last few decades to
comply with ever-stricter air pollution regulations. Lower sulfur levels have also enabled the
development and economic operation o f more efficient engines. Before 1990, sulfur
concentrations above 3000 mg/kg were common. During the 1990s, maximum permitted sulfur
contents in gasoline and diesel fuels were brought down to 500 mg/kg in most technologically
advanced countries and as low as 50 mg/kg in some. "Ultra low sulfur" (ULS), usually defined
as a sulfur content less than 50 mg/kg, is now mandated in the USA, EU, UK, and Japan. The
goal b y 2010 is "sulfur free" fuels, that is, fuels containing less than 10 mg/kg sulfur. The
obvious benefit to air quality will be a greater than 99 % reduction in sulfur dioxide (SO2)
emissions. Other benefits, no less important, include improved engine efficiency and fuel
economy, reduced particulate emissions from diesel engines, and longer-lived catalytic
converters (many sulfur compounds are notorious catalyst poisons).

Analytical Techniques and Instrumentation


The reduction in sulfur levels b y over two orders o f magnitude has significantly impacted the
analytical techniques and instruments employed for sulfur measurement both in laboratory
analysis and process monitoring. For example, Energy Dispersive X-ray Fluorescence (EDXRF),
probably the most widely used method up to now because o f its relative simplicity, low cost,
reliability, and easy sample conditioning requirements, may not be sensitive enough to measure
ultra low sulfur levels. All other techniques employ to a greater or lesser extent more
complicated instrumentation and sample conditioning than EDXRF. Table 1 compares values o f
precision at 20 mg/kg S in ULS diesel obtained b y the analyzer described in this paper with
Manuscript received 27 September 2004; accepted for publication 25 January 2005; published September 2005.
Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future
Prospects on 6-8 December 2004 in Tampa, FL.
1Rhodes Consulting, P.O.Box 201784, Austin, TX 78720-1784.

Copyright 9 2005 by ASTM International, 100 Ban" Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

152

RHODES ON PYRO-ELECTROCHEMICAL

153

values published in two recent Interlaboratory Comparisons [1,2]. Precision is defined here as
the 95 % confidence limit on the difference between two test readings of the same sample (or
about 2.8 standard deviations). In the case of the Intercomparisons the precisions are the "pooled
within laboratory precisions" as defined in paragraph 23.1.3 of ASTM Standard Practice for Use
of the Terms Precision and Bias in ASTM Test Methods (ASTM E 177-90a, reapproved 2002),
usually called repeatability, r. The precision value for the present analyzer was obtained using
the method described in paragraph 23.1.1 of E 177-90a. That is, a single operator performed
measurements on a succession of aliquots drawn from the same homogeneous bulk sample on
the same day using the same instrument. The ASTM methods employed are as follows: ASTM D
6920, "Standard Test Method for Total Sulfur in Naphthas, Distillates, Reformulated Gasolines,
Diesels, Biodiesels and Motor Fuels by Oxidative Combustion and Electrochemical Detection";
ASTM D 3120, "Standard Test Method for Trace Quantities of Sulfur in Light Liquid Petroleum
Hydrocarbons by Oxidative Microcoulornetry"; ASTM D 5453, "Standard Test Method for
Determination of Total Sulfur in Light Hydrocarbons, Motor Fuels and Oils by Ultraviolet
Fluorescence"; ASTM D 2622, "Standard Test Method for Sulfur in Petroleum Products by
Wavelength Dispersive X-ray Fluorescence Spectrometry"; and ASTM D 4294, "Standard Test
Method for Sulfur in Petroleum Products by Energy-Dispersive X-ray Fluorescence
Spectrometry". In Table 1, Pyro-EC (APS) and Pyro-EC (Antek) refer to results obtained by two
different instruments (from APS Technologies, now Analytik-Jena AG, Jena, Germany and
Antek Instrumehts, Houston, TX, respectively) which had significantly different repeatabilities.
WDXRF(HP) means that a high power X-ray tube (greater than 3 kW) was employed for
excitation, and WDXRF(LP) means that a low power tube (less than lkW) was used. The
European Round Robin [2] called out European or International standard methods, but they are
similar to the corresponding ASTM methods, so they are not detailed here. Since the precisions
were measured by different procedures, exact comparisons are not justified. However, the data in
Table 1 can be used as a guide to the performance of the various techniques.
TABLE 1--Methods and precisions for ULS in diesel.
Technique
,
Pyro-EC
On-Line Analyzer(this paper)
Pyro-EC(APS)
Pyro-EC (Antek)
Pyro-mieroC
Pyro-microC
Pyro-UVF
Pyro-UVF
WDXRF (HP)
WDXRF (LP)
EDXRF
EDXRF (polarized)

ASTM
Precisionat
Method 20mg/kg,S (m~Jkg)
D6920
0.39
D6920
D6920
D3120
D3120
95453
95453
92622
92622
D4294
In process

0.41 [1]
3.89 [1]
1.76 [l ]
2.02 [2]
1.34 [1]
1.66 [2]
2.04 [2]
2.46 [2]
7.67 [2]
1.59 [2]

It is clear that any method whose precision is worse than about 1 mg/kg S will have difficulty
performing satisfactory analyses of ULS diesel at, say, 15 mg/kg S. The results in Table 1
indicate that for most methods significant improvement is necessary. In fact, the US

154

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Environmental Protection Agency has published performance specifications (Code of Federal


regulations, 40CFR part 80.584(a)(1)) which include a required standard deviation of 0.72 ppm S
or better, in the range 5-15 ppm sulfur in diesel when repeat measurements of the same sample
are made over a time interval of 20 days. A new Interlaboratory Crosscheck Program is being
conducted by ASTM in 2004, targeting the sulfur content range 0--15 mg/kg S in diesel [3]. The
following ASTM methods are represented: D 2622, D 3120, D 5453, and D 6920.
An on-line analyzer must be packaged much more robustly than a laboratory instrument to
withstand the process environment which, for example, may have an explosive atmosphere and
significantly variable ambient temperature. It must also be capable of continuous, unattended
operation over long periods of time. Clearly, the simpler the instrument the better. Of the
methods listed in Table 1, WDXRF, polarized EDXRF, and Pyro-mierocoulometry have not
been adapted to on-line process instrumentation, whereas the other methods have. The relative
simplicity of Pyro-EC makes it particularly suitable for adaptation to process instrumentation.
The sulfur dioxide sensor is a small, plug-in, low cost electrochemical cell, easily replaceable
and with an expected lifetime of over one year. The UV lamp, UV optics, and photomultiplier
used in PYro-UVF are not required. The X-ray tube (or radioactive source), X-ray detector, and
X-ray optics used in all the XRF instruments are not required.
Description of the Instrument
Figure 1 shows a simplified flow diagram of the Model 682P-EC (Spectro Analytical
Instruments, Marble Falls, Texas) [4]. The essential subassemblies are: means for sampling from
the process stream; sample injection into the pyrolysis furnace; thermal oxidation at 1050~
conversion of any nitrogen dioxide (NO2) to nitric oxide (NO); control of flow rate, relative
humidity and temperature of the resulting gas, quantitative detection of SO2 concentration. In
addition (not shown in the Figure), there is a means for providing calibration liquids, a series of
diagnostic alarms, and a programmable logic controller (PLC) to control instrument operation,
calculate the sulfur concentration output, and provide alarm signals. The instrument is housed in
two NEMA 4X (IP-65) stainless steel enclosures purged to either Class 1 Div 2 Groups C&D
(Type Z purge) or Type X purge for Class I Div 1 Groups C&D and CENELEC Eex p II T4
explosive atmosphere specifications. LEL1 in the figure is a combustible gas sensor to provide
an alarm in the event of a leak of combustible fluid into the cabinet. The Scrubber shown in the
figure is an activated charcoal-filled cartridge to absorb any noxious gases, such as oxides of
sulfur, nitrogen, and carbon before exhausting the sample gas to atmosphere. The other
components shown are parts of the essential subassemblies, described below.

Sampling
Liquid from the process stream is sampled by a two position, six port rotary valve with a 520 I.tL sample loop connected between two ports. The sample may be preconditioned to remove
any gas bubbles, entrained particulates, and water. Since this part of the instrument is outside the
enclosures and could be in an explosive atmosphere, all valves are pneumatically operated, and
the sample is driven by compressed air or inert gas.

RHODES ON PYRO-ELECTROCHEMICAL

155

FURNACE
SAUPLE
~
ROTARY
VALVE

PROCESS RETURN
PROCESS INPUT

EXHAUST
EXHAUST

OXYGEN INPUT
ARGON INPUT

FIG. 1-- Flow diagram of model 682P-EC.


In one position of the rotary valve, the process sample flows through the sample loop, thus
filling it (and continues back to the process), while the carrier gas flows through the valve and
the injector, ensuring that the injection path is clear and ready for the next sample injection. In
the second position, the carrier gas is switched to push the sample out of the sample loop,
through the injector, and into the pyrolysis tube, while the process fluid continues to flow though
the valve and back to the process (bypassing the sample loop). In this way, the rotary valve and
sample loop combination "cuts" and dispenses a sample of fixed volume from the process stream
for analysis. The sample volume is usually chosen to be 10 IxL and is constant to better than one
percent. The sampling frequency is usually selected to be once every 5 min. Since the sample has
a constant volume, the analysis must be corrected for sample density if results are reported on a
wt/wt basis. Note that the flow rate between the process and the rotary valve is not limited by the
analyzer and may be quite high, so each analyzed sample is contemporaneous with the process.

Injection
The purpose of the injection subassembly is to inject the sample at an average rate of about
1~tL/s and an instantaneous rate that never exceeds that which would starve the thermal oxidation
process of oxygen and so cause incomplete combustion. Also, any carbonization of the sample
within the injector must be avoided. The sample is driven by a flow of inert carrier gas (argon)
that is controlled by a combination of a flow restricter (MMV1) and a pressure regulator (LPR1).
The injector, which is a stainless steel tube, penetrates the pyrolysis tube to a depth where the
temperature is high enough to ensure sample volatilization within the injector while avoiding

156

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

carbonization. The pyrolysis tube is supplied with oxygen flows to carry the injected sample and
completely oxidize it.

Pyrolysis
Pyrolysis takes place in a standard quartz tube inside a tube furnace controlled at a
temperature of 1050~ as specified in the ASTM method. The output gases are passed through a
quartz wool plug and a membrane filter to capture any soot particles.

Conversion
The sample gas stream then passes through a catalytic converter that converts any NO2 to NO
without affecting the SOz content. Nitrogen in the fuel sample is theoretically completely
oxidized to NO at the pyrolysis temperature of 1050~ However, some NOz is formed due
possibly to reconversion of NO to NO2 in the cooler parts of the furnace downstream. The
electrochemical sensor has a negative response to NO2 but is insensitive to NO. Hence, it is
important to convert any NO2 to NO before it reaches the sensor. Another reason for making sure
that all nitrogen is in the form of NO is that the instrument can potentially be used for total
nitrogen determination by simply adding an NO sensor.
The effective conversion efficiency (that is the conversion efficiency allowing for the fact
that only some of the nitrogen ends up as NO2 in the first place) has been found to exceed 99 %.
Long term tests show that the converter lifetime is at least one year. The converter is an easily
replaceable cartridge.

Sample Conditioning
The relative humidity (RH) of the sample gas stream is such that the dew point is about 55~
during pyrolysis of the fuel sample but fails to very low levels when no liquid is being thermally
oxidized. The dryer shown in Fig. 1 is a permeation dryer (Perma Pure Inc., 8 Executive Drive,
Toms River, NJ 08754), which removes water vapor without affecting the gases of interest and
so prevents water condensation when the sample has cooled down. It employs a purge gas of dry
argon that is controlled through a metering valve (MV1) and monitored by an electronic
flowmeter (FM1). It is important that the RH of gas incident on the electrochemical sensor is in
the range 10-90 % to prevent the sensor getting too dry or too wet. To achieve these conditions,
the sample gas is passed through a permeation humidifier (Perma Pure Inc., 8 Executive Drive,
Toms River, NJ 08754)just upstream of the sensor.
The flow rate of the sample gas stream must be constant, otherwise the SO2 content will vary
due to changes in gas dilution. The flow rates are controlled to better than 1 % by metering
valves (MV2 and MV3) and pressure regulators (not shown). Electronic flowmeters (FM2 and
FM3) provide signals to the PLC that are used for alarm and diagnostic purposes. They can also
be used to correct the SO2 concentration for the above dilution effect, if necessary.
All parts "wetted" by the sample gas stream must be inert to SO2. This is achieved by the use
of AISI 316 stainless steel or fluoropolymer plumbing where necessary.

RHODES ON PYRO-ELECTROCHEMICAL

157

Electrochemical Sensor and Response


The electrochemical sensor is an off-the-shelf SO2 sensor (City Technology Ltd., City
Technology Centre, Walton Road, Portsmouth, PO6 1SZ, England). Typical sensor response
profiles, or "pyrolysis spectra," are shown in Figs. 2 and 3. Each peak represents the response, as
a function of time, to a single constant volume sample injection. The liquid injection rate rises
rapidly from zero at the beginning of the injection cycle, remains more or less constant during
the 10-s injection, then falls as the last of the sample is cleared out of the injector. Thermal
oxidation takes place almost immediately since the residence time in the pyrolysis tube is only
about 10 s. The gas transit time from pyrolysis furnace to sensor is only a few seconds, so there
is not much dispersion of the SO2 concentration profile. It follows that the volume concentration
of SO2 is proportional to the peak integral. It is also seen that the sensor time-dependent response
is quite rapid.
Performance

This section describes the analyzer performance to ULS in diesel and gasoline.

Nitrogen Interference
Nitrogen is present in some fuels in concentrations up to, but rarely exceeding, the sulfur
content. The approximate nitrogen content of a given fuel is often known, so that small
interferences can be minimized by calibrating the instrument with samples that contain
approximately the same amount of nitrogen as the fuel to be analyzed. However, it would be
preferable to reduce any nitrogen interference to negligible proportions.
Initial tests were performed using samples of diesel and gasoline containing zero or 20 mg/kg
sulfur and either zero or 50 mg/kg nitrogen. Figure 2 shows the effect of nitrogen when the
converter is not installed. The nitrogen content was changed from zero to 50 mg/kg and then
back to zero, keeping the sulfur content at 20 mg/kg. The effect of the nitrogen is to considerably
reduce the electrochemical cell response over a period of several sample injections. When the
nitrogen content goes to zero, the cell response recovers slowly back to its original level. Figure
3 shows the effect when the converter is installed. Any nitrogen interference is now not
observable. Table 2 gives results of quantitative measurements made on similar gasoline
samples.
Further tests have been made to determine the long term stability and lifetime of the
converter. An example of these results is shown in Fig. 4, where the sample has an extremely
high nitrogen content compared to the sulfur content. Experience over the past two years of
operation with samples containing varying amounts of nitrogen have verified that in practice
nitrogen interference does not occur, and the converter lifetime is in excess of one year.

158

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

to

FIG. 2--Effect of nitrogen at 20 mg/Kg sulfur in diesel, without converter.

~000

2000

3000

4000

Saeont~

FIG. 3--Effect of nitrogen at 20 mg/Kg sulfur in diesel, with converter.


TABLE 2--Test resultsfor nitrogen in gasoline.
mg/kg S

mg/kg N

20
20
blank
blank

50
zero
50
zero

mean S
reading
20.2
20.5
1.2
1.5

two
sigma
0.5
0.6
0.5
0.4

5OOO

RHODES ON PYRO-ELECTROCHEMICAL

159

~ma~s
Sb:l [ ] e v = O,

50

100

150

2B)
Msam~.~;

FIG. 4

25O

3]O

3~0

Number

On-line analysis of approximately 7 mg/Kg S, 120 mg/Kg N in diesel (31-h run).

Precision and Sensitivity


Measurements o f precision at different concentrations o f sulfur in diesel yielded the
following value as a function of concentration:
p = 0.016088"X + 0.068238
where X is the sulfur content in mg/kg, and p is defined as the 95 % confidence limit on the
difference between two test results.
An accepted measure o f sensitivity is the minimum detectable limit (MDL). This is equal to
the value o f p at zero concentration, which is seen to be 0.068 mg/kg S.
The above values o f precision and sensitivity were obtained by one operator on a single
instrument during a time interval o f a few days. However, many similar measurements have been
made at different times over the past two years with comparable results.

Long Term Stability


A continuous 26-day unattended run with diesel containing 20 mg/kg sulfur was performed
to test the long term stability. No calibration checks or recalibrations were done. Figure 5 shows
the plot obtained o f sulfur readout versus measurement number. It is seen that the long term
stability is excellent, with no significant up or down drift. Since drift is negligibly small, the
standard deviation o f all the measurements (sigma = 0.17 mg/kg S) is another indicator o f the
precision at 20 mg/kg sulfur.

Linearity
Linearity tests were performed using diesel samples with known sulfur content in the range
zero to 500 mg/kg. Typical results are shown in Fig. 6. It is possible that the responses o f
different electrochemical cells will be slightly different from each other. However, nonlinearities of the degree shown here are negligible over the smaller concentration ranges used in
practice. Table 3 gives examples of non-linearity error to be expected over practical
concentration ranges. The maximum errors are well below the normal instrument precision.

160

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

25

20!

m m

~wmlm

~ P

us run wilhc t caJibratior update.


ents at 5 rri ~tes per rnE
15
:=+0.13%

Tot~ r
E

~verage re~ :ling: 20.6 "g/Kg S

10

Precision = ).83% O" 0.1 ' rrg/Kg S

1000

30OO

2000

4OOO

6OOO

5O0O

70OO

8OO0

Mea~-=T=t

FIG. 5 - - L o n g term stability.


8O0OOOO

7OOOOOO
6OOOOOO
,5OOO00O

Net :~eak = 5. 1917"X^2 + 1227ff X - 19421


std clev = 2.8 L~JK.q

jf

"4000000
Z

3OOOOOO

,J

z /

2OOOOOO
1000000

50

100

150

200

250

300

350

S~Xmg~g

FIG. 6 - - N e t p e a k versus sulfur content.

4OO

450

5C0

RHODES ON PYRO-ELECTROCHEMICAL

161

TABLE 3--Examples of non-linearity error.


Range (mg/kg S)
0-20
0-30
10-50
50-100
100-200
200-300
300--400
400-500

Maximum Error (at mid-range)


mg/kg
%
-0.04
-0.41
-0.09
-0.61
-0.16
-0.54
-0.24
-0.33
-0.92
-0.61
-0.86
-0.34
-0.80
-0.23
-0.75
-0.17

Speed of Response
The speed of response is generally measured by plotting the instrument output as a function
of time in response to step changes in sample concentration. Such a plot is shown in Fig. 7.
3~)

250

200

l ~OnVKg

15o
E
100

6Omoj~
50

ZSrrg<g

9149149149149149

10 rn, I / K g

eeee~, 9149149

~eeoe,~

10

10 rrg/l~

[
15

20

25

30

35

40

45

50

FIG. 7--Response to step changes in concentration.


The response is nearly but not quite "instantaneous," where instantaneous is defined here as a
100 % response to a step change by the next measurement after the change. The response speed
is influenced mainly by the degree to which, during the injection of the previous sample, the next
sample flushes out the previous one in the Rotary Valve and associated plumbing. This, in turn,
is a function of the sample flow rate and the magnitude of the step change in concentration. So

162

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

the response can be made faster by increasing the sample flow rate, which is a choice the
operator can make. The flow rate used in this test was about 40 ml/min.

Temperature Dependence
There are three main temperature effects. One is due to the variation of sample gas density
with temperature, according to the gas law. This causes a change of sulfur concentration in the
sample gas stream since the volume of liquid sample injected does not change significantly with
temperature. This effect is common to all the pyrolysis techniques. The second effect is the
temperature sensitivity of the electrochemical sensor, which is small but not negligible. The third
effect is the influence of changes in total water vapor content of the gas sample stream. All three
effects have been minimized by proprietary design. Figure 8 shows the results of varying the
ambient temperature over a range greater than that likely to be experienced in the instrument
shelter of a refinery or other plant. It is seen that the concentration output does not vary when the
temperature is changed over the range 17-29~ I f a temperature dependence is ever observed,
the instrument has a built-in algorithm that can be used to correct the output for measured cabinet
temperature.

FIG. 8--Response to temperature changes.


Conclusions

9
9
9
9
9

Pyro-electrochemical technology is viable for ultra low. sulfur determination.


Suitable instrumentation is available for on-line or at-line monitoring of petroleum fuels.
The sensor is a small, low-cost plug-in component, leading to a greatly simplified
detection subsystem compared to those of other techniques.
Nitrogen interference has been eliminated.
Sensitivity, repeatability, stability, linearity, and speed of response are equal to or better
than those for other on-line ULS analyzers.

RHODES ON PYRO-ELECTROCHEMICAL

163

References

[1] ASTM Lnterlaboratory Crosscheck Program, Low Level Sulfur, ASTM International, West
Conshohocken, PA, 2002.
[2] CEN TC19 WG27, "Test Methods for the Determination of Sulfur Content. Report of the
2001 Round Robin," European Committee for Standardisation, 36 rue de Stassart, B-1050,
Brussels, Belgium.
[3] ASTM Interlaboratory Crosscheck Program, Ultra-Low Sulfur Diesel Fuel, ASTM
International, West Conshohocken, PA, 2004.
[4] U.S. Patent #6,830,730, Dec 14, 2004, and U.S. Patent #6,867,047, March 15, 2005.

Joumal of ASTM International, July/August 2005, Vol. 2, No. 7

Paper ID JA113090
Available online at www.astm.org

Ronda L. Gras, 1 B.S., Jim C. Luong, 1B.S., Robert V. Mustacich, 2 Ph.D., and Randy L. Shearer, 3
Ph.D.

DP-SCD and LTMGC for Determination of Low Sulfur Levels


in Hydrocarbons
ABSTRACT: A novel measurement technique has been developed for the rapid determination of low
levels of sulfur in hydrocarbon matrices. The technique employs low thermal mass gas chromatography
(LTMGC) and a dual plasma sulfur chemiluminescence detector (DP-SCD). Highly sensitive total
volatile sulfur measurement can be made in less than 30 s with detection limits in the 20-30 parts per
billion range. Response is linear over at least three orders of magintude sulfur with excellent repeatability.
By taking advantage of LTMGC's rapid temperature programming and cool down time, it is also possible
to measure individual sulfur species, analogous to ASTM D 5623. Conditions have also been developed
using the same analytical apparatus to detect the major elemental sulfur allotropes ($2, $3.$6, $7, and $8) in
gasoline at levels that can cause silver corrosion.
KEYWORDS: sulfur, elemental, sulfur chemiluminescence detection (SCD), low thermal mass gas

chromatography(LTMGC), dual plasma SCD


Introduction

The United States Environmental Protection Agency (U.S. EPA) and other environmental
regulatory agencies, particularly those in Europe, Japan, and other industrialized nations, have
developed regulations aimed at further reducing air pollution from mobile sources. Sulfur plays a
key role in this since it is known to poison catalytic converters that are used to reduce
concentrations o f carbon monoxide, hydrocarbons, nitrogen oxides, and particulates in engine
exhausts. The allowable levels o f sulfur in liquid fuels, such as gasoline and diesel fuels, are
continuously being driven to lower levels. For example, in California the total average allowable
level o f sulfur in gasoline has been 30 ppm total sulfur since 1996 [1], and the U. S. federal
limits for sulfur in on-road diesel fuel will be 15 ppm (mg/kg) starting in 2010 [2]. This is a
significant development, considering that in the past average sulfur levels for gasoline were
higher, often b y a factor o f 10 times or more.
Numerous techniques and instrumental methods can be used to measure sulfur in gasoline
and diesel fuels [3]. It is beyond the scope o f this paper to cover any o f the standard techniques
in detail. Unfortunately, many o f the traditionally used techniques are out o f consideration for
analyzing fuels at low levels because o f inadequate sensitivity or lack o f precision at those levels.
For instance, a study group comprised o f 69 industry participants has recommended that Energy
Dispersive X-ray Fluorescence (EDXRF) and some other older technologies be dropped from
consideration for analyzing fuels at the 30 ppm level [3].

Manuscript received 7 October 2004; accepted for publication 14 December 2004; published July 2005. Presented
at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
l Dow Chemical, Hwy 15, P.O. Bag, Bldg. 158, Fort Saskatchewan, Alberta, Canada, T8L 2P4.
2 RVM Scientific, 5511 Ekwill St. # A, Santa Barbara, CA 93111-2398, USA.
3Ionics Instruments, 6060 Spine Rd., Boulder, CO 80301, USA.

Copyright 9 2005 by ASTM lntornational, 100 Barr Harbor Drive, PO Box C700, West Consbohockcn, PA 19428-2959.

164

GRAS ET AL. ON DP-SCD AND LTMGC

165

Sulfur measurements from process and product quality standpoints are extremely important
and necessary. Nevertheless, total sulfur measurements alone may provide little information
about the best manner to process a feed or even to determine whether a process unit is
functioning as designed. For example, in a process that converts mercaptans to disulfides, total
sulfur measurements would show little or no difference between the feed and the product, while
a sulfur distribution measurement would show if the intended chemical conversion is taking
place. Furthermore, there may be unexpected or unintended consequences from process changes
that affect product quality that are related to individual trace sulfur species. For example, it has
recently become apparent that low ppm levels o f elemental sulfur can contribute to a product
quality issue o f corrosion of automobile fuel gauge sensors [4]. This is especially surprising and
ironic considering that sulfur in fuel levels are lower than ever before. There is speculation that
some other sulfur species may contribute synergistic or even inhibitive effects on elemental
sulfur formation and subsequent silver corrosion. Nevertheless, at least 45 000 vehicles have
been affected by this issue to date, and the financial impact caused is apparent. Repair costs
alone are on the order of at least $200-600 per vehicle.
In this paper, the use of LTMGC to attain high speed heating and unprecedented cool down
time and a new commercially available detection system, Dual PlasmaT M technology, based on
sulfur chemiluminescence detection (SCD) is described. The system is able to provide
quantitative measurements of individual sulfur species, such as the allotropes responsible for
silver corrosion, and also can be operated in various modes to provide a rapid total sulfur
measurement by itself or a total sulfur measurement in conjunction with speciated sulfur
analysis.
The new Dual Plasma technology for the SCD (DP-SCD) has been developed to improve
performance in several areas, including stability and selectivity [5]. The SCD has become a
commonly used sulfur selective detector in petroleum and petrochemical laboratories and has
been described in detail [6,7,8,9,10,11,12]. As of this writing, there are three ASTM methods
currently being considered that utilize SCD technology. These are ASTM D 5504 (Standard Test
Method for Determination of Sulfur Compounds in Natural Gas and Gaseous Fuels by Gas
Chromatography and Chemiluminescence), D 5623 (Standard Test Method for Sulfur
Compounds in Light Petroleum Liquids by Gas Chromatography and Sulfur Selective
Detection), and D 7011 (Test Method for Determination of Trace Thiophene in Refined Benzene
by Gas Chromatography and Sulfur Selective Detection) [13]. The high sensitivity, selectivity,
absence from hydrocarbon interferences, linea~rity of the SCD, and equi-molar response are well
established. It may be operated in series with a flame ionization detector (FID), or the column
may be connected directly to a combustion furnace. Briefly, detection of sulfur compounds in the
SCD is based upon the two-step mechanism shown below:
S-compound + 02 --> SO + Other Products

(1)

SO + O 3 ~ SO2" + 02 -9 SO2 + hv
(2)
In Eq 1, sulfur compounds are combusted to sulfur monoxide (SO) and other products. In Eq
2, the second step of the mechanism, light energy (hv) in the blue region of the spectrum is
emitted from the excited species resulting fi:om the ozone reaction. The basic mechanism of the
DP-SCD is the same as that described above, but two plasmas (flames) instead of one are
provided to improve selectivity and the ability to measure lower sulfur levels without
hydrocarbon interferences. A conceptual drawing of the flow dynamics used in the Dual Plasma
burner is shown (Fig. 1).

166

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 1--Conceptual flow diagram of Dual Plasma burner.


A schematic block diagram illustrating an entire DP-SCD detection system is shown in Fig.
2. An analytical system consists of a gas chromatograph equipped with a split/splitless injector
with the option of a Pressurized Liquid Injection System (PLIS), with or without low thermal
mass gas chromatography apparatus, for sample introduction and sulfur speciation (if required);
an electrically heated burner with an interface that controls the burner gas flows and temperature;
and a detector that contains a chemiluminescent reaction cell, ozone generator, optical filter,
amplifier, and electronics. Lastly, a vacuum pump is used to keep the reaction cell under low
pressure conditions to prevent loss of chemiluminescent species and to reduce collisional
quenching.
Gas chromatography ( G C ) a s an analytical technique is accepted and widely used,
particularly in the petroleum industry. Its application is amenable to gaseous, liquid, and even
solid samples using a variety of sample introduction techniques. In the case of gaseous and liquid
samples, introduction of the sample is usually based on a volumetric injection using a valve or
syringe, either of which can be automated. Because of the volumetric nature of sampling, it is
important that density be taken into account when calculating results on a mass basis. In some
cases, density values will not affect calculated results, for instance when an instrument is
calibrated on a mass per unit volume basis, or if the sample matrix and calibration matrix are
similar or have the same density. In addition, there are intricacies with regard to the injector

GRAS ET AL. ON DP-SCD AND LTMGC

167

technique that is used for sample introduction, such as split and on-column injection [14]. Even
though GC could be considered to be a mature technique, significant improvements are being
made in several areas besides detection, such as approaches to improve separation or analysis
speed.

25psigmax~
25 psig ma

F.C.
Dual Plasma Controller

Dual
PlBurner
asma r-"-IInjInlector
et

Detector

Une

Transfer

Column

Gas Chromatograph

Vacuum
Pump~ent
FIG. 2--Block diagram of GC system with DP-SCD.
The process of increasing column temperature during a GC run is referred to as temperature
programming gas chromatography (TPGC). For a particular solute, the increase of temperature
during temperature programming leads to a decrease of retention factor, retention volume, and
retention time. Key benefits of TPGC include better separation of wide boiling point range
solutes, improved detection limits, peak shapes, and precisions, especially for late eluting peaks.
It also provides an excellent means of column cleaning. A drawback for temperature
programming is that it requires considerable cool down time (several minutes or longer), which
lengthens the analysis cycle time compared to isothermal chromatography. Cryogenic fluids can
be used to lower oven temperatures and speed cooling, but this adds additional expense.

168

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

TPGC of capillary columns is typically conducted by electronic control of the temperature of


an oven containing the GC column. The winding of the columns on the wire frame support
provides extensive surface contact of the capillary gas chromatography column with the heated
air in the air-circulation oven for fast temperature equilibrium. Using this technique, depending
on the size of the oven, reproducible temperature programming rates ranging from 30-70~
can be attained.
Resistive heating elements combined with GC columns have been demonstrated to provide
effective and rapid temperature control, but higher thermal effieiencies, accommodation of a
wide range of column dimensions, and immunity to electrical shorts at high temperatures were
still needed. New technologies that address these needs have been recently developed and
implemented. Recently, RVM Scientific of Santa Barbara, California introduced an advanced
low thermal mass gas chromatographic accessory (LTMGC) that can be easily integrated to
Agilent GCs 4 for high throughput chromatography.
A typical LTMGC module is comprised of: an analytical column, a temperature sensor, a
nickel alloy heating wire, and a metal tray to support the module and its interface. Mounted
under the tray is an electric fan for heat removal during the cooling down cycle of a run. The low
thermal mass design, coupled with high surface area provides rapid temperature programming
rates in excess of 1800~
with unprecedented cool down time. Power consumption for
temperature programming is only about 1 % of that required for conventional gas
chromatographs. Fewer than 60 s were required to cool a 5 m column in an LTMGC module
from 280~ to 40~
LTMGC takes advantage of many of the key fundamental principles of
high-speed gas chromatography [15]. LTMGC provides the capability to either collapse the
separation power of the analytical column or the capability to maximize its separation in the case
where speciation is required. This technique, coupled with a specialized pressurized liquid
injection system (PLIS) [16] and conventional injection, with DP-SCD are discussed.
In GC, highly volatile liquefied hydrocarbons are commonly injected using devices such as
high-pressure syringes, piston valves, liquid rotary valves, or vaporizing regulators. Although
these techniques are adequate in some cases, there are known deficiencies. The PLIS is a small
injector that can be coupled directly to an injector port without the need of pre-injection
vaporization and transfer lines and sample sizes range from 0.1-2 uL for liquid, with the valve
capability to be helium leak-free of up to 1200 psig. These characteristics make the PLIS
attractive for introduction of sulfur containing samples into a GC.
Materials and Methods

The work covered in this paper was carried out in two laboratories with two different sets of
instruments. At the Dow Chemical Canada, Western Canada Operations, Analytical Sciences
Laboratory, an Agilent GC, equipped with a split/splitless injector, a Transcendent Pressurized
Liquid Injection System (PLISrU), a RVM 5 Scientific LMTGC system, and a Sievers6 SCD with
a prototype dual plasma burner was used. Data collection was done using PerkinElmer
AccessChrom Chromatography System. Various columns were used including an 18 m, 0.25
mm id, 0.25 micro Varian VF-1 column; a 50 m, 0.32 mm id, 5 micron CP-Sil 5 CB column; and
a 5 m, 0.25 mm id, 0.25 micron DB-1 column.
4Agilent 6890 and HP 5890 GCs, AgilentTechnologies,Litflefalls,DE, USA.
5RVM Scientific, SantaBarbara, CA USA.
6SieversModel 355 SCD, IonicsInstruments,Boulder, CO USA.

