Talanta 62 (2004) 389–394

Gas diffusion–flow injection determination of total inorganic carbon inwater using tungsten oxide electrode

L. Monsera,∗, N. Adhouma, S. Sadokb

a Département de Chimie, Centre Urbain Nord, Institut National des Sciences Appliquées et de Technologie, B.P. 676, 1080 Tunis Cedex, Tunisiab Institut National des Sciences et Technologies de la Mer, 28 Rue 2 Mars 1934, 2025 Salammbo, Tunisia

Received 14 April 2003; received in revised form 3 July 2003; accepted 12 August 2003

Abstract

A novel gas diffusion–flow injection method has been developed for the rapid and sensitive determination of total inorganic carbon (TIC)in water. The method is based on the diffusion of CO2 across gas permeable membrane from a donor stream containing 0.1 M HCl to anacceptor stream of sodium acetate (10−5 mol l−1 and pH 10). The CO2 trapped in the acceptor stream passes through an electrochemicalflow cell contains a tungsten oxide wire and a silver/silver chloride electrode, where it was sensitively detected. The parameters affecting thesensitivity of the electrode such as buffer concentration, pH, flow rate and injected volume were studied in detail. The electrode response waslinear in the concentration range from 5 to 100�g ml−1 CO3

2− with a correlation coefficient (R2) of 0.998. Precision (R.S.D.) was 1.42% for20�g ml−1 standard solution of CO32− (n = 10). The detection limit was 0.20�g ml−1 CO3

2−. The method was evaluated by the injectionof real natural water samples and an average recovery of 100.1% was obtained. The sampling rate was 30 samples h−1. The method is simple,feasible with satisfactory accuracy and precision and thus could be used for monitoring TIC in water.© 2003 Elsevier B.V. All rights reserved.

Keywords: Flow injection; Gas diffusion; Tungsten oxide electrode; Carbonate; Water

1. Introduction

Carbon dioxide determination is an important parame-ter particularly in monitoring the concentrations of totalinorganic carbon (TIC) in natural, waste and seawaters,beverages, physiological fluids and soil. Carbonate playsan important role in regulating the chemical compositionof natural waters[1]. Different approaches have been usedfor the determination of TIC in waters including the titri-metric method[2], Fourier transform infrared (FTIR)[3],gas sensing probes[4–6] and gas diffusion–flow injectionmethods[7–19]. Flow injection analysis (FIA) based on gasdiffusion separation, has proven useful in complex matricessuch as sea and natural waters, blood and beverages. Inthis technique, the sample was injected into a flowing acidstream in which the carbon dioxide formed diffuses acrossa hydrophobic gas permeable membrane into a receivingstream. The trapped carbon dioxide is then passed into adetector where it is quantified by spectrophotometry[7–13],

∗ Corresponding author. Tel.:+216-71-703729; fax:+216-71-704329.E-mail address: lotfi.monser@insat.run.tn (L. Monser).

FTIR [14], conductometry[15–18] or potentiometry[19].The electrochemical detection in combination with flow sys-tems has been widely used due to its low detection limit andhigh selectivity. Recently, tungsten oxide electrode has beensuccessfully used for the potentiometric detection of car-boxylic acids[20] and trimethylamine[21] in flow injectionsystems. The potential changes observed when acidic so-lutes was injected into the system is based on the followingreaction:

WO3 + 6H+ + 6e− → W + 3H2O

While the electrode response is governed by the differencebetween the carrier pH and the pH of solute injected, there-fore, this electrode could be used for the determination ofTIC as carbon dioxide. To our knowledge, there is no pub-lication that combine FIA with tungsten oxide electrode forthe determination of TIC.

This work describes the development of a novel FIAmethod, based on the combination of tungsten oxide elec-trode with gas diffusion–flow injection technique for thedetermination of TIC in water samples. The performanceof a tungsten oxide wire electrode for the determination of

0039-9140/$ – see front matter © 2003 Elsevier B.V. All rights reserved.doi:10.1016/j.talanta.2003.08.008

390 L. Monser et al. / Talanta 62 (2004) 389–394

carbonate was evaluated under different physical and chem-ical flow injection parameters.

2. Experimental

2.1. Reagents and standards

All chemicals used were of analytical reagent gradepurchased from Fisher Scientific (Loughborough, UK).Tungsten wire 99.9% (0.4 mm) was obtained from AdventResearch Materials (Oxon, England). Sodium acetate solu-tion (0.1 M) and hydrochloric acid (1 M) were prepared inultrapure water (resistivity >18 M�). Carbonate stock so-lution (1000�g ml−1 CO3

2−) was prepared by dissolving1.767 g sodium carbonate in 1 l of ultrapure water. Standardcarbonate solutions (2–100�g ml−1) were prepared by ap-propriate dilution of the stock solution. Solutions pH wasadjusted with 10−2 M orthophosphoric acid and 10−2 Msodium hydroxide and measured with a Mettler-ToledopH-meter (MP 225).

