Talanta 51 (2000) 537 545 www.elsevier.

com/locate/talanta

A ow-through uorescent sensor to determine Fe(III) and total inorganic iron

P. Pulido-Ton o, J.M. Barrero-Moreno, M.C. Pe rez-Conde *
Departamento Qu mica Anal tica, Facultad de Ciencias Qu micas, Uni6ersidad Complutense, Cuidad Uni6ersitaria, 28040, Madrid, Spain Received 5 July 1999; received in revised form 28 September 1999; accepted 14 October 1999

Abstract A ow-through uorescent sensor for the consecutive determination of Fe(III) and total iron is described. The reactive phase of the proposed sensor, which has a high afnity for complexed Fe(III), consists of pyoverdin immobilized on controlled pore glass (CPG) by covalent bonding. This pigment selectively reacts with Fe(III) decreasing its uorescence emission. Total inorganic iron is determined as Fe(III) after on-line oxidation in a mini-column containing persulphate immobilized on an ion exchange resin. The developed method allows the determination of Fe(III) in the 3200 mg l 1 range. The relative standard deviations of 10 determinations of 60 mg l 1 of Fe(III) and 20 mg l 1 of Fe(III) + Fe(II) are 3 and 5%, respectively. The sensor has been satisfactorily applied to speciate iron in synthetic, tap and well waters and wines. There were no signicant differences for total inorganic iron determination between this new method and the atomic absorption spectroscopy reference method at the 95% condence level. The sensor allows the concentration of Fe(II) to be calculated as the difference between total inorganic iron and Fe(III). The lifetime of the sensor is at least 3 months in continuous use or the equivalent of 1000 determinations. 2000 Elsevier Science B.V. All rights reserved.
Keywords: Iron(III/II) speciation; Siderophore; Flow-through uorescent sensor; Water analysis

1. Introduction Speciation is the determination of the individual physico-chemical forms of an element. Specia-

* Corresponding author: Tel.: + 34-91-394-4219; fax: + 3491-394-4329. E -mail address: cpconde@eucmax.sim.ucm.es (M.C. Pe rezConde)

tion is necessary to evaluate the toxicity, bioavailability, bioaccumulation and transport of a particular element [1]. Iron is one of the most important elements in metabolic processes, being indispensable for plants and animals and therefore it is extensively distributed in environmental and biological materials. Owing to the active role of iron in aquatic redox processes, the determination of the two oxidation states of this element, Fe(II) and Fe(III), is of great importance for

0039-9140/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 9 - 9 1 4 0 ( 9 9 ) 0 0 3 0 8 - 2

538

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545

estimating properties relevant to biological or geochemistry, such as the solubility, photochemistry and colloidal chemistry of the element [2,3]. The interactions of Fe(II) and Fe(III) with inorganic anions in natural waters have been studied using the specic interaction and ion pairing models [4]. Various methods of metal speciation have been introduced in recent years, including most notably ow injection (FI), which is an attractive analytical technique because of its exibility and ease of automation [5,6]. Most of these methods are based on spectrophotometric detection and require a preconcentration step to enhance their sensitivity [7,8]. Different kinetic approaches have also been used for speciation purposes [9]. Obata et al [10] determined Fe(III) and Fe(II) with high sensitivity and selectivity by chemiluminescence detection, but the method suffers from very restrictive working conditions prior to the preconcentration step. Haghighi and Safavi [11] determined simultaneously the two redox species in spiked tap water with opto-electrochemical detection, with detection limits of about 20 and 200 mg l 1, respectively. An integrated retention/spectophotometric method for iron speciation based on the Fe(III) thiocyanate complex has been described by Lopes da Conceic a o [12], however, it was not sensitive enough to determine concentrations at the mg l 1 level. The development of optical-chemical sensors by immobilizing a reagent on suitable matrixes has provided a technique for in situ analyte monitoring without reagent consumption and a very short analysis time [13]. In a previous work, we developed a sensor selective for Fe(III) by using pyoverdin, a natural uorescent siderophore biosynthesised by Pseudomonas uorescens, immobilized on controlled pore glass (CPG) and sol-gel glass as solid active phases [13,14]. This paper describes how the optical sensor has been modied and applied to iron speciation. The addition of a mini-column containing the oxidising reagent in the loop of an injection valve allows the determination of Fe(III) and total inorganic iron Fe(II + III).