GRAS ET AL. ON DP-SCD AND LTMGC

169

In the Ionics Instruments R&D laboratory, an Agilent GC was used with a Sievers SCD with
Dual Plasma Burners. Agilent ChemStation software was used for data acquisition and
instrument control for most experiments. Both on-column and split injectors with electronic
pressure control were used. Work covering the use o f several GC columns, such as a CP-Wax
from Varian and HP-5 for Agilent will be described in greater detail. The typical sample size was
0.1-1 pL, split 1:10 to 1:50, and chromatographic separations were done with a variety o f
capillary columns. A PLIS was used for some liquid injections. The oven was typically operated
in the range o f 40-340~ at programming rates o f around 10~
for conventional GC and
about 200~
for the RVM module. In some cases, the run was ended at a lower temperature
when complete elution of the sample was ensured. For fast chromatographic analysis, an SPB-1,
30-m x 0.32 mm ID capillary cohnnn with a 4/~m film thickness or an equivalent SPB-1 of 5-m
length was used with a helium cartier gas at a constant flow rate o f 6 mL/min. An uncoated 30m, 0.32 mm ID deactivated fused silica capillary acted as a transfer line for single-peak totalsulfur determinations in conjunction with a Dual Plasma burner for some experiments. Pertinent
experimental details are given with the appropriate results.
Solvents used in the work, such as hexane, iso-octane, toluene, and benzene, were obtained
from Aldrich Chemical and Fisher Chemical Canada and were analyzed prior to use. Benzene
was treated to remove thiophene as per ASTM D 7011 [12]. Sulfur compounds and elemental
sulfur (flake) were obtained from Aldrich Chemical and used as received for preparation of
gravimetric blends. Silver wool (0.05 mm OD fibers) was obtained from Alfa Aesar. A
California Phase II reference fuel analyzed by D 5453 (Standard Test Method for Determination
o f Total Sulfur in Light Hydrocarbons, Motor Fuels and Oils by Ultraviolet Fluorescence) [12]
and found to contain 28 mg/kg sulfur was obtained from Spectrum Quality Standards. Elemental
sulfur was first dissolved in benzene as a stock solution, and this stock solution was
volumetrically added to the reference gasoline fuel to generate low ppm elemental sulfur levels.
In addition to the instrumental sulfur analysis, approximately 10 mg o f silver wool was placed in
about 1:5 g o f the reference gasoline with and without elemental sulfur. The fuel sample was
then placed in a small high-pressure all-glass (except for aluminum foil seal) reaction vessel in a
secondary container and oven and heated to 150~ for about 1 h. After the samples were cooled,
the silver wool was removed and inspected with a photomicroscope at 10x and 40x
magnifications.
Results

The detection o f trace levels o f sulfur in complex hydrocarbon matrices, such as fuels, is
dependent on sensitivity (the ability to detect trace sulfur levels) and selectivity (the ability to
detect sulfur without interference from potentially interfering hydrocarbons). Hydrocarbon
interference can manifest itself positively or negatively or even both in some instances. It is most
desirable to have no interference whatsoever, however, if the interference is known and
repeatable, then calibration can compensate for it. In most analytical techniques designed to
measure trace levels, interferences are likely to exist, and selectivity is most likely to be the
ultimate dominant factor in determining detection limits.
It is relatively straightforward to determine selectivity by using GC because the hydrocarbon
effect can be decoupled from the sulfur response via chromatographic separation. Figure 3 is
illustrative o f both the sensitivity and selectivity of the GC Dual Plasma SCD for trace sulfur
species when used in GC. In this example, the matrix consisted o f about 99.9 % toluene with

170

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

sulfur species added gravimetrically to a stock solution, which was further diluted. The
chromatogram shows that the toluene solvent does not cause any significant deviation of the
baseline, either positive or negative. The ratio of sulfur response to hydrocarbon response
indicates a selectivity sulfur: carbon selectivity of > 2 x 10 7. This, coupled with the sensitivity of
< 0.5 pg S/s, allows detection of low ppb levels of sulfur compounds.

180-

Carbon disulfide

140-

99.9%

120-

Toluene Solvent
Virtually No Response

Ethylmethyl sulfide

100-

30 ppb Sulfur
3-methiophene

00Bo-

Thiophene

I;E

O0

0.5

t ,0

1.5

2.0

25

3.0

:3.5

4.0

4.5

5.0

5,5

I~.0

Time - Minutes

FIG. 3--GC Dual Plasma SCD chromatogram of trace sulfur species in toluene. Injection
volume of 1 ILLsplit 1:10 using a 0.32 mm ID CP-Wax column.
High sensitivity and selectivity are required to measure trace sulfur species in a complex fuel
sample like gasoline. This is especially so for elemental sulfur, which can form several allotropes
and which at ppm levels can contribute to the fuel gauge sensor issue described previously [4].
Using the DP-SCD, chromatographic conditions were chosen to elute all of the volatile sulfur
species in gasoline. Analysis of the California reference fuel (28 mg/kg total sulfur) showed no
evidence of containing the major elemental sulfur allotropes, $6, $7, and $8. Chromatographic
separation of the minor allotropes, $2 and $3, from naturally occurring sulfides (typically present
at low ppm levels) found in the gasoline was not possible using the conditions described here,
but their addition results in detectable differences in the chromatograms. The procedure is readily
capable of detecting $6 and $8 as shown in Fig. 4, however, even at the low level that can
contribute to fuel gauge sensor corrosion.
Though not shown here, it is possible to deduce which peaks of a sulfur chromatogram
contribute to silver corrosion by comparing chromatograms of a sample prior to and after
reaction of the sample with silver wool. In this type of procedure, the reaction vessel must be
capable of withstanding the high pressure that the gasoline will produce when heated, and nonreactive volatile sulfur compounds must not be allowed to escape. It was hoped that standard
laboratory GC auto-sampler vials could serve this purpose, but they were found to leak when
heated above 100~ and a temperature of 150~ was necessary to react with the elemental
sulfur in a reasonably short time, such as 1 h. Some researchers have noted that elemental

GRAS ET AL. ON DP-SCD AND LTMGC

171

mercury rapidly reacts with elemental sulfur in gasoline at room temperature, and this could be
used for the same purpose [4]. However, use of silver is more representative of the materials
used in the fuel gauge sensors that corrode. Furthermore, mercury is more toxic and is subject to
greater concerns with regard to material disposal and industrial hygiene issues.
AIB1B, (JUN04~SULFUR13.D)

15uV
~1

2.6 ppmelementalsulfurin CA Refgasoline1:12.5split

20O

" r

"1

AIB1B, (JUN04\SULFUR16.D)
15~1

10 pprnelementalsulfurin CA Refgasoline1:12.5sptit

400
2O0

, 10

lfi

?fl

.A

FIG. 4---Sulfur chromatograms of a California reference gasoline that had 2. 6 ppm (top) and
10 ppm (bottom) of elemental sulfur added. (GC Conditions: 1 IzL split1:25; 9.8 psig headpressure and 280~ injection temperature; column: HP-5, 30 m, 0.32 mm ID with a 0.25 pm
film; 50 ~ for I min programmed to 260 ~ at 8 ~
and held at the upper temperature for 4
min.)
Formation of black silver sulfide is visible on silver wool when 10 mg of silver wool was
placed in about 1.5 g of the reference gasoline containing 2.6 ppm elemental sulfur and heated as
described above. No visible formation of silver sulfide occurred with silver wool in the reference
fuel prior to the addition of elemental sulfur. This is illustrated in Fig. 5, where "A" represents
silver wool heated in the reference fuel versus "B," which represents silver wool heated in the
reference fuel that contained 2.6 ppm elemental sulfur. Even though the pictures in Fig. 5 were
obtained under magnification, the silver corrosion was readily apparent to visual inspection
without magnification. The surface of the silver wool became even darker when exposed to
higher levels of elemental sulfur as one would expect, and it appeared that the entire surface
became coated at about the 10-15 ppm level. Further experimentation was not pursued in this
area, as it was demonstrated that the more qualitative silver wool test and quantitative GC
analysis appear to be correlated. One could expect to find an optimum ratio between the amount
of sample and silver wool to maximize the visible contrast that is observed, however.
It is apparent that elemental sulfur species that are suspected to cause silver corrosion can be
detected at the appropriate levels that contribute to the fuel gauge issue. Of course, the ability to
make a rapid total sulfur measurement is extremely useful in many circumstances. For instance,
in a process plant upset, speciation of sulfur may be desirable only when a total sulfur
specification is exceeded. In those circumstances, it is desirable to be able to simply change
operating conditions back and forth between total sulfur and speciated sulfur analysis, and this is
achievable using LTMGC.

172

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 5--Silver wool corrosion; control sample ",4 "" (left) heated with reference fuel without
elemental sulfur and sample "B'" (righ 0 reference fuel containing 2.6 ppm elemental sulfur.
(Top ~ l Ox magnification; bottom approximately 40x magnification).
A standard containing 13 sulfur compounds, each in the range of 170-300 ppm, was used to
determine a number of analytical figures of merit including short-term precision and response as
shown in Table 1. The system demonstrated good short-term precision with most compounds
yielding a relative standard deviation of response less than 4 % RSD (Relative Standard
Deviation) for number of injections, n=4. The results also indicate that the system responds to
sulfur on an equi-molar basis within experimental error (:t:5 %), meaning that sulfur response is
equivalent for different sulfur compounds on a molar or mass sulfur basis. This is a desirable
property because it enables quantitative analysis of unknowns, and it allows for easier
calibration. A chromatogram for the 13 sulfur compounds that were used is shown in Fig. 6. The
results are particularly good considering that mercaptans and disulfides are very reactive.
Figure 7 shows a chromatogram of an unleaded gasoline sample obtained with the above
LTMGC DP-SCD system. This demonstrates that the system is amenable to sulfur analysis of
petroleum products, described by ASTM D 5623 [12]. By operating the LTM GC module
isothermally at a high temperature, e.g., 250~ instead of temperature programming the module,
it is possible to measure total sulfur as a single unresolved peak. It should be noted that peak
shape will be dependent on several factors, such as sample size, sample matrix volatility, the
column used, and the temperature of the module. In this part of the study, the focus was on light
hydrocarbons used in a chemical plant, for which the PLIS injection valve was a good choice.
The effect of split ratio of the GC injector was adjusted so that no hydrocarbon interference was
observed (at ca. 1:5). Figure 8 illustrates sensitivity, peak shape, and repeatability for a butane
standard that contained about 1.5 ppm total sulfur. The detection limits obtained with the LTM
GC in the total sulfur mode were between 20-30 ppb S for a signalto noise ratio of 3:1. Even
though variation of the peak height is readily apparent, peak area was used to calculate
repeatability. For the 15 injections, the area repeatability was 3.3 % RSD.

GRAS ET AL. ON DP-SCD AND LTMGC

173

TABLE 1--Short-term precision and response for sulfur compounds.

400000-

11

==
o

12

13

10

5771

A__JU U ~L._.J)U UL__/t


y~
2'
4'
0'
Time (min)

L
i

h..__..,,,.,_,
I0'

12'

FIG. 6---Standard chromatogram of 13 sulfur compounds in the range of 170-300 ppm


sulfur. The column was an 18 m, 0.25 mmid, 0.25 micron PDMS, programmed at 30~ for 2
rain to 200~ at 30~ ~rain; Split rate: 10:1, Helium carrier, 30 cm/sec [Peak identifications: (1)
Dimethyl Sulfide; (2) Isopropyl mercaptan; (3) Tert-butyl mercaptan; (4) Propyl mercaptan; (5)
Sec-butyl mercaptan; (6) lsobutyl mercaptan; (7) Diethyl sulfide; (8) Butyl mercaptan;
(9)Dimethyl disulfide; (10) Dipropyl sulfide; (11) Diethyl disulfide; (12) Diisopropyl disulfide;
(13) Dipropyl disulfide].

174

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Benzothiophene

2&3-Methylthiophene=
t-0

Methylbenzothiophenes
Thiophene

5804

lo

Time (rain)

FIG. 7--Sulfur chromatogram of an unleaded gasoline sample obtained with the LTM GC
SCD system operated in a fast chromatographic mode.
6500

5400.
0.~

~.5'
3.0'
Time (min)

3.5'

4.0'

4.B'

~.O'

FIG. 8--Sulfur chromatogram of an unleaded gasoline sample obtained with the LTMGC
SCD system operated in afast chromatographic mode.
Figure 9 demonstrates that a linear response to total sulfur is obtained with the system over a
concentration range of ca. 1.2 ppm to about 1200 ppm. The linear least squares analysis
produced a correlation coefficient > 0.9999, and the y-intercept was essentially zero. Lack of a yintercept is a good indicator that there is no positive or negative interference. Still, caution
should be exercised when making measurements close to the detection limit because of the
difficulty of the potential for interferences and contamination at such low levels. If one is going
to make measurements in that region, then it would be advisable to check for interferences and
contaminants using the standard addition technique, for example. Keep in mind that not only is it
possible for "pure" solvents to contain trace levels of sulfur, but the possibility of sulfur

GRAS ET AL. ON DP-SCD AND LTMGC

175

adsorption and even sulfur release from surfaces could affect analytical results. Several
injections were made at each concentration level. Nevertheless, in this experiment good
repeatability was observed over the entire concentration range, for example 2.2 % RSD at the
120 ppm level. These results are particularly good considering the volatile nature and difficulty
of sampling this liquefied butane matrix.

FIG. 9--Plot of sulfur response as a function of concentration for a liquefied butane matrix
over the range of approximately 1.2-1200 ppm PLIS injection valve and Dual Plasma SCD with
LTMGC module at 250~
Discussion

The main advantage of using the LTMGC approach is that the LTM can easily give
speciated or total sulfur (single peak) results without changing hardware. That is, only the
temperature programming profile of the LTMGC is changed for the particular type of analysis.
This does not preclude the use of a standard GC for total sulfur or speciated sulfur analysis. In
fact, obviously a standard GC is designed for speciated analysis, but it can also be used for
dedicated total sulfur analysis as well. This can be accomplished with an uncoated fused silica
tube, as described previously, attached to a conventional split/splitless injector. Again, no
interference to hydrocarbons was observed with the dual plasma burner (injection size, 1 ~tL;
split 1:10), and detection limits and speed of analysis are similar to those obtained with LTM
GC. Using an auto-sampler, short-term precision for ppm levels of sulfin: is similar to that
obtained for a typical GC analysis, 1-2 % RSD. These results are acceptable for a typical
laboratory analyzer. Stability and longer-term precision were evaluated using a gasoline fuel that
contained ca. 32 ppm sulfur, making an analysis approximately every 20 rain over a period of
100 h. Individual data points collected over this period are plotted in Fig. 10. Stability of
response and precision for the above gasoline was about 1.7 % RSD over the entire test period.
These results compare favorably with other accepted methodologies, such as combustion UV
fluorescence, ASTM D 5453 [12]. This approach is attractive, especially because of the wide
availability of GC instrumentation and its relative low cost.

176

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. lO--Stability of response to ca. 32 ppm sulfur gasoline fuel over 100 h using a
conventional GC configured for total sulfur analysis.
A fast sulfur speciation can also be achieved with a conventional GC by sacrificing some
resolution using a short, e.g., 5 m, 0.32 mm ID 4 #In PDMS column. This is illustrated in Fig. 11
as an isometric plot of a six replicates of a gasoline sample that contained approximately 32 ppm
total sulfur. Benzothiophene is eluted at about 5.2 min in the "fast" mode compared to about 23
min using conventional conditions. Of course, this gain in speed results in some loss in
chromatographic efficiency as readily seen by the loss of resolution of the methylbenzothiophene isomers. The precision of total sulfur values obtained by this technique was
about 1.6 % RSD for number of injections, n = 6. These results illustrate the applicability of the
technique for low sulfur gasoline that meets most current regulations.
150

j~t

.n
o 400

10X Dilution

Oveday of 6
Replicate Injections

0 .O

1.0

2.0

'

3.0
Time

4.0

5.0

8.0

7.0

8.0

- Minutes

FIG. 11--Isometric plot of six replicates of the California reference gasoline sample that
contained ca. 32 ppm total sulfur. The small insert is the response to ca. 3 ppm total sulfur
obtained by 10 x dilution of the reference sample.

GRAS ET AL. ON DP-SCD AND LTMGC

177

To demonstrate the potential for application at even lower level sulfur fuels, the California
reference gasoline was diluted by 10 times in a "sulfur-free" toluene solvent and re-analyzed.
Results from this test are illustrated in the small insert of Fig. 11. The results show that this
technique is capable of detecting and speciating sulfur at levels even lower than the 5 ppm total
sulfur level. On the other hand, it is likely that regulations may mandate sulfur concentrations in
fuels at even lower levels. The insert also illustrates that the dilution of the sample to ~90 %
toluene has little effect on the sulfur distribution, and clearly no hydrocarbon interference is
observed. Loss of resolution of the 2 and 3-methylthiophene isomers (eluted just prior to 2 min)
was observed due to a chromatographic "solvent effect," but otherwise this did not significantly
impact the results.

Summary
Gas chromatography with sulfur chemiluminescence detection has become a useful
technique to help refiners meet their sulfur challenges by providing sulfur speciation and
analysis. This work shows the potential for extending a new version of the SCD (Dual Plasma
technology) to chromatographic speciation and total sulfur detection in at least two modes using
a LTM GC or conventional GC as a single sulfur peak measurement approach. In addition, this
work shows that both of these approaches are able to measure the low sulfur levels required
today, and they are capable of meeting even lower sulfur measurement requirements that are
anticipated. Future work involving this technology will be directed toward lower concentration
ranges, near the detection limits of 20-30 ppb to about 100 ppb, and examining the technology
for heavier sample matrices, such as diesel fuel. Significant findings include:
9

The DP-SCD demonstrates excellent sensitivity and selectivity (which is frequently more
important) for measurement of sub-ppm levels of sulfur.
9 This technique is able to detect the major elemental sulfur allotropes at levels that
contribute to silver corrosion involved in the fuel gauge issue.
9 A LTMGC module can be successfully used with this sulfur detection technology in both
a chromatographic and non-chromatographic mode for trace sulfur detection. The use of
LTMGC is attractive because, at least for volatile matrices, the same hardware without
physical modification can be used to make chromatographic and non-chromatographic
sulfur measurements.
9 The technique for a non-chromatographic total sulfur measurement using a conventional
gas chromatograph has also been demonstrated.

Acknowledgments
The authors wish to acknowledge the support of Dr. Mary Fairhurst, contributions and
helpful discussions from Myron Hawryluk, Dr. Joaquin Lubkowitz, Paul Cecil, Dr. Bruce
Quimby, and others.

References
[1] Title 13, California Code of Regulations, Motor Vehicles Division, Section 2281 and 2282, 2
Rules.

178

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

[2] Title 40, CFR, Part 80, Compilation of Federal Regulations of Fuel and Fuel Additives,
Including Diesel Fuel.
[3] Nadkarni, R. A. K., "Determination of Sulfur in Petroleum Products and Lubricants: A
Critical Review of Test Performance," American Laboratory, Vol. 32, 2000, pp. 16-22.
[4] Special session to address reactive sulfur in gasoline, ASTM D2 Meeting, June 2004, Salt
Lake City, UT.
[5] Gras R. L., Luong, J. C., Shearer, R. L., and Mustacich, R.V., "Dual Plasma Sulfur
Chemiluminescence Detector and Low Thermal Mass Gas Chromatography (LTMGC) for
ppb Level of Sulfur Analysis in Hydrocarbons," Poster 1.17, June 2, 2004, the 27 th ISCC
Conference Proceedings, Rival Del Garda, Italy.
[6] Benner, R. L. and Stedman, D. H., "Universal Sulfur Detection by Chemiluminescence,"
Analytical Chemistry, Vol. 61, 1989, pp. 1268-1271.
[7] Hutte, R. S., Johansen, N. J., Plato, M., and Legier, M. F., "Column Selection and
Optimization for Sulfur Compound Analyses by Gas Chromatography," Journal of High
Resolution Chromatography, Vol. 13, 1990, pp. 421-426.
[8] Shearer, R. L., "Development of Flameless Sulfur Chemiluminescence Detection:
Application to Gas Chromatography," Analytical Chemistry, Vol. 64, 1992, pp. 2192-2196.
[9] Shearer, R. L., Poole, E. B., and Nowalk, J. B., "Application of Gas Chromatography and
Flameless Sulfur Chemiluminescence Detection to the Analysis of Petroleum Products,"
Journal of Chromatographic Science, Vol. 31, 1993, pp. 82-87.
[10] Chawla, B. and Di Sanzo, F., "Determination of Sulfur Components in Light Petroleum
Streams by High-Resolution Gas Chromatography," Journal of Chromatography, Vol. 589,
1992, pp. 271-279.
[11] Di Sanzo, F. P., Bray W., and Chawla, B., "Determination of the Sulfur Components in
Gasoline Streams by Capillary Gas Chromatography with Sulfur Chemiluminescence
Detection," Journal of High Resolution Chromatography, Vol. 17, 1994, pp. 255-258.
[12] Shearer, R. L. and Meyers, L. M., "Simultaneous Measurement of Hydrocarbons and Sulfur
Compounds using Flame Ionization and Sulfur Chemiluminescence Detection for Sulfur
Simulated Distillation," Journal of High Resolution Chromatography, Vol. 22, 1999, pp.
386-390.
[13] Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA. USA.
[14] Modern Practice of Gas Chromatography, 4th Addition, John Wiley and Sons, Grob, R. L.
and Barry, E. F., Ed., New York, USA, 2004.
[15] United States Patent 6,217,829, Mustacich, R. V., et al., "Reduced Power Consumption Gas
Chromatograph System," April 17, 2001.
[16] Luong, J., Gras, R., and Tymko, R., "Innovations in High Pressure Liquid Injection
Techniques for Gas Chromatography: Pressurized Liquid Injection System," Journal of
Chromatographic Sciences, Vol. 41, 2003, pp. 550-559.

MERCURY DETERMINATION

Journal of ASTM International, October 2005, Vol. 2, No. 9


Paper ID JAI12985
Available online at www.astm.org

S. Mark Wilhelm, Ph.D.,1 David A. Kirchgessner, Ph.D., 2 Lian Liang, Ph.D., 3 and Peter 1t.
Kariher 4

Sampling and Analysis of Mercury in Crude Oil


ABSTRACT: Sampling and analytical procedures used to determine total mercury content in crude oils
were examined. Three analytical methods were compared with respect to accuracy, precision, and
detection limit. The combustion method CLT.S.EPA 7473 hybrid) and a commercial extraction method
(non-standard) were found adequate to provide a good combination of sensitivity and accuracy, while
instrumental neutron activation analysis was found to suffer from interferences from elements other than
mercury (Ni, Se) that are typically found in crude oil. Aliquot removal and processing steps were found to
be important to accuracy and precision of the combustion method. Potential errors caused by incomplete
homogenization and loss of volatile mercury were investigated in detail. Homogenization of oils that
contain suspended mercury compounds was found important to allow retrieval of representative aliquots
for analysis. Ultrasonication at slightly elevated temperature yielded the more reproducible and higher
magnitude results. It was discovered that volatile mercury can be lost from sample containers when they
are opened to remove aliquots. Alternative sample containers and aliquot removal techniques were
developed that minimized loss of volatile mercury. Speciation experiments were utilized to determine the
identity of volatile species and the rates at which they are lost from containers, Both elemental mercury
and dimethylmercury were identified as volatile mercury components of one freshly sampled crude oil.

KEYWORDS: petroleum, mercury, oil, volatile mercury, speciation, refining


Introduction

The United States Environmental Protection A g e n c y (U.S. EPA), in cooperation with the
American Petroleum Institute (API) and the National Petrochemical and Refiners Association
(NPRA), has initiated a project to determine the mercury content o f crude oil processed in the
U n i t e d States. The focus o f the project is to determine the m e a n concentration and range o f
concentrations o f total mercury in crude oil accurately and with statistical justification. Data
generated in the project will be used to calculate the m a x i m u m potential contribution o f crude oil
mercury to anthropogenic mercury emissions in the United States.
Samples o f crude oil are acquired at locations immediately upstream o f refinery tank farms
and analyzed for total mercury concentration (THg). Data are generated in the form o f total
m e r c u r y concentration in liquid aliquots (THg, gg/kg) o f market-named oil streams obtained at
particular points in time. The project is divided into three parts. Part 1 involved comparison o f
analytical methods and has been reported previously [1]. Part 2 involved investigation o f
sampling and analysis procedures and is reported here. Part 3 is in progress and involves
sampling and analysis o f the majority o f crude oils processed in the U.S. in years 2003-2005.

Manuscript received 1 October 2004; accepted for publication 9 March 2005; published October 2005. Presented at
ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
i Principal Scientist, Mercury Technology Services, 23014 Lutheran Church Rd., Tomball, TX 77377.
2 Senior Research Scientist, U.S. EPA, Mail Drop E305-02, Research Triangle Park, NC 27711.
3 Principal Scientist, Cebam Analytical, Inc., 3927 Aurora Ave. N, Seattle WA 98103.
4 Research Scientist, ARCADIS, 4915 Prospect Dr., Durham, NC 27713.
Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

181

182

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Data for the oil ensemble and their statistical treatment will be reported at the completion of the
project.

Technical Background
One goal of the API/EPA/NPRA project is to obtain an estimate of the amount of mercury
that may be emitted to the environment from crude oil processed in the United States. The
analogous question for coal has been answered [2]. The average concentration of mercury in
U.S. coal is approximately 100 gg/kg. U.S. coal combustion contributes on the order of 70 tons
of mercury to the atmosphere annually or about 60 % of the total amount from all U.S.
combustion sources [3]. It is equally important to understand the concentration of mercury in
crude oil because the mass of petroleum processed in the U.S. annually is roughly the same as
the mass of combusted coal (1012 kg) [4].
Most mercury in coal is substitutional in pyrites found in the coal matrix, and thus coal
presents mostly non-volatile species to the analyst [2]. ASTM Method D 6722-01 (Standard Test
Method for Total Mercury in Coal and Coal Combustion Residues by Direct Combustion
Analysis) and U.S. EPA Method 7473 (Mercury In Solids and Solutions by Thermal
Decomposition, Amalgamation and Atomic Absorption Spectrophotometry) were developed
primarily for analysis of mercury in coal and are now widely practiced. A number of mercury
species have been identified in crude oil (Table 1) [5]. Some evidence exists that other species
may also be present in crude oil, but those listed in the table are thought to encompass all the
major species. Of those listed, elemental mercury and mercuric sulfide dominate the others.
Mercuric sulfide is a generic term that really means mercury bonded to sulfur such that the
resultant compound is thermally stable and non-volatile in the primary distillation of crude oil.
Actual (meta-) HgS has been identified in crude oil conclusively, but other Hg-SR species, where
R is an organic substituent, are likely present as well. Aside from mercuric sulfide, most of the
mercury species are volatile to one degree or another. Ionic mercury is not thought to be a major
species in the reservoir because elevated mercury concentrations are not measured in produced
water associated with either crude oil or gas condensates having high concentrations of total
mercury. The prevalence of dialkylmercury (DMHg) in oil is a matter of considerable debate, but
DMHg is thought to be a very low percentage of the total mercury content of crude oil and gas
condensates based on existing data.
Historically, a variety of methods has been applied to measuring the total mercury
concentration in crude oil. Notable in the evolution of technologies is the early work of Filby and
colleagues [6,7] with instrumental neutron activation analysis (INAA), Bloom [8] with
extraction/cold vapor atomic fluorescence (CVAF), Liang [9] with combustion - CVAF, and
Kelly [10] with isotope dilution. The only standard method for mercury in crude oil is a method
developed by Universal Oil Products identified in industry as UOP 938. UOP 938 is a
combustion-amalgamation method specific to a particular instrument (Nippon Instruments
Corporation). Mercury in crude oil historically has been routinely determined using some form
of digestion followed by cold vapor atomic absorption (CVAA), but mercury losses in the
digestions and extractions have been reported [I 1]. More recently, many industry analysts have
adopted some variation of the combustion/amalgamation method because of its relative
simplicity and demonstrated accuracy for mercury in a variety of hydrocarbon matrices.
The initial experiments conducted in the EPA/API/NPRA project examined analytical
method and compared INAA, combustion/CVAF, and extraction/CVAF [1]. Three separate
laboratories participated, each analyzing a series of constructed single-species standards using

WILHELM ET AL. ON CRUDE OIL

183

dodecane or toluene as the oil surrogate. It was found that certain mercury species were
problematic to routine analysis. Suspended HgS, for example, was difficult to homogenize in
samples, and thus the analysis o f a liquid suspension of HgS in mineral oil was not reproducible.
Likewise, problems were encountered when attempting to measure solutions o f elemental
mercury in dodecane and toluene leading to the experimental work reported here.
TABLE 1--Properties of mercury and mercury compounds found in petroleum. _
Property
Hg~
HgC12
HgS
CH3HgC1
(CH3)2H~__
Melting
-39
277
584
167
Point (~
(sublimes)
(sublimes)
""
Boiling
357
303
96
Point (~
Vapor
0.180
0.899
n.d.
1.76
8.3 x 103
Pressure (Pa)
(20~
(20~
(25~
(25~
Water Solubility
49.4 x 10~
66
2 x 10-24
5-6
2.95
(gL")
(20~
(20~
(25~
(25~
(25~
Alkane
2 x 10-3
-2.5 x 10-z
Solubility (gkg"l)
(20~
(25~
miscible
Henry's Law
729
3.69 x 10-5
1.6 x 10"5
646
Coefficient
...
(Pa mamol~)
(20~
(20~
(15~
(25~
Octanol - water
4.2
0.5
n.d.
2.5
180
partition ratio
Crude oil is known to contain volatile mercury because mercury is found concentrated in the
liquefied petroleum gas (LPG) and naphtha fractions of the atmospheric distillation when oil is
refined. It is known conclusively that elemental mercury is at least one o f the volatile mercury
species in crude oil because it sometimes is found condensed in trays in refinery distillation
towers and condensed in cryogenic heat exchangers that liquefy petroleum gases. What is not
known is whether elemental mercury is the only volatile species.
The rate at which volatile mercury evaporates from samples o f crude oil is much greater than
most analysts realize, leading to the possibility o f a prevalent low bias in oils routinely analyzed
for total mercury. The logical path that leads to this conclusion is the principal subject o f the
experimental work to be discussed.

Experimental
Three independent laboratories, each using a different method, were selected to analyze
crude oil samples. One objective o f the project was to identify methods and procedures that
could be used by industrial or academic laboratories to analyze crude oil for mercury in a
practical fashion. The practicality of a method is a function o f sample processing steps, reagent
availability, instrumental analysis time, instrument expense, detection limit, and accuracy o f
result. Crude oils exhibiting total mercury concentrations (THg) below 1-2 ~tg/kg are not
problematic to refining, but some specifications for refined products (naphtha) are set close to
1 ~tg/kg; thus a method detection limit (MDL) o f 0.1-0.3 ~tg/kg was deemed acceptable for
routine quality assurance (QA) purposes. Differentiating mercury concentrations below 1 ~tg/kg
may have academic utility, but an MDL o f 10 ng/kg (parts per trillion) is not necessary for oil
quality determinations or for assessment of atmospheric emissions attributable to petroleum.

