Chem. Eng. Comm.

, 190: 1
14, 2003

Copyright # 2003 Taylor & Francis
0098-6445/03 $12.00 + .00
DOI: 10.1080=00986440390171071

REFRACTIVE INDEX OF SUPERCRITICAL

CO2-ETHANOL SOLVENTS

YONGDA SUN
BORIS Y. SHEKUNOV
Drug Delivery Group, School of Pharmacy,
University of Bradford, Bradford, United Kingdom

PETER YORK
Drug Delivery Group, School of Pharmacy,
University of Bradford, Bradford, United Kingdom;
Bradford Particle Design plc,
Bradford, United Kingdom

In-line analysis of refractive index is required for efficient design and mon-
itoring of supercritical fluid extraction and precipitation processes. In the
present work, a robust method has been developed based on measurements of
laser beam deviation using an interferometer and image processing system.
Data on refractive index of CO2-ethanol mixtures were obtained at pressures
between 70 and 200 bar and temperatures between 308 and 363 K, for con-
tinuous flow of premixed solvents and, in addition, for equilibrium gas phase
below the mixture critical pressure. The refractive index of a mixture is a
linear function of ethanol mole fraction and can adequately describe mixing
and phase behavior in the vessel. For pure CO2, refractive index was
determined as a function of pressure and density and its Lorentz-Lorenz
functions determined.

Keywords: Refractive index; Interferometry; Supercritical fluids; Compressed

gases

Received 6 December 2000; in final form 5 June 2001.

Address correspondence to Boris Y. Shekunov, Ferro Corporation, Technical Center,
7500 East Pleasant Valley Road, Independence, OH 44131. E-mail: shekunovb@ferro.com

1
2 Y. SUN ET AL.

INTRODUCTION
There is currently a growing interest in using supercritical fluids (SCF) in
chemical and pharmaceutical industries as a real alternative to existing
process technologies to improve product quality and reduce usage of
chlorinated and nonchlorinated solvents. Current application areas
include nutraceutical extraction, reaction chemistry and polymerization,
polymer and food fractionation, waste recycling, precision cleaning for
electronic and optical parts, impregnation of paper and polymers, aero-
gels, and crystallization and particle formation of pharmaceuticals
(Eckert, 1996; McHugh and Krukonis, 1994). Carbon dioxide (CO2) is by
far the most important SCF because of its relatively low critical tem-
perature (304.3 K) and pressure (73.8 bar) and its low cost. It is emerging
as a widely accepted ‘‘green’’ solvent, nontoxic and nonflammable.
Solvent refractive index, ns, is very important for characterization
and monitoring of SCF processes because it is related to such funda-
mental thermodynamic properties as solvent density, phase composition,
solute concentration, and interfacial tension (Reid et al., 1987). This
parameter can be directly and accurately measured using in situ optical
techniques. In addition, detection of ns is an attractive option for HPLC
separation due to the universal nature of such detection (Synovec and
Renn, 1991). However, for high pressure compressible solvents such as
CO2, there are significant experimental problems associated with the
intricate cell design, reproducibility, and complexity of such measure-
ments. For example, pass-length interferometry (Achtermann et al., 1991;
Achtermann and Baehr, 1989; Bose et al., 1986; Thomas and Tayag,
1988; Michels and Hamers, 1937; Adjoury et al., 1997) is capable of
measuring the absolute refractive index with a precision of ca. 10
5.
Although this technique is highly accurate, it is complex to operate and
also unsuitable for dynamic processes such as mixing or separation in
which ns can fluctuate. Minimum-angle-of-deviation method (Besserer
and Robinson, 1971, 1973) makes use of an autocollimating refracto-
meter with overall possible error of approximately 10 7 4. This method
requires relatively long measurement time (minutes per data point) and
very elaborate optical cell design with an internal mirror. A beam-
displacement technique (Kirby and McHugh, 1997) gives the advantage
of simple experimental design but may, however, lack accuracy, as a
typical error is about 10 7 2. To date, the refractive index of pure CO2 has
been measured using these techniques at temperatures between 298 and
373 K and at pressures up to 2400 bar (Michels and Hamers, 1937),
temperatures between 323 and 373 K and pressures up to 230 bar
(Adjoury et al., 1997), and temperatures between 310 and 394 K and
pressures up to 102 bar (Besserer and Robinson, 1973). In addition, ns of
CO2-n-butane system was measured at 310 K and pressures up to 80 bar
REFRACTIVE INDEX OF SUPERCRITICAL CO2-ETHANOL SOLVENTS 3