2. Experimental

2.1. Instrumentation
A PerkinElmer LS-50 spectrouorimeter controlled by an IBM, model 55 SX computer and uorescence data manager (FDM) was used to measure the uorescence of the reactive phase, which is located in a Hellma quartz ow cell (18 ml and 3 mm light path). The bandwidths for the emission and excitation monochromators were all in cases of 5 nm. Solutions were pumped from a Gilson Minipuls two pump and pH was controlled by a Crison 2001 pH meter. The oxidation reactor consisted of a minicolumn (Omnit, 50 mm length and 3 mm inner diameter), containing the oxidising reagent, which was immobilized on an anionic resin and placed in the loop of an injection valve between the peristaltic pump and the reactive phase. The experimental set-up and details of the ow cell are shown in Fig. 1.

2.2. Reagents and materials

The biosynthesis of pyoverdin by Pseudomonas uorescens strain AR-11 and its immobilization on CPG have been described elsewhere [13,15]. , , particle size 3774 microns obCPG 460 A tained from Pierce, was used as an inert support for pyoverdin immobilization, 3-amino propyltriethoxysilane from Fluka and glutaraldehyde (Sigma) were employed during the immobilization process. Ion exchange resin Dowex 1 2 from Sigma was used as the support on which the oxidising agent was immobilized. A 1000 mg l 1 Fe(III) stock solution was prepared by dissolving Fe(NO3)39H2O in 1% HNO3 and titrating it with potassium permanganate. A 50 mg l 1 Fe(II) standard solution was freshly prepared by dissolution of Fe(NH4)2.(SO4)2.6H2O (Carlo Erba) and further dilutions as required were employed. Certied water (SLRS-3, NRC-CNRC) for total iron (100 9 2 mg l 1) was used for validation of the method.

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545

539

The optimum pH for formation of the Fe(III)pyoverdin complex was obtained with a solution of 0.01 M phthalic acid/biphthalate (Probus) (pH 4.5). HCl (1 M) from Merck was used to regenerate the reactive phase. Ammonium persulphate (Carlo Erba), cerium sulphate (Panreac) and hydrogen peroxide (Panreac) were tested as oxidising agents. Analytical reagent-grade chemicals were employed for the preparation of all solutions. Freshly prepared Milli Q (Millipore) ultrapure water was used in all experiments.

2.3.2. Pyo6erdin immobilization on CPG Immobilization was unchanged from the previously described procedure [13]. 2.3.3. Determination of Fe(III) The response curve of the reactive phase to Fe(III) was obtained by successively pumping through the cell: buffer (0.01 M phthalic acid/biphthalate), different concentrations of solution containing Fe(III) ( B 200 mg l 1) and nally the regenerator solution. The contact between the reactive phase and the buffer provides a constant uorescence signal. The uorescence intensity was monitored at 550 nm (excitation 500 nm) and recorded versus time. The Fe(III) solution, pumped for 2 min, was responsible for the observed decrease in the uorescence signal, whereas the 1 M HCl regenerator, releases the iron from the reactive phase and the buffer solution restores the initial uorescence signal of the solid phase, which is now ready for another iron determination. The calibration graph was obtained by plotting the slope of the decrease of the uorescence intensity versus Fe(III) concentration. 2.3.4. Determination of total inorganic iron The response curve of the solid phase to the mixture of Fe(III) and Fe(II) was evaluated by

2.3. Procedures 2.3.1. Immobilization of persulphate on polymeric support One gram of Dowex 1 2 resin was washed several times with Milli-Q water, oven-dried at 5060C for 10 h and then triturated in an agate mortar until it passed through a 0.160 mm sieve. Ten milliliters of 5% persulphate were added to the resin and the solution was mechanically stirred for 2 h [16]. After this contact time the resin was ltered, washed with water and nally placed in a glass minicolumn.