184

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

LAB 1 employed the extractive method developed by Bloom [8] in which an aqueous
oxidant, BrC1, is used to oxidize and thus extract all mercury species in an aliquot without
chemical decomposition of the oil matrix. Mild heating (100~
was required to promote
complete oxidation o f mercury species. The mercuric ion in the aqueous extract was treated in
the same manner as acid digests that are measured using U.S. EPA Method 1631. The sequence
o f steps involved reduction of aqueous phase mercuric ion using stannous chloride (SnC12),
sparging o f elemental mercury produced by the reduction step, and trapping o f the mercury
vapor from sparging on gold (Au). The amalgamated mercury was then thermally desorbed from
the gold amalgam and detected using cold vapor atomic fluorescence (CVAF). Bloom developed
these extractive techniques for operational speciation of mercury in crude oils and condensates
[8]. Data quality parameters obtained separately from this project are shown in Table 2.
References [1] and [8] provide details on the operational speciation procedures also utilized in
this project (see Table 5 and discussion).
TABLE

2--Data qualityparametersfor Hg extractions in petroleum samples [8].

Mercury Concenlrations, lag/kg(ppb)


Parameter
Total*
Hg(II)
Hg~
Mean method blank
0.10
0.007
0.16
SD blank
0.02
0.003
0.01
N blanks
4
4
4
eMDL at blank
0.06
0.01
0.04
Low spike level
0.30
0.12
0.26
Mean low spikes
0.24
0.07
0.24
SD low spikes
0.05
0.03
0.03
N low spikes
7
8
8
Measured MDL
0.17
0.10
0.09
Matrix spike level
487
14.3
93.8
N matrix spikes
4
4
4
Mean spike recovery
100.3
93.1
98.7
% recovery range
98-102
76-101
94-103
* Same figures apply to total dissolvedHg.
t Results are matrix spike recovery averagecorrected (63.7 % recovery).

CH3Hgt
0.001
0.005
4
0.02
0.58
0.47
0.09
8
0.25
16.87
4
105.5
99-109

LAB 2 used a combustion method developed and constructed in-house [9,12]. The
combustion-trap system was composed o f a quartz temperature controlled combustion column, a
gas scrubbing section, and gas flow apparatus. The combustion column was divided into two
independently controlled heating zones. The initial segment vaporized the sample, and the
second segment combusted the hydrocarbon. Liquid aliquots were injected directly into
vaporization section o f the instrument as opposed to using a ceramic boat as in UOP 938 or U.S.
EPA 7473. The carrier gas was air purified to remove any ambient mercury using a gold-coated
sand trap. Mercury vapor exiting the combustion chamber was scrubbed to remove partially
combusted hydrocarbon and then collected on gold traps (double amalgamation). Mercury vapor
detection was by cold vapor atomic fluorescence (CVAF).
LAB 3 employed INAA and used procedures developed by Filby and Musa [6,13,14]. Oil
aliquots were irradiated in a uranium fission reactor, and the flux and energy spectrum of emitted
gamma radiation was measured. Mercury concentrations were determined by comparison to
standards that were irradiated at the same time as the samples. Oil aliquots were removed from
sample bottles by pipette and placed in quartz vials that were then torch-sealed prior to
irradiation.

WILHELM ET AL. ON CRUDE OIL

185

When analyzing actual crude oil samples, the INAA method suffers from interferences
caused by nickel and selenium. Total mercury was determined by correction for the selenium
interference as discussed by Filby and Shah [7]. The principal 7 photopeak of 80Hg2~ overlaps
the 7 photopeak of selenium (345e75). The quantification of 80Hg2~ was accomplished by
subtracting the 279.6 keV 348e 75 overlap of the 279.2 keV 80Hg2~ line with knowledge of the
348e 75 264.6/279.2 keV intensity ratio. The 34Se75 264.6 keV gamma line is not subject to
interference and was determined independently. The 34Se75 264.6/279.2 keV intensity ratio is
fixed by isotopic abundance and known independently. This correction procedure provided
reproducible results in samples containing both Hg and Se, but the resultant detection limits were
elevated substantially. Nickel in oil was likewise problematic to mercury determination INAA
due to a general increase in background 7 radiation levels. The combination of the Se and Ni
interferences produced high detection levels for oils that contained either or both nickel and
selenium.
Three crude oils were selected for analysis. The oils were selected to demonstrate a range of
density (20-50 API gravity), sulfur content (0.05-3.5 wt. %), and metals content (Ni + V; 0.05300 ppm). Grab samples of crude oil were acquired from sampling stations at the terminus of
pipelines to refineries or at stations adjacent to tanker unloading facilities. Initially, oil samples
were obtained in the field by mostly filling 125 ml glass bottles. The headspace in each bottle
varied, but was generally in the range of 15-30 ml (10-25 % of sample bottle volume).
Experimental work conducted separately from the project identified borosilicate glass as the
most suitable sample container material. Somewhat ironically, the industry standard for crude oil
samples historically has been tin-plated metal cans because metal cans are necessary for
compliance with air transport regulations. Because mercury amalgamates with tin, the
concentration of mercury measured in oil contained in a tin-plated metal can is biased (very) low.
The loss of elemental mercury in any type of polymeric container has been observed in several
studies [8].
Results and Discussion

Issues related to variations in measured THg associated with actual crude oil samples
acquired using field sampling techniques from actual pipelines and from tankers during
discharge to refinery tanks were examined. The items of concern were organized by task with
each task designed to answer specific questions. Tasks were added and analytical procedures
changed as data were acquired in the course of the experimental work.

Method Comparison
Total mercury MDLs, as defined by U.S. EPA [15], were determined by analysis of reagent
dodecane using each laboratory's standard procedures, which differed because of instrumental
differences. Data are recorded in Table 3. LAB l ' s extractive method was blanked using
extraction reagents and reflects reagent background. Subsequent analyses of dodecane by LAB 1
were blank corrected. LAB 2's blanking method was the detector response from a combustion
cycle and reflected instrument and carder gas cleanliness. LAB 2's MDL determinations using
dodecane are not blank corrected. LAB 3's MDL was determined from the background emission
of irradiated vials.
Because of the volatile nature of crude oil and the associated safety restrictions on
transportation of samples, significant time intervals can exist between when oil is sampled and

186

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

when samples are analyzed in laboratories. One hypothesis concerning sample stability was that
certain species o f mercury could partition to bottle walls or sediment irreversibly over time. This
would result in lower than actual measured concentrations, if the adsorbed mercury could not be
redistributed to the liquid phase by a simple homogenization procedure (heat and shake).
Secondly, it was thought that volatile mercury could be lost when sample bottles were opened.
Lastly, it was thought possible that species' conversion by oxidation due to exposure to air might
occur, which would produce a redistribution o f species being more inclined to attach to bottle
wails or sediment and thus become elusive to aliquot removal. All of these hypotheses would
translate into measured amounts o f mercury in aliquots that would be sensitive to
homogenization procedure or frequency of bottle openings, and thus concentrations could
diminish over time or in successive analyses.
TABLE 3--MDL determination,
LAB 1
LAB 2
(/ag/kg)
____(Egz/kg) _ ( ~ g / k g )
(~g/kg)_
Method
Method
Dodecane*
Dodeeane*
Blank
Blank
0.22
0.01
<0.11
0.095
0.47
0.03
<0.11
0.182
0.15
-0.01
<0.11
0.153
0.18
0.00
<0.11
0.100
0.22
0.00
<0.11
0.158
0.11
-0.03
<0.11
0.172
0.11
-0.07
0.172
-0.03
0.206
-0.04
0.195
0.21
-0.02
0.159
Meall
0.12
0.03
0.039
SD
0.14
0.11
MDL
*Corrected for method blank.
tNot corrected for methodblank.

LAB 3
(~tg/kgQ____
Method
<2.3
<2.4
<1.3
<1.6
<1.6
<1.1
<1.1
<1.3
<1.1
~2

Data for the method comparison study are reported in Table 4. Method performance and
time-dependant variation o f measured total mercury concentrations were compared using three
field-sampled crude oils (US4, US19, and MX4) that were selected based on crude oil assays of
gravity, metals, and sulfur content. Crude oils were sampled in the field according to a uniform
sampling protocol (125 ml glass bottles with perfluorocarbon lined lids; three grab samples not
composited; 25 mL headspace in sample bottles). Individual sample containers were distributed
to selected laboratories within one week o f when the samples were taken. Replicate analyses of
oil from a single sample container from the first sample event were performed at each of the
three laboratories (LABs 1, 2, 3) repetitively at approximately one month intervals. For each
analysis, triplicate aliquots were removed from samples that had been warmed (30-40~ and
manually shaken prior to bottle opening.
The first sample event was repeated three additional times. In Table 4, the sample event is
indicated by the number after the oil code. Sample events did not follow a time sequence but
were independent. Independence was defined as samples taken fi'om different tankers or at least
two weeks apart for oil delivered by pipeline.

WILHELM ET AL. ON CRUDE OIL

187

TABLE 4--Method comparison-total mercury in commercial crude oils.


Sample ID

LAB 1
(SD) 0tg/kg)

LAB 2
LAB 3
(SD) (~g/kg) (SD) (pg/kg)

Mean
(~g/kg)

At
(days from sampling)
LAB 1/2/3
46/41/43
73/67/78
99/97/121
_
42/45/46
69/71/81
78/101/124
16/11/5
42/37/40
68/67/53
27/23/32
27/23/35
19/19/14
27/23/12
23/29/41
67/67/56

US4-1*
1.8 (0.7)t
2.4 (0.4)
<10s
2 (mean labs 1, 2)
US4-1
0.7 (0.2)
2.4 (0.1)
10
4
US4-1
2.2 (0.2)
2.0 (0.I)
<10
2 (mean labs 1, 2)
US19-1
33.9 (0.6)
31.4 (0.6)
41 (2)
35
USI9-1
36.1 (2.5)
32.8 (2.8)
44 (5)
38
US19-1
33.6 (3.3)
33.3 (1.0)__
50 (3)
39
MX4-1
0.1 (0.5)
4.5 (0.5)
<20
2 (mean labs 1, 2)
MX4-1
1.1 (1.2)
1.7 (0.1)
<20
1 (mean labs 1, 2)
MX4-1
1.1 (0.5)
1.7 (0.1)
<20
1 (mean labs 1, 2)
US19-2
8.2 (0.6)
11.2 (0.6)
9 (4)
9
US19-3
31.3 (5.6)
35.9 (1.1)
34 (3)
34
US19-4
32.5 (2.6)
34.3 (1.7)
48_(_3)
38
MX4-2
-0.3 (0.6)
9.8 (1.3)
<20
5 (mean labs 1, 2)
MX4-3
3.0 (0.7)
5.1 (0.4)
<20
4 (mean labs 1, 2)
MX4-4
1.4 (0._5)
2.1 (0.4)
<20
2 (mean labs 1, 2)
Homogenization
40
40
40
T (~
Aliquot Taken
Hot
Warm
Warm
* US4 is the code name for the oil. Numbers after the dash refer to sampling events. For example, US4-1 refers to
oil acquired in the first sampling event. Three grab samples were obtained at each sampling event.
t Reported data are the mean of triplicate analyses. Standard deviations are in parenthesis.
< indicates less than MDL due to interference.
The 1NAA method suffered from interferences caused b y nickel and selenium discussed
previously. The selenium correction procedure provided reproducible results in samples
containing both Hg and Se, but the resultant detection limits were substantially elevated. Nickel
in oil was likewise problematic due to a general increase in background ~ radiation levels. The
combination o f the Se and Ni interferences produced high detection levels for oils that contained
either or both nickel and selenium. For these reasons the I N A A method was not used in any
additional tasks.
The variation in mercury content o f oils having the same trade name was examined. Samples
were obtained from independent sampling events. Independence was defined as different tankers
or at different times from pipelines (weeks apart). Two o f the three oils examined were found to
exhibit significant variation in mercury concentration from one sample event to another. The
US19 oil exhibited concentrations o f roughly 38, 9, 34, and 38 Ixg/kg (average of 3 laboratories)
in 4 separate sampling events and the MX4 values o f 2, 5, 4, and 2 ~tg/kg (average o f 2
laboratories). The variation in mercury concentration o f market-named oils over time was
important to project objectives, in that, i f the temporal variation o f mercury content in oils was
found to be large, the number o f individual samples required to define an exact mean and range
for a particular market-named oil likewise would be large. It was not surprising that oil
concentrations were found to vary from tanker to tanker and in pipelines given the blending o f
oils that typically takes place in production systems.

188

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Speciation
Speciation and bottle wall determinations (Table 5) were performed using procedures
developed by Bloom [8]. Fresh aliquots were removed from the same sample bottles used for
method comparison experiments (Table 4). Data in Table 5 are independent. Separate aliquots
were used for each determination.
TABLE 5--Speciation data (LAB 1).
Dissolved
Suspended
Hg2
Hg~
Walls
Sample ID
(~tg/k~)
(btg/kg)
0-tg/kg) (lag/kg) (~tg/kg)US4-2 rep 1
1.4"
0.0
1.0
0.2
0.06
US4-2 rep 2
1.6
-1.2
1.0
-0.4
mean
1.5
-0.6
1.0
-0.1
SD
0.1
0.8
0.0
0.4
0.04
US19-2 rep 1
3.3
4.6
2.4
-0.4
US19-2 rep 2
2.9
5.1
2.6
-0.2
mean
3.1
4.9
2.5
-0.3
SD
0.3
0.4
0.1
0.1
MX4-2 rep 1
0.6
-1.1
-0.2
-0.5
0.07
MX4-2 rep 2
1.2
-0.9
-0.2
0.2
mean
0.9
-1.0
4).2
-0.2
SD
0.5
0.1
0.0
0.5
* All data corrected for method blank.
The speciation measurements did not find volatile mercury (assumed to be Hg ~ [8]) in any of
the samples. Mercury in samples consisted o f ionic and suspended forms. The fact that low THg
concentrations were found in the oils examined was unfortunate, given the goal of understanding
species effects. Partition o f mercury to bottle walls was not discovered to any significant degree
for any of the three oils examined.
The partition of mercury to sample bottle walls is a known complicating effect to
measurement o f THg in some crude oil samples based on published investigations [8]. It
certainly may be the case that some oils demonstrate container wall adsorption to a much greater
extent than that discovered in the small number o f oils examined in these experiments. Likewise,
it was suspected that any volatile mercury may have been lost or oxidized by the time the
speciation experiments were performed. The samples used for the speciation experiments had
been opened several times for analysis and density measurements prior to speciation
experiments.

Sample Homogenization
For the data reported in Table 4, samples ofoil in 125 ml bottles were warmed, shaken, and
then aliquots were removed using a syringe (LAB 2) or micro-pipette (LABs 1, 3). The
temperature at which the aliquot was removed varied between hot (40~ LAB 1) to an
intermediate temperature between 40~ and ambient (warm, LABs 2 and 3). It was thought that
procedures used to remove hot aliquots may have provided opportunities for loss o f volatile
mercury or may have allowed reactions to take place possibly with oxygen in air when bottles
were opened for aliquot removal.

WILHELM ET AL. ON CRUDE OIL

189

Variations in measured THg were examined as a function of the manner in which samples
were homogenized prior to removal of aliquots. LAB 2 examined the homogenization procedure
by measuring concentrations in unheated samples followed by homogenization at 40~ and vice
versa. As opposed to the situation associated with data in Table 4, the sequence of experiments
started with a sample bottle that had not been opened previously. Care was taken to minimize the
opportunity for evaporation by removing aliquots quickly and by closing containers between
aliquot removals. Data are reported in Table 6. Samples analyzed first exhibited higher
concentrations, irrespective of whether the samples were heated (40~
or not. Close
examination of quality assurance data and standard deviations obtained from replicate analyses
suggested that the downward trend in measured concentrations was real and not the result of
random variation.
..

TABLE 6--Homogenization procedure (LAB 2).

.....

Sample

Homogenization Temperature (~
25~
80/40~
40~
40oc
(lSt
o ~ _ g _ ~ 0 / 0 2 _ ) _ _ _ _ ( 2 "d opening - 7 / ~ o E e _ n i n g
- 8/16/~10/15/02)
Otg/kg)
(~tg/kg)
(g/kg)
(gg/kg)
US4-1
5.6 (0.2)*
2.4 (0.4)
2.4 (0.1)
2.0 (0.1)
MX4-1
12.4 (1.5)
4.5 (0.5)
1.7 (0.1)
1.7 (0.1)
US19-1
51.0 3~.4)
31.4 (0.6)
32.8 (2.8)
33.3 ( 2 . 9 )
40~
25~
. _ _ _ _ _ ( 1 ~topem~g 8/16/02)
(2~do~_nin~ 8/16/02)
MX4-2
9.8 (1.3)
2.5
US19-2
11.2 (0.6)
7.5
* Reported data are the mean of triplicate THg analyses. Standard deviations are in parenthesis.

The reason for the observed decline in measured mercury concentration in successive
measurements is thought to lie in the fact that volatile (elemental) mercury in the headspace
(25 ml) of the sample bottle is lost when the bottle is opened. When the bottle is closed, volatile
mercury in the liquid phase redistributes to the headspace of the bottle, and thus the first aliquot
reflects a higher concentration (in the liquid phase) than aliquots obtained in successive
removals. If volatile mercury is lost when bottles are opened, one would expect the measured
concentration in liquid aliquots to diminish the more times a bottle was opened until most of the
volatile mercury had escaped and only non-volatile mercury forms remained. The data suggest
that only a relatively small number of bottle openings is sufficient to deplete a significant portion
of the volatile mercury in the sample bottle.
The homogenization procedure was further examined using an oil sample of HgS suspended
in paraffin oil. This sample suffered .from inconsistent results when analyzed previously [1].
Lack of reproducibility was attributed to the inability to produce a homogenous HgS suspension
prior to aliquot removal, at least using simple techniques (heat and shake). Table 7 demonstrates
that ultrasonic treatment is much more effective to homogenize samples than a heat and shake
approach.

Headspace Mercury
Loss of volatile mercury as might occur due to partitioning of volatile mercury to the
headspace of sample containers was examined. The term "headspace" refers to the portion of the
sample bottle not occupied by liquid. The hypothesis was that the total initial measured

190

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

concentration o f mercury in a freshly taken and homogenized sample was closer to the actual
sum of volatile and non-volatile mercury components in the sample and that volatile mercury
components can escape when bottles are opened.
TABLE
Aliquot Conditions
20~ shake
40~ shake
30~ ultrasonic
30~ ultrasonic

THg n g / g
(lag/kg)
16.9
611
1312
1470

7--Homogenization procedure (LAB 2).


Duplicate
(~tg/kg)
7.9
397
1361
1414

Mean
(I.tg/kg)
12.4
504
1338
1442

RPD
72
42
2
2

Target
(~tg/kg)
1659
1659
1659
1659

Reanalysis
Recovery(%)
1
30
81
87

Elemental mercury (Hg ~ spiked dodecane (C 12) was added to a crude oil sample that had
been previously analyzed repetitively (numerous bottle openings). Forty mL of US19-4 and 40
mL of C12/Hg~ were mixed in a 125 mL glass bottle, and the mixed sample was identified as
US19-C12/Hg~ Each of three bottles, C12/Hg ~ US19-4, and US19-C12/Hg ~ had about 80 mL
of sample (~68 g) and 45 mL o f headspace. In addition, a H~ ~ spiked toluene sample was also
constructed. Initial concentrations are shown in Table 8, HgU (liquid phase) was measured by
sparging an aliquot and trapping mercury in the sparge gas [16]. In Table 8, the amount of
mercury (ng) in the liquid phase of the constructed samples was measured (THg in liquid phase).
The concentration o f mercury in US 19-C12/Hg ~ was calculated from added mercury (oil density
0.85).
TABLE

S--Mercury in samples usedfor headspace experiments (LAB 2).

Hg~
THg
Hg~in the sample bottle
(liquidphase) (liquidphase)
(~tg/kg)
(~tg/kg)
(ng)
Hg~spiked C12
124.4 (n = 2)
145.4(n = 2)
8000
Hg~spiked toluene
14.5
14.8 (n = 2)
1000
US19-4
ND
40.0
0
US19-C12/Hg~ 60 (calculated) 90 (calculated)
5000
Date - 1/13/03

The experimental procedure for headspace Hg measurements involved removing headspace


vapor using a volumetric syringe (20 mL) without opening the sample bottle. The syringe was
discharged onto a gold trap to capture the mercury in the gas volume extracted by the syringe.
Liquid phase samples (same bottle as headspace samples) were also removed with a syringe and
analyzed for THg. The volume o f the syringe used to measure headspace mercury concentration
was thought to approximate a bottle opening.
Headspace mercury was sequentially analyzed as a function of time, temperature, and bottle
equilibration procedure (Table 9). The 40 ~tg/kg THg (liquid) concentration exhibited by oil
US 19-4 represented its residual non-volatile mercury concentration as demonstrated by the low
values o f mercury found in headspace measurements. In constructed samples, the mercury that
partitioned to the bottle headspace and then retrieved in successive headspace analyses originated
from elemental mercury added to the sample. The initial attempt to measure mercury in the
headspace o f the US19-C12/Hg ~ sample bottle produced an off scale response by the detector,
and thus the amount lost in the first attempt at measurement is not known but is certainly high.
Leaving the bottle open for 24 h was sufficient to deplete the elemental mercury originally added
to the bottle.

WILHELM ET AL. ON CRUDE OIL

191

TABLE 9--Sequential measurements of heads ace and liquid phase THg in spiked oil (LAB2).
US19US19US19-4
C12/Hg0 C12/Hg0 Toluene/Hg~
Order of
Headspace
Headspace Headspace Liquid
Date
Conditions
Experiments
Hg
Hg
Hg
THg
(rig/20 rot)
(ng/20 ml) (ng/20ml)
(p.g/kg)
Shake/settle/no heat
...*
> 100
...
9 1/13/63
1st
5th
Shake/settle/no heat
18.8
28.9
1/14/03
Ist
No shake/no heat
01i
25.4
73.'9
50~
25.6
72.9
4th
70.9
No shake/go heat
NDt
16.1
1/15/03
1st
40.8
2nd
Shake/no heat
0.2
19.6
53.8
3rd
50~
0.5
43.7
67.9
115.2
1/16/03
1st
No shake / no heat
ND
15.8
44.7
70~ heat/shake/settle
1.5
44.6
2nd
117.6
Bottle left open with
1/17/03
1st
stirring (20~ for 25 h
4.3
34.2
70~ shake/settle
*not measured.
t not detected.
temperature of vapor aliquotremoval.
The amount of mercury in the syringe should be proportional to the concentration o f mercury
in the liquid phase, if equilibrium was established. While the toluene sample was constructed at a
lower concentration (14.5 ~tg/kg) than the US19-C12/Hg ~ sample (124 lag&g), it consistently
provided more mercury to the vapor phase. This demonstrates the effect o f liquid phase
composition on mercury volatility. It is not known if complete equilibrium was established
between mercury in the vapor and liquid phase, but the qualitative trends are apparent. Volatile
mercury partitions to the vapor space o f sample containers at a significant rate.
Figure 1 shows a compendium o f THg measurements for both constructed and real crude oils
on a (natural) log scale for clarity of trend in time dependence. Real, as opposed to elemental
mercury spiked crude oils, decay to a seemingly constant value that is thought to represent the
non-volatile fraction. The elemental mercury-spiked dodecane decreased in mercury
concentration from approximately 170 ppb originally to approximately 7 ppb after numerous
bottle openings for various purposes.

Identification of Volatile Species


The procedure to identify volatile mercury species followed the methods o f Liang [16] for
determination o f dimethylmercury and elemental mercury in vapor. An open oil sample vial (40
mL) was placed in a Teflon chamber (250 mL, see Fig. 2). The chamber was then purged with
nitrogen for varying periods of time at 165 mL/min. The purge gas exiting the chamber flowed
sequentially through a Tenax trap and gold trap. The Tenax trap adsorbs organic or
organometallic (DMHg) compounds but not elemental mercury. The gold trap amalgamates both
DMHg and Hg ~ After purging the chamber, the gold trap was analyzed by thermal
desorption/CVAF. The Tenax trap was heated under flowing argon to desorb mercury species
onto a gas chromatography column with CVAF detector (column 1 m; 85~ carrier gas flow
rate 50 mL/min.; DMHg peak retention time 2.7-2.8 min.; Hg ~ peak retention time 1.7 min). The
particular crude oil selected for the experiment had a natural concentration of total mercury close
to 600 ppb and was selected so as to have a sufficient quantity o f evaporated mercury for

192

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

measurement. Data (Table 10) show the amount of mercury collected on the traps as a function
of purge time. Both Hg~ and DMHg are clearly identified. Hg~ is dominant in the purge gas, but
the relative amounts of Hg~ and DMHg in the oil cannot be inferred because the rate of
evaporation of each species is unknown.

FIG. 1--THg decay in solvent and crude oil samples.

FIG. 2--Apparatus for volatile mercury species determination.

WILHELM ET AL. ON CRUDE OIL

TABLE l O---Vo(atile mercury species (LAB 2)"


Purging Time . . . . . . .
2.5 rain
10 rain
DMHg, ~tg(Tenax)
N.D.
0.0023
H~~ ~tg(sold) . . . . . . .
0.048
0.045

193

~30 rain
0.0050
0.092

Conclusions

Mercury concentrations in constructed and actual crude oil samples were measured using
three analytical methods that were compared with respect to accuracy, precision, and detection
limit. The combustion method (U.S. EPA 7473 hybrid) and a commercial extraction method
(non-standard) were found adequate to provide a good combination of sensitivity and accuracy,
while instrumental neutron activation analysis was found to suffer from interferences from
elements other than mercury but typically in crude oil. In the combustion method, direct syringe
injection of aliquots to the combustion chamber was found advantageous in that it minimized
opportunities for loss of volatile mercury.
Mercury in crude oil consists of both volatile and non-volatile forms. The study identified
two volatile species (Hg~ and DMHg) that partition to sample bottle headspace. The volatile
forms of mercury are difficult because they can be lost from sample containers when containers
are opened, lost from aliquots that are removed from containers, and lost in aliquot processing
steps prior to element detection.
The non-volatile forms were not investigated for identity but could include suspended HgS,
mercury associated with asphaltenes, and mercury chemisorbed to normal sediment.
Homogenization of oils that contain suspended mercury was found important to allow retrieval
of representative aliquots for analysis. Ultrasonic agitation at slightly elevated temperature
yielded the more reproducible and higher magnitude results. Dilution of some viscous oils may
be necessary to assist homogenization or to assist aliquot removal with a syringe, but dilution
could allow the possibility of loss of volatile mercury when bottles are opened for dilution.
Speciation experiments on oils in sample bottles previously opened numerous times did not
discover any elemental mercury. The lack of elemental mercury in speciated oils likely was due
to the loss of this component during prior bottle openings as evidenced by the lower total
mercury concentrations measured in the speciation experiments compared to those obtained
previously upon first bottle opening.
The volatility of elemental mercury was demonstrated by analysis of headspace vapor in
sample bottles containing oil and solvents spiked with elemental mercury. The amount of
mercury that partitions to bottle headspace depended on temperature and on liquid phase
composition, as expected. Based on these results, one would expect significant variation in the
rate of mercury evaporation depending on oil composition.
Oils that contain a volatile mercury component typically do not analyze reproducibly unless
precautions are taken to prevent escape of the volatile mercury. Oils that analyze reproducibly
irrespective of bottle openings or other precautions likely have lost their volatile component and
may be unrepresentative. Such oils demonstrate reproducible concentrations over time.
The ongoing project did not investigate sampling procedures per se, but the principles that
have been elucidated no doubt apply to sampling as well. Sampling hot oil or using methods that
are slow and allow prolonged contact of oil with the atmosphere while filling containers provide
opportunities for the volatile mercury component to escape. The sampling methods now
employed in the project entail filling 40 mL glass containers with oil flowing from a conditioned

194

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

sample port as quickly as possible (a few seconds), leaving a very small portion of the sample
container as headspace.
We have adopted a 40 mL glass sample vial with a Teflon septum as an improvement to
standard glass bottles utilized previously. The vial is filled almost full (1-2 mL headspace) and is
not refrigerated. Samples are homogenized ultrasonically. The ultrasonic energy warms the
sample slightly. Aliquots are removed using a pre-conditioned volumetric micro-syringe and
injected directly into the vaporization/combustion chamber of the apparatus for determination.
Using these techniques, reproducible and higher magnitude concentration values are found for
the majority of actual crude oil samples so far investigated in the project.
To meet the objectives of the EPA/APUNPRA project, measuring the sum of dissolved
volatile, dissolved non-volatile, and suspended mercury compounds is important as each type
enters the refinery in crude oil. The newly developed procedures for sample homogenization, for
retrieving representative liquid aliquots, and for preservation of volatile components are thought
to eliminate a major portion of the negative bias that can exist in routine sampling and analysis
of crude oil for determination of total mercury content.

Acknowledgments
The U.S. Environmental Protection Agency through its Office of Research and Development
funded and collaborated in the research described here under contract 68C 02 088. The material
in this document has been subjected to Agency editorial and policy review and approved for
publication. The views expressed by individual authors, however, are their own, and do not
necessarily reflect those of the U.S. Environmental Protection Agency. Mention of trade names,
products, or services does not convey, and should not be interpreted as conveying, official EPA
approval, endorsement, or recommendation.
The participation and cooperation of the American Petroleum Institute (Karin Ritter) and the
National Petrochemical and Refiners Association (Norbert Dee) are gratefully acknowledged.
The donation of crude oil samples by API and NPRA member companies is sincerely
appreciated.
References

[1] Wilhelm, S. M. and Kirchgessner, D. A., "Mercury in U.S. Crude Oil: A Study by U.S. EPA,

API and NPRA," SPE/EPA/DOE Exploration and Production Environmental Conference,


San Antonio, TX, Society of Petroleum Engineers Paper 80573, 2003.
[2] Brown, T. D., Smith, D. N., Hargis, R. A., and O'Dowd W. J., "Mercury Measurement and
Its Control: What We Know, Have Learned, and Need to Further Investigate," J. Air & Waste
Manage. Assoc., June 1999, pp. 1-97.
[3] Mercury Study Report to Congress, U.S. EPA, Washington, DC, EPA-452R/R-97-004, 1998.
[4] Wilhelm, S. M., "An Estimate Mercury Emissions to the Atmosphere from Petroleum,"
Environ. Sei. Tech., Vol. 35, No. 24, 2001, p. 4704.
[5] Wilhelm, S. M. and Bloom, N. S., "Mercury in Petroleum," Fuel Processing Technology,
Vol. 63, 2000, pp. 1-22.
[6] Shah, K. R., Filby, R. H., and Hailer, W. A., "Metals in Crude Oil by Neutron Activation
Analysis,",/.. Radioanal. Chem., Vol. 6, 1970, p. 413.
[7] Filby, R. H. and Shah, K. R., "Neutron Activation Methods for Trace Metal Analysis of
Crude Oil," The Role of Trace Metals in Petroleum, T. F. Yen, Ed., Ann Arbor Science

WILHELM ET AL. ON CRUDE OIL

195

Publishers, Ann Arbor, MI, 1975.


[8] Bloom, N. S., "Analysis and Stability of Mercury Speciation in Petroleum Hydrocarbons,"
Fresenius'J. Anal, Chem., Vol. 366, No. 5, 2000, p. 438.
[9] Liang, L., Horvat, M., and Danilchik, P., "Novel Analytical Methods for Determination of
Low Level of Total Mercury in Petroleum and Its Products by Gold Amalgamation CVAFS,"
Proceedings of the 4 th International Conference on Mercury as a Global Pollutant,
Hamburg, Germany, August 4-8, 1996.
[10] Kelly, W. R., Long, S. E., and Mann, J. L., "Determination of Mercury in SRM Crude Oils
and Refined Products by Isotope Dilution Cold Vapor ICP-MS Using Closed-System
Combustion," Anal, Bioanal. Chem., Vol. 376, 2003, pp. 753-758.
[11] Liang, L., Horvat, M., Fajon, V., Prosenc, N., Li, H., and Pang, P., "Comparison of
Improved Combustion/Trap Technique to Wet Extraction Methods for Determination of
Mercury in Crude Oil and Related Products by Atomic Fluorescence," Energy & Fuels, Vol.
17, No. 5, 2003, pp. 1175-1179.
[12] Liang, L., Lazoff, S., Horvat, M., Swain, E., and Gilkeson, J., "One Step Determination of
Mercury in Crude Oil and Related Products Using a Simple Lab Built Thermal
Decomposition System," Fresenius' J. of Anal, Chem., Vol. 367, 2000, pp. 8-11.
[13] Hitchon, B., Filby, R. H., and Shah, K. R., "Geochemistry of Trace Elements in Crude Oils,
Alberta, Canada," The Role of Trace Metals in Petroleum, T.F. Yen, Ed., Ann Arbor Science
Publishers, Ann Arbor, MI, 1975.
[14] Musa, M., Markus, W., Elghondi, A., Etwir, R., Hannan, A., and Arafa, E., "Neutron
Activation Analysis of Major and Trace Elements in Crude Petroleum," J. Radioanal. Nucl.
Chem., Vol. 198, No. 1, 1995, p. 17.
[15] 40 CFR Part 136, Appendix B.
[16] Liang, L., Kichgessner, D. A., Wilhelm, S. M., Kariher, P. H., and Pang, P., "Occurrence of
Dimethylmercury in Organic Solvents," Materials and Geoenvironment, Vol. 51, No. 2,
2004, pp. 1968-1971.