(Besserer and Robinson, 1971). There is very little data available on the ns
of organic solvent-CO2 mixtures. This is a considerable knowledge gap
because CO2 is miscible (or partially miscible below the mixture critical
point) with most organic solvents, forming homogeneous binary and
ternary solvent-CO2 systems. The ability to form such mixtures greatly
increases the solvating power and polarity range of supercritical CO2. In
particular for precipitation processes, solvent-to-CO2 mole fraction and
its fluctuation about average value is one of the most important para-
meters that define supersaturation during mixing (Shekunov et al., 2001).
Refractive index sensors can provide dynamic data and control
parameters for such processes.
The main aim of this work is to develop an accurate analytical
method for refractive index measurements in SCF suitable for dynamic
applications. CO2-ethanol is chosen as a model system. The measure-
ments were carried out for premixed solvents in a through-flow config-
uration and, in addition, for the gaseous CO2-ethanol phase at
equilibrium below the mixture critical pressure.

EXPERIMENTAL METHODS
SCF System
The principal scheme is shown in Figure 1. A high-pressure stainless steel
optical cell (about 30 cm3 volume) was manufactured by Thar Designs
Co. (Pittsburgh, U.S.). It had two parallel silica windows. The cell alu-
minum jacket included four electrical cartridges and a platinum resister
thermometer connected to a digital controller (2132 PID, RS, U.K.). This
allowed the cell temperature to be measured and controlled within
0.1 K. Pressure in the optical cell was controlled by a computerized
back-pressure regulator (26
1761 with ER3000 controller, Tescom, U.S.)
and maintained within  0.2 bar. The pressure transducer was calibrated
against a certified pressure transducer within  0.25%. Ethanol flow rate
was provided by a metering pump (Jasco PU-986, Japan) and varied
between 0.1 and 10 cm3=min. The CO2 flow, supplied from a surge tank
by a water-cooled pump (Milton Roy B, U.K.), was typically kept con-
stant at 18 mL=min (assuming 1 g=cm3 liquid CO2 density) using a
metering valve. This flow corresponded to expanded gas flow of about
9.5 Nl=min, which was monitored using a gas flow meter (SHO-Meter,
Brooks Instruments B.V., Holland). Both the ethanol and CO2 flows
were preheated before entering the optical cell (Figure 1).
The flows were premixed in a separate 50 cm3 vessel using a coaxial
twin-fluid nozzle (Figure 1). Above the mixture critical pressure, Pm, the
flows could be premixed at all ratios, whereas for the experiments below
Pm, excess of CO2 was supplied to form homogeneous gaseous phase at a
4 Y. SUN ET AL.

Figure 1. Schematic of the high-pressure system for measurement of the refractive index of
supercritical flow.

given mole fraction. Measurements of the two-phase system in equili-

brium were carried for the gaseous phase by pumping both ethanol and
CO2 in the optical cell. Pressure and temperature were kept between 60
and 200 bar and 313 and 363 K.

Optical System
A schematic diagram of the optical system is shown in Figure 2a. The
beam from a polarized, 5 mW, 632.8 nm He-Ne laser (Coherent Co,
U.K.) is expanded to a diameter of about 12 mm using lenses L1 and L2.
The Mach-Zehnder interferometer is built using fully reflecting mirrors
M2 and M3, and the beam splitters BS1 and BS2. The optical cell with a
glass prism is inserted into one arm of the interferometer. In contrast to
direct pass-length measurements (Achtermann et al., 1991; Achtermann
and Baehr, 1989; Bose et al., 1986; Thomas and Tayag, 1988; Michels and
Hamers, 1937; Adjoury et al., 1997), this interferometer is used to mea-
sure the angle deviation between the object and the reference beams,
which is defined by the distance between the interferometric fringes. Such
REFRACTIVE INDEX OF SUPERCRITICAL CO2-ETHANOL SOLVENTS 5

Figure 2. (a) Optical scheme; (b) laser beam deviation in the optical cell.

a configuration is more stable to refractive index fluctuations and more

reliable than counting the number of passing interferometric fringes. The
method is also resistant to scattering by particles in the measuring
channel because the beam intensity has no effect on measurements.
6 Y. SUN ET AL.