2.2. Apparatus

A schematic diagram of the flow injection system em-ployed is shown inFig. 1. The manifold was equipped witha peristaltic pump (Gilson Minipuls, Anachem, Luton, Bed-fordshire, UK) which propelled the reagents at variable flowrates. A low pressure injection valve (Omnifit, Anachem)with a variable sample loop volume and a gas diffusioncell were also used. The gas diffusion cell consisted of twoidentical methacrylate blocks each with S-shape grooves(channel dimensions: 240 mm×1.5 mm×0.2 mm) betweenwhich a microporous PTFE membrane of 75�m thickness,0.1�m pore size and 50% porosity (Nastro, ProfessionalGas System, Italy) was placed. The two blocks were pressedtogether by eighth screws. A silicone rubber tubes (0.8 mmi.d.), PTFE joints and tubing (0.8 mm i.d.) were used forconnections. The online detection was carried using a poten-tiometric flow cell consisting of a tungsten oxide electrodeand Ag/AgCl reference electrode. The potential difference

W

HCl

Acetate buffer

P

V

GDC

E1

E2

Fig. 1. Schematic representation of FIA manifold used for determination of total inorganic carbonate: P, pump; V, injection valve; GDC, gas diffusioncell; E1, tungsten oxide electrode; E2, reference electrode; W, waste.

between the electrodes was monitored using a Chesel BD4040 chart recorder (Kipp and Zonen, The Netherlands).

2.3. Preparation of tungsten oxide electrode

The tungsten oxide electrode was prepared as previouslyindicated[21], by polishing a tungsten wire (0.4 mm) with agrit paper and then washed with acetone. The wire was thenheated in an oven at 500◦C for at least 1 h. After heating, thecolour of tungsten wire became yellow-green which indi-cates a successful operation. The electrode was then soakedin sodium hydroxide solution (10−3 M) for 1 day before use.

3. Results and discussion

In the proposed gas diffusion–flow injection system(Fig. 1), the potential difference (WO3 against Ag/AgCl)of the solution flowing through the flow cell is directlyrelated to the composition of the acceptor solution and con-centration of the total CO2 trapped in it. Thus, in FIA, theresponse of the proposed electrode to CO2 concentration isinfluenced by a number of parameters. These parameterswere studied using a univariate method.

3.1. Optimisation of the reagent concentrations

3.1.1. Composition of the donor streamIn order to optimise the reagent concentrations, the as-

sembly shown inFig. 1was used with HCl as carrier streamand sodium acetate (10−5 M) as acceptor stream. Initially,an overall flow rate of 1.0 ml min−1and a sample volume of200�l were used.

Standard solutions were injected into a donor stream ofHCl, where dissolved carbonate was converted into CO2,which can penetrate through the PTFE membrane. Becausethe conversion efficiency is dependent upon pH, the FIAsignal was observed at a HCl concentration ranged from 0.01to 0.6 M. As shown inFig. 2, the electrode potential increaseswith the increase in the concentration of HCl until it reachesa maximum at a concentration of 0.1 M, then it starts to

L. Monser et al. / Talanta 62 (2004) 389–394 391

0

20

40

60

80

100

0 0.2 0.4 0.6 0.8

HCl concentration (M)

∆E (

mV

)

Fig. 2. Influence of HCl concentration in the donor stream on the electroderesponse.

decrease slowly as the concentration of HCl increases. Insubsequent experiments, 0.1 M HCl was used as the reagentstream to ensure the maximum conversion of carbonate intoCO2.

3.1.2. Composition of the acceptor streamThe acceptor stream composition was studied by varying

its concentration and pH. Initially, a buffer solution contain-ing sodium acetate (10−5 M) was selected in an attempt toinvestigate the effect of the pH on the detector response. pHvalues ranged from 7.5 to 10.5 were studied by injectinga standard carbonate solution (5�g ml−1). Results obtained(Fig. 3) showed that the electrode potential increased rapidlyby increasing the pH of the acceptor stream until it reachesa maximum at pH 10. Therefore, the pH of the acceptorstream was adjusted to 10 for further experiments.