Journal of ASTM International, June 2005, Vol. 2, No. 6


Paper ID JAI13089
Available online at www.astrn.org

Brian S. Fox, 1 Kelly John Mason, Ph.D., 2 and Frank C. McElroy, Ph.D. 2

Determination of Total Mercury in Crude Oil by Combustion


Cold Vapor Atomic Absorption Spectrometry (CVAAS)
ABSTRACT: Unlike many heavy metals, mercuryis a long-rangeair and water pollutantwhose load on
the environmentis increasing. Accordingto publishedreports, a major source of atmosphericmercuryis
the burningof coal for electricalpower generation. However,certain crude oils may euntainup to several
hundred ng/g of total mercury. Accurate analysis of total mercury in crude oil is the first step in
determiningthe value of a crude oil and one's ability to process the material within the refinery,while
generating finishedproducts. This paper details the use of CombustionCold Vapor Atomic Absorption
Spectrometry (CVAAS) technology for the analysis of crude samples and distillation cuts. The
instrument described in the report is one of several instrument methods currently being evaluatedby
groups in industry and government. The use of a boat-inlet sample introductionsystem allows crude oil
and other types of material to be analyzed without acid digestion. The strengths and weaknesses of the
combustion-CVAAStechnique,along with a year's worth of operating experience, will be presented from
the user's perspective.
KEYWORDS: mercury, crude oil, combustionCVAAS

Introduction
World consumption of hydrocarbons is steadily increasing. To meet current and future
demand, crude oils that possess less desirable characteristics from a processing standpoint must
be utilized. These are sometimes referred to as challenged crudes. Certain crude oils may
contain significant concentrations o f arsenic, selenium, and mercury. These elements can affect
how a crude oil is processed and what steps must be taken to mitigate its presence in product and
waste streams. Mercury is a particular challenge, given its nature as a high profile pollutant with
so many negative effects.
When present in liquefied natural gas (LNG) product, elemental mercury may cause
embrittlement of aluminum vessels and heat exchangers [1]. The potential for catastrophic
equipment failure could be a direct outcome of processing mercury-containing LNG. Even when
corrosion does not occur, elemental mercury can pose an exposure hazard to refinery workers
during turnaround operations o f process units. Mercury may also find its way into fuel gas and
wastewater streams with subsequent release to the environment. Once in the environment,
mercury may be transformed by bacteria into organic forms, moving up the aquatic food chain
where it reportedly causes neurological damage to fish-eating adults and birth defects and
developmental problems in children [2,3].
Based on its serious environmental impact, the determination o f total mercury in crude oil is
a priority. Researchers have used numerous detection techniques in the pursuit of total mercury
Manuscriptreceived 15 October 2004; accepted for publication2 December2004; publishedJune 2005. Presented
at ASTM Symposiumon ElementalAnalysis of Fuels and Lubricants:Recent Advancesand FutureProspects on 6-8
December 2004 in Tampa, FL.
1AdvancedResearch Associate,ExxonMobilResearch and Engineering,600 BillingsportRoad, Panlsboro,New
Jersey 08062.
2 ExxonMobilResearch and Engineering.

Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

196

FOX ET AL. ON ATOMIC ABSORPTION

197

measurement. They include Cold Vapor Atomic Absorption Spectrometry (CVAAS), Cold
Vapor Atomic Fluorescence Spectrometry (CVAFS), and Isotope Dilution Cold Vapor
Inductively Coupled Plasma Mass Spectrometry (ID-CV-ICP-MS) [4].
To accurately measure total mercury concentration in crude oil, several criteria were deemed
important. Since both the analyte and the matrix are volatile, minimal sample handling and
preparation are key attributes. Researchers have previously developed acid digestion schemes,
using both open and closed vessel microwave systems to prepare food and biological samples for
mercury analysis [5]. Petroleum products are often more difficult to digest than these sample
types, requiring even more rigorous conditions. The hot acid digestions may contribute to
inaccuracy from either thermal evolution of the volatile analyte or contamination.
Extraction techniques have also been used to determine total mercury and its species in
petroleum products [6]. The technique is effective, but it appears to be more labor-intensive than
what one may desire for a refinery laboratory. An instrument that provided a direct determination
of mercury in crude oil by thermal decomposition followed by CVAFS was built [7]. This
approach appears to hold the promise of uniquely low detection limits, but unfortunately a fully
automated instrument based on this design has not been commercialized.
An instrument that uses thermal sample decomposition followed by gold amalgamation and
CVAAS was specified in EPA Method 7473 [8]. A few commercial instruments based upon this
technique have been used to analyze crude oil samples [9,10].
Another important attribute of the candidate instrument is that its response to mercury should
be independent of the form in which it is found in the sample. The NIC instrument thermally
decomposes the sample and converts all of the mercury to the elemental form, thereby ensuring
that instrument response is independent of the mercury species. This enables the analyst to use a
stable ionic form of mercury in an acidified aqueous medium for the purpose of instrument
standardization.
The focus of this paper is on the technique of Combustion CVAAS and its applicability to
some of the samples encountered in the petroleum industry. It describes the equipment and test
procedure adopted by our laboratory for the determination of total mercury in crude oil,
condensate, and other process stream samples. Validation of this technique using standard
reference materials (SRM), certified standards, and spiked samples is addressed. Instrument
response to various organic mercury species and precision data derived from actual crude oil
samples is presented.
Instrumentation

The instrument used to measure total mercury by Combustion-CVAAS was the SP-3D
mercury analyzer manufactured by the Nippon Instruments Corporation (NIC), Osaka, Japan.
The boat inlet design of the instrument allows for analysis of either solid or liquid samples. With
a separate attachment, LPG and gas samples may be analyzed. The instrument uses air that is
supplied by an on board pump as the carrier gas. Before use, the air is dehumidified and passed
through an activated carbon filter to scrub away background mercury.
A computer-controlled furnace in the ceramic tube thermally decomposes the material in the
boat. A choice of six different heating cycles is available for combusting various sample
matrices. Mercury released by the sample passes through a second 850~ decomposition furnace
that has a copper catalyst insert. This step converts all of the mercury to elemental form.
Mercury vapor and combustion gas pass through a scrubber filled with an aqueous phosphate
buffer solution, then through a dehumidifier before reaching the first mercury collector. Mercury

198

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

is selectively retained by gold-coated Chromosorb, while combustion by-products pass through a


multi-port valve and onto vent.
The first mercury collector is heated, and the mercury vapor passes through a six-port Teflon
valve and onto a second gold-coated Chromosorb trap. This second step concentrates mercury
again and minimizes the presence of interfering species. This trap is heated, and the mercury
vapor is swept into the cell of a CVAAS, where measurement is performed at 253.7 nanometers.
The amount of energy absorbed in the cell increases proportionally to the number of ground state
mercury atoms present. The concentration of mercury in the sample is quantified by comparison
to the signal from a previously established calibration curve. The exhaust from the CVAAS cell
passes through an activated carbon filter, where it is scrubbed of mercury before release into the
atmosphere.
Data from the analyses are collected by a PC running "Mercury MD-I" data analysis
software. This software can store multiple calibration curves and raw data from the samples.
The mercury response can be measured using either peak height or integration over time. The
PC software is strictly for data acquisition and does not control the instrument start or any
operating parameters.

Reagents and Materials


The cleanliness of equipment and reagents used in the test is of paramount concern. Mercury
is a ubiquitous contaminant that can show up in surprising ways, especially at the range of
concentrations in the samples. An example of potential contamination is the mercury compounds
that may be used as preservatives in certain phosphate buffer solutions. The analyst must be
continuously vigilant when choosing reagents and supplies.
All water used in the analysis is deionized and polished with a Millipore Simplicity system,
available from Millipore Corporation, Billerica, MA, USA. Two reagents are used in the sample
boats to assist the combustion of organic materials. Additive B is a coarse aluminum oxide.
Additive M is a 35:50 combination of calcium hydroxide and sodium carbonate. The additives,
specially prepared by Wako Pure Chemical Industries, Ltd., Osaka, Japan, are available for
purchase through NIC. The phosphate buffer in the scrubber vessel was prepared from a
powdered concentrate (Wako).
A stock standard aqueous solution of mercury chloride, 997 mg I"1 (Wako) was used to
prepare a series of working standard solutions. The working standards are prepared in an
aqueous L-Cysteine, 10 mg 1-1 in 0.2 % nitric acid. The L-Cysteine, 97 % was obtained from
Aldrich Chemical Company, Inc. Milwaukee, W-I, USA. The nitric acid, concentrated, Trace
Metal Grade was obtained from Fisher Scientific, Pittsburgh, PA, USA.
Mercury free solvents were used to dilute heavy hydrocarbon samples, when necessary.
HPLC Grade toluene, Optima Grade hexanes, and low odor kerosene (Fisher Scientific) were all
found to have suitably low mercury concentrations for most work. Samples were caught in precleaned, 40-mL Volatile Organic Analysis (VOA) vials with Teflon lined septa (Fisher
Scientific). The ceramic sample boats and high-form crucibles are available from NIC or other
suppliers. A Hamilton Gastight Model 81000 100-#1 syringe (Fisher Scientific) was used to
withdraw crude oil sample from the VOA vials.
To perform the instrument validation and determine analytical figures of merit, stock
standard solutions of dimethyl mercury, methyl mercury chloride, and ethyl mercury chloride
were prepared by dissolving the pure compounds in HPLC Grade toluene. Each solution was
prepared to approximately 100 ~tg/g of the mercury compound. These solutions were obtained

FOX ET AL. ON ATOMIC ABSORPTION

199

from Spectrum Quality Standards, Sugarland, TX, USA. To prepare sample spikes, 10 /xg/g
certified mercury in oil standard was obtained from Conostan Division, Conoco Specialty
Products, Ponca City, OK, USA. Aqueous mercury check standards, prepared in 10 % nitric acid
were also obtained from SCP Science, Champlain, NY, USA. Finally, standardized reference
materials SRMs and CRM were obtained from NIST, Gaithersburg, MD, USA and from the
Canada Centre for Mineral and Energy Technology, Ottawa, Ontario, Canada.
Test Procedure

A general test procedure that describes instrument standardization and analysis of various
sample matrices, including liquid hydrocarbons, is found in the NIC instruction manual for the
SP-3D [11]. A second, nearly identical procedure that is specifically designed for the analysis of
mercury in liquid hydrocarbons may be found in UOP Method 938-00 [10].

Instrument Preparation and Standardization


Ceramic crucibles that contain the necessary amount of reagent for the expected analyses are
placed in a muffle furnace at 750~ the prior day. Ceramic sample boats are similarly stored in
the muffle furnace to keep them free of ambient mercury contamination. The sample boats and
reagents are allowed to cool before use in a covered pan that is over-pressured by dry, mercuryfree, filtered air from the instrument. It is important to allow the surfaces that the sample
contacts to reach ambient temperature to minimize loss of elemental mercury.
After choosing the appropriate instrument range and working standard, as shown in Table 1,
the analyst measures aliquots of standard into the cooled boats using an adjustable pipettor with
disposable tips. Instrument heating mode "1" is chosen for the aqueous standards or samples of a
similar nature. Several heating modes are available, and examples of the samples for which each
mode may be used are shown in Table 2. A calibration curve is established from the instrument
response to various known mercury concentrations.
TABLE 1--Working standards and injection sizes.
MeasuringRange, n g .
2
20
200
.....
1000

Hg Standard, mg/L
0.01
0.1
1.0
5.0

...

InjectionVolumes,.~L
50,100,150,200
50, 100, 150, 200
50,100,150,200
50,100,150,.200

TABLE 2--NIC SP-3D Heating Cycles and Sample Types


Mode
1
2
3
4
5
6 ......

Sample Heating Cycle


.
1 min - 350~ 4 min - 700~
4 min - 350~ 6 min - 700~
10 rain - 350~ 6 min - 700~
2 min - 350~ 4 min - 700~
4 rain - 600~ 6 min - 900~
Decomposition heating element off

Sample Types
Aqueous standards, drinking water
Coal, fly ash, wastewater, soil
Crude oil, heavy naphtha, lube oil
Light naphtha, gasoline
Sulfuric acid
LPG, LNG, air

200

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Sample Boat Preparation


A reagent blank value must be determined, since the process of filling the sample boats leads
to trace contamination by ambient mercury sources. The effect of the reagent blank value is
most pronounced for samples with mercury concentrations at, or close to the minimum detection
limit. Since the sample boats vary in shape and capacity, there is no hard rule for the amount of
reagent to add. Approximately 2-3 g of Reagent B will fill the lower third of the sample boat.
On top of this layer, about 3-5 g of Reagent M is used to bring the level of the reagents up to the
top of the boat.
The sample boat with reagents is placed at the entrance of the combustion tube and slid into
the heated zone with a long fork. The cap at the entrance to the combustion tube is screwed into
place, heating mode "3" is selected, and the start button is pressed. The mercury response values
from this measuring cycle are discarded. While the first boat is being heated, a second sample
boat is prepared with reagents in a similar manner. At the end of the cycle, the first boat is
removed and allowed to cool in a covered tray. The second boat is also subjected to one heating
cycle.
After the second boat is removed, the first boat is placed into the heated zone of the
combustion tube. The cap is screwed into place, heating mode "3" is selected, and start is
pressed. The mercury response from this run is captured by the data acquisition software and
used to calculate a reagent blank. The second boat is treated in a similar manner. It is important
to allow sample boats to cool to ambient temperature before proceeding with sample
measurements.

Sample Preparation and Analysis


Crude samples normally arrive for testing in 40-mL VOA vials with Teflon-coated silicone
septa. Mercury loss needs to be minimized by opening containers as briefly as possible for
sample transfer. If the crude oil sample flows into a syringe at ambient temperature, then all
sampling may be done through the septum. For heavy oil samples, the containers must be
opened briefly to withdraw an aliquot.
Samples are placed in an ultrasonic bath for 1 h immediately prior to analysis. This step was
recommended to us by the NIC representative to increase mercury recovery in a sample. Ice
chips are added to the ultrasonic bath to prevent the temperature from rising more than 3~
above ambient laboratory temperature.
The crude oil aliquot is obtained by piercing the septum on the container and slowly drawing
the sample into a 100- #1 syringe. The container is inverted to dispel bubbles from the syringe.
A cooled sample boat, loaded with reagent, is placed at the entrance of the combustion tube. The
crude sample is injected beneath the surface of the reagent, and the boat is quickly pushed into
the combustion tube. The cap is replaced, heating mode "3" has been pre-selected, and the start
button is quickly pressed.

Heavy Crude Sample Preparation


Heavy crude samples do not flow at room temperature. Their highly viscous nature makes
them impossible to withdraw from a sample vial with a syringe. These samples must be diluted
before injection. The vial of heavy crude is placed in the ultrasonic bath for 1 h. Allowing the
bath temperature to rise above ambient temperature often is enough to cause the heavy oil to

FOX ET AL. ON ATOMIC ABSORPTION

201

flow. The vial is quickly opened, and a weighed aliquot is transferred to an empty, tared 40-mL
VOA vial. Toluene is added, a total weight is recorded, and the vial is re-capped. Typically, a
1:2 mass/mass ratio of sample to toluene is enough to lower the viscosity for withdrawal using a
syringe. From this point, diluted samples are treated similarly to light crude oil samples.

Sample Mass Determination


The mass of the crude sample that was injected into the boat is determined by measuring a
second aliquot of the crude oil sample. A capped 2-mL GC vial is tared on a balance. A 100-#1
aliquot of the crude oil sample is slowly drawn into the same syringe that is used to inject the
sample into a boat. Bubbles are carefully excluded from the syringe. The sample is then
injected into the tared GC vial and weighed again. This procedure allows the mass of an
injection to be measured directly, rather than calculated based upon a separate density
determination.
Results and Discussion

The response of the instrument is linear using the recommended standard solutions prepared
in nitric acid and L-cysteine. Figure 1 shows a calibration curve for the low range. The
correlation coefficient is typically 0.9995 or higher. Several organic mercury compounds were
tested, and the instrument response was found to be independent of the mercury species in the
samples. This allows the analyst to choose more stable inorganic mercury standards in acid
solution. Inorganic standards stored in tightly capped, FEP Teflon bottles appear to last for a few
months.

FIG. 1--Typical low range standardization curve o f the instrument.

202

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Certified materials were used to evaluate the accuracy o f the technique. These included coal,
coal fly ash, soil, and crude oil. Table 3 shows that measured concentrations of mercury in the
SRMs agree well with certified values. With only one exception, the Relative Percent Difference
(RPD) is within 5 % of the certified values. The one exception is for a value that is less than half
o f the claimed detection limit for the published method. Six repetitions were made for each
material. The expanded uncertainty o f each measurement is calculated by multiplying the value
for o f one experimentally derived standard deviation o f the data times the critical value. The
critical value o f 2.57 is determined from a Student's t-distribution [12] with 5 degrees of freedom
and a confidence level o f 95 %.
TABLE 3--Total mercury results of standard reference materials.
SRM Name
Certified Hg, ng/g
Found Hg, ng/g ..... RPD,%
10.9+ 1
11.1+ 1
2
NIST 1635 Subbituminous Coal
NIST 2683b Bituminous Coal
90.0 + 3.6
94.5 + 4.7
5
NIST 2685b Bituminous Coal
146.2 + 10.6
152.2 + 1.4
4
0.0417 + 0.0057
0.047 + 0.031
12
NIST 2721 Light-Sour Crude Oil
0.129 + 0.013
0.135 + 0.06
5
NIST 2722 Heavy-Sweet Crude Oil
160
158 + 13
1
NBS 1633a Coal Fly Ash
CRM SO-1 Reference Soil 2
22
21 + 2
5
i This is an inactive SRM obtained from NBS. The mercury value was obtained from the containeriabel.
2Current status of CRM is unknown. The CRM was obtained from CANMET, Ottawa Ontario, Canada. The
mercury value was obtained from the container label.
The NIC operating instructions call for weighing about 100 #g of sample onto a layer o f
Reagent B in the sample boat and then covering that with a layer o f Reagent M. Other
researchers had determined that this step causes the loss of volatile analyte when compared to
sampling the crude oil through the Volatile Organic Analysis (VOA) vial septum using a syringe
[13]. To minimize volatile analyte loss, we inject a volume o f sample below the layered reagents
in the ceramic boat. The mass is obtained by measuring a separate 100-~1 aliquot o f the sample
on a balance. After repeatedly testing a material of known density, it was observed that
uncertainty around using a mass determined by this procedure was less than 0.5 % of the value.
Some of this observed uncertainty may have been generated by temperature change, due to
continuous handling o f the syringe.
There are two factors that influence the minimum detection limit (MDL) for this method.
The first is the maximum sample size for a crude oil sample. In the interest of combustion
efficiency, it is recommended that no more than 100 mg o f an oil sample be used. The second
limiting factor is the magnitude and variability o f the reagent blank signal. By carrying out a
pre-heat cycle o f the reagents after they are loaded into the boat, the blank response is reduced
along with its variability. Figure 2 shows how the blank response varies before and after a preheat cycle. With multiple heating cycles immediately followed by the reagent blank
measurement, it is possible to drive the reagent blank down to the level o f the instrument blank.
However, this will result in a low bias for the reagent blank signal. The best procedure is to
allow the sample boat to cool to ambient temperature, thereby causing the analyst to treat the
reagent blank measurement in exactly the same way as a sample measurement.

FOX ET AL. ON ATOMIC ABSORPTION

203

FIG. 2--Effect of a pre-heat cycle on the response of the reagent blank.


The positive outcome of taking this extra step is that the mercury concentrations in the two
available NIST crude oil SRMs that straddle the instrument detection limit are practically
resolved. However, the uncertainty around the measured mercury values is about five times
greater than what NIST obtained with ICP-MS. Many of the crude oils samples that have been
tested to date in our laboratory have less than 2 ng/g of mercury. The use of a syringe to inject a
volume of sample below the level ofpre-treated reagent has resulted in less variability between
repeat injections being observed, along with the possibility of determining mercury
concentrations at or near the claimed detection limit. The sensitivity and precision of this
Combustion-CVAAS technique is compatible with published results of Liang et al. [7], who used
CVAFS after thermal decomposition of crude oil samples.
Crude oil samples that had very low measured mercury concentrations were spiked with a
variety of organic mercury species to test the recovery. The results of spiked samples are shown
in Table 4. Elemental mercury was not used to spike crude oil samples, due to the difficulty in
preparing a known, stable concentration in an organic matrix.
TABLE 4--Total mercury results of spiked samples.
Sample ID
Crude 1
Crude 2
Crude 3
Crude 4
Crude 5
Conostan Base Oil
Conostan Base Oil

Mercury Spike Compound


Methyl Mercury Chloride
Methyl Mercury Chloride
Dimethyl Mercury
Ethyl Mercury Chloride
Conostan Hg Standard
Conostan Hg Standard
Conostan Hg Standard

Total Hg, ng/g


15.9 + 0.4
32.3 + 1.1
13.2 + 1.1
8.7 + 0.7
33.7 + 1.1
11.0 + 1.6
104.5 + 2.1

Recovery, %
105
105
93
104
97
100
104

204

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Three analyses were made of each spike sample. The uncertainty was determined in a
similar manner as before. The critical value of 4.3 is determined from a Student's t-distribution
[12] with 2 degrees of freedom and a confidence level of 95 %. With just one exception, the
recoveries were within 5 % of the expected concentrations. The fact that dimethyl mercury was
the most volatile of the analyte species used in the study may have contributed to the slightly
lower recovery of its spike sample.
Our laboratory has tested over 80 crude oil and condensate samples to date. Three external
laboratories also tested some of the samples as part of a cooperative effort to develop standard
procedures for handling and analysis. The laboratories used techniques other than combustionCVAAS. Details of the outside labs' testing statistics are not known. Table 5 shows a
representative subset of these results for comparison. Each analysis by our laboratory was done
three times and averaged. The same statistics apply as those discussed in the preceding
paragraph. The critical value of 4.3 with 2 degrees of freedom from the Student's t distribution
table [12] exaggerates the small differences that were actually observed between the three
mercury determinations.
TABLE 5--Total mercury results of actual crude oil and condensate samples.
Sample Type
Crude 1
Crude 2
Crude 3
Crude 4
Crude 5
Crude 6
Crude 7
Crude 8
Condensate 1
Condensate 2

Total Hg, ng/g


In-house Laboratory
0.3 + 0.4
3.9 + 1.3
82 + 2
1.5 + 0.4
0.5 + 0.4
0.6 + 0.4
1.4 + 0.4
10.1 + 1.7
450 + 15
51 + 3

Average Total Hg, ng/g


Outside Laboratories
0.2
4.5
. . . .
1.6
0.2
0.6
1.3
. . . .
. . . .
. . . .

RPD,
%
20
14
.

.
6
86
0
7

.
.
.

.
.
.

Precision is good, and agreement with the outside laboratories is acceptable, considering the
fact that many of the samples in the comparison have sub-ppb mercury concentrations that are
just above the stated minimum detection limit of 0.1 ng/g for the technique. Since these samples
were rtm over the course of a year, some did not benefit from accumulated experience with
reagent blanks and the special handling required for low ppb mercury samples.
Observations

Several observations have been gleaned from more than a year of using Combusfion-CVAAS
to analyze total mercury in crude oil samples and several other matrices.
The instrument flow path must be kept clean. Carbon residue, no matter how slight, can
cause dramatic precision and recovery problems. Teflon connectors and valves must be handled
with care to prevent damage. Regular leak checking also helps maintain mercury recovery and
precision. Additionally, laboratory cleanliness is imperative for accurate trace mercury

FOX ET AL. ON ATOMIC ABSORPTION

205

measurements. All sources of contamination must be controlled. The utensils used to dispense
reagent and handle the boats must be kept covered.
The boats themselves should be treated in a muffle furnace, before use. Their condition must
be monitored, since the caustic Reagent M will erode the glazing from ceramic after a period of
use. This may lead to precision problems with aqueous standards, since water has been observed
being wicked over the side of the boat. Scrupulous flushing of the syringe with clean solvent
minimizes analyte carryover from high mercury concentration samples. Running a blank
between the high calibration standards and the samples also helps to flush residual mercury out
of the instrument flow path.
Container integrity can be a problem with certain samples. The Teflon surface of the septum
on the 40-mL VOA vial serves as a barrier between hydrocarbons and the soft silicone rubber of
the septum. Just a few punctures of this barrier by a syringe needle may allow the vapor to swell
the septum. Both light crude oil and condensate samples can swell a septum to a point where it
splits and releases material. These problems of container integrity and the volatility of both the
matrix and the analyte make it a challenge to find a suitable Statistical Quality Control (SQC)
sample for this test.
Once it had been determined that instranaent response is independent of mercury species or
matrix, any stable mercury sample, even a coal SRM, can be used to verify instrument
performance on a daily basis. The solid samples seem to be far more stable than any mercury
standard in a hydrocarbon matrix that we have prepared to date.
The boat inlet design of the NIC Combustion-CVAAS instrument system precludes the
addition of an auto-sampler. Due to the necessity of minimizing the residence time of the sample
in the boat prior to the analysis, an auto-sampler may not be a practical accessory for the analysis
of crude oil.
Conclusion
A mercury analyzer based upon complete sample combustion followed by CVAAS has
proven to be a relatively simple instrument to operate and maintain. The technique requires
practically no sample preparation for a light crude oil; therefore loss of analyte or contamination
by mercury sources in the laboratory is minimized. By testing a variety of mercury species and
sample types, we have demonstrated that instrument response is independent of the form of
mercury and the sample matrix. This makes standardization with stable mercury species an easy
and reliable procedure.
The improvements in techniques for the trace level determination of difficult analytes, such
as mercury in crude oil, serve to focus future efforts on proper sampling and preservation
procedures. The inherent simplicity of a total combustion procedure for hydrocarbon samples
removes many of the analytical variables that would otherwise confuse the issue.

Acknowledgements
The authors would like to thank Amy S. Brown of EMRE for providing the variety of
samples used in the study.

206

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

References

[1] Edmonds, R., Moorwood, A. S., and Szczepanski, R., "Mercury Partitioning in Natural
Gases and Condensates," Paper Presented to the GPA European Chapter Meeting, London,
March 21, 1996.
[2] Weise, E. and Watson, T., "Mercury in Many Lakes, Rivers," USA Today, Wednesday,
August 25, 2004, p.1.
[3] Raines, B., "Mercury Testing Planned in Gulf," Mobile Register, Friday, March 22, 2002, p.
4A.
[4] Wilhelm, S. M. and Bloom, N., "Review - Mercury in Petroleum," Fuel Processing
Technology, Vol. 63, 2000, pp. 1-27.
[5] Microwave Enhanced Chemistry, Kingston, H. M. and Haswell, S. J., Eds., American
Chemical Society, 1997, pp. 380-381.
[6] Bloom, N. S., "Analysis and Stability of Mercury Speciation in Petroleum Hydrocarbons,"
Fresnius J. Anal Chem, 2000, Vol. 366, pp. 438-443.
[7] Liang, L., Lazoff, S., Horvat, M., Swain, E., and Gilkeson, J., "Determination of Mercury in
Crude Oil by In-Situ Thermal Decomposition Using a Simple Lab Built System," Fresnius
J. Anal Chem, 2000, Vol. 367, pp. 8-11.
[8] US EPA Method 7473 - Mercury in Solids and Solutions by Thermal Decomposition,
Amalgamation, and Atomic Absorption Spectrometry.
[9] Mercury Application Note #6 Rev. 01/20/04 - "Determination of Mercury in Crude Oil
Using A Direct Mercury Analyzer," Milestone Incorporated, Monroe, CT 06468.
[10]Total Mercury and Mercury Species in Liquid Hydrocarbons, UOP Method 938-00, UOP
LLC, Des Plains, IL, 2004.
[11] Instruction Manual for SP-3D, Nippon Instruments Corporation, Takatuki-shi, Osaka, Japan.
[12]Anderson, R. L., Practical Statistics for Analytical Chemists, Van Nostrand Reinhold
Company Inc., 1987, p. 299.
[13] Teleconference with Dr. S. Mark Wilhelm.

"Journal of ASTM International, January 2006, Vol. 3, No. 1


Paper ID JA112988
Available online at www.astm.org
Professor Peter B. Stockwell, 1 Dr. Warren T. Corns, 2 and Dr. Derek W. Bryce 3

Mercury Measurements in Fossil Fuels, Particularly


Petrochemicals
ABSTRACT: The dangers of mercury and its derivatives, especially organomercury compounds have
been well documented for almost half a century. Concerns remain regarding mercury contamination in
aqueous ecosystems. Significant quantities of mercury are present in fossil fuels, particularly
petrochemicals.
We still manage to pollute the atmosphere and environment by the use of coal and petrochemicals. The
impact of mercury in natural gas has been such that there have been at least six experiences of plant
failure which have been traced to corrosion due to mercury attacks on aluminium rotors in the plant
fabric. Very little attention is being given to mercury removal strategies, and in order for these to be
successful there is an urgent requirement to determine the speciation profile of the natural gas or liquid
hydrocarbon condensates. Atomic fluorescence spectrometry is an ideal analytical technique to determine
the levels of mercury both before and after removal systems. The authors will outline achievements of
both laboratory and online measurement systems for this important area of research. These systems have
been operating at the required levels of performance and in the case of some of the online systems, have
been in place operating 24 hours per day 7 days per week for more than 2 years. This has encouraged
other companies to install similar systems.
KEYWORDS: mercury, fossil fuels, petrochemicals
Introduction

The dangers o f m e r c u r y (Hg) and its derivatives, especially organomercury compounds, have
been well documented for almost h a l f a century. Significant concerns remain regarding mercury
contamination in aqueous ecosystems. The U S E P A has advanced water quality criteria [1] for
the protection o f organisms native to water environments. The criterion for mercury in fresh
water ecosystems is 12 nanogram/liter (ng/1), and the mercury chronic criterion for salt water is
25 ng/1. These extremely low criteria present significant demands for the analyst, and atomic
fluorescence spectroscopy provides a viable option for the measurements.
W e still manage to pollute the atmosphere and environment b y the continual use o f coal and
petrochemicals as our m a i n sources o f energy. The impact o f m e r c u r y in natural gas has been
such that there have b e e n at least six experiences o f plant failure which have been traced to
corrosion due to m e r c u r y attacks on aluminium rotors in the plant fabric. The six experiences o f
plant failure have had significant financial impact on the petrochemical companies. Since very
Manuscript received 8 October 2004; accepted for publication 14 June 2005; published January 2006. Presented at
ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
i Managing Director, P S Analytical Ltd, Arthur House, Crayfields Industrial Estate, Main Road, St Paul's Cray,
Orpington, Kent, BR5 3HP, U.K.
z Research and Development Manager, P S Analytical Ltd, Arthur House, Craytields Industrial Estate, Main Road,
St Paul's Cray, Orpington, Kent, BR5 3HP, U.K.
3 Application & Technical Support Specialist, P S Analytical Ltd, Arthur House, Crayfields Industrial Estate, Main
Road, St Paul's Cray, Orpington, Kent, BR5 3HP, U.K.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

207

208

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

little attention is being given to mercury removal strategies and in order for these to be
successful, there is an urgent requirement to determine the speeiation profile of the natural gas or
liquid hydrocarbon condensates.
Burning fossil fuels as a source of electricity also creates a source of mercury pollution. Hg (~
and Hg 2+ are both produced from electricity generation, but the fate of these species is totally
different. Atomic fluorescence has long been available as a measurement technique, however, it
suffers from matrix interference effects, which have limited its use. The measurement of mercury
at the levels of interest and legislation mandate the use of vapor generation techniques as a
measure of sample introduction. These techniques overcome the majority of the matrix
interference effects and exploit the sensitivity and linearity of atomic fluorescence. The
sensitivity gained is equivalent to or better than that achieved by the more expensive ICP-MS
instrumentation and is two orders of magnitude better than those obtained by atomic absorption.
It is therefore also important to know the speciation profile of the coal-fired stack emissions. The
US Environmental Protection Agency is actively pursuing an ambition to regulate the levels of
mercury that can be emitted from the chimneystacks. Therefore, there is a requirement for fast,
reliable measurements, firstly to determine the levels of mercury species so that removal
strategies can be determined. Both of these requirements are important for the control of
environmental pollution. However, the measurement of mercury and its products in these
complex matrices is extremely difficult and requires considerable attention to detail. The
reliability of these measurements can be facilitated by the atomic fluorescence technique; it
provides a simple but sensitive solution with linearity over several orders of magnitude.
Measurement

The use of atomic fluorescence for the determination of mercury was first reported by
Thomson and Reynolds in 1971 [2]. Since then, several authors [3-6] have described
enhancements to the technique that have reduced formal instrument detection limits (IDL) for the
fluorescence technique to the 1-10 ng/1 range. Knox et al. [7] report on the use of atomic
fluorescence detection limit for mercury to less than 1 ng/1. A European standard EN 13506 was
published in 2001. This uses vapor generation coupled to direct atomic fluorescence
measurement. The most recent version of the US EPA standard 1631 utilizes an additional gold
amalgamation step. The amalgamation provides the potential for an additional order of
sensitivity but also requires considerable attention to detail and cleanliness to avoid
contamination.
Although many improvements have been made to the measurement techniques over the last
30 years, there has been little attention paid to the Achilles heel of analysis, that of the sampling
procedures. The regulatory requirements put a heavy burden on the analyst even for water
analysis, and proper sampling strategies and quality reagents are required.
Extending the procedures for the measurement of mercury levels in natural gas and coal fired
stack emissions places far more difficult demands on the skills of the analyst. Both present very
complex matrices, which although different in many respects, provide similar requirements to
overcome the effects of the make up within the matrix.
Measurement of the mercury levels is not in itself the main issue; firstly, it is important to
obtain a sample, which is fully representative of the stream being examined without modifying it
in any way. Secondly, the mercury present should be removed from its complex matrix and
transferred into a stream of argon carrier gas. Passing the sample over a gold substrate, which
collects all forms o f mercury prior to revaporizing into a stream of argon, has been well

STOCKWELL ET AL. ON FOSSIL FUELS

209

documented for these matrices. Requirements for an instrttment to achieve this are shown in
Table 1 and are represented schematically in Fig. 1. The development of offline and online
systems for measurement of mercury in process gas streams are discussed here. The applications
are focused in the petrochemical industry and in the coal fired utilities operation.
TABLE 1--Analyzer spec~cation. .....
0.1 pg detection limit
Easy to use
Suitable for both on and offline operation
3000 ~tg/m3 range
Typical 5 min per.sample

REMOTE:::: "~g~ I

PERMANENTTR/P

ToAFS

FIG. 1--Schematic diagram for samplinj mercury levels in gaseous samples.