Fig. 1. Flow-injection manifold for the determination of Fe(III) and Fe(II).

540

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545

2.3.5. Quantication of Fe(III), total Fe and Fe(II) A calibration graph for Fe(III) was prepared by plotting the slope of the decrease of the uorescence signal versus Fe(III) concentration. Once the calibration curve has been prepared, the reading of the slope value of the sample allows the quantication of Fe(III), if we avoid the solution passing through the oxidation column. The slope obtained when the solution is passed through the oxidation reactor gives the amount of total iron in the sample (determined as Fe(III)). The amount of Fe(II) in the sample is calculated by subtraction of Fe(III) from total iron. 2.3.6. Determination of iron in water and wine The analysis of samples was performed after the addition of different known amounts of Fe(II) Samples were diluted with buffer solution to obtain the optimum pH for measurement. The calculation of Fe(III), total iron and recovery of Fe(II) were calculated as explained above.
Fig. 2. Chemical structure of pyoverdin siderophore and its complex with Fe(III) (AA, aminoacids). Table 1 Effect of Fe(II) on the determination of Fe(III) Fe(III) added (mg l1) Fe(III)/Fe(II) 60 80 120 1:50 1:100 1:50 1:100 1:50 1:100 Slope ( 9 S )a 0.082 9 0.03 0.085 9 0.04 0.090 9 0.04 0.119 9 0.05 0.120 9 0.04 0.125 9 0.05 0.159 9 0.03 0.162 9 0.05 0.165 9 0.04

3. Results and discussion The proposed method is based on the high selectivity and sensitivity of pyoverdin in the determination of Fe(III). This compound is a natural uorescent pigment with a high afnity for iron, which forms an Fe(III) complex [17]. This complex is non-uorescent, so the amount of iron can be quantied by the decrease in the uorescent signal. Fig. 2 shows the structure of the pigment and its iron complex. The immobilization process has been described previously and is based on the formation of covalent bonds between the glass and the reagent. This process can be divided into different steps, the rst of which is to provide a functional group (amine) on the glass to enable it to react with pyoverdin. Glutaraldehyde (a bifunctional agent) is used for the reaction between the aminated glass and the pyoverdin. The parameters affecting Fe(III) determination with pyoverdin, the effect of the presence of interfering elements and merits of this method were investigated in previous work [13].

a Slope of the decrease of uorescence intensity; s = standard deviation (n = 10).

pumping the mixture through the oxidation reactor. Contact with the reactive phase placed in the ow cell was observed as a decrease in uorescence intensity.

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545

541

Its specicity, its high sensitivity by Fe(III) and the fact that the presence of Fe(II) at concentrations 100 times higher than that of Fe(III) does not interfere (Table 1), thus, making it possible to evaluate the concentration of Fe(II) after oxidation of this species.

3.1. Oxidation step

The most important parameter to be xed was the oxidising agent because it had to be able to oxidise Fe2 + to Fe3 + (E : 0.77 Fe3 + /Fe2 + ) and not interfere in the determination of Fe(III) with pyoverdin. Different oxidising agents including (E : Ce4 + (E : 1.78), H2O2 (E : 1.78) and S202 8 2.01) were tested. To determine the effect of the presence of the oxidising agent on the reactive phase, solutions containing the same amount of Fe(III) (60 mg l 1) and different concentrations of the oxidising agent were prepared. These solutions were passed through the ow cell and the response was compared with that obtained when a solution of Fe(III) in the absence of oxidising agent was analyzed. This study yielded the following results: the use of Ce4 + interferes in the determination of Fe(III) for a Fe3 + /Ce4 + ratio of 1/10, and moreover the pyoverdin is less sensitive to iron in further determinations. The presence of H2O2 at the same concentration as the Fe(III) interferes in the iron determination and this interference is signicantly more severe for persulphate. Contact between the pyoverdin and the oxidising agent produces, in all cases, a decrease in the sensors sensitivity to iron, probably because pyoverdin is readily affected by an oxidation/reduction process, modifying the intrinsic characteristics of the pigments response to Fe(III) complex. Immobilization of the oxidising agent on an inert support could prevent the contact between the reactive phase and this agent and therefore impede the decrease in sensitivity or total destruction of the solid phase. Persulphate was chosen for immobilization since it has a higher oxidation potential, E , than the other agents tested. The oxidation reactor was placed in a loop of the injection valve as shown in Fig. 1. When the

