Abstract. Non-suppressed ion chromatography (NSIC)has been optimized to permit the determination of chlo-ride, nitrate and sulfate in the low mg/l concentrationrange within 3 min. Using conductometric detection theextraneous (positive) injection peak was found to ad-versely affect the early eluting chloride signal when sam-ples with high amounts of total cations are analyzed. Theserial arrangement of a downstream potentiometric detec-tor with a chloride ion selective electrode, which does notrespond to other alterations of the composition of the elu-ent than the change in chloride concentrations, is shown tobe suitable for interference-free, sensitive and reliablechloride evaluation. Application of the proposed systemto the determination of chloride in extracts of filter col-lected airborne particulates revealed (for those samplesthat could be evaluated by both detection methods) highprecision and no systematic deviations between conducto-metric and potentiometric detection.

Introduction

Since its conception in the mid-seventies [1] ion-chro-matograpy (IC) has received broad acceptance in variousfields of chemical analysis and is now established as aleading technique for the determination of many inor-ganic and organic ions in aqueous solution [2, 3]. The un-rivalled feature of IC is that automated simultaneous de-termination of several ions is feasable with high selectiv-ity and sensitivity in a reasonably short analysis time. Thealmost universal response of conductometric detection toionic species can be advantageously utilized in both,chemically suppressed and non suppressed IC and is mostwidely applied. Other detection principles like direct andindirect UV-absorption measurement, amperometry, po-tentiometry with ion selective electrodes (ISE) and the use

of diverse spectrometric methods following post-columnderivatization are, however, a matter of continuous con-cern since the one or the other of these methods offershigher sensitivity or better selectivity [2–7]. Moreover,weakly ionized species (e.g. cyanide, thiocyanide, iodide,sulfide) at the pH typically used in IC of anions whichevade sensitive conductometric detection become accessi-ble [2, 3, 8–10]. In order to avoid the preselection of a particular detector, the use of different detectors connectedin series or in parallel has been proposed [1, 2, 11–16].The major aims of such approaches are to overcome prob-lems associated with low-resolved pairs of signals or toextend the number of ions that can be determined in a sin-gle run. Typical examples of applications are given in thepaper of Talasek [15] who used a fluoride ISE in combi-nation with conductometric detection to avoid the inter-ference of carboxy anions eluting along with fluoride, andthe work of Slanina et al. [11] who applied a multiple de-tector with serial arrangement of conductometric, directUV-absorption and ISE cell (for detection of fluoride andbromide) to investigate mutual interferences of pairs ofions insufficiently resolved in their chromatographic sys-tem.

Sometimes the use of short separation columns for fastNSIC [17–19] is an economically (in time and money) in-teresting alternative to commonly employed costly high-performance IC columns. Though the latter, particularlyin combination with gradient elution offer outstandingseparation efficiency they appear overdeveloped for prac-tical applications where only a few ions have to be deter-mined.

Recently we were confronted with the need of ana-lyzing thousands of filter-collected air samples, chloride,nitrate and sulfate being the only anionic species of inter-est [20]. The low budget of the project and the require-ment of fast data presentation led to the development of a suitable NSIC-method with conductometric detection[21]. During elaboration of the procedure we were repeat-edly faced with the problem that the early eluting chloridesignal was adversely affected by the extraneous injectionpeak. A solution of this problem is reported using the ser-

Fresenius J Anal Chem (1995) 353 :123–127

Dedicated to Professor Dr. Peter Brätter on the occasion of his60th birthdayCorrespondence to: W. Frenzel

Serial potentiometric and conductometric detection in fast non-suppressed ion chromatography applied to the analysis of filter collected airborne particulates

Wolfgang Frenzel, Annette Rauterberg-Wulff, Dietrich Schepers

Institut für Technischen Umweltschutz, Fachgebiet Luftreinhaltung, Technische Universität Berlin, Strasse des 17 Juni 135, D-10623 Berlin, Germany

Received: 20 March 1995 / Revised: 19 June 1995 / Accepted: 20 June 1993

© Springer-Verlag 1995

Fresenius’ Journal of

ial arrangement of a common conductometric detectorand a potentiometric detector cell with a chloride ISE. Tothe best of our knowledge this combination has not yetbeen employed for similar purposes. The experimentalconfiguration of the system and its performance charac-teristics are outlined and results obtained for chloride infilter extracts of airborne particulates using the two detec-tion modes presented.