It is important that all forms of mercury are trapped and measured because the form of
mercury in natural gas and liquid hydrocarbons changes in relation to the geographical location
of the product. In South East Asia, for example, organomercury compounds are present. Not
recognizing these as mercury can cause the mercury embrittlement referred to previously, since
the processing arrangement can transfer the organic mercury into metallic mercury, which
creates the problem and causes a plant shutdown. Table 2 shows that the Amasil trapping
arrangement is efficient at trapping all forms of mercury and that it is ideal for control and
measurement purposes, confirming the work ofDumarey et al. [8]
. . . . . . TABLE
.

F10w Rate
1/min
1.0
2.5
5.0

HgO
99.8~0.3
101.0
99.9

2--Trapping e,,ffilcien c y [8] .

HgC12

(CH3)2Hg

CH3HgC1

(C2Hs)2Hg

C2HsHgC1

99.3~0.4
99.7
99.8

98.9
99.7~0.5
99.5~0.4

99.9
99.1~0.6
100.7~0.8

99.7~0.4
100.3
99.9~0.4

98.8
99.1
99.9~0.3

Natural gas is often pressurized above 3000 psig, and to take a sample at atmospheric
pressure it is necessary to design a suitable pressure let down system, which avoids liquefaction
of the hydrocarbon gases. For coal fired gas sampling, it is important to avoid any condensation
of moisture present in the stack gas, since with the matrix containing sulfur dioxide, nitrous
oxide, and carbon dioxide, these may result in acidic solutions.
The reason for the measurement of mercury is often for regulatory requirements, but
basically if you cannot measure the levels correctly, then you cannot control the processing or

210

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

produce a reliable product, nor can you sell it. The latter is particularly important in the
petrochemical industry where mercury embrittlement has caused several plant shutdowns over
the years, the latest being in Australia in January 2004.
Natural Gas
Figure 2 shows a schematic diagram o f a pressure let down system, which conforms to the
European Standard ISO 6978 (ASTM 6350) for natural gas sampling. By careful choice o f
components and temperature of operation it is possible to provide reliable data on the mercury
levels present in the natural gas samples. By injecting defined samples of mercury into the
sample line prior to the sample collection, it is possible to check the validity o f the measurements
by testing the spike recovery. Recoveries in the region o f 95-105 % will show that there is little,
if any, loss with the system, and the analyst can be confident of the levels found in the gas
streams.
Typical data for Hg in export natural gas are shown in Table 3. The reproducibility for
triplicate measurements was found to be less than 4 %.
For these measurements two traps are operated in series, and the total mercury is the sum of
that collected on each of the traps. Since the levels are very low it is common practice to measure
the levels from a trap without sample collection, i.e., a blank level, and to correct measurements
for these blank levels.
PRIMARY
BY-PASS

PIPELINE

RV1
VENT

FIR 1 FIR 2
V E N T VENT

RV 1
L

ENCLOSURE

25 PSIG
"~"
RELIEF VALVE L

P1

(OPTIONAL)
PROBE

P2
PRV 1

SAMPLE

FIR 2
i

ELECTF ICALLY
HEATED
REGULATOR

~ i~J

5 SERIESBYPASSTO FIR 2 iN.

FIG.

I
INJECTION
PORT

--O

ONE ADSORBER LINKS:


3 SAMPLEPORT 1 TO FIR 1 iN.

4~

uJ

FIR I

I 1i

ill i

'~t~APII= ~
~IESBYP./~$$FIR1 IN FIR21N
PORT2 PORT1

2--Schematic diagram for commercial pressure let down system.

STOCKWELL ET AL. ON FOSSIL FUELS

.......
Sample
Reference
Export Gas
(V4500)
Export Gas
(V4500)
Export Gas
(V4500)
Export Gas
(V4100)
Export Gas
(V4100)
Export Gas
(V4100)

TABLE

211

3--Determination of mercury in export naturalgas.

Gas
Volume (I)
1.53 (T = 19~
1.53 (T = 19~
1.53 (T= 19~
1.52 (T=22~
1.51 (T = 23~
1.50 (T = 25~

Tube
No

ng
Hg/tube

ng Hg/tube
blank
corrected

ng
Hg/sample

#g Hg/ra3

1
2
1
2
1
2
1
2
1
2
1
2

41.88
2.897
42.59
2.056
38.79
2.738
45.20
2.182
42.63
2.103
44.98
1.438

41.38
2.293
42.04
1.644
38.43
2.255
44.83
1.305
42.29
1.465
44.108
0.873

43.673

28.54

43.684
40.685

28.55
26.59

46.135

30.35

29.77 + 0.71

43.755

28.98

RSD = 2.4 %
(n = 3)

44.981

29.99

Mean #g/rn3
27.89 + 1.12
RSD = 4 %
(n -- 3)

Challenges for Online Natural Gas Measurements


Online systems can also be devised, but these provide additional challenges due to the safety
aspects required for process stream measurements at petrochemical facilities. Sampling systems
based on carefully designed and tested speed loops will be required, as will explosion-proof
enclosures. A speed loop is a configuration where sample is taken out o f the process line at one
point and returned to the process line at a lower pressure down the process line. This provides a
sample that is continuously updated and representative o f the sample being processed.
The online measurement o f mercury in natural gas presents additional problems: operating on
a continuous basis put instruments on the design of the instrumentation, and the safety issues
involved put an additional requirement on the suppliers.
The mercury level in natural gas streams can be quite varied and the problems caused by high
levels quite catastrophic. As such, online measurement can qae essential in spotting potential
problems allowing a measured response to increased Hg levels, potentially saving millions o f
dollars. A schematic block diagram o f the online system is shown in Fig. 3.
Practical experience from a number o f operating companies has greatly assisted the process
o f designing fully functional online systems that operate 24 hours a day on a continuous basis.
Collecting a representative sample is a major challenge. Fundamental to the design is the
ability to collect a fully representative sample and to do this from sampling streams which have
pressures o f up to 3500 psig. The sampling must be carried out onto the traps at atmospheric
pressure, so it is necessary to avoid any liquefaction in the traps. A series of speed loops is used
to present the sample to a selection system, which introduces the relevant samples to the analyzer
for measurement. The requirements for the sampling system are set out in Table 4.

212

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Industrial
PC

O~m~rto
(Can b e ~ p ~

~oams)
$k
Probe

PSA
1O64O
Filter

s~m

--

Stream
Selector

Genenl~
FIG.

9
9
9
9
9

3--Block diagram of typical system.

TABLE 4 Sampling requirements.


Sample is delivered using speed loop.
A normally closed high pressure solenoid valve is included in the speed loop to stop the
sample flow under alarm conditions.
Heated regulator reduces pressure of sample to 6 psig.
Pressure switch and relief valve are used to prevent high pressure entering enclosure.
Pressure switch activates solenoid in alarm conditions.
All components are explosion proof.

Calibration Accessory
In order to ensure that the measurements are valid, the authors have developed a simple
calibration accessory that is traceable to NIST standards [9]. This allows a set value of mercury
concentration to be introduced at different stages of the measurement/sampling setup to check
the integrity of the measurements. All of the components in the sampling system must be tolerant
to mercury being present and not absorb or desorb any mercury which might distort the
measurements.
It is essential to ensure that measurements are correct, and that reliable data are provided.
This requirement is facilitated by specifically designed process control software which has
the following features:
9
9
9

The online process control software provides a simple interface between the analyzer, the
user, and the control room.
The software allows the user to define not only the sampling and calibration frequency
but also actions to be performed under alarm conditions
The software connects to other systems either by 4-20mA, RS232, RS485, RS422,
MODBUS, TCP/IP, or contact closure.

STOCKWELL ET AL. ON FOSSIL FUELS

213

The software is fully password protected to allow the user the ability to define who can
do what.
Figure 4 shows a typical installation on a plant where mercury removal beds have been
installed. The mercury level has to be measured both at the inlet and outlet of the system. This
requires that the measurement system can cover a wide concentration range, from a few ng/m3 to
parts per trillion/m 3, typically up to three orders of magnitude. The Sir Galahad Online Natural
Gas Analyser handles this reliably using only one detector.
Icier S ~ ,

4D"

MweuW Removal Beds

41'-

FIG. 4--Schematic of typical plant installation.

Summary
Using the above criteria, a number of online and offline explosion proof sampling systems
for Hg in natural gas applications have been designed, and these consist of a multi-stream
process analyzer, which performs inlet and outlet measurements across Hg removal beds.
To validate these measurements, an automatic calibration system was used and this has been
tested by NIST [9] and found to give values within 5 % of theoretical concentrations.

Online Hg Monitor for Stack Emissions


Mercury has become a major concern with respect to the emissions from coal-fired power
stations. Systems have been operated in the field since 2001 and have provided consistent and
reliable data.
Oxidized Hg is water soluble and is generally washed out of the atmosphere or adsorbed onto
particles which are deposited locally with precipitation. In contrast to this, elemental Hg has an
atmospheric lifetime between six months to two years [10]. Mercury emissions are therefore a
global issue, since Hg can travel long distances with no distinction of boundaries.
As a result of health concerns, there is an increasing number of international programs
concerned with reducing mercury emissions to the environment. One of the most recent

214

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

estimates [11] for global emissions of Hg from human activities was 5000 tons/year, two-thirds
of which are due to the emissions from the combustion sources worldwide. Other models predict
differing levels, but all provide the urgency to solve problems [12]. The complex behavior of
mercury species in the atmosphere and the continuous cycling of mercury between land, water,
and air complicate the development of accurate models and estimates. Without reliable
measurement data, research into understanding the fate and transportation of mercury within the
environment and its potential control from combustion sources will be severely hindered.
Durham et al. [13] have summarized the numerous problems of manual sampling:
9
9
9
9
9
9
9

They are slow, complex and elaborate


They characterize emissions during a sampling period on processes that may be variable
over time.
They are man power intensive.
They are prone to interferences.
Speciation is difficult.
They are not real time or online and are therefore limited in their use for process control
or compliant monitoring.
They are expensive.

Current methods for sampling and analysis are known to be problematic and prone to error.
Several wet chemical impinger-based methods have been approved for the determination of total
Hg in flue gas. These include US EPA method 29 and 101A. More recently, these methods have
been modified to enable the speciation of particulate, oxidized, and elemental mercury. The most
commonly used procedure is the ASTM Standard Test Method for Elemental, Oxidized, ParticleBound and Total Mercury in Flue Gas Generated from Coal-Fired Stationary Sources (D 678402) known as the "Ontario Hydro Method." This method was developed by public consultation
with ASTM membership.
There is a need, therefore, to develop online continuous emission monitors to enable accurate
mercury speciation measurements in flue gas. Real time continuous monitoring of mercury is
essential for numerous reasons. Control of mercury emissions from combustion sources is
currently being considered, and when implemented it is likely to cost billions of pounds per year.
Most of the mercury control strategies being proposed include some sort of sorbent injection.
Prior to installing a Hg control device, long-term mercury speciation data will be required by the
legislators and utilities to make more informed decisions concerning their needs and control
options. Real time continuous monitoring of mercury would also provide options for advanced
process control feedback, as well as monitoring the performance of control technology. This
would minimize the sorbent usage and thus lower the operating costs. The development of Hg
emission monitors would also benefit other stationary sources such as municipal and sewage
waste incinerators, commercial and industrial boilers, medical waste incinerators, crematoria and
smelters.
In 1996, the authors' research team was invited by the Energy and Environmental Research
Center (EERC), University of North Dakota to participate in a comparison study of mercury
speciation measurements in flue gas. This proved to be a very difficult application given the low
mercury concentration (1-4 ppbv) and difficult flue gas matrix. During these tests we discovered
that the Amasil tube could be used to collect Hg species at elevated temperatures so that the
condensation of acid gases could be avoided. Using a manual collection procedure, we achieved

STOCKWELL ET AL. ON FOSSIL FUELS

215

reasonable agreement (+/-20 %) with the wet chemical methods. However, we discovered later
from Benchscale studies performed at the EERC that this approach had some limitations. The
combination of nitrogen oxides (NOx) with hydrogen chloride (HCI) to form nitrosyl chloride
(NOC1) or the presence of chlorine (C12) at low ppm concentrations in the flue gas severely
lowered the collection efficiency of rig on the gold substrate.
With the increasing demand for mercury speeiation measurements, we developed an online
wet chemical speciation module. In this case, the hot flue gas was separated into two streams
performing different chemistries. The first stream represents total Hg, since an alkaline starmous
chloride (SnC12) reagent was used to convert divalent mercury to elemental mercury. The second
channel utilized a reagent of potassium chloride (KC1) to complex divalent mercury, whereas the
elemental fraction was un-retained. The difference between the two streams was attributed to
oxidised Hg. Figure 5 summarizes the process in the form of a schematic diagram. The system
was used to develop the first fully automated online speciated measurements of mercury in flue
gas. As a research tool in the development and optimization of mercury control technologies, the
system has proved invaluable.

I
LU

3
U.

Prim~l~Q~s

aVp~

FIG. 5--Schematic diagram of wet chemical speciation module and amalgamation - AFS.
The majority of research to date has been completed on bench and pilot scale facilities.
Testing in the field for long-term duration has proved to be somewhat more difficult. Numerous
measurement issues have come to our attention, which need to be addressed before the full
potential of rig emission monitors can be realized.
One of the major disadvantages of the mercury measurement is the industry standard Ontario
Hydro method. Whereas most mercury CEMs provide data every few minutes, the results from
the Ontario Hydro method are often not available for 1-3 months. Experience from our research
shows that the system described above will operate reliably over long periods of time and will
provide data that compare well with the industry standard method. Figure 6 shows results of the
series tests comparing the wet impinger based system with the Ontario Hydro method. The
mercury CEM data agrees to within 10 % of the Ontario Hydro method, and it is available within
a few minutes of the measurement being taken.

216

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 6--Comparison of Ontario Hydro and PSA Hg CEM.

Recent Developments for Continuous Emissions Monitors for Mercury in Stack Emissions
Regulations relating to the levels of mercury emissions are considered in the USA and
elsewhere. Whereas the problems of measurement are similar to those for mercury in natural gas,
the matrix is quite different and often more challenging. On the one hand, you have a
hydrocarbon matrix, and in stack emissions you have a set acidic matrix which contains sulfur
dioxide, sulfur trioxide, and other by-products from burning fossil fuels. The standard methods
using the Ontario Hydro method and some commercial systems based on wet chemistry provide
data that can correlate well. These are demanding on the involvement of staff to ensure that the
instruments remain operational. A system that requires less intervention, essentially a dry-based
system, is desirable. The challenges of this requirement are being address by several groups, but
the inherent problems of a suitable reference method that can be easily operated is hampering
this achievement.
A schematic diagram of the dry based system developed by the authors is shown in Fig. 7.
This allows the measurement of both speciated and total mercury levels in a continuous manner.
Comparative results for the wet and dry based systems are shown in Fig. 8. This shows
promise, and the next priority is to challenge the instrurnental system with a range of coal types
and to ensure that comparative results are obtained across the full spectrum of available fuels.

STOCKWELL ET AL ON FOSSIL FUELS

217

FIG. 7--Schematic diagram for a dry based mercury speciation module.

FIG. 8--Comparison of data provided by both wet and dry based speciation module using a
continuous emissions monitor for mercury.
Conclusions

Atomic fluorescence, coupled to innovative sample conditioning arrangements, offers a


significant advantage over other approaches. In the petrochemical industry, understanding the
fate of mercury is vital to the efficient production of the range of products. The presence of
mercury will create financial implications, and therefore there is the motivation both to measure
and control the mercury levels. In the coal fired utilities area, where the possibility of regulatory
control is an issue, the commitment for the samples' potential to provide solutions at a fast time
scale is not really evident, and this is delaying developments in this area. Wherever we look,
mercury is around us; its measurement and removal from the environment are goals that will

218

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

only be beneficial if we continue to monitor the use of the fossil fuels. Atomic fluorescence
spectrometry offers an affordable approach to the measurement of mercury, and several
procedures to provide accurate methods have been described above.
References

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]

EPA, Quality Criteria for Water, US Environmental Protection Agency, Office of Water
Regulations and Standards, EPA 440/5-86-001, 1986.
Thompson, K. C. and Reynolds, G. C., Analyst, Vol. 96, 1971, p. 771.
Bloom, N. S., Fresenius JAnal Chem, Vol. 355, 2000, pp. 438-443.
Stockwell, P. and Godden, R., Journal of Analytical Atomic Spectroscopy, Vol. 4, 1989, p.
301.
Kammin, W. and Knox, R., Environmental Laboratory, August/September 1992.
Potter, B. et al., US EPA Method 245-7, Revision 1-1, Environmental Monitoring and
Systems Laboratory, Cincinnati, Ohio, 1994.
Knox, R., Kammin, W. R. and Thomson, D., Journal of Automatic Chemistry, Vol. 17, No.
2, pp. 65-71.
Dnmarey, et al., Anal. Chem., Vol. 57, 1985, pp. 2638-2643.
"Mercury in Nitrogen Gas Research Gas Mixture," EPA Reference: DW13939860-01-0,
G. D. Mitchell and W. D. Dorko, Anal. Chem. Div., Chemical Science & Technology Lab,
National Institute of Standards and Technology.
Ohnez, I., Ames, M. R., Gullu, G., Environmental Science and Technology, Vol. 32, 1998,
pp. 3048-3054.
Schindler, D. W., Proceedings for USEPA/NERL/NACEC Science experts" workshop on
Hg, Oct 1998, Las Vegas, pp. 20-30.
Mason, R. P., Fitzgerald, W. F., Morel, F. M. M., Geochimica et Cosmochimica Acta, Vol.
58, No. 15, August 1994, pp. 3191-3198.
Durham, M. D., Schlager, R. J., Hyatt, D. E., Fuel Processing Technology, Vol. 39, 1994,
pp. 285-299.

OTHER HETEROATOMS

Journal of ASTM International, October 2005, Vol. 2, No. 9


Paper ID JAI12989
Available online at www.astm.org

Frank P. Di Sanzo I and John W. Diehl 1

Recent Advances in Gas Chromatographic/Atomic Emission


Hetero-Atom Selective Detection for Characterization of
Petroleum Streams and Products
ABSTRACT: The selective detection of organo-compounds containing hetero-atoms is used extensively
in the petroleum industry for process control (sulfur, nitrogen, etc.) and additives in fuels (sulfur,
nitrogen, oxygen, phosphorous, etc.). The most common selective atom detection is performed with
atomic emission (AED), sulfur/nitrogen chemiluminescence (SCD, NCD), and photometric (FPD, pFPD). Atomic emission detection has an advantage over the other selective detectors since in an
automated analysis scheme several hetero-elements may be detected for rapid sample screening. This
paper describes the unique capabilities of atomic emission detection of fuels and petroleum fractious
using several applications. In addition, a simplified version of comprehensive GCxGC that enhances
hetero-atom selectivity by significantly reducing the matrix interference from the large hydrocarbon
background is coupled with a microwave induced helium plasma atomic emission detector
(GCxGC/AED). The GCxGC uses simple valve based flow modulation optimized for samples with a
boiling point in the gasoline and jet fuel range. The technique is useful for screening fuels additives,
contaminants, etc.
KEYWORDS: Gas chromatography, atomic emission detection, betero-atom speciation, sulfur,
comprehensive two-dimensional gas chromatography, comprehensive two-dimensional gas
chromatography/AED, GCxGC/AED, GCxGC, fuels, gasolines, jet fuels

Introduction
The selective detection o f hetero-atom containing organo-compounds is used extensively in
the petroleum industry for process control (sulfur, nitrogen, etc.) and additives in fuels (sulfur,
nitrogen, oxygen, phosphorous etc.). The most c o m m o n selective atom detection is performed b y
gas chromatography with atomic emission (AED) [1-3], sulfur and/or nitrogen
chemiluminescence [4-8] and flame photometric (FPD, p-FPD) [9].
Furthermore, in some cases, these detectors m a y be used in conjunction with a higher
chromatographic resolution/selectivity to m i n i m i z e matrix interference. Comprehensive twodimensional gas chromatography (GCxGC) has b e e n under development for more than ten years
[10(and references contained therein)] as an enhanced resolution technique. The technique
involves continuously transferring or modulating short time periods from a primary column, the
first dimension, to a second c o l u m n o f different polarity or selectivity, which serves as the
second dimension. The modulation period is typically less than the peak widths observed on the
Manuscript received 15 October 2004; accepted for publication 11 April 2005; published October 2005. Presented
at ASTM Symposium on ]~lementalAnalysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
l ExxonMobil Research and Engineering, Analytical Science Laboratory, Paulsboro, NJ 08066.
This paper was also part of an invited presentation at the Second International Conference on GCxGC, Atlanta, GA,
Sept. 1-3, 2004.
Copyright9 2005by ASTMInternational,100Ban"HarborDrive,POBoxC700,WestConshohonkon,PA 19428-2959.
221

222

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

primary column, and the total elution time from the second dimension is also less than or equal
to the modulation period. Because of the narrow peak widths of the compounds eluting from the
second dimension, detectors for the technique have been limited to those with very fast
acquisition rates such as flame ionization and time of flight mass spectrometry. In this paper the
GCxGC process is 'slowed down' by use of 0.53 mm i.d. Megabore columns and relatively long
modulation times using valve flow modulation [11], thus potentially allowing a wider variety of
detectors. In addition, this paper describes the unique capabilities of atomic emission detection of
fuels and petroleum fractions. A simplified version of comprehensive two-dimensional GCxGC
to enhance hetero-atom selectivity is coupled with a microwave induced helium plasma atomic
emission detector (GCxGC/AED) using a total sample transfer valve based flow modulation for
samples boiling in the gasoline and jet fuel range. The technique is useful for screening fuels
additives, contaminants, etc.

Experimental
An Agilent Technologies 6890N gas chromatograph/G2350 atomic emission detector (note: a
current supplier of GC/AED is Joint Alytical Systems, Moer, Germany) was configured as
follows:
On-Column
Injector
Injector Temperature Program
50~ (0.1 min) 30~
to 260~ (until end of oven
program).
0.1 #L with a 7683 autosampler.
Injection Volume

Columns
First Dimension
Second Dimension
Carrier first dimension
Carrier second dimension
Oven Temperature Program

Agilent Technologies 60 m x 0.53 mm i.d. 5.0 micron


film HP-1.
Agilent Technologies 60 m x 0.53 mm i.d. 1.0 micron
film DB-WAX
Helium at 2 mL/min
Helium at 50 mL/min
40~ (0 min) 2~
to 250~ (15 min or until all
compounds of interest eluted).

Detector (AED)
Cavity temperature
Transfer Line Temperature
Helium Make-up flow
Reagent Gases

260~
260~
20 ml/min
Research grade hydrogen, oxygen, and 10 volume %
methane in nitrogen. Flows set per factory specifications.

Elements Detected
Carbon
Sulfur
Nitrogen
Oxygen
Phosphorous
Data Acquisition Rate

Wavelength (rim)
179
181
174
171
186
10 points/s (10 Hz)

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

223

GCxGC Modulator
General Set-Up--A Valco DC8WT 1/16 in. 8 port valve (maximum temperature = 350~
was configured with two 5 mL gas loops as shown in Figs. la and lb. Turning the valve on and
off allowed effluent from the first dimension to be collected alternately in one of the two gas
loops, while the other loop was being swept into the second dimension. The valve was pulsed on
and off with an Agilent valve actuator controlled by a Valco digital valve interface and a Valco
digital valve sequence programmer. The modulation interval was controlled by the sequence
programmer, which was started and stopped by timed contact closures from the gas
chromatograph. Columns were connected to the valve with Restek fused silica lined stainless
steel tubing, Swagelok 1/16 in. low-dead volume unions and Supelco M2-A 1/16 in. graphitevespel ferrules. The modulator was synchronized by adjusting the start signal (in this case to 1.70
min) until pentane, relatively unretained on Carbowax column, eluted near the beginning of its
respective second dimension interval. Modulator discrimination was tested with n-paraffins from
C5 through C12. As expected with first dimension column flow, loop sizes, and modulation
interval used, no loss was found and no sample was vented during modulation. The valve was
mounted at 330~ in a standard valve oven on the 6890 GC. Gas lines from the actuator on this
oven were connected to the Valco digital valve interface.
Data System
A Hewlett Packard Vectra Pentium IV personal computer operating under Microsoft
Windows NT 4.0 and Agilent Technologies A.07.01 ChemStation software was used for data
acquisition and instrument control. Two dimension plots were obtained with Fortner Software
Transform Version 3.4 after the signal files were exported in CSV file format by the
ChemStation software.
Pure compounds were purchased from Aldrich Chemical Co., Milwaukee, WI.
Results and Discussion

The system described above was used to screen a number .of fuel samples for bulk
composition, additives, and potential contaminants. The following describes qualitatively a
number of these applications.
Figure 2 shows the carbon specific C-GCxGC chromatogram o f a fluidized catalytic cracking
(FCC) gasoline. Note that the compounds present were resolved into distinctive bands for nonaromatics, mono-aromatics, and di-aromatics. This can also be seen in Fig. 3, which shows the
C-GCxGC separation of a steam-cracked naphtha (SCN). Because of the process used to produce
this gasoline, additional aromatic compound classes such as styrenes and indenes were identified.
Figures 4 and 5 show the sulfur specific chromatograms (S-GCxGC) of the two above FCC
and SCN samples. The sulfur compounds were also resolved by the GCxGC process into sulfur
classes with distinct bands for mercaptans + sulfides, thiophenes, and benzothiophenes. Note that
there was a sulfur compound in the steam cracked naphtha at approximately X=55 min, Y=110 s.
This unknown compound was also found to contain carbon, hydrogen, and oxygen by further
GCxGC/AED analysis. Further characterization of the compound may be possible by GC/Mass
Spectrometry (GC/MS) using a comprehensive GCxGC/GC/MS set up very similarly to that
described in this paper.

224

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIGURE 1A.
Flo~ Modulation Valve
Diagram
Second
Dimension
(DB-~AX)
First
Dimension
(HP-1)

Loop 1 - -

Loop 2
FIGURE lB.
Flow Hodutation Valve
Diagram
~

"I

~
_________+ /

Second
Dimension
(DB-iIAX)

First
Dimension
(HP-1)-ON

Laap 1

>
I

Loop 2

FIG. 1--Flow modulation valve diagrams.

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

FIG. 2--Carbon GCxGC/AED of a fluid catalytic cracked (FCC) naphtha.

FIG. 3--Carbon GCxGC/AED of a steam cracked naphtha.

225

226

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 4--Sulfur GCxGC/AED of a FCC naphtha.

FIG. 5--Sulfur GCxGC/AED of a steam cracked naphtha.

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

227

Figure 6 shows the C-GCxGC separation of a racing fuel. Note that this mixture blended for
high engine performance had a very distinct fingerprint compared to the full range naphthas
shown in Figs. 2 and 3. Benzene was absent, but toluene was very abundant. Figure 7 shows the
oxygen specific separation (O-GCxGC) of this same fuel and the presence of methyl-tbutylether.
Figure 8 shows the C-GCxGC of a jet fuel. In this separation, the oven temperature program
rate was reduced from 2~
to I~
As with gasoline, the hydrocarbon aromatic classes
were readily resolved, and their carbon number distributions are readily distinguishable from C7
to C13. As expected, this sample did not contain benzene. Note that cleanly separated
compounds at approximately X = 85 min, Y = 90 s (labeled 'phenolics') were almost as retained
as the di-aromatics (naphthalenes). The O-GCxGC separation (Fig. 9) revealed the presence of
oxygen in these compounds, which were identified as phenolics.
Figure 10 shows the C-GCxGC of a hydraulic fluid. Phosphorous (Fig. 11) P-GCxGC and
oxygen O-GCxGC (Fig. 12) specific detections were consistent with the fact that the most
abundant compound at approximately X = 105 rain, Y = 30 s was a phosphate ester. Two
additional compounds at approximately X = 115 rain, Y = 80 s contained only carbon and
oxygen, which was consistent with their identification as epoxides.
The effect of the increased selectivity and detection limits using GCxGC over single
dimensional GC/AED [1-3 and references therein] for the elements in such samples was not
determined in this study. However, it is expected that the detectability of many of the elements is
enhanced by the removal of a significant amount of the hydrocarbon matrix background, such as
the saturated, olefinic, and much of the aromatic hydrocarbon compounds, by using the second
dimension separation of GCxGC. Therefore, it is expected that in many cases the detection limits
of the elements will approach those determined with pure compounds using such a system.

FIG. 6--Carbon GCxGC/AED of a racing fueL

228

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 7--Oxygen GCxGC/AED of a racing fuel.

FIG. 8--Carbon GCxGC/AED of a jet fuel.

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

FIG. 9--Oxygen GCxGC/AED of a jet fuel.

FIG. 1O---Carbon GCxGC/AED of a hydraulic fluid.

229

230

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 11--Phosphorous GCxGC/AED of a hydraulic fluid.