valve was in the load position, total iron could be determined in the ow cell because the sample passed through the oxidation reactor before reaching the ow cell. When the valve was in the inject position, the oxidation step was prevented and Fe(III) was determined in the ow cell when the solution was in contact with the solid phase. The performance of the oxidation reactor in a ow system was studied. For this purpose a solution of Fe(II) was passed through the oxidation reactor at 1.5 ml min 1. The presence of Fe(III) in the efuent was observed qualitatively by a positive reaction with thiocyanate. The oxidising agent was present in considerable excess with respect to the Fe(II), but the oxidation step was performed under dynamic conditions, therefore, it was necessary to evaluate the oxidation efciency. A decrease in the oxidation efciency of the reactor with increasing Fe(II) concentration was observed, and 60 mg l 1 was the maximum amount of Fe(II) that could be quantitatively oxidized at 1.5 ml min 1 The ow rate used for the oxidation step should be compatible with the ow rate used to determine the iron in the ow cell. In evaluating the inuence of this parameter on the oxidation step, the upper limit of the ow rate is limited by pressure problems in both the ow cell and the minicolumn. The lower limit is also restricted by the poor sensitivity obtained for Fe(III) determination [13], although the extent of the Fe(II) oxidation increases. When low ow rate is used the amount of iron passing through the cell is small, the slope of curve response is minimum and the sensitivity of the method decreases drastically. Thus, a ow rate of 1.5 ml min 1 as a compromise was chosen. An important parameter affecting the oxidation process in a ow system is the solution residence time in the oxidation reactor. This parameter was tested using a stop-ow mode. Once the solution reached the reactor, the ow was stopped for 110 min. In Table 2 the uorescence intensity of the reactive phase, after different residence times in the oxidation mini-reactor, is summarized and compared with the response obtained for a solution of Fe(III). From this study we concluded that

542

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545 Table 5 Effect of NaCl on the Fe(III)-Pyoverdin reaction % NaCl 0 0.05 0.1 0.5 1.0 1.5 2.0
a

for 60 mg l 1 of Fe(II) the oxidation process is complete even when the ow is not stopped. However, the complete oxidation of a solution with a higher Fe(II) concentration requires a residence time of 10 min, indicating slow oxidation reaction kinetics rather than saturation of column capacity. The optimum conditions for Fe(II) and Fe(III) determination are summarized in Table 3.
Table 2 Relative signala after different residence times of Fe(II) in the oxidation micro-reactor t (min) Fe(II) 60 mg ll Fe(II) 100 mg ll Fe(II) 150 mg l1 0 1 5 10 101 9 3 99 9 4 100 9 3 101 9 4 87 9 2 93 9 3 101 9 3 100 9 3 70 9 2 84 9 4 95 9 3 101 9 4

Signala 0.1082 0.1071 0.1092 0.1071 0.076 0.054 0.043

Relative signal 100 99 101 99 70 50 40

Slope of the decrease of uorescence intensity.

a Expressed as the ratio of the signal of Fe(II) to that of Fe(III) at the same concentration.

Table 3 Optimum conditions for iron determination using immobilized pyoverdina Parameters Excitation maximum Emission maximum PH Buffer Regeneration Persulfate cone Flow rate Fe(III+II) determination 500 nm 550 nm 4.5 0.01 M (biphthalate/phthalate) 1 M HCl 2 mmol g1 resin 1.5 ml min1

a The experiments were carried out at 60 mg l1 Fe(III) and 20 mg l1 Fe(II).