Then, the influence of the buffer concentration on theelectrode response was investigated by varying the concen-tration of sodium acetate from 10−2 to 0.5 × 10−5 M andmaintaining the pH at 10. As shown inFig. 4, a decrease

0

20

40

60

80

100

120

0 100 200 300 400 500 600 700 800 900 1000 1100[Acetate] / moll-1 x 10-5

∆E(m

V)

Fig. 4. Influence of buffer concentration in the acceptor stream on the electrode response.

∆E (

mV

)

0

20

40

60

80

100

120

6 7 8 9 10 11 12pH (acceptor stream)

Fig. 3. Influence of buffer pH in the acceptor stream on the electroderesponse.

in the detector response was observed as the concentrationof sodium acetate increases. Similar observations were re-ported in previous work[21] for the detection of trimethy-lamine using a tungsten oxide electrode. This was attributedto the increase in the buffer capacity and reduction of themagnitude of the change in [H+], which is responsible onthe response of tungsten oxide electrode. For this reason, aconcentration of 10−5 M was chosen as optimum for car-bonate detection.

3.2. Optimisation of the FI variable

The FI variables, such as injection volume and flow rateswere studied under the above mentioned optimum chemicalvariables.

3.2.1. Injection volumeThe effect of the sample injection volume on the detector

response was studied in the range from 100 to 550�l (Fig. 5).

392 L. Monser et al. / Talanta 62 (2004) 389–394

20

60

100

140

0 200 400 600Injection volume (µl)

∆E(m

V)

Fig. 5. Effect of sample injection volume on the electrode signals.

It can be seen that the detector response increases with theincrease in the injection volume where it reaches a plateauat a sample volume of 200�l, where the increase in thesample volume did not induce any further increase in thesignal. Therefore, a sample injection volume of 200�l wasselected for further experiments.

3.2.2. Flow ratesThe influence of the overall flow rates (acceptor and donor

streams) on the detector response was investigated in therange from 0.4 to 1.2 ml min−1. Fig. 6shows that the detectorsignal decreases as the overall flow rate increase. This isdue to two reasons; the first was the diffusion kinetic ofCO2 through the PTFE membrane and the second was theoptimum equilibrium potential in which at high flow ratescauses the sample zone to flow past the electrode too quicklybefore it reaches its equilibrium potential. An overall flowrate of 0.7 ml min−1 was selected for further experiments asit gave the best compromise between the sensitivity and thedetector response as at lower flow rates the analysis timewill increase.

0

20

40

0 0,4 0,8 1,2 1,6

Flow rate (ml/min)

∆E (

mV

)

Fig. 6. Effect of the overall flow rate on the electrode signals.

3.3. Analytical applications

The optimum FI parameters (flow rate: 0.7 ml min−1,injection volume: 200�l, [HCl]: 0.1 M, [acetate]: 10−5 M,and pH 10) were applied for preparing the calibration graph.Results obtained show that tungsten oxide electrode gavesensitive response to changes in carbonate concentrations(Fig. 7). The potential change (peak height) versus thelog[CO3

2−] was found to be linear over the wide dynamicrange from 2 to 100�g ml−1 with a regression coefficient of0.998. The equation of the line (�E = 54.44 log[CO3

2−] −28.39) was used for the determination of carbonate in watersamples. The method provides satisfactory precision for theanalysis of carbonate (20�g ml−1) with a relative standarddeviation (R.S.D.) of 1.42% (n = 10). The limit of detec-tion corresponding to a signal-to-noise ratio of three was0.20�g ml−1 carbonate and the sampling rate was about30 samples h−1.

The method accuracy was evaluated from the recoveryassay, by spiking water samples with three addition levels (5,10 and 15�g ml−1) of standard carbonate solutions. Goodrecoveries were obtained (Table 1) indicating that TIC couldbe analysed in water with good accuracy.

Table 1Recoveries obtained from various water samples

Sample CO32− (�g ml−1) Recovery (%)

Added Found

Safia mineral water 5 4.90± 0.08 9810 10.1± 0.1 10115 15.0± 0.1 100

Seawater 5 5.10± 0.08 10210 10.0± 0.2 10115 14.8± 0.2 98.7

L. Monser et al. / Talanta 62 (2004) 389–394 393

1

2

3

4

5

6

mV

0

20

40

60

80

100

0 1 2

log([CO32-])

∆E/m

V

Fig. 7. Typical potentiometric response of the tungsten oxide electrode to carbonates using the optimum operating conditions: (1) 5�g ml−1; (2)10�g ml−1; (3) 20�g ml−1; (4) 30�g ml−1; (5) 40�g ml−1; (6) 50�g ml−1.