The interferograms were magnified using a microscope (Leica AG,

MZ6, Switzerland), then captured by a monochrome coupled charge
device (CCD) camera (Cohu Inc., U.S.) focused onto the optical cell and
recorded using a time-lapse VCR (Panasonic, Japan). The camera mag-
nification varied dependently on the measured fringe width. Real-time
image processing was carried out using a PCI video digitizer and inte-
grated high-performance VGA card (Bandit, Coreco Inc., Canada) cap-
able of displaying captured pictures in a resizable window and measuring
the fringe width with high resolution.
As shown in Figure 2a, if all optical components are of perfect
quality and aligned to produce parallel object and reference beams, these
beams arrive in the recording plane in phase, so no fringes appears in the
basic position of the interferometer. This is the ‘‘infinite fringe width’’
alignment. The finite fringe width alignment is produced by tilting one
plate (e.g., M3) by a small angle j. Tilting M3 is equivalent to inserting a
transparent wedge of angle j into the path of the object beam. This angle
can be measured using a goniometer or directly from the fringe width, D:

l
D¼ ð1Þ
j

where l is the laser wavelength. Provided that all angles are very small
(Figure 2b), the total deviation angle, j2 , is described by a simple
equation derived from the Snell’s law:

ðnp
ns Þa
j2 ¼ ð2Þ
na

where np, ns, and na are the refractive indices of the wedge prism, the SCF
fluid inside the optical cell, and the air outside the optical cell, corres-
pondingly. The measurement can be performed to obtain ns in respect to
air or relative to a previous measurement point, for example, between ns
of pure and modified CO2. A systematic error may be introduced if the
optical windows (or the flats of the same window) are not parallel or by
systematic error in the prism angle, a. These errors are examined below.

Operating Procedure and Accuracy

An optical cell with a fixed wedge prism was placed in the interferometer
(Figure 2a), the reflecting mirror M3 was adjusted to the ‘‘infinite fringe
width’’ alignment as a referential point, and the flow was pumped
through the optical cell. For the measurements in two-phase system,
systematic drift in the image with time was checked during equilibration.
REFRACTIVE INDEX OF SUPERCRITICAL CO2-ETHANOL SOLVENTS 7

This time depended on the pressure and temperature and was typically
between 30 and 60 min. Results from a minimum of three measurements
were used to calculate ns.
The experimental accumulated error of ns mainly depends on preci-
sion of the fringe width measurement. This precision can be estimated
using Equation (1) assuming that the linear error of the wavelength half-
width, l=2, is equal to the CCD size with characteristic dimension
D ¼ 12 mm. Thus, taking into account that the prism angle, a ¼ 0.211 rad,
the precision calculated using Equations (1) and (2) is Dns ¼
 1.2  10 7 4. In addition, the experimental error contains a contribu-
tion from the nonparallel alignment of the optical cell windows, the
magnitude of which is a function of ns (1<ns<1.2) and sample-to-image
plane distance (or magnification). The window alignment test was carried
out under pressure by measuring the beam deviation without the prism.
The windows were found to be parallel within 1.7  10 7 4 rad for the
whole interval of pressures and ns. The prism angle was measured with
accuracy 2  10 7 4 rad. Thus the combined systematic error of ns may
reach 2  10 7 3. This error, however, can be taken into account during
cell calibration and validation. Such calibration=validation experiments
were carried out using water as a test substance. Changes in the optical
properties of the wedge prism at pressures and temperatures applied were
negligible. The refractive index of air, na, was corrected to account for the
wavelength of the laser and temperature using a relationship given by
Edlen (1996). The ns results obtained for water at different pressures and
temperatures showed a good agreement with the tabulated data within
the errors specified.

RESULTS AND DISCUSSION

Pure CO2
Refractive index of the pure carbon dioxide was measured at 313 K and
363 K and pressures up to 150 bar (Figure 3). In Figure 4 these data are
plotted as a function of CO2 density and show a linear relationship.
Theoretically, the relationship between ns and density of gases can be
expressed by the Lorentz-Lorenz function:

ðn2s
1Þ
¼ RLL ð3Þ
ðn2s þ 2Þr

where r is the gas density in moles per unit volume and RLL is the molar
refraction. For a given substance, this function is almost constant,
showing only a slight decrease with increasing density and an almost
negligible decrease with increasing temperature. Some theoretical
8 Y. SUN ET AL.