FIG. 12--Oxygen GCxGC/AED of a hydraulic fluid,

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

231

Conclusion

Atomic emission detection coupled with a simplified version of comprehensive GCxGC to


reduce hydrocarbon matrix interference using simple and rugged modulation, together with wide
bore capillary columns, may be used successfully for the analyses and screening of light boiling
fuels, such as gasolines and jet fuels. The technique, together with other spectroscopic techniques
such as GC/MS, can provide information on many selective elements in compounds that may be
added to fuels as additives or be present as a result of unwanted contaminations.
References

[1]
[2]
[3]
[4]

[5]
[6]
[7]
[8]
[9]
[10]

[11]

Albro, P. A., Dreifuss, P. A., and Worrnsbecher, R. F., "Quantitative Determination of


Sulfur Compounds in FCC Gasoline by GC/AED," J. High ResoL Chromatog. Chromatog.
Commun, Vol. 16, 1993, pp. 13-17.
Quimby, B. D., Grudoski, D. A., and Giarrocco, V., "Improved Measurement of Sulfur and
Nitrogen Compounds in Refinery Liquids Using Gas Chromatography-Atomic Emission
Detection," J. ChromatographicSci., Vol. 16, 1998, pp. 435--443.
Diehl, J. W., Finkbeiner, J. W., and Di Sanzo, F. P., "Determination of Ethers and Alcohols
in Reformulated Gasolines by Gas Chromatography/Atomic Emission Detection," J. High
Resol. Chromatog. Chromatog. Commun,Vol. 18, 1995, pp. 108-110.
Di Sanzo, F. P., Bray, W., and Chawla, B., "Determination of Sulfur Components of
Gasoline Streams by Capillary Column Gas Chromatography with Sulfur
Chemiluminescence Detection," J. High Resol. Chromatog. Chromatog. Commun,Vol. 17,
1994, pp. 255-258.
Chawla, B. and Di Sanzo, F. P., "Determination of Sulfur Components in Light Petroleum
Streams by High-Resolution Gas Chromatography with Chemiluminescence Detection," J.
of Chromatography, Vol. 589, 1992, pp. 271-279.
Wang, F. C., Robbins, W. K., and Greaney, M. A., "Speciation of Nitrogen-Containing
Compounds in Diesel Fuel by Comprehensive Two-Dimensional Gas Chromatography,"J..
Separation Science, Vol. 27, 2004, pp. 468-472.
Chawla, B., "Speciation of Nitrogen Compounds in Gasoline and Diesel Range Process
Streams by Capillary Gas Chromatography with Chemiluminescence Detection," J.
Chromatographic Science, Vol. 35, 1997, pp, 97-104.
Wang, F. C., Robbins, W. K., DiSanzo, F. P., and McElroy, F. C., "Speciation of SulfurContaining Compounds in Diesel by Comprehensive Two-Dimensional Gas
Chromatography,"J. ChromatographicScience, Vol. 41, 2003, pp. 519-523.
Chambers,L. and Duffy, M. L., "Determination of Total and Speciated Sulfur Content in
Petrochemical Samples Using a Pulsed Flame Photometric Detector," ./. Chromatographic
Sci., Vol. 41, 2003, pp. 528--534.
Gorecki, T., Harynuk, J., and Panic, O., "The Evolution of Comprehensive TwoDimensional Gas Chromatography (GCxGC)," J. Separation Sci., Vol. 27, Nos. 5-6, 2004,
pp. 359-379.
Gorecki, T., Harynuk, J., and Panic, O., "The Evolution of Comprehensive TwoDimensional Gas Chromatography (GCxGC)," d. Separation Sci., Vol. 27, Nos. 5-6, 2004,
p. 373.

Journal of ASTM International, November/December2005, Vol. 2, No. 10


Paper ID JAIl 2990
Available online at www.astm.org
L i s a J. Nash l

Improvements in the Determination of Fluorine in Fuel and


Lubricants by Oxidative Combustion and Ion-Selective
Electrode Detection
ABSTRACT: Oxidative combustion combined with ion-selective electrode detection is one of several
techniques available for the analysis of fluorine in fuels and lubricants to the petroleum and
petrochemical industry. This technique was first developed in the mid-1990s to provide an alternative to
the widely used Wickbold method, which is very time consuming and extremely hazardous. The early
designs were marginally successful due to the lack of sensitivity and stability, the use of large volumes of
solvents, inconsistent delivery of fluids, plugging of catch and hold valves, numerous sources of
contamination, and long analysis times. The design has been refined to enhance the sensitivity, stability,
and accuracy of the fluoride determination. Enhancements include improved ion-specific electrodes with
increased sensitivity, the use of circulating baths to stabilize temperature fluctuations, improved sample
introductionand flow systems, and redesigned electronic circuitry.
KEYWORDS: fluorine, fluorine, fluoride, ion-selective electrode, fuel analysis, oxidative combustion,
elemental analysis
Introduction

The determination o f fluorine in various liquid and gaseous hydrocarbons is vital at many
points in the refining process primarily in any blend component that has been sourced from the
hydrogen fluoride (HF) Alkylation Unit. Fluorinated compounds poison process catalysts;
therefore, it is essential that process feeds be as free o f fluorine as possible. As an example,
butane is used to produce methyl tertiary-butyl ether (MTBE). The butane must be fluorine free
prior to butane isomerization to prevent the poisoning o f the process catalyst. In addition, any H F
acid or its combustion products m a y be extremely destructive in any environment. Therefore,
any finished hydrocarbon product or synthesized material that is utilized in the presence o f
sufficient heat (i.e., car engine), such as fuel and lubricating oils, must be flee o f fluoride.
Classical measurement o f fluorine in inorganic and organic materials included sample
collection, preparation, and determination. For the reliable determination o f fluorine in any type
o f sample, the fluorine: 1) was separated from the interfering substances, or the interferences are
suppressed; 2) the experimental blank was low, and the final concentration o f fluorine was
adequate for measurement with the method used; and 3) the results were corroborated b y an
independent method o f analysis [1]. Preparation involved one or more o f the following stages:
washing, drying, ashing, fusion, acid extraction, distillation, or diffusion. Ashing and fusion were
sometimes necessary to oxidize organic matrices and to release fluorine from refractory
compounds, respectively. Separation was used to avoid interference and/or as a means o f
concentration. Separation techniques included distillation, diffusion, solvent extraction,
Manuscript received 8 October 2004; accepted for publication 23 May 2005; published November 2005. Presented
at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
l Applications Laboratory Manager, Antek/PAC, 300 Bammel Westfield Road, Houston, TX 77090.
Copyright9 2005byASTMInternational,100Ban"HarborDrive,POBoxC700,WestConshohocken,PA 19428-2959.
232

NASH ON FLUORINE IN FUEL AND LUBRICANTS

233

chromatography, adsorption, pyrolysis, and pyrohydrolysis. Once the fluorine was released, it
was typically measured in the form of fluoride ions (F).
Many methods were available for the determination of fluoride in suitably prepared samples
including titrimetry, spectrophotometry, fluorometry, gas chromatography, ion chromatography,
mass spectrometry, and potentiometry. However, the advent of the fluoride specific electrode,
developed by Frant and Ross in 1966 [2], greatly decreases the use of spectrophotometric and
fluorometric techniques because it offers speed and relative freedom from interference; and in
some circumstances, separation may not be necessary. The ion-selective electrode (ISE) is a
variation on the pH electrode with a different sensing membrane and internal solution. Figure 1
illustrates the measurement of potential difference between an ISE and a reference electrode. The
membrane of a fluoride specific electrode consists of a single lanthanum crystal which has been
doped with europium fluoride. The electrode responds to the activity of fluoride ions in aqueous
solutions and is directly proportional to concentration. It is very selective, and its only interferent
is the hydroxide ion (OH-). The pH is adjusted to >5.5 to eliminate this interference. The
electrode can determine fluoride ion activity over a wide concentration range and down to
0.1 #g/g F" in the presence of several other ions.

FIG. 1--Measurement with an ion-selective electrode.


The fluoride selective electrode is routinely used for the determination of fluoride in drinking
water, industrial effluents, sea water, air, aerosols, flue gases, soils and minerals, urine, serum,
plasma, plants, and other biological materials. The following discussion focuses on the
development of analogous techniques for the determination of fluorine in hydrocarbons.
Early Methods

Several methods were developed prior to 1990 for the combustion of organic fluorine
compounds like those found in hydrocarbons. The Parr bomb method was first developed in the

234

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

early 1950s. The Parr combustion apparatus consists of a sealed steel reactor that is pressurized
to 40 atm (4053 kPa) with oxygen. The sample is ignited and the combustion process releases
covalently bound fluoride. The fluoride is converted to its anion and analyzed by titration or
other wet chemical techniques. This method works for most organic fluorines except in the case
of highly volatile compounds and compounds with high fluorine content where there may be
difficulty in the decomposition step due to incomplete combustion and the formation of salt
residue.
The oxy-hydrogen flame combustion method was developed in the late 1950s by R.
Wickbold [3,4,5] to address the critical decomposition step during combustion. The method,
commonly referred to as the Wickbold method, is used for the analysis of liquids and liquefied
petroleum gases but may be used for viscous liquids and solids if the sample can be dissolved in
an appropriate combustible solvent. In the present method, UOP Method 619-83 Fluorine in
Hydrocarbons by Oxy-Hydrogen Combustion, the sample is passed through an oxy-hydrogen
burner in combination with a Wickbold quartz-tube combustion apparatus which decomposes the
sample to carbon dioxide and HF. The combustion products are absorbed in 1 % aqueous boric
acid and treated with formaldehyde to remove peroxides. The fluoride content can be determined
by several means including measuring the extent of bleaching action by the fluoride on an
intensely red zirconium-SPADNS (4,5-dihydroxy-3-(p-sulfophenylazo)-2,7-naphthalenedisulfonic acid trisodium salt) complex or potentiometric titration. The elapsed time for a single
analysis is given as 2.8 h in the written method but is often closer to 4 h in actual practice. After
burning a sample of high fluoride concentration, the Wickbold system must be purged free of
entrapped fluoride by running consecutive blank determinations until a constant low level of
fluoride is obtained before a sample of low fluoride can be accurately determined. The procedure
can be hazardous due to the use of hydrogen and oxygen and the potential for peroxide
formation.
The oxygen flask method (also called the Schrniger flask combustion method) involves the
ashing of the organic material in a closed flask containing oxygen and an absorbing solution in
which the fluoride is subsequently determined by ISE or titration. This method is relatively
simple; however, the size of the combustion flask limits the amount of the sample that can be
used, occasionally causing incomplete combustion. Also, due to the absence of a concentration
step, this method has a higher limit of detection.

Oxidative Combustion/ISE Method Development


All of the previously described methods have a separate combustion step prior to detection.
A combination of combustion and detection in one instrument would greatly enhance the
determination of fluorine, making it simple and rapid to operate and suitable for automation. The
goal was to develop a rapid, sensitive, and accurate method with safe operation for the analysis
of fluorine in liquids, solids, and gases with little to no sample preparation.
In the mid-1990s, development of a technique began which combined oxidative combustion
with ISE detection in a flow through system to provide an instrumental alternative to the
aforementioned methods. In this method, the organic fluoride-containing sample is delivered to a
quartz pyrotube in an 800-1100~ furnace. The sample is burned in the presence of excess
oxygen and steam (pyrohydrolysis). Excess oxygen and high temperature are needed to ensure
complete combustion of the sample and to form free fluoride ions. The steam provides sufficient
hydrogen ions to react with the released fluoride ions rapidly to form HF, thus preventing the
fluoride from reacting with the quartz walls of the pyrotube. The combustion gases and the HF

NASH ON FLUORINE IN FUEL AND LUBRICANTS

235

are bubbled through a solution of Total Ionic Strength Adjustment Buffer (TISAB). The TISAB
serves to dissociate the hydrogen from HF, to prevent the formation of H E and HF2", which
cannot be detected by ISE, and to adjust the pH to >5.5 to prevent any interference from
hydroxide ions. The dissociated fluoride ion concentration in the solution is measured directly by
ISE. The reaction chemistry is shown below.

R-F + H20 + O~ . . ~ . ~
HF T ~

HF + CO 2 + 1t20 + oxides
F"

The first fluoride instrument design employed catch and hold valves which served to capture
and concentrate the combustion products prior to detection. The results achieved with this
oxygen combustion/ISE method were compared with the commonly used Wickbold method. It
was shown that the results corroborated and, furthermore, the relative standard deviations were
decreased with the oxygen combustion/ISE method (see Table 1).
TABLE 1--Com arison of Wickbold and oxidative combustion/ISE.
Method
Wickbold
Oxidative
Combustion
Sample
Conc.
RSD
Conc.
RSD Variance
(#g/g)
(%)
(ppm)
(%)
Between
Methods
LPG 1
6.1
10.7
6
3.8
2%
LPG 2
18.3
8.3
17.1
2.6
7%
LPG 3
0.3
8.8
0.3
7.6
0%
Fluorobenzene
116
10.4
114
3.5
2%
in Toluene Std.
Fluorobenzene
1048
3.8
1013
0.1
4%
in Toluene Std.
The initial design proved to be marginally successful due to the use of large volumes of
solvents, inconsistent delivery of fluids, plugging of catch and hold valves, numerous sources of
contamination, and long analysis times. The catch and hold valves were eliminated in favor of a
direct delivery of the solution containing the dissociated fluoride ions and TISAB to the ISE.
This arrangement required the solution to be cooled prior to the ISE to prevent damage to the
electrode and an attempt to introduce the sample to the ISE at a constant temperature. A
receiving vessel/condenser downstream o f the pyrotube was added to accept the combustion
gases and also served to mix the gases with the TISAB. A cooling block and fan were installed
that served to reduce the temperature of the mixture prior to the ISE. This design was more
successful than the first, but temperature fluctuations were experienced with the air-cooled
condenser, resulting in varying potential readings from the temperature sensitive electrode.
Beginning in 2000, numerous modifications were made to the fluoride instrument design to
greatly enhance the sensitivity, stability, and accuracy of the fluoride determination. Several
electrodes were evaluated to determine the most sensitive and durable fluoride electrode
available. The ELIT 8221TM, a solid state sensor with a mono-crystalline membrane, was chosen.

236

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

The electrode specifications include a working temperature range of 5-80~ a linear measuring
range of 0.1-1 900 ppm F, and a detection limit of 0.02 mg/kg F. Use of this electrode greatly
increased the sensitivity of the instrument as can be seen in Fig. 2. Figure 2 shows the
comparison of results of a fluoride determination of 5 mg/kg in isooctane between the previously
used electrode and the new electrode.
The existing air-cooled condenser was replaced with a vertical, thermostatically controlled
water condenser located directly at the exit of the pyrotube where the TISAB is added to mix
with the exiting combustion gases. This arrangement greatly reduced the temperature of the
effluent of the pyrotube and effectively eliminated one meter of heated transfer tubing and
potential fluoride losses. Furthermore, another thermostatically controlled condenser was added
after the vertical condenser to maintain the condensate temperature at 40~ while flowing
through the electrode cell. In this way, the temperature fluctuations were alleviated, and there is
significantly less drift in the electrode readings (see Fig. 3). Additionally, the time required to
stabilize the instrument before use and between analyses was significantly reduced.

FIG. 2--Comparison of sensitivity achieved with initial (top) and current (bottom) design.

FIG. 3--Comparison of resulting baseline from initial (top) and current (bottom) design.

NASH ON FLUORINE IN FUEL AND LUBRICANTS

237

Other upgrades included a redesigned fluoride detector board with three attenuation levels
and improved sample introduction techniques. Currently, the software is being updated to
incorporate automation and to use a variety of integration techniques. Although the initial
oxidative combustion/ISE designs were able to achieve comparable results to existing
techniques, the culmination of recent improvements to the instrument design has greatly
enhanced the sensitivity and stability of the fluoride determination and decreased the overall
analysis time.
Results and Discussion

The oxidative combustion/ISE instrumental method is used routinely for the analysis of fuels
and liquefied petroleum gases (LPG). A standard method, UOP Method 975-02 Fluorine in
Liquid Hydrocarbons and LPG by Oxidative Combustion with 1on-Selective Electrode, is
available through ASTM International. The method is applicable to liquid hydrocarbons with
fluorine in the range of 1-500 ng/~d and LPG in the range of 0.5-500 ng/~tl with the appropriate
calibration standards and materials. Higher concentrations can be analyzed by dilution in an
appropriate solvent or higher concentration standards can be used as the calibration is linear
through several orders of magnitude (see Fig. 4).

FIG. 4 Calibration with Fluorohexane in lsooctane.


Currently, the method is used predominantly for the analysis of liquid and gaseous process
feeds and automotive lubricants. Results are shown below for an automatic transmission fluid
(ATF) and butane (see Figs. 5 and 6, respectively). These results serve to illustrate the sensitivity
and repeatability of the fluoride measurement that can be achieved with the current design. With
the improved detector response, the level of detection has been shown to be less than that stated
in the UOP Method 975-02. Work has started to develop the method as an ASTM standard for
the measurement of fluoride in LPG and to establish the lower limit of detection, repeatability,
and reproducibility.

238

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 5--Analysis of 30 mg/kg F in automatic transmission fluid (in duplicate).

FIG. 6--Analysis of l Omg/kg F in Butane (in triplicate).


Conclusions

The oxidative combustion with ion-selective electrode method described here can be used to
accurately determine fluorine in liquid hydrocarbons and liquefied petroleum gases at levels
down to 0.5 ng/Ixl. Results obtained are in good agreement with those achieved with the
Wickbold Oxy-Hydrogen Combustion method. The oxidative combustion with ion-selective
electrode method is less time consuming, less hazardous, and more sensitive than the Wickbold
method. The analysis is also applicable to a range of sample types not discussed here, including
aqueous based systems and solids.
References

[1] Venkateswarlu, P., "Evaluation of Analytical Methods for Fluorine in Biological and Related
Materials," Journal of Dental Research, Vol. 69, Special Issue, 1990, pp. 514-521.
[2] Ross, J. W. and Frant, M. S., Science, Vol. 154, t966, pp. 1553.
[3] Wickbold, R., Angew. Chemistry, Vol. 69, 1957, pp. 530-533.
[4] Wickbold, R., Angew. Chemistry, Vol. 64, 1952, pp. 133.
[5] Wickbold, R., Angew. Chemsitry, Vol. 66, 1954, pp. 173.

Journal of ASTM International, October 2005, Vol. 2, No. 9


Paper ID JAI12977
Available online at www.astm.org

T. W. Selby, 1 R. J. Bosch, 2 and D. C. Fee 2

Phosphorus Additive Chemistry and Its Effects on the


Phosphorus Volatility of Engine Oils
ABSTRACT: The most widely used and effective anti-wear/anti-oxidation additives in engine oil
contain phosphorus that can partially volatilize during engine operation. Unfortunately, volatile
phosphorus in the exhaust stream degrades the fimction of the exhaust catalyst in reducing air pollution.
Earlier studies in a special volatility bench test using the Phosphorus Emission Index have shown that
phosphorus volatility is not related to engine oil volatility or to phosphorus content in the unused engine
oil. At the time, it had been speculated that this unexpected lack of correlation with initial phosphorus
concentration could be explained by 1) the effects of other engine oil additives and/or 2) variations in the
phosphorus additive chemistry. The first speculation was relatively recently confn-med by taxi fleet
studies of catalyst degradation by phosphorus-containing oils by the Ford Motor Company. It remained to
determine if differences in phosphorus additive chemistry were also a factor, and this is the subject of the
present study.
KEYWORDS: phosphorus volatility, Phosphorus Emission Index, PEI, ZDDP, zinc dithiophosphates,
exhaust catalyst, exhaust emissions

Nomenclature
ZDDP
PEI
API
ILSAC
EGR

zinc di(organo) di(thio) phosphate


Phosphorus Emission Index
American Petroleum Institute
International Lubricant Standards and Approval Committee
exhaust gas recirculation system

Introduction and Background


Importance o f ZDDPs
The class o f anti-wear, anti-oxidation additives c o m m o n l y k n o w n as ZDDPs has been a
mainstay o f engine oil formulation and performance for more than 60 years [1]. Engine designers
and metallurgical engineers have found significant help in overcoming problems b y utilizing the
protection brought b y the chemists and lubrication engineers forming and formulating with these
additives. Concentrations o f Z D D P phosphorus in the formulated engine oil have often been
above 0 . 1 % to ensure the desired level o f protection, even though somewhat lesser amounts
have been shown to suffice. More recently, with emphasis on smaller, more powerful engines
(carrying ever longer warranties), the circumstances would seem ideal for increasing use o f
ZDDPs.
Manuscript received 9 November 2004; accepted for publication 11 April 2005; published October 2005. Presented
at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
1Director of R&D, Savant, Inc., Midland, Michigan.
2 Technical Service Fellow and Process Improvement Specialist, respectively, Astaris LLC, Webster Groves,
Missouri.
Copyright9 2005byASTMInternational,100Ban"HarborDrive,POBoxC700,WcstConshohoeken,PA 19428-2959.
239

240

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Phosphorus and Catalyst Interaction


A problem demanding creative solutions has arisen over the last few decades. Automotive
exhaust pollution control requires expensive exhaust catalysts, and as a result of government
mandates today, these catalysts must remain highly effective for up to 120 000 miles of service.
The problem is that it became evident that these catalysts were degraded by phosphorus in the
exhaust stream [1-8].
Association of Phosphorus Emission with Oil Consumption and Volatility--Oil consumption
caused by oil volatility has always been a concern for automotive engineers because of the
associated deposits often formed on piston rings, combustion chambers, and valves. Concern that
oil volatility could cause phosphorus in the exhaust stream adds to the desire to reduce oil
volatility. Consequently, progressively lower formulated engine oil volatility (as measured by the
classic Noack volatility test [9-12]) was required for oils meeting API and ILSAC GF-2 (22 %
loss) and GF-3 (15 % loss) specifications.
Phosphorus Concentration Reduction and Concerns--A more challenging and debated
direction in limiting phosphorus volatility was to limit the concentration of ZDDPs in engine oil.
Understandably, this brought intense discussion among automotive and lubrication engineers
reagarding:
1. lowering levels of ZDDP in engine oils without considerable field experience in modem
high-performance engine durability, at a time when
2. engine oil operating viscosities and levels of hydrodynamic lubrication were being
reduced to gain fuel efficiency, as well as concern about
3. the impact of such changes on older, used, and presumably more vulnerable automobiles.
Despite these concerns, phosphorus levels were reduced first to a maximum of 0.10 % in lower
viscosity grade passenger car engine oils designated API SJ and ILSAC GF-2 in 1995 and then
in 2004, to a level of 0.08 % for API SM and ILSAC GF-4.

A New Noack Testfor Determining Phosphorus Volatility


Selby-Noack Volatility Test--The toxic nature of the Woods Metal used for heat transfer in
the classic Noack volatility test [9-12] led to the development of a new approach [13-17]. This
approach, while retaining all Noack test conditions, 3 was designed to heat the test oil safely using
noble metal resistive heating and to collect 99+ % of all the material volatilized. Figures la and b
show a picture of the special Noack instntment and a sketch of the arrangement of the essential
components, respectively.
As indicated in Fig. 2, early studies with the special Noack showed that more than 99 % of
the oil mass lost from the volatilization chamber was accounted for either in the collector or the
~1 g on the walls of the tubing leading to the collector [16]. Such efficiency is important to be
sure that the volatilized material collected is essentially identical to all material volatilized
instead of forming a questionable composition.
Analyses of phosphorus in the collected volatiles from the special Noack in these early
studies [13-17] showed that phosphorus was present. Moreover, the limited data also suggested
that phosphorus volatility was not closely related to oil volatility.
3The Noack volatilitytest determinespercent loss of volatile materialfrom 65 gramsof the test oil when exposedto
a small vacuumof 20 mm water at 250~ for one hour.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

241

FIG. 1--(a) Selby-Noack instrument and (b) essential components of Selby-Noack.

Mass Balance of Amount Collected and


AmountVola'dlized
Grams Volatiles Recovered
20

t5

Rz = 0.998
Slope : 0.993 g/g
Intercept: .1.224g

10

Ideal
~

~ j ~ r

j ~ J
j ~ r "
JJ"

~/__~d~'~Actual

10

16

2O

26

GramsEn~lineOilVolatilized
FIG. 2--Mass balance of loss and collection of volatilized material.

Effects of Oil Volatility and Concentration on Phosphorus Volatility


This special Noack approach - permitting collection of all volatiles - was opportune. Among
other applications, it presented the ability to evaluate the relationship of phosphorus volatility to
oil volatility and/or phosphorus concentration in the fresh oil.
A study of data published by the Institute of Materials (IOM) [ 18] for the years 1999 and
2000 of over 1200 oils was used. Phosphorus volatilized from fresh engine oils was compared to:
1) their oil volatilities and 2) initial phosphorus concentrations. Phosphorus volatility was first
shown no__Atto be dependent on engine oil volatility, as can be seen in Fig. 3 where phosphorus
volatility varies independently of the associated engine oil volatility.
In a further analysis of the IOM data, it was surprisingly found that phosphorus volatility is
unrelated to phosphorus concentration in the fresh oil [19,20] as shown in Fig. 4.

242

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 3 Showing lack of relationship between the grams of an engine oil volatilized in the
special Noack test (from an initial charge of 65 g for the tesO and the grams of phosphorus
volatilized during the same test.

FIG. 4---Showing lack of relationship between the concentration of phosphorus in a fresh


engine oil and the concentration of phosphorus volatilized from the particular engine oil in the
special Noack test.
Phosphorus Emission lndex
Considering the lack of relationship of volatilized phosphorus with either oil volatility or
initial phosphorus concentration in the oil, it seemed important to have a measure of the
comparative phosphorus volatility of engine oils. This led to a concept of ranking engine oils
based on the volatility of their phosphorus-containing additive(s) [20]. The ranking method was
called the Phosphorus Emission Index, or PEI.
PEI is defined as milligrams of phosphorus volatilized per liter of fresh oil based on results
obtained in the special Noaek test. Repeatability was tested and shown in past papers [20,21].

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

243

Considering the surprising lack of correlation with phosphorus concentration in the fresh oil,
it was thought to be likely that phosphorus volatility might be dependent on ZDDP variations of
chemistry and/or effects of other additives in the formulation. However, at the time, there was
no evidence whether the Phosphorus Emission Index and the volatile phosphorus to which it was
related had any correlation with the degeneration of exhaust stream catalysts.

Correlation of PEI with Field Study


Further work testing the significance of the PEI concept was undertaken as a consequence of
taxi fleet tests of phosphorus volatility and catalyst degradation conducted by Ford Motor
Company [22,23]. Interest in the PEI concept led Ford engineers to submit three blind-coded,
fresh engine oils of the several that had been used in the taxi fleet tests.
On completion and submission of this blind-sample PEI study to Ford, it was revealed that
the ZDDP, concentration of ZDDP, and the base oil used in each engine oil were the same.
Moreover, one of the oils had no additives present other than the ZDDP.
Results are shown in Table 1 and Figs. 5 and 6, in which four of the Ford test oils are listed.
(Oil X, which contained no phosphorus, was not submitted for PEI evaluation by Ford but has
value in the comparisons made.) The figures show the additives present as well as the level of
phosphorus deposits and the NOx emissions for each oil.

TABLE 1--Comparison of PEI to phosphorus found on exhaust catalyst and NOx emitted at
the end of Ford Motor field taxi tests.

FIG. 5--Correlation of PEI and catalyst.

FIG. 6---Correlation of PEI and NOx.

244

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

This study of field test oils resulted in three important findings on the questions raised in the
earlier work [19-21] regarding the significance of PEI and the effects of other oil additives:
1. The PEI level was found to correlate with the phosphorus found on the catalyst.
2. The presence of other additives reduced the PEI value considerably.
3. PEI also correlated with the grams/mile of NOx emitted.
Thus, the PEI analysis of the Ford Motor Company field study resolved the previous posed
question [20] of whether other additives in the oil formulation can influence the volatility of
phosphorus. More specifically and importantly, it seems apparent that engine oil formulation can
be tailored to control phosphorus volatility.
Interestingly, Ueda and his associates [8] found much earlier that both calcium and
magnesium diminished phosphorus effects on the catalyst. (If these two types of additives were
capable of forming volatiles, they might curtail adverse phosphorus effects.)
The major remaining question was whether the various chemistries of ZDDPs also affect the
volatility of phosphorus, and answering this question was the focus of the present paper.
Studies of the Chemistry and PEI of ZDDPs

Instruments and Techniques


Noack Test Procedures--The Selby-Noack is the only instrument capable of generating the
Phosphorus Emission Index and was consequently used to obtain the PEI data of this study.
Volatile and residual materials were also used to provide samples for the associated NMR
determinations presented in a conjoined paper [24].
As briefly described in the previous section, the routine special Noack technique applies the
same conditions as the original Noack procedure. (For greater detail, this procedure is fully
described in ASTM Method D 5800c [12].) However, in generating data for this study, in some
experiments the technique was modified to be run either for longer temperature ramps to the
operating temperature of 250~ or for shorter intervals than one hour at 250~ In these modified
tests, other than collecting volatiles for longer or shorter time periods, the overall technique was
identical to the routine special Noack test.
Elemental Analyses--A Leeman Model PS 1000 inductively coupled plasma (ICP)
spectrometer was used to determine all elemental concentrations of phosphorus required in these
studies. ASTM Method D 4951 [25] was followed using cobalt at 200 PPM as an internal
standard.
Nuclear Magnetic Resonance Procedure--Small samples of the original oils and the volatile
materials and residues obtained during the special Noack tests were analyzed by 31p NMR at
Washington University (St. Louis, MO). The spectra were obtained either on a 500 MHz Varian
NMR equipped with a 10 mm probe or on a 600 MHz Varian NMR equipped with a 5 mm
probe. Samples were diluted with 10-15 % chloroform-d (CDC13), which also served as an
internal reference for establishing spectral positions. Average data accumulation time for these
spectra was one hour. As mentioned, a conjoined paper [24] covers this aspect of the study.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

245

Initial Studies

Reference Oils
Formulated and Reference Oils--The fully formulated oil, RO 780, was classified as SAE
5W-30 engine oil known to be in APUILSAC categories SL/GF-3. This oil - minus any ZDDP was the primary Reference Oil, RO 779, providing blending stock for specially prepared ZDDPs.
PEIs of the Comparator and Reference Oils--Volatilities and PEIs using the special Noack
were obtained on these two oils, and results are shown in Table 2. As expected, the formulation
without ZDDP, RO 779, shows no phosphorus in either the formulation or the volatiles. In
comparison, RO 780 has a PEI of 4.0 - considered relatively low among the engine oils collected
around the world for the international engine oil database of the IOM.
TABLE 2--0il and phosphorus volatilities of fully formulated Oil RO 780 and ZDDPdeficient Reference Oil 779.

PEI Analysis of Engine Oils Selected from the Market


For purposes of determining the effect of large differences in PEI on the NMR signature of
the volatile and residual material, four oils having similar oil volatilities but widely different
phosphorus volatilities were requested from the North American IOM engine oil database. These
samples were kindly supplied, and results are shown in Table 3 and Fig. 7 (also including RO
780 for comparison).
Compared to phosphorus concentrations of the fresh engine oils, Table 3 and Fig. 7 show that
the volatilized material of the four oils has widely different concentrations of phosphorus and,
thus, considerably different levels of associated PEIs. It was considered reasonable that these oils
would provide good opportunity in later NMR studies [24] to determine the cause(s) of
differences in phosphorus volatility and the related PEI.
TABLE 3--Comparative oil and phosphorus volatilities of four marketed engine oils of
considerably different PEIs.

246

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 7--Variation of PEI among four marketed North American engine oils of different
formulations with somewhat similar overall volatility loss and initial phosphorus concentrations.

Preliminary Study of Temperature Ramping on Phosphorus Volatility


As a preliminary test to determine when phosphorus volatiles were generated during the
special Noack test, the ZDDP-containing RO 780 was given a one-hour ramping time to the
temperature of 250~ followed by the normal hour duration at that temperature. Results are
shown in Table 4, contrasted with the normal ramp time of about nine minutes to 250~
TABLE 4 Effects on phosphorus volatilization using one-hour ramp versus normal, 9-rain
ramp.

The data show that, although the total volatile material increased by about 13 % because of
the longer test time, the phosphorus volatilized increased by slightly more than 100 %. From
these results, it would seem that the longer test encouraged the volatilization of more phosphorus
and/or the slower, longer ramping interval affected the nature of the phosphorus being
volatilized. The questions raised by these considerations led to further tests reported in the next
section of the paper.

Second Stage of Studies

General Considerations
The questions of when phosphorus volatilizes, as well as what volatile compounds are
produced and what compounds are left in the residual oil during the special Noack test were of
considerable interest. It was recognized by the authors that the study of these various

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

247

relationships would constitute a sizable effort, and it was decided to resolve the questions
progressively. Some of these efforts are reported in this paper, others in the conjoined second
paper [24].

Dependence of PEI on Residence Time in the Special Noack Test


One of the first studies was a modified special Noack test on RO 780 (having a PEI of 4) and
a formulated engine oil having much higher PEI. Modified special Noack tests were run in four
progressive time intervals of 10, 20, 30, and 50 min exposure after the normal 9-min ramp to
volatility-inducing conditions. A special vial-filling drain for collecting small amounts of the
volatilized oil was attached to the bottom of the collection vessel. Normal 65-g samples were
used for the test. Following the test, the collected samples were analyzed by ICP to determine
their phosphorus content and the related PEI. Table 5 and Figs. 8 and 9 show the data obtained
(including the 60-min data given earlier in Table 2).
TABLE 5--First sequence of exposure-modified special Noack results on the low-PE1
formulated Reference Oil 780 compared to the results from a marketed oil having a relatively
high value of PEI.

FIG. 8--Dependency of oil volatility on exposure.

FIG. 9--Dependency of PEI on exposure.