3.3.1. Interference studyCompounds frequently encountered in natural and sea-

waters that can diffuse through the membrane and hencecan interfere with analytical signal of CO2 were mainly sul-phides. These compounds were present in smaller amountwhen compared to the high concentration of carbonate andtherefore, the analytical signals were slightly affected bythe presence of these potential interfering species. H2S wasconsidered to interfere when the relative error reaches valueof approximately 5%. The errors were below 5% when H2Swas added to a level of 10% of carbonate concentration. Ifthe relative error exceeds 5%, therefore, this type of inter-ference could be eliminated online or offline by the additionof lead nitrate to the donor stream or to the sample.

3.3.2. Analysis of real samplesThe described method was applied to the determination

of TIC in water samples, such as natural, sea and tap water.

Table 2Determination of TIC in water samples using the proposed and the officialmethods

Sample CO32− (�g ml−1) Recovery (%)

Proposed method Official method

Safia mineral water 202± 2 206± 2 98Marwa mineral water 216± 1 220± 2 98.2Cristal mineral water 163± 2 161± 1 101Seawater 455± 2 445± 3 102

Results obtained (Table 2) were compared with an officialmethod[22] in which TIC was determined after acidifica-tion of the sample (pH<2.0), the evolved CO2 was thenmonitored online using an infrared spectrophotometer. Thismethod[22] was applicable for the determination of TIC,total carbon and total organic carbon in water. Values ob-tained with the proposed method are comparable to thoseobtained by the official method, where they prove the ap-plicability of this technique for the determination of TIC indifferent matrixes.

4. Conclusion

The use of tungsten oxide electrode combined to gasdiffusion–flow injection system showed high sensitivitytowards carbonate in aqueous solutions. This system hasbeen optimised and it was successfully applied to thedetermination of TIC in natural and seawater with a sam-pling rate of 30 samples h−1. The method developed underthis approach is sensitive, rapid and versatile and thuscould be used as an alternative to the already existingmethods.

Acknowledgements

The authors are thankful to the British Council LINKProgramme and the University of Hull for the support ofthis work.

394 L. Monser et al. / Talanta 62 (2004) 389–394

References

[1] J. Bisogni, J.R. Arroyo, Water Res. 25 (1991) 185.[2] R. Moosbrugger, Water 19 (1993) 11.[3] A. Perez-Ponce, S. Garrigues, M. de la Guardia, Vib. Spectrosc. 19

(1998) 61–67.[4] R. Faschung, F. Keplinger, G. Hanreich, G. Jobst, G. Urban, F. Kohl,

R. Chabicovsky, Sens. Actuators B 78 (2001) 291–297.[5] J. Bubendorfer, R. Kunz, G. Jobst, I. Moser, G. Urban, Sens. Actu-

ators B 50 (1998) 210–219.[6] J. Shim, D. Sakong, H. Nam, G. Cha, Anal. Chem. 68 (1996) 221.[7] G. Atanassov, R. Lima, R. Rangel, I.V. Toth, Analysis 28 (2000)

77–82.[8] E. Ljunggren, B. Karlberg, J. Autom. Chem. 17 (1995) 105.[9] W. Van Der Linden, Anal. Chim. Acta 151 (1983) 359.

[10] S. Motomizu, T. Kyoji, T. Kuwaki, M. Oshima, Anal. Chem. 59(1987) 2930–2932.

[11] X. Su, L. Nie, S. Yao, Anal. Chim. Acta 349 (1997) 143–151.

[12] V. Kuban, P. Dasgupta, Talanta 40 (1993) 831.[13] M. Oshima, Y. Wei, M. Yamamoto, H. Tanaka, T. Takayangi, S.

Motomizu, Anal. Sci. 17 (2001) 1285–1290.[14] A. Perez-Ponce, S. Garrigues, M. de la Guardia, Anal. Chim. Acta

358 (1998) 235–243.[15] J. Rohwedder, C. Pasquini, Analyst 116 (1991) 841.[16] T. Aoki, Y. Fujimaru, Y. Oka, K. Fujie, Anal. Chim. Acta 284 (1993)

167.[17] P. Hall, R. Aller, Limnol. Oceanogr. 37 (1992) 1113–1119.[18] L. Ma, M. Oshima, T. Takayanagi, S. Motomizu, J. Flow Injection

Anal. 17 (2000) 188.[19] P. Linars, M. Luque de Castro, M. Valcarcel, Anal. Chim. Acta 225

(1989) 443.[20] Z. Chen, P. Alexander, Anal. Chim. Acta 332 (1996) 187–192.[21] N. Adhoum, L. Monser, S. Sadok, A. El-Abed, G. Greenway, R.

Uglow, Anal. Chim. Acta 478 (2003) 53–58.[22] European Standard, EN 1484, La Qualité de l’eau, AFNOR, 1997,

p. 191.