Figure 3. Refractive index of pure CO2 as a function of pressure.

Figure 4. Refractive index of CO2 as a function of density. The dashed line corresponds
to Lorentz-Lorenz Equation (3) with molar refractivity obtained using Equation (4) and
empirical constants given by Michels and Hamers (1937).
REFRACTIVE INDEX OF SUPERCRITICAL CO2-ETHANOL SOLVENTS 9

improvements have been proposed to account for these variations. For

example, the polarization effects were taken into account in the models of
Raman and Krishhan (1928) and Kirkwood (1936). Buckingham and
Pople (1956) have suggested that the progressive variation of the
Lorentz-Lorenz function can be approximated by the following
relationship:

B C
RLL ¼ R0LL þ þ ð4Þ
V V2

where R0LL is the limit of the molar refractivity at zero density, B and C
are the second and third viral refractivity coefficients, respectively, and V
is the molar volume. The theoretical refractive index of pure CO2, cal-
culated using Equations (3) and (4) is given in Figure 4. The refractivity
R0LL ¼ 6:6447, for CO2 at the wavelength 632.8 nm, was obtained from
reference (Stoll, 1922). Coefficients B ¼ 1:10 and C ¼
265 were
obtained by analyzing data given in references (Michels and Hamers,
1937; Besserer and Robinson, 1973) by plotting ðRLL
R0LL ÞV versus
1=V. The CO2 density is taken from International Thermodynamic
Tables of the Fluid State 3 (1976). These results are in very good
agreement with the data obtained in the present work. The molar
refractivity calculated on the basis of present experimental results using
Equation (3) is shown in Table I. Clearly, RLL decreases with increasing
density in agreement with Equation (4). However, over the temperature
and pressure range in the present experiments, RLL changes insigni-
ficantly, and for ns values close to unity there is a simple linear rela-
tionship ðns
1Þ ¼ 1:5rRLL that follows from Equation (3) and was
observed experimentally (Figure 4).

Table I Refractive Index of Pure CO2 and Calculated Lorentz-Lorenz Function, RLL , at
313.15 K

Pressure (bar) Refractive index Density (g cm 7 3) RLL (cm3 mol 7 1)

60 1.0342 0.1490 6.6858

70 1.0453 0.1976 6.6682
80 1.0639 0.2784 6.6605
90 1.1120 0.4828 6.6636
100 1.1469 0.6287 6.6608
110 1.1606 0.6850 6.6614
120 1.1689 0.7193 6.6578
130 1.1750 0.7445 6.6564
140 1.1798 0.7645 6.6526
150 1.1839 0.7814 6.6484
10 Y. SUN ET AL.

Figure 5. Refractive index of gaseous CO2-ethanol mixture below the mixture critical
pressure.

Ethanol-CO2 Mixtures
The refractive indices of ethanol-CO2 mixtures at selected pressures and
temperatures are given in Figure 5 (two-phase region, partial miscibility,
P < Pm ) and Figure 6 (single-phase region, complete miscibility,
P > Pm ). The selected choice of experimental conditions is dictated by the
CO2 density, which varies from a relatively low ‘‘gaseous’’ state
(approximately 0.17 g=cm3 at 90 bar, 363 K) through the state close to the
critical (approximately 0.23 g=cm3 at 75 bar, 313 K) to a relatively high,
‘‘liquid’’ state (approximately 0.79 g=cm3 at 200 bar, 323 K). Importantly,
these conditions correspond to different mixing behavior between CO2
and ethanol solvents. The obtained dependencies can be approximated by
the following linear relations:

ns ¼ 0:423x þ 1:049 ð5Þ

ns ¼ 0:651x þ 1:039 ð6Þ

ns ¼ 0:160x þ 1:151 ð7Þ

where Equations (5), (6), and (7) correspond to 313 K, 75 bar; 363 K,
90 bar (Figure 5); and 323 K, 200 bar (Figure 6). The slope of these
dependencies decreases with increasing CO2 density. It is noted that in the
REFRACTIVE INDEX OF SUPERCRITICAL CO2-ETHANOL SOLVENTS 11