248

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

As would be expected, the oil volatilities rise with increasing time of exposure, as shown in
Fig. 8. Oil 780 may be coming to equilibrium in about 50-60 min, while the volatility of EO
7450 is still rising, although slowing.
Effects of the forgoing study on the Phosphorus Emissions Index are shown in Fig. 9. With
Reference Oil 780, the PEI remains essentially constant and may even be interpreted as falling
somewhat. In contrast, the ZDDP producing the much higher volatilizing Oil 7450 increases
during the exposure. These dissimilar results raise further questions regarding the nature of the
chemical composition of the phosphorus additives as well as effects of the other additives in the
two oils - questions appropriate for the NMR study in the conjoined paper [24].

Modification of the Chemistry of ZDDPs on PEI


Chemistry and Test Setup---Four ZDDP samples were prepared in the Astaris labs. The
alcohols employed were 2-ethylhexanol, a primary alcohol, and 4-methyl-2-pentanol, a
secondary alcohol. Two P2S5 samples were employed, one that consists of slightly lower (at
27.77 %) and the other of slightly higher (at 28.07 %) levels of phosphorus than the
stoichiometric value of 27.87 % phosphorus theoretically calculated for P2S5. These four ZDDPs
were blended into a typical GF-3 oil at Savant using the non-ZDDP-containing RO-779. These
blends, identified as Test Oils A through D, were again exposed to the time-interval approach.
In this portion of the work, exposure times of 10, 20, 30, and 50 min in the special Noack
instrument were used with the routine 9-min ramp to an operating temperature of 250~
Test Results--Results are shown Table 6 and in Fig. 10. As evident from Fig. 10, PEI values
of each of the Test Oils A through D show major differences between the primary and secondary
ZDDPs when added in similar concentrations to the Stock Oil RO 779. That is, primary ZDDPs
A and B increase in PEI with increasing exposure time in the test, while secondary ZDDPs C and
D show little change in their PEIs.
TABLE 6--Effect of ZDDP chemistry and reaction stoichiometry on the level of phosphorus
volatility in exposure-modified special Noack tests.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

249

FIG. l O--Plot of PEI data from Table 6 showing the clear differences of type of alcohol
chosen and the chemistry of preparation of ZDDP.
A response showing little or no increase of volatile phosphorus indicates that ZDDPs C and
D emit all volatile phosphorus components in less than 10 min exposure in the special Noack
test. More important, however, is the fact that the one-hour PEI values of the secondary ZDDPcontaining oils are comparatively low and do not increase, whereas the primary forms of ZDDP
investigated in this study are considerably higher and increase further in PEI values with
exposure time. For reasons offered earlier, increase in PEI implies continuing generation of
phosphorus volatiles during the test.
It will also be noticed in Table 6 that the primary and secondary alcohol preparations are
made with two samples of P2S5 that contain different levels of phosphorus. Of these two pairs of
ZDDPs, little difference is shown between the primary ZDDPs A and B. However, significant
differences are shown between the secondary ZDDPs C and D. That is, at closely similar
concentrations (991 versus 1064 PPM, respectively), ZDDP C has about half the value of PEI
shown by ZDDP D.
Table 7, obtained from the Institute of Materials Engine Oil Database, shows that both of
these secondary ZDDPs are at relatively low levels of PEI compared to other oils around the
world.
Discussion

Relevance of the PEI to Observed Field Catalyst Deterioration


On the basis of the Ford Motor Company field studies, the correlation of the PEI generated
using the special Noack shown in Table 1 and Figs. 5 and 6 is evident. The degree of correlation
suggests that the volatility of phosphorus is perhaps the most important source of such catalyst
contamination, whether or not it is generated from the crankcase and passed to the exhaust
stream through the EGR valve or from oil leaking directly into the combustion zone from piston
ring bypass or down the valve stems.

250

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TABLE 7--Comparison of PEI of specially prepared primary and secondary alcohol


ZDDPs to marketed oils from around the world. (Data from IOM Engine Oil Databases, used
with permission.)

Moreover, the work clearly shows that engine oil formulation is a critical aspect o f
controlling phosphorus volatility and raises the question o f whether certain additives are more
effective than others or whether some o f the other additives actually enhance phosphorus
volatility.

General Comments on PEI Technique Used in These Studies


In this work a special technique was used to drain the contents of the collection cup o f the
special Noack instrument into a small vial after a shortened run where it was desirable to catch
small amounts o f volatile material. This was successful in obtaining PEIs over the series of time
intervals. Considering the results obtained on those samples continuing to generate phosphorus
to the end of the 60-min test, it is evident that longer exposure periods should be even more
revealing.

Comparison of Two Special Reference Oils


As previously noted, two special reference oils, RO 779 and RO 780, were provided by
Chevron Oronite Company LLC. RO 779 had no ZDDP added but otherwise was identical in
base stock and additive content to RO 780. Table 1 showed that the responses o f the two
reference oils were as expected, with RO 779 evidencing no volatile phosphorus and RO 780
showing a PEI o f 4 - a comparatively low value. RO 779 was thus considered to be acceptable
as a stock for the subsequent blending studies o f specially made ZDDPs.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

251

Comparison of Four Marketed Engine Oils


In an effort to determine the differences between engine oils having similar oil volatilities
and fresh oil phosphorus concentrations, but markedly different phosphorus volatilities shown by
the PEI, four oils were chosen from the IOM Engine Oil Database. Sufficient samples were
received through the courtesy of IOM to run comparative special Noack tests, from which the
volatile material and the residual oil for NMR studies were collected. These data, shown in Table
3 and Fig. 7, show the similarities of both oil volatilities and fresh oil phosphorus concentrations
and are in clear contrast to the different levels of PEI, which are as great as 3700 %.

Effect of Different Temperature Ramps


A preliminary effort was made to determine whether ramping time from ambient to 250~
for the special Noack test would show differences in the phosphorus volatiles collected. Data
presented in Table 4 using RO 780 indicate that significant differences in ramping time do have
an effect and raised the questions of whether the effect is related to a difference in phosphorus
decomposition products or merely the extended time of exposure of the sample.
In view of the later experiments shown in Table 5 and Fig. 9 conducted by varying exposure
time for Oil 780 (in which the full content of volatilizable phosphorus was delivered to the
special Noack collector in less than 10 min), it is more likely that there is a difference in the path
of phosphorus decomposition. That is, for Oil 780, the difference in oil volatility increases with
and without extended ramping time - about 13 % - for the longer ramping test is considerably
less than the 125 % increase in volatilized phosphorus. This preliminary finding emphasized the
value of investigation using 31p NMR spectrometry to determine what differences were present.

Effects of Exposure Time on Otherwise Normal Test Technique


First Evaluation of Technique--Desire to determine the progression of the breakdown and
volatilization of ZDDP in the fresh oil led to the technique of varying the length of exposure of
the oil to 10, 20, 30, 50, and 60 min in the special Noack test. In this initial study, RO 780 was
compared to IOM's EO 7450, a formulated oil from the market having relatively high PEI.
First results were shown in Table 5 and Figs. 8 and 9. Not surprisingly, increasing exposure
showed progressive increase in oil volatilization. The technique, however, showed dissimilar
results in progressive values of the PEI since RO 780 gave results indicating that all volatile
forms of phosphorus from the ZDDP additive in this oil were generated within the first 10 min of
exposure at 250~
Volatile Phosphorus, a Small Part of Available Phosphorus--As noted previously, the PEI
data of RO 780 showed no further generation of phosphorus with more lengthy exposures (and,
in fact indicated some decrease in PEI at 50 and 60 rain exposure). Thus, whatever volatile
decomposition products were produced before 10 rain exposure comprised only a small fraction
of the phosphorus-containing additive(s) present in the fresh oil. That is, 850 g of the fresh RO
780 contains 966 PPM phosphorus or approximately 820 mg/L, whereas the PEI is 9 mg/L about 1% of the phosphorus in the fresh RO 780 volatilized in this test.
In contrast, the IOM EO 7450 contains 1198 PPM or approximately 1020 mg/L. In this case,
the PEI rises to 60 at 60 min of exposure, and thus, about 4 % of the phosphorus material in the
fresh oil is capable of being volatilized at this length of exposure. Greater length of exposure
might increase this level of phosphorus volatility of IOM EO 7450, although it does seem to be

252

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

reaching a plateau.

Application of Progressive Exposure Technique to Special ZDDPs--The four special ZDDPs


made by Astaris to generate further understanding of the causes and forms of phosphorus
volatilization were added to the stock oil, RO 779, at levels of about 1000 PPM. Using the
technique of progressive time exposure, the results are shown in Table 6 and Fig. 10.
The PEI values obtained demonstrate obvious differences in phosphorus volatility between
the particular primary- and secondary-alcohol ZDDPs. Specifically, both higher and lower
stoichiometric variants of this primary alcohol's ZDDPs show continuing release of volatile
phosphorus during the one-hour special Noack test with the rate of release decreasing with
increasing time of exposure. At 50 min, about 2 % of the original phosphorus in the oil has been
volatilized. Both preparations of the primary ZDDP at slightly higher and lower P2S5
stoichiometries are closely similar in their PEI response to increasing exposure.
In contrast, the particular secondary alcohol ZDDPs show little or no change in PEI with
exposure time indicating that all the volatile phosphorus entities that will be formed are
volatilized early in the test. However, in this case there is a difference in whether the ZDDP is
formed from P2S5 that contains a higher or lower than stoichiometric level of phosphorus. The
higher-than-stoichiometric ZDDP C gives phosphorus release of about 1 % of the original
phosphorus in the oil. In contrast, the lower-than-stoichiometric ZDDP D gives phosphorus
release of about 0.6 % of the original phosphorus - a considerable difference in phosphorus
release.
Resolution of the Question of the Volatility Role Played by ZDDP Formation--Most
important, these latter data answer the question of whether the chemical form of ZDDP can
markedly affect its tendency to volatilize. As to how well this statement applies to other forms of
ZDDP, this is a matter of fiarther study. Certainly, from the viewpoint of controlling phosphorus
volatility, there is a strong need for much more information regarding the relationship of the
chemical form of ZDDP and its dependence on the other additives used in formulating engine
oils.
Conclusions

This series of studies of the variation of the Phosphorus Emission Index generated from
several applications of the special Noack instrument has been effective in revealing some of the
characteristics of the phosphorus volatilization from ZDDPs made from a primary and a
secondary alcohol. Thus, the primary objective of this study has been met. That is, it has been
shown that phosphorus volatilization and the associated PEI are affected strongly by the
chemistry of the ZDDP. Whether or not these differences are caused by thermodynamic effects
and/or some form of steric interference remains to be determined, as well as the range of
differences that may exist among the different alcohols.
Moreover, using the PEI as an investigative tool for this initial study, differences between the
primary and the secondary alcohols chosen are also found in the immediacy and ultimate levels
of forming volatile forms of phosphorus species. Variation in PEI of four marketed engine oils in
North America has shown marked differences of 3700 %. Even comparing the two alcohols used
in this study, the differences in PEI were about 400 %. To be sure, at this point, the latter
observations are limited to the two alcohols used to make the ZDDPs for this study, but the
technique is obviously readily applicable to all other variations of ZDDPs.
The study also generated sufficient volatilized samples to permit the generation of 31p NMR
spectra, and this work is reported in a conjoined paper [24].

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

253

One area of question that was not part of this study but should be considered is the impact of
phosphorus volatilization on the residual oil from the special Noack test. That is, to what extent
is the anti-wear and anti-oxidation performance of a ZDDP compromised by phosphorus
volatilization? This information is obviously of high interest and importance given the primary
purpose of ZDDPs in engine oils.

Acknowledgments
The authors would like to thank those who dedicated their efforts in preparing, blending, and
analyzing the various oils used in this study. Among these were Skip Ramsey of Astaris for his
assistance in preparing the ZDDPs, George Schaller of Savant Labs for running the special
Noack tests, and both he and Mike Habitz of Savant Labs for blending the test samples and
analyzing the results. For the marshalling of samples and data, the help of Amy Parker and Joyce
Gee were indispensable. Special thanks are extended to Chevron Oronite Company LLC for
providing typical GF-3 oil samples with and without ZDDP.
References

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]

Spikes, H., "The History and Mechanisms of ZDDP," Tribology Letters, Vol. 17, No.3,
October 2004, pp. 469--489.
Minamitani, A., et al., "Effects of Engine Oil Additives on the Activity of Exhaust Gas
Oxidation Catalysts, J. Japan Petroleum Institute, Vol. 21, 1978, p. 116.
Caracciolo, F. and Spearot, J. A., "Engine Oil Additive Effects on the Deterioration of a
Stoichiometric Emissions Control System," SAE Paper #79094, 1979.
Monroe, D. R., "Phosphorus and Lead Poisoning of Pelleted Three Way Catalysts," SAE
Paper #800859, 1980.
Miyoshi, N., "Poisoning Deactivation of Automotive Catalysts by Lead and Phosphorus,"
SAE Paper #852219, 1985.
Brett, P. S., et al., "An Investigation into Lubricant Related Poisoning of Automotive Three
Way Catalysts and Lamda Sensors," SAE Paper #890490, 1989.
Dmry, C., et al., "The Effect of Lubricant Phosphorus Level on Exhaust Emissions in a
Field Trial of Gasoline Engine Vehicles," SAEPaper #94075, 1994.
Ueda, F., et al., "Engine Oil Additive Effects on Deactivation of Monolithic Three-Way
Catalysts and Oxygen Sensors," SAE Paper #940746, 1994.
Noack, K., Angewandt Chemic, Vol. 49, 1936, p. 385.
Determination of Evaporation Loss of Lubricating Oils, (Noack Method), DIN 51-581,
1981.
Evaporation Loss of Lubricating Oils, CEC L-40-T-87, 1987.
ASTM Standard D 5800, "Evaporation Loss of Lubricating Oils by the Noack Method,"
Annual Book of ASTM Standards, Vol. 5.03, ASTM International, West Conshohocken,
PA, 2004, pp. 352-366.
Selby, T. W., et al., "A New Approach to the Noack Test for Volatility Measurement," SAE
International Fuels & Lubricants Meeting and Exposition, Philadelphia, PA, 1993.
Reichenbach, E., et al., "A New Approach to the Noack Volatility Test," Proceedings of
Ninth International Colloquium, Esslingen, Germany, 1994; Tribotest Journal, Vol. 1, No.
1, Leaf Coppin Publishing, September 1994, pp. 3-29.

254

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

[15] Selby, T. W., et al., "Base Oil Characterization Techniques Using a New Approach to the
Noack Volatility Test," ACS Symposium on "The Processing, Characterization, and
Application of Lubricant Base Oils: Part 1I, San Diego, CA, March 1994.
[16] Selby, T. W. and Reichenbach, E. A., "Engine Oil Volatility Studies - Generation of
Phosphorus," Proceedings of the International Tribology Conference, Yokohama, Japan,
1995, pp. 813-816.
[17] Cluff, B., et al., "A Laboratory Method for Measuring Bulk Volatility of Engine Oils Comparative Results," SAE Paper #961227, International Spring Fuels and Lubricants
Meeting, Dearborn, MI, May 1996.
[18] Institute of Materials Engine Oil Database, Institute of Materials, Midland, MI, issued
yearly from 1984.
[19] Selby, T. W., "Analysis of Engine Oil and Phosphorus Volatility - Development and Use
of the Selby-Noack Apparatus to Recover and Study Phosphorus Volatiles," 11th
International Colloquium, Ecological and Economic Aspects of Tribology, Esslingen,
Germany, 1998.
[20] Selby, T. W., "Development and Significance of the Phosphorus Emission Index of Engine
Oils," Proceedings of 13th International Colloquium - Lubricants, Materials, and
Lubrication Engineering, Esslingen, Germany, 2002, pp. 93-102.
[21] Selby, T. W., "Phosphorus Volatility of Lubricants - Use of the Phosphorus Emission
Index of Engine Oils," 9th F&L Asia Conference, Singapore, January 21-24, 2003.
(available on CD from F&L Asia, Inc., Manila, Philippines).
[22] Johnson, M., "Catalyst Contamination by Phosphorus from Engine Oils - Emission Effects
and Transport Processes," ILSAC-Oil Meeting, Romulus, MI, July 10, 2002.
[23] Johnson, M., "Effects of Engine Oil Formulation Variables on Exhaust Emissions in Taxi
Fleet Service," SAE Paper #2002-01-2680, SAE Powertrain Meeting, 2002.
[24] Bosch, R. J., et al., Analysis of the Volatiles Generated during the Selby-Noack Test by 31p
NMR Spectroscopy, to be published by ASTM International, West Conshohocken, PA.
[25] "Determination of Additive Elements in Lubricating Oils by Inductively Coupled Plasma
Atomic Emission Spectrometry," Annual Book of ASTM Standards, Vol. 5.02, ASTM
International, West Conshohocken, PA, 2004, pp. 974-980.

Journal of ASTM International, November/December 2005, Vol. 2, No. 10


Paper ID JAI 12984
Available online at www.astm.org

R. J. Bosch, 1 D. C. Fee, 1 and T. W. Selby 2

Analysis of the Volatiles Generated During the Selby-Noack


Test by 31p NMR Spectroscopy
ABSTRACT: A series of engine oil samples collected during a study of their Phosphorus Emission Index

values were analyzed by 31p Nuclear Magnetic Resonance Spectroscopy. (The PEI analyses themselves
were presented in an associated paper [1]). NMR spectra were generated to obtain and explain the mode
of formation and identity of the phosphorus-containing species in the volatiles generated during the
Selby-Noack volatility test and to compare these species to those found in both the fresh oil and the
residual oil remaining after the volatility test.
KEYWORDS: phosphorus volatility, exhaust catalyst, exhaust emissions, ZDDP, zinc dithiophosphates,
Phosphorus Emission Index, PEI, 31p NMR, NMR spectroscopy
Nomenclature

ZDDP
PEI
NMR

Zinc Di(organo) Di(thio) Phosphate


Phosphorus Emission Index
Nuclear Magnetic Resonance

Introduction

Zinc Di(organo) Di(thio) Phosphates, commonly referred to as ZDDPs, are the most widely
used and effective anti-wear/anti-oxidation additives in engine oil. Phosphorus contained in the
ZDDP molecules has been shown to partially volatize during engine operation [2].
Unfortunately, volatile phosphorus in the exhaust stream degrades the function o f the exhaust
catalytic converter, and as a consequence, there has been pressure to reduce the amount o f ZDDP
in engine oil.
One o f the new analytical test procedures developed during the 1990s is the Selby-Noack
instrument [3]. This instrument uses the earlier protocol developed b y Noack [4-6] to measure
the volatility o f motor oil but, in addition, collects all o f the volatile material generated in the
test. In Noack-protocol tests, 65 g o f o i l are heated in air in a metal or glass cup for 1 h at 250~
under a slight vacuum o f 20 mmw~t~, and the percent loss o f sample mass is measured.
During the development and early use o f the Selby-Noack test, it was found that phosphoruscontaining compounds were present in the volatiles obtained from the procedure [7]. With the
growing interest in the subject o f volatile phosphorus, the Institute o f Materials incorporated the
Selby-Noack test as part o f the database they maintain on worldwide oil analyses [8] and also
determined the concentration o f phosphorus in the volatile material generated (initial phosphorus
concentration o f the fresh oil was already a part o f the database).
Manuscript received 15 January 2005; accepted for publication 28 April 2005; published November 2005. Presented
at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8
December 2004 in Tampa, FL.
J Technical Service Fellow and Technology Fellow, respectively, Astaris LLC, Webster Groves, Missouri.
2 Director of R&D, Savant, Inc., Midland, Michigan.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

255

256

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

In 2002, a paper was published on the phosphorus volatility results that appeared in the IOM
database from 1999-2001 [9]. Contrary to expectations, it was found that phosphorus volatility
was neither related to engine oil volatility nor to the phosphorus content in the unused engine oil.
It was speculated that this lack of correlation with initial phosphorus additive concentration
could be explained by either or both: 1) effects of other engine oil additives and 2) variations in
the chemistry of the phosphorus additives. A 2002 field study conducted by Ford Motor
Company of catalyst degradation by phosphorus-containing oils [10] was subsequently shown to
correlate with the PEI data generated from the Selby-Noack bench test and, further, showed that
phosphorus volatilization was strongly affected by other oil additives [11].
One of the objectives of the present studies is to examine the second premise concerning
whether variation in the chemistry of phosphorus-containing ZDDP additives also affects
phosphorus volatility from the formulated engine oil.

Experimental
Small samples of the original oils and the volatile materials and residues obtained during the
Selby-Noack tests in the associated paper [1] were analyzed by 31p NMR at Washington
University (St. Louis, MO). The spectra were obtained either on a 500 MHz Varian NMR
equipped with a 10 mm probe or on a 600 MHz Varian NMR equipped with a 5 mm probe.
Samples were diluted with 10-15 % chloroform-d (CDC13), which also served as an internal
reference for establishing spectral positions.
The individual phosphorus-containing species are identified by a "chemical shift" in the
phosphorus-31 peak because of the different chemical environment around the phosphorus atom
in phosphorus-containing oil components (or decomposition fragments) compared to the
phosphorus atom in phosphoric acid (H3PO4). This shift is reported in units of ppm (which term
has no relation to concentration). By definition, the shift for the phosphoric acid standard is zero
ppm. Average data accumulation time for these spectra was 1 h.

Results
First Study---Comparison of Formulated Oils with Different PEIs
A series of five motor oil samples, four from the Institute of Materials database (samples EO7450, EO-7459, EO-7538, and EO-7551) and one of a typical GF-3 motor oil supplied by
Chevron Oronite Co., L.L.C (Sample RO-780) was subjected to the Selby-Noack test. The first
four engine oils were chosen from the database because of similar volatilities but dissimilar
phosphorus contents in the volatilized oil. The fifth oil was matched to a second sample of oil
that contained no ZDDP (Sample RO-779).
A series of 31p NMR spectra was taken of each of the fresh oils and of their volatiles and
residue generated by the Selby-Noack test (see Tables 1-5 and Figs. 1-5). Also included in the
Tables are the Selby-Noack oil volatility results and the PEI values for each of the starting oils.
TABLE 1--Summary of 31 P NMR Results for E0-7450.
Selby- Noack PEI,
Sample: EO-7450
Volatility
mg/L
Chemicalshifts(ppm)
EO-7450 Fresh Oil
19.4 %
46
104.3, 99.5, 86.4, 77.8, 75.6
EO-7450 Volatiles
97.1, 69.4, 69.3, 29.6, 29.4, 29.3, 12.4
EO-7450 Residue
97.3, 3 (broad)

BOSCH ET AL. ON PNMR SPECTROSCOPY

Sample 7450

Volatlllm

1
/

Original 011

Residue

...........................
'

x~,

~.." 'V ~ '~: ~" :~'~;: ~ ~IV-:'..:Z~:~,,'~.,'~'~":.IT '~/":i

............................

"*-', .... ,',

d.."

~'a

~',

*,

s'*

*'~

" ~', .... z', .... ~',

-' ' -i,

-"~
~.~'

FIG. 1--E0-7450, Oil.


TABLE 2--Summary O f 31 P NMR Results for E0-7459.
Sample: EO-7459

Selby
Noack
Volatility
18.05 %

EO-7459 Fresh Oil


EO-7459 Volatiles
EO 7459 Residue

Sample

PEI,
mg/L

Chemical shifts (ppm)

1
-

93.2, 83.9, 83.5, 81.2, 74.6, 74.2


125.6, 94.8, 94.6, 94.3, 94.0, 93.6, 83.2, 48.2
2 (broad)

7459

Volatlhm

II

Odglnll Oil

l~Iduo

lZO

1i|

lee

se

8|

7e

oo

se

~e

FIG 2--E0-7459, Oil

3e

zo

le

-It

ppe

257

258

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TABLE

3--Summary of ~1P NMR Results for E0-7538.

Sample: EO-7538

Selby - Noack
Volatility
22.55 %

EO-7538 Fresh Oil


EO-7538 Volatiles
EO-7538 Residue

PEI,
mg/L
31

Chemical shifts (ppm)


93.3, 83.7, 82.6, 82.5
97.1, 96.9, 96.3, 95.8, 95.7, 95.5, 69.3,
69.2, 29.5, 29.4, 29.3, 28.0
4 (broad)

Sample
7538

J.

Volatiles

Original Oil
i..,,L.,

.u.,,

|[

l..~Jh~,llkl~ll~

i ~ a l ~ J j l l ~ t L , ld.. L I . J ~ a . a . , L i . , l . L J a d t ~ . , . . l . . i , h , * l d ~ A t a ~ . - - a . ~ . ~ / h . . l l ~ , ~ , . l A

Ii,Ia...

~'"'"';~ ""'11~" 7;';...... ,';'"" "j," *'1~;;'1' " "k'0........j, ...... ~',.... '3';...... W ' " ';, . . . . : . . . . -;i" "';;-"'
FIG.

TABLE
Sample: EO-7551
EO-7551 Fresh Oil
EO-7551 Volatiles
EO-7551 Residue

3--E0-7538, Oil.

4--Summary of 31P NMR Results for E0-7551.

Selby Noack
Volatility
17.34
-

PEI,
mg/L
8

Chemical shifts (ppm)


103.6, 98.1, 95.8, 93.3, 86.7, 86.6, 78.0, 77.8
93.7, 93.3, 92.6, 25.7, 25.6
0 (broad)

BOSCH

Sample

ET

AL.

ON

PNMR

SPECTROSCOPY

~5~

7551

Volatlles

i-

Prll

Residue

~ o

" ~;o " " ~ o

eb . . . . 7b . . . .

~ . . . . ~ . . . . 4~ . . . .

....

~b ....

6'

~b ....

" -;b-~

/~IG. 4---E0-7551, Oil.


TABLE 5--Summary of 3t P NMR Resultsfor R0-780.
Sample: Typical GF-3 Selby - Noack PEI,
Chemical shifts (ppm)
oil (RO-780)
Volatility
mg/L
RO-780 Fresh Oil
14.23 %
4
100 (broad), 93.8, 84.3, 83.7, 83.6, 83.5, 82.8, 74.6, 67.9
RO-780 Volatiles
101.7, 95.1, 94.6, 94.5, 93.9, 72.4
RO-780 Residue
2 (broad)

Sample

RO-780

Volatlies

, . . . . ,

. . . .

. . . .

, . . . . ,

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

Residue

FIG. 5--R0-780, Oil.

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

260

ELEMENTALANALYSIS OF FUELS AND LUBRIGANTS

Second Study---Rate of Heating Dependence


As a preliminary test to determine when phosphorus volatiles were generated during the
Selby-Noack test, RO 780 was given a 1-h ramping time to the temperature of 250~ followed
by the normal hour duration at that temperature. The volatiles generated during this study were
subjected to 31p NMR analysis. The results o f this study are presented in Fig. 6 and are
summarized 1'n Table 6 . The 3 1 P N M R analysis o f the volatiles generated during a typical SelbyNoack test (normal ramp time o f about 9 rain to 250~ followed by a 60-min hold) are also
presented in Fig. 6 and are also summarized in Table 6.
Both residue samples were also analyzed by 31p NMR and were found to consist o f a large
broad peak centered at approximately 2 ppm.
:Sample 780
Standard vs t Hour R a m p

1- I
Ramp

Standard

~'P

tl

_.,....Jl ...........

t .........................................................................
"TT" 7 : " : " : T ' 2 T
Jl
~g
ig
]1
41
34
ii
IJ
i

,~2"-T'~,'?T ~"?T 7:~'7 2 V T ; 7 S4"IL'T" 7 " t : 1 : ' T T T 2 T T T - T


log

ij

? "'T2TTT',
-11
p}

FIG. 6--R0-780, 1-h ramp versus normal ramp, volatiles.


TABLE 6--Summary of the 31p NMR results of a 1 h ramp versus the standard ramp in the
Selby-Noack Test.
Sample: Typical GF-3 oil
(RO-780)
1Hr Ramp volatiles
Standard Ran~-volatiles

Selby - Noack
PEI,
Volatility
rng/L
16.03 %
9
14.23 %
.... 4

Chemical shifts (ppm)


102.5, 95.0, 94.5, 93.8, 93.7, 92.7
102.5, 95.0, 94.5, 93.8, 83.2

Third Study--Time Dependence


Modified Selby-Noack tests were performed on EO-7450 and RO-780. These experiments
were performed using a normal ramp heat-up time o f about 9 min to 250~ followed by a
variable hold o f 10, 20, 30, or 50 min for each progressive test. The volatiles and residues were
all subjected to 31p NMR analyses, however only the results of the 10-min and 50-rain runs were
included in this paper. They are presented in Figs. 7 and 8 and are summarized in Table 7 for
EO-7450 and Fig. 9 and are summarized in Table 8 for RO-780.

BOSCH ET AL. ON PNMR SPECTROSCOPY

261

Sample 74r~, 10 rain.

VolaUles
. . . . .

' I

"

'

1 .

Elm

FIG. 7--E0-7450, t = 10 min, volatiles and residue.


Sample 7450, t:50 rain.

VolaU~s

' " ' " t~t

ill

"

1-'1

Ii

a,

Ti

,i

sl

tl

st

z*

t,

-xl

-~l

pl..

FIG. 8 - - E 0 - 7 4 5 0 , t = 50 min, volatiles and residue.


TABLE 7--Summary o f 31 P NMR results f o r E0-7450, 10 and 50 min runs.
Chemical shifts (ppm)
Selby - Noack
PEI,
Sample: EO-7450
Volatility
mg,/L
104.3, 99.5, 86.4, 77.8, 75.6
19.4 %
46
Original oil (from Table 1)
97.0, 91.1, 77.8, 75.6, 69.3, 65i3, 29.6, 29.4,
6.12 %
39
Volatiles, T=10min
29.3, 28.4, 14.5, 14.1, 12.5, 12.4,
97.1, 29.4, 29.2, 0 (broad)
Residue T=10 rain.
97.2 91.1, 75.5, 69.5, 69.4, 61.0, 29.5, 29.3,
12.97 %
53
Volatiles, T=50min
28.4, 12.4
97.3, 4 (broad)
Residue T=50 rain.

262

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Sample750,
10 and ISO mln.

-I/'~

SO min. Volatiles

....... lat,~ ..........


I. . . . . . . . . . . .
SO mill. Residue

......

~W,.~] ..................
r

] . . . . . . . .

i . . . . . . . . .

~,-L,, ..................................
1,,~

. . . . . . . .

,. . . . . . . . . .

A.L, L..Ld.,.,la ,, i[i xalR


I I

. . . . . . . . . . .

,1 . . . . . . .

~,.r,.

...... ill ....... iii ....... :, ......... ,; ......... g,. . . . . . . . ,',......... ; ........ ;i, ' p,'

t0 rain. Volatlles

[[

,L
~ j ~]L&,.I
t. . . . . . . . . . . . . . . . . . . . . . . . . .
......
;;":. ?V
. . .??L"~"
. . . . .";
. . 'Y
. .?.~. . .."?
. T[.tlZ"Y77.'
. . . . . .7.77":
. .7 "T',Z"." : Y;" Y. .';", ~ 77 ~ 7: T ~ "?TE" ~"l', Y,~ .' "~'. . .'~ ,'7: ' ': '!
Y'
7!'~"Y~,
10mln.Res due
.t

. . . . . . .

FIG. 9---R0-780, t = 10 min volatiles and residue and t = 50 min volatiles and residue.

TABLE 8--Summary of 31 P NMR results for R0-780, 10 and 50 min runs.


Sample: RO-780 oil
Selby - Noaek PEI,
Chemical shifts (ppm)
Volatility
mg/L
Original oil (from Table 5)
100 (broad), 93.8, 84.3, 83.7, 83.6, 83.5,
82.8, 74.6, 67.9
Volatiles, T=10 min
4.03 %
9
103.0, 102.4, 94.4, 94.0, 93.8, 93.3, 88.9,
83.2, 67.6, 54.1
Residue, T=10 min.
54.1, 0 (broad)
13.62 %
5
95.1, 94.6, 93.9, 83.0, 54.1
Volatiles, T=50 rain.
Residue, T=50 min.
4 (broad)

Fourth Study--Effects of ZDDP Chemistry and Formation


Four Z D D P samples were prepared in the Astaris labs. The alcohols employed were 2Ethylhexanol, a primary alcohol, and 4-Methyl-2-pentanol, a secondary alcohol. Two P2S5
samples were employed, one with a phosphorus content o f 27.77 % (Low Phos. P2S5) and the
second with a phosphorus contend o f 28.07 % (High Phos. PzSs). These four Z D D P s were
blended into a typical GF-3 oil at Savant using a non-ZDDP-containing version o f RO 780
provided b y Chevron Oronite C o m p a n y LLC.
ZDDP
ZDDP
ZDDP
ZDDP

A
B
C
D

2-Ethylhexanol
2-Ethylhexanol
4-Methyl-2-pentanol
4-Methyl-2-pentanol

Low Phos. P2S5


High Phos. P2S5
Low Phos. P2S5
High Phos. P2S5

These samples were subjected to a series o f Selby-Noack tests in which the runs were
terminated at 10, 20, 30, and 50 m i n progressively. The resulting volatiles and residues, along
with the starting oil samples, were subjected to 3~p N M R analyses.