Table 4 Analytical characteristics for determination of Fe(III) and Fe(II) using pyoverdin immobilized on CPG Parameters Linear range (mg l1) Detection limit (mg l1) Relative standard deviation Lifetime of reactive phase
a

Fe(III) 3200 3 3% 1000 determinations

Fe(II) 360 (100)a 3 5% 100 determinations

With stopped-ow of 5 min.

The analytical characteristics of the two sensors are shown in Table 4. The linear range, expressed as the slope of the response curve versus concentration, allows the determination of 3200 mg l 1 of Fe(III) with a linear regression y = 0.00922.34 10 3x (r 2 = 0.9952) and the determination of 3 60 mg l 1 of Fe(II) with a linear regression y = 4.6 10 4 2.31 10 3x (r 2 = 0.9947). Comparing the two calibration graphs, no signicant difference was found at the 95% condence level, which allowed the use of a single calibration graph for total iron inorganic determination. The detection limit of Fe(III) is 3 mg l 1 calculated as three times the standard deviation of the blank. The relative standard deviations of 10 determinations of 60 mg l 1 of Fe(III) and 20 mg l 1 of Fe(III) + Fe(II) are 3 and 5%, respectively. The detection limit of Fe(II) is also 3 mg l 1 and it relative standard deviation is 15% calculated as the difference (n = 10, 20 mg l 1 Fe(II)). The lifetime of the sensor was at least 3 months, which was equivalent to about 1000 Fe(III) determinations and the lifetime of the oxidation reactor was about 100 determinations. Different ions were selected to evaluate their effect on the sensor. The study showed that none of the ions tested interferes [13], which signies that the sensor is highly selective for Fe(III). An ionic strength higher than 0.5 M decreases the sensitivity of Fe(III) (see Table 5). We have chosen as ionic strength lower then 0.1 M in order to increase the reproducibility of the oxidation step of Fe(II).

Table 6 Recovery of Fe(III) and Fe(II) added to synthetic watersa [Fe(III)/Fe (II)] Fe(III) added Fe(II) added Total Feb found 54.7 88.1 110 59.8 118 177 118 131 152 Fe(III)bfound Fe(III) recovered (%) Fe(II) recovered (%) Total Fe recovered (%) 134 9 20 126 9 18 80 9 12 108 9 16 85 9 13 97 9 14 95 9 14 65 9 9 50 9 7 99 9 3 100 9 3 100 9 2 99 9 3 98 9 3 98 9 2 98 9 3 81 9 2 76 9 2

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545

10/1

5/1

1/1

50 80 100 50 100 150 60 80 100

5 8 10 10 20 30 60 80 100

48 9 1 78 9 2 102 9 2 49 9 1 101 9 3 148 9 3 61 9 1 79 9 2 102 9 2

96 97 102 98 101 99 101 99 102

a b

Concentrations expressed in mg l1. Average of ve determinations 9 s.

543

544

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545

3.2. Sample analysis

The proposed method using immobilized pyoverdin was applied to the determination of Fe(III) and Fe(II) in synthetic, tap and well waters and wines. Samples were analyzed after addition of Fe(II) and/or Fe(III) without any pre-treatment. The recovery of Fe(III) and Fe(II) added to the synthetic samples is shown in Table 6. As can be seen, there are no signicant differences between the amount of Fe(II) added and the amount found for all the ratios tested. The mean recovery of total iron was 99% when the amount of Fe(II) present in the sample was lower than 60 mg l 1. If the amount of Fe(II) is higher than 60 mg l 1, the on-line oxidation step should not be used and an oxidation step with a stop-ow method is recommended in order to obtain quantitative oxidation. All analyzed samples (different waters and wines) were spiked with Fe(II). The total iron content was also determined by ame or electrothermal vaporization atomic absorption spectroscopy (FAAS or ETVAAS) as alternative techniques. The results of analysis of real samples by the proposed method and by the reference techniques are shown in Table 7. No signicant

differences between the methods were found at the 95% condence level.