Figure 6. Refractive index of CO2-ethanol mixtures above the mixture critical pressure.

ideal case of the Lorentz-Lorenz law (3), ns is a function of both the

ethanol and CO2 densities. These densities have not yet been determined
for different ethanol mole fractions, x. At small x, the simplified form of
the Lorentz-Lorenz equation (above) can be rewritten as:

ðns
1Þ ffi 1:5rðRLL ð1
xÞ þ R LL xÞ ð8Þ

where RLL and R LL are the molar refractivities of CO2 and ethanol vapor,
correspondingly. Equation (8) also assumes that the molar density of
ethanol at small x is proportional to the molar density of pure CO2, r, at
a given pressure and temperature. Equation (8) is similar in form to
Equations (5)
(7). However, the fact that the slope of the experimental
dependencies decreases with increasing density indicates that the molar
refractivity of ethanol vapor, R LL , also decreases with increasing CO2
density. A strong dependence of R LL on pressure and temperature or, in
combination, on the fluid density, reflects changing interactions between
CO2 and ethanol molecules and between ethanol molecules.
It is well known that SCF mixtures have a nonideal behavior that
originates from the clustering and aggregation phenomena, affecting the
dielectric constants and the molar refractivity. This mechanism is not
fully understood. Molar refraction of binary liquid mixtures has been
previously estimated by applying the linear rule, on the basis of the mole
fraction and the molar refraction of pure liquids that were calculated
from atomic and structural contributions (Dutt and Prasad, 1996).
12 Y. SUN ET AL.

Figure 7. Refractive index of the equilibrium gaseous CO2-ethanol phase as a function of

pressure and temperature.

The deviations of molar refraction in subcritical organic mixtures were

fitted to a polynomial-type equations (Arce et al., 2000; Moumouzias and
Ritzoulis, 2000). Different equations such as Lorentz-Lorenz, Gladstone-
Dale, and Arago-Biot were applied for prediction of the refractive indices
using the molar refractivity (Tojo and Diaz, 1995). Molar refractivity was
also used to develop an equation of state to predict the behavior of
hydrocarbon liquids and vapors (Riazi and Mansoori, 1993). However,
according to the present experimental data, the compressibility of
supercritical CO2 and nonideal behavior of the cosolvent may lead to
different mixing rules for prediction of the molar refraction at different
pressures and temperatures. From a practical viewpoint, empirical linear
equations such as (5)
(7) can be obtained for each pressure and tem-
perature, above and below Pm . This simple approach allows the ethanol
mole fraction to be determined on the basis of refractive index
measurements.

Ethanol-CO2 Equilibrium
The refractive indices of the equilibrium gaseous CO2-ethanol phase are
shown in Figure 7 as a function of pressure and temperature. The
numerical data are included in Table II. Such a two-phase system can be
obtained only at pressures below Pm . For the phase equilibrium of
REFRACTIVE INDEX OF SUPERCRITICAL CO2-ETHANOL SOLVENTS 13

Table II Refractive Index of the Equilibrium Gaseous CO2-ethanol Phase

Pressure 313.15 K 323.15 K 333.15 K 343.15 K 353.15 K 363.15 K

(bar) Ref. index Ref. index Ref. index Ref. index Ref. index Ref. index

60 1.0545 1.054202 1.052036 1.050993 1.049683 1.049615

65 1.058394 1.057164 1.055546 1.054021 1.05106 1.05073
70 1.063051 1.061695 1.059104 1.05724 1.053145 1.052531
75 1.072061 1.066146 1.063491 1.060183 1.056644 1.054801
80 1.069821 1.067104 1.06389 1.060353 1.057833
85 1.075896 1.073269 1.068852 1.063493 1.060923
90 1.084436 1.080829 1.073651 1.06681 1.063393
95 1.092144 1.087881 1.079147 1.068733 1.066592
100 1.093952 1.086815 1.072491 1.069009
105 1.094446 1.07767 1.072848
110 1.10263 1.082475 1.075665
115 1.03046 1.087268 1.078941
120 1.092067 1.083712
125 1.089918

CO2-ethanol, the ethanol mole fraction and the density was estimated
using the Peng-Robinson equation of state (data not included). The
analysis shows that ns cannot be represented by the Lorentz-Lorenz
equation, which is similar to the result, obtained for undersaturated
mixtures below Pm (above).