BOSCH ET AL. ON PNMR SPECTROSCOPY

263

Figures 10-14 contain the most relevant 31p NMR spectra; however, Tables 9-12 contain
summaries of the results of all of the NMR spectra.
The 31p NMR spectra for the experiments involving the two 2-Ethylhexyl ZDDPs were
integrated, which are summarized in Tables 13 and 14.

Original OII

FIG. 10---ZDDP A, Original Oil and t = 10 volatiles and residue.

!
Sample ZDDP A

50 mln. Volatltes
l

,.

FIG. 11

.~

Is

-1.

-t,

p,.

ZDDP A, Original Oil, and t = 50 volatiles and residue.

264

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

T A B L E 9--Summary Of 31 P NMR results for


Sample: ZDDP A
Selby - Noack
....
Volatility
Oil
Volatiles, T=10 min.
3.31%
Volatiles, T=20 rain.
7.09 %
Volatiles, T=30 min.
8.43 %
Volatiles, T=50 min.
11.82 %
Residue, T=10 min.
Residue, T=20 min.
Residue, T=30 n'fin.
Residue, T=50 rain.

Sample

ZDDP

ZDDP A (2-Ethylhexanol + Low Phos. P2S5).


PEI,
mg/L
6
13
14
19
-

Chemical shifts (ppm)


104, 98.1, 69.5, 65
104, 96.7, 84.4, 69.0, 29.6
103.3, 97.2, 84.2, 69.5, 29.9
103.2, 97.2, 69.6, 30.0
103.5, 97.2, 69.6, 30.0
97.3, 69.6, 29.7, 2
97.2, 69.6, 29.7, 2
97.3, 69.7, 29.7, 2
97.3, 69.7, 65, 2

1
I

,,

. . . . . . . . . . . . . . . .

. . . . . . . . . .

9 .

" "

50 MIn. Residue

7*

le ,

Wl

,J

* ~ ' ~ ' Zl

~X ,

- i*

" " - .~l

Pm

'

FIG. 12--ZDDP t?, Original Oil, and t = 50 volatiles and residue.

T .ABLE lO--Summary of 31P NMR results for ZDDP B (2-Ethylhexanol + High Phos. P2S5).
Sample: ZDDP - B
Selby - Noack
PEI,
Chemical shifts (ppm)
Volatility
mg/L
Oil
104.1, 103.5, 98.5, 69.7, 65
Volatiles, T=10 rain.
3.59 %
6
103.9, 96.8, 84.3, 69, 29.6
Volatiles, T=20 rain.
5.97 %
10
103.3, 97.2, 84.2, 70, 29.9
Volatiles, T=30 min.
8.55 %
13
103.5, 97.1, 69.4, 29.8
Volatiles, T=50 rain.
13.72 %
18
103, 97.3, 69.6, 30.0
Residue, T=10 rain.
97.2, 69.6, 65.3, 29.6, 2
Residue, T--20 min.
97.3, 69.7, 65.2, 29.7, 3
Residue, "1"--30rain.
97.2, 69.6, 65.3, I
Residue, "I"--50min.
97.3, 69.7, 65.4, 2

BOSCH ET AL. ON PNMR SPECTROSCOPY

265

SampleZDDPC
mln.VolMlkll

20

.... ' ' " " "

1 ~ .........

''

F~'I rl'''

'r] ~-

", ~ " , ' " r r -

' ,']l'

,, i,p,~ H I ' , - -1[~1 I, , r I ~ I' 1' '~"1,'

FIG. 13--ZDDP C Original Oil and t = 20 volatiles and t = 10 residue.


TABLE, 1 X--Summary of3'P NMR results for ZDDP C (4-Methyl-2-1aentanol+Low Phos. PeSs).
Sample: ZDDP-C
Selby - Noack PEI,
Chemical shitts (ppm)
Volatility
m~/L
99.2, 93.1, 77.3
Oil
Volatiles, T=10 rain. 4.28 %
5
103, 97, 94.1, 89.1, 83.2, 66.4
Volatiles, T=20 rain. 6.23 %
5
102.7, 101.6, 95.0, 94.5, 94.2, 93.5, 90.2, 89.2, 89.1, 88.1, 85.3,
83.1, 66.6, 26.5
Volatiles, T=30 rnin. 9.80 %
5
102.5, 101.5, 94.3, 93.5, 89.3, 88.1, 83.1, 66.7,
Volatiles, T=50 rain. 15.29%
4
102.3, 101.2, 95.3, 94.7, 94.4, 93.7, 93.5, 83.1, 66.8, 26.8
Residue, T=10 rain.
3 (broad)
Residue, T=20 rain.
4 (broad)
Residue, T=30 rain.
3 (broad)
Residue, T=50 rain.
3 Ibroad)
SIl~pleZDDPD

,ir.l~-ir Hi ~ "r ['" .i,i- ~ . 1~-. U,, .,.[r,T ...... i...1.~.,rni..,p~.. ,], ~..,n ,n , l~li?vl'" T

Orlgln~O~OII

FIG. 14---ZDDP D - Original Oil and t = 10 volatiles and residue.

266

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

T A B L E 12--Summary of31P NMR results for ZDDP D(4-Methyl-2-pentanol+High phos. PeSs).


Sample: ZDDP D
Selby - Noack PEI,
Chemical shiRs (ppm)
Volatility
mg/L
100.4, 93.1, 77.3, 5.5
Oil
Volafiles, T=10 rain. 3.98 %
102.9, 101.9, 94.2, 93.9, 83.2, 66, 26,5, 5.9, 5.5
Volatiles, T=20 rain. 5.83 %
102.4, 101.3, 83.1, 67.5, 66.7, 26.6, 5.9, 5.4, 4.8
Volatiles, T=30 rain. 8.65 %
102.6, 94.5, 83.1, 67.8, 66.7, 5.5, 4.8,
Volatiles, T=50 min. 13.99%
102.2, 95.4, 94.5, 83.1, 66.8, 5.9, 5.4, 4.8
Residue, T=10 rnin.
4 (broad)
Residue, "1"--20rain.
4 (broad)
Residue, T=30 rain.
3 (broad)
6 (broad)
Residue, T=50 rain. . ,7
TABLE 1 3 - - I n t e g r a l s of the 31p NMR results for ZDDP A (2-Ethylhexanol + Low Phos. PeSs).

(Reported as relative mole %.)


104pprn 97ppm

ZDDPA
Sample
V, t=10
V, t=20
V, t=30
V, t=50
R, t=10
R, t=20
R, t=30
R, t=50

7
5
5

84pprn

69ppm

30ppm

2ppm

6
3

28
30
27
28
6
4
5
5

5
5
4
3
2
1
1
1

55
69
67
69

54
57
64
70
38
26
28
25

trace

T A B L E 14---Integrals of the 3Jp NMR results for ZDDP B (2-Ethylhexanol + High Phos. P2S5).

(Reported as relative mole %.)


zDDPB
Sample
V, t=10
V, t=20
V, t=30
V, t=50
R, t=10
R, t=20
R, t=30
R, t=50

104ppm 97ppm
9
4
3
2

53
62
67
67
32
24
23
19

84ppm

69ppm

6
2

26
28
25
27
5
4
4
7

65ppm

2
2
2
1

30pprn
6
5
5
5
2
1

2ppm

59
69
71
73

Discussion

General
From the foregoing studies
engine oils decompose during
species, some portion o f which
ZDDPs also break down under
in an engine.

it is evident that the ZDDPs in all the engine oils and simulated
the Selby-Noack test and are converted to different phosphate
are volatile. Similarly, it has been shown in other studies [2] that
analogous conditions o f temperature, time, and oxygen exposure

BOSCH ET AL. ON PNMR SPECTROSCOPY

267

First Study
The NMR analyses shown in Figs. 1-5 and Tables 1-5 indicated that the various samples of
volatiles from these Selby-Noack volatilization tests contain multiple phosphorus species and
that most of these species are different from those present in the original motor oil. Perhaps more
interesting, the first study gave clear evidence that different formulations of engine oils with
different values of PEI gave considerably different phosphorus breakdown products. The residue
oil left in the heated cup after completion of the Selby-Noack volatility test also contains totally
different phosphorus species than those in the original motor oil. More specifically, all of the
residue samples were observed to consist of a broad peak centered at 0-4 ppm, which is
indicative of inorganic and/or simpler organophosphate compounds than ZDDP. In these typical
GF-3 oils, a PEI of 8 corresponds to approximately 1 % of the phosphorus volatilizing from the
engine oil sample.
In three of the five motor oils tested, the ZDDPs completely decomposed under test
conditions. In samples 7450 and 7538, small amounts of ZDDP remained in the residual oil at
the end of the test.

Second Study
In the experiments on RO-780 (typical GF-3 oil) in which the effect of extending the time of
the heat up ramp from 9 min to 60 min was examined, it was found that although the PEI
increased when the time to operating temperature was considerably extended, the NMR spectra
did not change significantly. That is, the same species were volatilized from the oil. This is
illustrated Fig. 6 and summarized in Table 6.

Third Study
The experiments performed on samples EO-7450 and RO-780 (see Figs. 7-gand Tables 7
and 8) were, as noted, run to determine what and when different volatile phosphorus species
were generated. The tests were run in 10-, 20-, 30-, 50-, and 60-rain (from the first and second
studies) exposure times to the Selby-Noack operating condition. From these samples the NMR
analyses of the residue and volatile samples collected from the samples at 10 and 50 min were
compared
Unexpectedly, in the RO-780, the decomposition of the motor oil was essentially complete
after 10 min. In addition, it was noted that the PEI of the 50-min sample was lower than that of
the 10-min sample. Although this might be associated with repeatability of the PEI at low values,
it may also indicate that some of the phosphorus-containing species in the collected volatiles
could be lost because of even further volatilization from the collection vessel during the SelbyNoaek test. This requires further investigation. The NMR spectra of the volatiles obtained in
these two runs were compared with Figs. 5, 6, and 9. Qualitatively, the respective spectra were
found to be very similar, which illustrates the repeatability of the volatilization and collection
technique as an analytical method.
As observed before, in sample 7450, the decomposition of the phosphorus containing species
is not complete after 50 rain. It is very interesting to note that both the 10-rain and 50-rain
residue spectra are significantly different from that of the original oil sample. It appears that this
oil first decomposes to an intermediate state, which then undergoes further reaction to generate
additional phosphorus volatile species. It is also noted that the NMR spectra obtained on the 10and 50-mi volatile samples qualitatively consisted of the same peaks. However, the proportionate
size of the peaks varied between the two samples.

268

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

Fourth Study---ZDDP Composition Effects


Evolution of volatile components was found to depend strongly on the alcohol used in ZDDP
manufacture, as illustrated in Figs. 10-14 and Tables 9-12. ZDDPs C and D made from the
secondary alcohols, 4-methyl-2-pentanol, appear to break down readily under the conditions of
the Selby-Noack test. Within the first 10 min, whatever volatile species were formed were given
off, and the only phosphorus-containing species in the residual oil were fully oxidized
phosphates. Interestingly, the PEI values were comparatively low and constant.
In contrast, ZDDPs A and B made from the primary alcohols, 2-ethylhexanol, have a longer
life. Volatile phosphorus-containing species were given off for up to 60 min. Concomitantly, the
residual oil contained some of the original ZDDPs, but consisted primarily of other phosphoruscontaining components in addition to fully oxidized phosphates.
However, in contrast to aforementioned results with the primary alcohol, the PEI values
increased with time and became comparatively high. These differences in rate and degree of
breakdown between ZDDPs made with primary and secondary alcohols reflect other findings in
the literature [2].

Fourth Study--Stoichiometric Effects


Results with ZDDPs A, B, C, and D---In the limited testing described in this paper, the
phosphorus/sulfur molar ratio of the PzS5 does not appear to have a very significant effect on the
composition of the volatile species generated from the ZDDP made from 2-ethylhexanol. Similar
results were obtained from ZDDP made from both High and Low Phos P2S5. However the PEI
values obtained on the ZDDPs manufactured form High and Low Phos PzS5 and 4-methyl-2pentanol differed significantly. Close examination indicates that some of the unique impurities in
ZDDP D are most likely volatilizing and thereby increasing the PEI value of this oil. These
impurities are observed in both the original oil and the volatiles at 3-8 and from 65-70 ppm.
Unpublished studies at Astaris have found that typical yields of dithioacids (the ZDDP
precursor prepared from the reaction of P2S5 and alcohol) are generally between 85 and 95 %.
These reactions are typically run with an excess of alcohol to drive the reaction to completion.
The phosphoms/sulfiar molar ratio in the P2S5 has been found to affect the trace components
generated during the reaction of PzS5 and alcohols. Use of P2S5 o f a slightly lower-thanstoichiometric phosphorus level generates some impurities that are rich in sulfur, while some of
the impurities generated in reactions performed with PzSs at a slightly higher-than-stoichiometric
level of phosphorus consist of p+3 compounds but do not include sulfur-rich impurities.
Similarly, the P+3 impurities of higher than stoichiometric reactions are not generally found in
the products generated from P2S5 with a lower than stoichiometric level of phosphorus. This was
one of the primary points of interest in performing the stoichiometric experiments with P2S5, i.e.,
to see if these different impurity mixtures lead to ZDDPs with different PEIs, which, in fact,
appears to be the case.
To generate more detailed information regarding the composition of the volatile products
from the PEI studies, the relative ratios of the various peaks found in Tables 13 and 14, which
were taken over the time interval of 10-50 rain of the Selby-Noack test, were multiplied by the
PEI values of these samples to calculate the PEIs of the individual peaks. That is, the total PEI
value for the sample was separated into the individual PEI components in the same relative ratio
as the 31p NMR peaks. The results o f this comparison are presented in Figs. 15 and 16 and show
interesting trends.

BOSCH ET AL. ON PNMR SPECTROSCOPY

269

FIG. 15--ZDDP A, PEI values by NMR peak.

FIG. 16--ZDDP B, PEI values by NMR peak.


Some peaks grow while some stay the same or slightly shrink during the duration of the test.
Two peaks, those at 104 and 84 ppm, remain the same or shrink. The peaks at 97 and 69 ppm
clearly grow, and the peak at 30 ppm appears to grow somewhat, but not dramatically. Before
the test, Oil A has peaks at 104, 98.1, 69.5, and 65 ppm, and Oil B has peaks at 104.1, 103.5,
98.5, 69.7, and 65 ppm. Therefore, it is possible that the peak at 104 ppm is an impurity that
volatilizes and the small peak at 69 ppm in the before-test oils is generated during the thermal
decomposition of the oils during the PEI test and subsequently is volatilized. The large peak at
approximately 98 ppm in the before-test oil is perhaps represented as the dithioacid in the
volatilized oil samples. Further study is required.

General Observations
Volatile components from ZDDPs may arise from the manufacturing process. Zinc
dithiophosphates are typically a mixture of phosphorus containing species. In addition to the
desired ZDDP, [(RO)2P(S)S]2Zn and basic ZDP salt, there are a number of other trace
components containing phosphorus, as shown in Table 15. Some of the trace components in the
unused oil were also observed in the volatiles.

Expedment

lab oven 180*C,


in air

lab; 200=to
260~ in air *~

lab; 150" to
200~ in air *

[,o3] i

m]

i r~toss]

40~

17o]

I
[691

(RO)zP(S)OR
[60 to 72]

[96to100)

I
J'40TO 45]

(RohP(s)o"
I

(RO)zP(S)O- I
~ all ZDDP gone in 4 hr at 200" and 260~

Feto8o]

I
I

I(RO)zP(O)SR
[20to26]

[0.5 to .12]

I
I
I

I
I

I '~"~"~

I ~~

I (RO}~P(O)
[13t0-13] I

I (RO)sP(O)

(RO)2P(O)SR
[24 to 30]
[13 to-13] I

The volatile components from ZDDPs may also arise from thermal degradation and/or oxidation. The mechanisms proposed for
ZDDP anti-wear and anti-oxidation performance in motor oil have included both the effects of ZDDP degradation and oxidation [2].
In addition, 3~p NMR studies at other Laboratories have identified ZDDP breakdown fragments in the used motor oil [12], as shown in
Table 15.
Moreover, ZDDP breakdown fragments have been observed using 3~p NMR in lab tests elsewhere [13-16]. These lab tests were
also done in the presence of air. So the relative contribution of thermal degradation versus oxidation is not clear. The direct oxidation
era ZDDP with hydroperoxide [17], as observed by 3~p NMR, also gives fragments that may be volatile.

*ZDDP gone in 1 hr at 200~ or 24 hr at 150~

~101to103]

-,,

(RO)2P(S)O" [(RO)zP(O)S]~
[49l
[22]

[37 to 551

'shift, ppm]

((RO)=P(S)S)eZn40((RO)2P(S),S)~Zn~(RO)zP(S)SR (RO)2P(S)SH[ (RO)2P(S)OR

1817-..ODPsyrlthesis

,]

ZDOP+ cumene
17 ~ hydropemxide ((RO)2P(S)S)eZn,OI ((RO)2P(S)S~Zn] (RO)zP(S)SR

16

15 I

""14,'1 lab4-ballin
air rigs

13

15000km ((RO)2P(S)S)eZn~O((RO)2P(S)S)2Zr (RO)2P(S)SR


12 ertgine
field test

Ref.

((RO)2P(S)S)sZr~C ((RO)2P(S)S)~.r RO)2P(S)SnR(RO)2P(S)SH (RO)zP(S)OR (RO)zP(S)O- [(RO)2P(O)S]2 (ROhP(O)SR (RO)3P(O) (OR)P(S)(SR)2 (S)P-(SRh
sulfide n =
basic ZDDP
neutral ZDDP
I r2r3
thio acid thiophosphate :hiophosphodcsai'
thiophosphate phosphate

Species Observed.,

Table 15 - Literature Survey of 31p NMR Analyses of ZDDPs

i1>
z
0
FC

C
m
r"

o-rl

ffJ

t'---

m
z

m
rm

",.4
0

BOSCH ET AL. ON PNMR SPECTROSCOPY

271

The evolution of volatile components has been found to depend strongly on the alcohol used
in ZDDP manufacture. ZDDPs made from the secondary alcohol 4-methyl-2-pentanol appear to
break down readily under the conditions of the Selby-Noack test, but only a small percentage of
the phosphorus break-down products were volatile. Within the first 10 min, all the volatile
species were given off, and the only phosphorus-containing species in the residual oil were fully
oxidized phosphates. In contrast, ZDDPs made from the primary alcohol 2-ethylhexanol appear
to have a longer life but produce more volatile phosphorus over their breakdown life. Volatile
phosphorus-containing species were generated for at least 60 min. In concert, the residual oil
contained ZDDPs and other phosphorus-containing components in addition to fully oxidized
phosphates.
The rapid breakdown of ZDDPs has also been observed elsewhere: Lab tests in air using 31p
NMR show that all the ZDDP is gone in 1 h at 200~ [15], 24 h at 150~ [15], 4 h at 200~ [16],
or 4 h at 260~ [16]. In the latter two tests, the only phosphorus containing species was a fully
oxidized phosphate.
A recently published study by scientists at MIT [20] has indicated how to quantify the
amount of oil lost by three possible mechanisms. These include thermal decomposition of the oil
in the engine, aerosol generation, and oil volatilization. The information gathered in the present
paper using the Selby-Noack and the related NMR spectroscopy testing methods can be used to
quantify phosphorus volatility in the same manner. Measurement of zinc in the volatiles can be
used to determine the amount of phosphorus generated via aerosol generation. In addition, it
appears that careful 31p NM-R analyses of the volatiles could be used to measure the phosphorus
generated by volatilization of the components of the motor oil versus the phosphorus homologs
generated by thermal decomposition.
Earlier studies by Savant showed that there were trace levels of zinc in the phosphorus
volatiles obtained from the Selby-Noack test. The combination of varying the length of the
Selby-Noack test, followed by 3lp NMR analysis of the volatiles and residues, has the potential
to be a very useful tool to further understand the mechanism of phosphorus volatility of motor
oil.
Conclusions

The 31p ~
data are reproducible using the Selby-Noack procedure. For the same
conditions of time and temperature, the information, though limited, shows that the composition
data are consistent when the experiments are repeated.
Different oils formulated with different ZDDPs and other additives have different volatile
components. Moreover, the composition of the volatile phosphorus containing components
depends on the specific ZDDP.
In most cases, the phosphorus forms found in the volatiles appeared to contain some of the
ZDDP additive. But the 31P NMR spectrum also indicated the presence of more species in the
volatiles than in the untested oil. That is, the volatile components appear to be generated
primarily by the decomposition of ZDDP. However, volatile phosphorus containing trace
components arising from the ZDDP manufacturing process may be significant in some cases.
It appears possible to meet engine oil performance standards using additives having low
phosphorus volatility. Although all the oils in the IOM Database for North America are
presumed to have met minimal standards for engine wear and oil oxidation, there is a large
difference in Phosphorus Emission Index (PEI) among these oils. By using PEI as an additional
criterion for additive selection, it would seem reasonable to provide protection for both the

272

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

engine and the emissions control system. Certainly, approaching the control of phosphorus
volatility by actual measurement is highly preferable to attempting to control phosphorus
volatility by limiting the phosphorus content of fresh engine oil.
Acknowledgments
The authors would like to thank Dr. Andre' d'Avignon at Washington University for
performing the 31p NMR studies, Mr. Skip Ramsey for synthesizing the ZDDP samples, and
Chevron Oronite Company, LLC for providing typical GF-3 oil samples with and without
ZDDP.
References

[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]

Selby, T. W., Bosch, R. J., and Fee, D. C., "Phosphorus Additive Chemistry and Its Effects
on the Phosphorus Volatility of Engine Oils," submitted to ASTM International.
Spikes, H., "The History and Mechanisms of ZDDP," Tribology Letters, Vol. 17, No. 3,
2004, pp. 469-489.
Selby, T.W., et al., "A New Approach to the Noack Test for Volatility Measurement," SAE
International Fuels & Lubricants Meeting and Exposition, Philadelphia, PA, USA, 1993.
K. Noack, Angewandt Chemie, Vol. 49, 1936, p. 385.
ASTM Standard D 5800, "Evaporation Loss of Lubricating Oils by the Noack Method,"
Annual Book of ASTM Standards, Vol. 5.03, ASTM International, West Conshohocken,
PA, 2004, pp. 352-366.
Determination of Evaporation Loss of Lubricating Oils, (Noack Method), DIN 51-581,
1981.
Selby, T. W. and Reichenbach, E. A., "Engine Oil Volatility Studies - Generation of
Phosphorus," Proceedings of the International Tribology Conference, Yokohama, Japan,
1995, pp. 813-816.
Institute of Materials Engine Oil Database, Published by the Institute of Materials,
Midland, Michigan, issued yearly from 1984.
Selby, T. W., Development and Significance of the Phosphorus Emission Index of Engine
Oils, Proceedings of 13th International Colloquium - Lubricants, Materials, and
Lubrication Engineering, Esslingen, Germany, 2002, pp. 93-102.
Johnson, M. D., McCabe, R. W., Hubbard, C. P., Riley, M. E., Kirby, C. W., Ball, D.J., et
al., "Effects of Engine Oil Formulation Variables on Exhaust Emissions in Taxi Fleet
Service," SAE Paper #2002-01-2680, SAE Powertrain Meeting, 2002.
Selby, T. W., "Phosphorus Volatility of Lubricants - Use of the Phosphorus Emission
Index of Engine Oils," 9th F&L Asia Conference, Singapore, January 21-24, 2003.
Peng, P., Hong, S. Z., and Lu, W. Z., "The Degradation of Zinc Dialkyldithiophosphate
Additives in Fully Formulated Engine Oil as Studied by P-31 NMR Spectroscopy,"
Lubrication Engineering, Vol. 50, No. 3, 1994, pp. 230-5.
Coy, R. C. and Jones, R. B., "The Thermal Degradation and EP Performance of Zinc
Dialkyldithiophosphate Additives in White Oil," Internationales Jahrbuch tier Tribologie,
Vol. 1, 1982, pp. 345-50.
Coy, R. C. and Jones, R. B., "The Degradation of Zinc Dialkyldithiophosphate Additives
in Rigs and Engines," 1Mech E Conference Publications, Vol. 1, 1982, pp. 17-22.

BOSCH ET AL. ON PNMR SPECTROSCOPY

273

[15] Fuller, M. L. S., Kasrai, M., Bancroft, G. M., F3ffe, K., and Tan, K. H., "Solution
Decomposition of Zinc Dialkyl Dithiophosphate and Its Effect on Antiwear and Thermal
Film Formation Studied by X-Ray Absorption Spectroscopy," Tribology International,
Volume Date 1998, Vol. 31, No. 10, 1999, pp. 627-644.
[16] Harrison, P. G. and Brown, P., "External Reflection FTIR, Phosphorus-31 MAS NMR and
SEM Study of the Thermal Decomposition of Zinc(II) Bis(O,O'-dialkyldithiophosphates)
on 316 Stainless Steel," Wear, Vol. 148, No. 1, 1991, pp. 123-34.
[17] Yagishita, K. and Igarashi, J., "31P NMR and Mass Spectrometric Studies of the Reaction
of Zinc Dialkyldithiophosphates with Cumene Hydroperoxide. (Part 1). Kinetics and
Mechanisms of the Initial Homolytic Reaction," Sela'yu Gakkaishi, Vol. 38, No. 6, 1995,
pp. 374-83.
[18] Zimmermann, V., Jaeger, G., and Meyer, H., "Characterization of Reaction Mixtures in
Zinc Bis(O,O'-dialkyl dithiophosphate) Synthesis by Phosphorus-31 NMR Spectroscopy,"
Chemische Technik (Leipzig, Germany), Vol. 38, No. 4, 1986, pp. 169-72.
[19] Oehler, R., Zimmermann, V., and Jager, G., "Preparation of Zinc Dialkyl
Dithiophosphates. Reaction and Reaction Products," Erdoel & Kohle, Erdgas,
Petrochemie, Vol. 40, No. 2, 1987, p. 87.
[20] Yilmaz, E., Tian, T. Wong, V. W., and Heywood, J. B., "The Contribution of Different Oil
Consumption Sources to Total Oil Consumption in a Spark Ignition Engine," SAE Paper,
2004-01-2909.

STP1468-EB/Sep. 2005

Author Index
A

Mason, Kelly John, 42, 196


McEIroy, Frank C., 42, 196
Mennito, Anthony S., 42
Mustacich, Robert V., 164

Anderson, Daniel E, 71
B

Beumer, Berry, 116


Bosch, R. J., 239, 255
Brouwer, P., 98
Bryce, Derek W., 207

Nash, Lisa J., 232


Ngobeni, Prince, 59

Chen, Z. W., 116


Corns, Warren T., 207
Cruko, John, i 37

Onyeso, Chris C., 17


P

Paszek, Joseph, 42
P6cheyran, C., 51
Prinzhofer, A., 51
Priola, Marco, 59

Diehl, John W., 221


Di Sanzo, Frank P., 221
Donard, O. E X., 51
Dreyfus, S., 51

R
F

Radley, Ian, I 16
Rhodes, John R., 152
Ricchiuto, Silvia, 59

Fee, D. C., 239, 255


Fess, Scott, 108
Fox, Brian S., 24, 196
G

Selby, T. W., 239, 255


Shearer, Randy L., 164
Stockwell, Peter B., 207

Gras, Rhonda L., 164


I-I
Horton, Meaghan, 33
Hwang, J. David, 33

T
Tarkanic, Steve, 137
Tittarelli, Paolo, 59

Kariher, Peter H., 181


Katskov, Dmitri A., 59
Kirchgessner, David A., 18 I
Kishore Nadkarni, R. A., I, 85

V
Vrebos, B., 98

Leong, David, 33
Liang, Lian, 181
Lienemann, C. P., 51
Lukas, Malte, 71
Luong, Jim C., 164

Wei, Fuzhong, 116


Wilhelm, S. Mark, 181
Wissmann, Dirk, 128
Wolska, J., 98
u

Magnier, C., 51

Yurko, Robert J., 71


275

Copyright9 2005by ASTMInternational www.astm.org

STP1468-EB/Sep. 2005

Subject Index
A
Additives, 17, 24, 221,239
Analysis, perfect, 1
Analytical laboratory, I
Asphaltenes, 51
ASTM D 5453, 137
ASTM D 5623, 164
ASTM D 6920, 152
ASTM D 7039, 116
Atomic emission detection, 221
Atomic fluorescence spectrometry, 207

Fluorine, 232
Fossil fuels, 207
Fuels, 108, 221
analysis, 98, 232
G
Gas chromatography, 221
Gasoline, 17, 59, 85, 137, 223
low sulfur content, 128
ultra-low sulfur, 152
Gasolines, 221
l-I

C
Combustion cold vapor atomic absorption
spectrometry, 196
Combustion/oxidation UV-fluorescence
detection, 137
Comprehensive two-dimensional gas
chromatography, 22 I
Contaminant particles, analysis, 71
Crude oil, 51, 108, 181, 196
D

Diesel fuel, 17, 59, 85, 116, 137


ultra-low sulfur, 116, 137, 152
Doubly curved crystal optics, 116
Dual plasma sulfur chemiluminescence
detector, 164

Hetero-atom speciation, 221


Hydrocarbons, sulfur determination, 164

Inductively coupled plasma atomic


emission spectrometry, 17, 24, 33,
71
Inductively coupled plasma-mass
spectrometry, 33, 42, 51
Interlaboratory crosscheck programs, 85
Ion chromatography, 42
Ion-selective electrode, 232

Jet fuel, 59, 221

EDXRF, 128
Electrochemical sensors, 152
Electrothermal atomic absorption, 59
Elemental analysis, 98, 232
Energy dispersive x-ray spectrometer, 98
Engine oils, 239
Exhaust catalyst, 239, 255
Exhaust emissions, 239, 255
F

Fast on-line/at process sulfur


determination, 137
Filter furnace electrothermal atomic
absorption spectrometry, 59
Fluoride, 232

L
Liquid chromatography, 42
Lower limits of detection, 98
Low thermal mass gas chromatography,
164
Lubricants, 232
Lubricating grease, 24, 98
M

Maltenes, 51
Mercury, 181, 196, 207
volatile, 181
Microflow nebulizer, 51
Microwave digestion, 33
Monochromatic wavelength dispersive x-ray
fluorescence, 116

276

INDEX

277

Neutron activation analysis, 181


NIST 1084a, 51
NIST 1085b, 51
NIST 1634c, 51
Nitrogen converters, 152
Nuclear magnetic resonance spectroscopy,
255
O
On-line process analyzers, 152
Organic solvents, 51
Oxidative combustion, 232

Sample preparation, 33
Selby-Noack volatility test, 255
Selenate, 42
Selenite, 42
Selenium, 42
Selenocyanate, 42
Speciation, 181
Spray chamber, 17, 51
Sulfur, 108, 116, 128, 164, 221
analysis, 85
ultralow, in diesels, 85
Sulfur chemiluminescence detection, 164
Sulfur/nitrogen chemiluminescence, 232
T

3zp Nuclear magnetic resonance


spectroscopy, 255
Petroleum and petrochemical products, 33,
183, 207
Phosphorus emission index, 239, 255
Phosphorus volatility, 239, 255
Pipeline, 108
Pneumatic nebulizer, 17, 51
Polarization, ! 28
Polarized energy dispersive x-ray
spectrometry, 128
Pooled limit of quantification, I 16
Pyro-electrochemical on-line ultra low
sulfur analyzer, 152

Thermal oxidation, 152


Trace metals, 51, 59
Transverse heated filter atomizer, 59
U
Ultrasonication, 181
Ultra-trace metals, 42, 51
U. S. EPA 7473, 181
Used oil analysis, 71
W
Wavelength dispersive x-ray spectrometer,
98
Wear metals, 17
Wear particles, analysis, 71

Quality management, 1
X
R

Rapid sulfur analysis, 137


Refinery, 108
Refining, 181
Rotating disc electrode spectrometer, 71
Rotrode filter spectroscopy, 71
Round robin test, 128

X-ray fluorescence spectrometry, 98


X-ray spectrometry, 98
X-ray transmission, 108

Zinc dithiophosphates, 239, 255

Вам также может понравиться