4. Conclusions The proposed ow-through sensor is useful for monitoring Fe(III) and total inorganic iron in several environmental matrices and wines. This method is very simple because it only requires a single calibration with Fe(III). The method is easy to adapt and allows 1015 iron determinations per h. Furthermore it can be used for 1000 determinations without deterioration of the active phase, (100 if the oxidation reactor is used) and can be regenerated in about 2 min by pumping 1 M HCl through the ow cell. The preparation of the oxidation reactor is very simple, which shows the advantages of the system. This reactor, in line, oxidises until 60 mg l 1 of Fe(II), higher concentrations can be oxidised quantitatively by a stop-ow step. Finally, the results show that the proposed method is accurate, precise, very selective and can be applied to the determination of Fe(III) and Fe(II) in a wide variety of materials.

Table 7 Application of the sensor to determine total Fe and Fe(III) in different natural waters and winesa Sample Tap water 1 Tap water 2 Well water 1 Fe(II) added 20 10 20 50 20 40 60 Total Fe foundc 578 9 2 179 9 3 198 9 4 160 9 3 169 9 2 181 9 2 203 9 3 101 9 2 114 9 3 45 9 4 137.9 9 0.4 157.0 9 0 4 177 8 9 0 5 198 0 9 0 9 Fe(II) found 21 9 3 991 18 9 3 48 9 5 21 9 2 40 9 4 61 9 6 Reference methodc 580 9 4 180 9 2 205 9 4 162 9 4 171 9 2 180 9 2 210 9 3 103 49 9 3 138 9 3 158 9 2 177 9 2 198 9 2

Certied waters SLRS-3 (100 9 2) Wine 1b Wine 2b Wine 3b

Concentrations expressed in mg l1. Dilution (1:100). c Average of 5 measurements, 9 s.

a b

P. Pulido -Ton o et al. / Talanta 51 (2000) 537545

545

Acknowledgements This work received nancial support from CICYT (AMB 98-1043-CO2-02). The authors wish to thank Max Gormann for revising the manuscript.

References
[1] T.M. Florence, Talanta 29 (1982) 345. [2] T.M. Florence, Analyst 111 (1986) 489. [3] J. Mortatti, F.J. Krug, L.C.R. Pessenda, E.A.G. Zagatto, Analyst 107 (1982) 659. [4] F.J. Millero, W.W. Yao, J. Aicher, Mar. Chem. 50 (1995) 21. [5] M.D. Luque de Castro, D.W. Bryce, A. Izquierdo, Talanta 42 (1995) 1215. [6] L. Campanella, K. Pyrzyuska, M. Trojanonowicz, Talanta 43 (1996) 825.

[7] S. Blain, P. Tre guer, Anal. Chim. Acta 308 (1995) 425. [8] D.W. OSullivan, A.K. Hanson, W.L. Miller, D.R. Kester, Limnol. Ocanogra. 36 (1991) 1727. [9] M. Endo, S. Abe, Fresenius J. Anal. Chem. 358 (1997) 546. [10] H. Obata, H. Karatani, E. Nakayama, Anal. Chem. 65 (1993) 1524. [11] B. Haghighi, A. Safavi, Anal. Chim. Acta 354 (1997) 43. [12] A.C. Lopes da Conceic a o, M.T. Tena, M.M. Correia dos Santos, M.L. Simoes Gonzalves, M.D. Luque de Castro, Anal. Chim. Acta 343 (1997) 191. [13] J.M. Barrero, M.C. Moreno-Bondi, M.C. Perez-Conde, C. Ca mara, Talanta 40 (1993) 1619. [14] J.M. Barrero, C. Ca mara, M.C. Perez-Conde, C. San Jose , L. Ferna ndez, Analyst 120 (1995) 431. [15] J.M. Meyer, M.A. Abdallah, J. Gen. Microbiol. 107 (1978) 309. [16] M. Faraldos, M.C. Moreno, C. Ca mara, Analusis 16 (1988) 87. [17] G. Anderegg, F. LEplattenier, G. Schwartzenbach, Helv. Chim. Acta 46 (1963) 1409.