CONCLUSION
A new method for measurements of solvent refractive index was devel-
oped that is based on the principle whereby a small glass prism is placed
inside a standard optical cell and refraction of a laser beam is then
analyzed using an interferometer. The method proved to be sufficiently
robust and accurate to measure the refractive index in dynamic flow
configurations as well as in stationary phase. The refractive index of pure
CO2 and ethanol-CO2 mixtures was determined at selected pressures and
temperatures. For pure CO2, the data obtained are in good agreement
with the published data and with the theoretical predictions based on
molar refraction and the Lorentz-Lorenz equation. For CO2-ethanol
mixtures, empirical linear dependences of ns on ethanol mole fraction are
shown; the slope of these dependences is a function of pressure and
temperature. This phenomenon is attributed to the interactions of ethanol
molecules in the supercritical state. Additional measurements are
required to understand this mechanism in the single- and two-phase
regions of CO2-ethanol mixture used as a model system in this study.
14 Y. SUN ET AL.

ACKNOWLEDGMENT
We gratefully acknowledge support of this work from the EPSRC. We
thank Dr. Andreas Kordikowski for providing computational data on
the phase behavior. Some results reported in this work were presented at
the 2000 AIChE Annual Meeting, Los Angeles.

REFERENCES
Achtermann, H. J. and Baehr, H. D. (1989). J. Chem. Thermodyn., 21, 1023.
Achtermann, H. J., Magnus, G., Hinze, H. M. and Jaeschke, M. (1991). Fluid
Phase Equilib., 64, 263.
Adjoury, C., Balu, N., Obriot, J. and Bose, T. K. (1997) J. Chem. Phys., 106, 7491.
Arce, A., Martinez-Ageitos, J., Rodil, E., Rodriguez, O. and Sato, A. (2000).
Fluid Phase Equilib., 170, 113.
Besserer, G. J. and Robinson, D. B. (1971). Can. J. Chem. Eng., 49, 651.
Besserer, G. J. and Robinson, D. B. (1973). J. Chem. Eng. Data, 18, 137.
Bose, T. K., St-Arnaud, J. M., Achtermann, H. J. and Scharf, R. (1986). Rev. Sci.
Instrum., 58, 2279.
Buckingham, A. D. and Pople, J. A. (1956). Discuss. Faraday Soc., 22, 17.
Dutt, N. V. K. and Prasad, D. H. L. (1996). Physi. Chem. Liq., 33, 171.
Eckert, C. A. (1996). Nature, 383, 313.
Edlen, B. (1996). Metrologia, 2, 71.
International Thermodynamic Tables of the Fluid State 3: Carbon Dioxide. (1976).
Pergamon Press, New York.
Kirby, C. F. and McHugh, M. A. (1997). Rev. Sci. Instrum., 68, 3150.
Kirkwood, J. G. (1936). J. Chem. Phys., 4, 592.
McHugh, M. A. and Krukonis, V. J. (1994). Supercritical Fluid Extraction:
Principles and Practice, Butterworth-Heinemann, Stoneham, 17955.
Michels, A. and Hamers, J. (1937). Physica, 4, 995.
Moumouzias, G. and Ritzoulis, G. (2000). J. Chem. Eng. Data, 45, 202.
Shekunov, B. Y. Baldyga, J. and York, P. (2001). Chem. Eng. Sci., 56, 2421.
Raman, C. V. and Krishhan, K. S. (1928). Proc. Royal Soc. Lond., A177, 589.
Reid, R. C., Prausnitz, J. M. and Poling, B. E. (1987). The Properties of Gases and
Liquids, 4th ed. New York: McGraw-Hill.
Riazi, M. R. and Mansoori, G. A. (1993). Oil Gas J., 91, 108.
Stoll, E. (1922). Ann. Physik, 69, 81.
Synovec, R. E. and Renn, C. N. (1991). Proc. SPIE, 1435, 128.
Thomas, M. E. and Tayag, T. J. (1988). Appl. Opt., 27, 3317.
Tojo, J. and Diaz, C. (1995). J. Chem. Eng. Data, 40, 96.