Experimental

Reagents and solutions. All reagents used were of analytical gradequality. Solutions were made up with bidistilled water from aquartz still. The chromatographic eluent used was prepared of 2mmol/l potassium hydrogen phthalate and 7% ethanol, the pH pre-cisely adjusted to 6.8 ± 0.1 by dropwise addition of NaOH. Prior touse the eluent was thoroughly degassed ultrasonically under re-duced pressure. Ion standards were prepared from Titrisol stocksolutions (Merck).

The ion strength adjustment buffer (ISAB) composed 0.2 mol/lpotassium nitrate acidified to pH 2 with nitric acid. A 0.2 mol/lpotassium nitrate solution spiked with 10 mg/l chloride was usedas reference solution.

Apparatus. The NSIC system (see Fig. 1) was built from individualcomponents and comprises a HPLC pump (Typ 64.00, Knauer,Berlin), a pneumatically driven six-port valve (Typ A 0258, Knau-er), an in-line protection filter (Upchurch), a PRP X-100 anion sep-aration column (100 × 4.1 mm, Hamilton, Bonaduz), a Waters con-ductivity detector (Model 430, Millipore, Milford) and a purpose-made potentiometric flow-through cell with a Cl-ISE (see below).PEEK capillaries of 0.25 mm i.d. were used for interconnections.The distances between the individual components were kept asshort as practically possible. If not stated otherwise the flow ratewas set to 4 ml/min.

The configuration of the potentiometric detector cell is de-picted in Fig.2. It is made of polyvinylidene difluoride (PVDF)and is constructed like a common y-connector employed in flowanalysis systems. Two identical metallic silver tubes (20 mmlength, 0.5 mm i.d., 2 mm o.d., Goodfellow, Cambridge) are press-fitted into threaded PVC male nuts. When mounting of the twoscrews the silver-electrodes are fixed in close proximity to eachother (distance between the two nearby ends < 2 mm). To producechloride selective electrodes the silver tubes were internally coatedwith a silver chloride layer by continuous pumping of 0.1 mol/lFeCl3, pH 2 through the tubes for about 10 min [22]. After rinsingwith water the electrodes are ready for use and are installed asdownstream detector in the IC-system. As shown in Fig.1 the elu-ent stream is mixed with the ISAB solution and directed throughone of the tubular electrodes, the other one (serving as the refer-ence electrode of the differential potentiometric cell) being contin-uously provided with a stream of 0.2 mol/l potassium nitrate solu-

tion containing 10 mg/l chloride for improved potential stability.The common outlet is directed to waste. A peristaltic pump (TypIPS-8, Ismatec, Zurich) is used to propel the ISAB and the refer-ence solutions at a constant flow rate of 2 ml/min each.

Potential measurements are made with an Orion ion-meter,Model 901. The conductometric and potentiometric detectors areinterfaced to a dual channel strip-chart recorder (Typ L 6512, Lin-seis, Selb) for simultaneous registration of the output. Signals areevaluated manually by peak-height measurements because of lackof suitable integrator.

Sampling of airborne particulates and sample preparation. The airsampling and sample preparation procedures were similar to thatdescribed previously [20]. Ambient air samples are taken with au-tomated dichotomous samplers (Ser. 245, Anderson, Atlanta) orhigh-volume samplers (Typ DHA-80, Digitel). PTFE membranefilters on polyester support (Gelman) are used in connection withthe dichotomous sampler whereas the high-volume samplers areequipped with quartz fiber filters (QAO 2500, Pall). After sam-pling the weighed filters are transferred into 50 ml polyethylenebottles, are wetted with 200 µl ethanol and covered with 2.8 ml ofwater. The extraction of the anionic species is accomplished by ul-trasonic treatment for 15 min. About 100 µl of the aqueous extractsare drawn off with a micropipette (care being taken to avoid aspi-ration of particulates) and are filled into the injection loop of theNSIC-system.

Results and discussion

Optimization of fast nonsuppressed ion chromatography(NSIC)

Fast NSIC has recently been reported [17–19] for simpleseparation problems using either separation columns ofreduced length (even guard columns have been employed)and/or eluents of higher ionic strength. Own investiga-tions were made using a 10 cm Hamilton PRP X-100 col-umn and phthalate as eluent. The ultimate aims were toachieve a sufficient separation of chloride, nitrate and sul-fate in as short as possible analysis time. The main vari-ables affecting resolution and residence time are the flowrate and the composition of the eluent with respect to thephthalate concentration and pH. In initial experiments the

124

Fig. 1. Schematic diagram of the IC system comprising seriallyarranged conductivity and potentiometric detectors. For details ofthe potentiometric detector see Fig.2

PRP X-100

Peristalticpump

Potentiometricdetector

ISAB

Ref. solution

Eluent

HPLC Sample SeparationConductivitypump injection column detector

waste

Fig.2. Design of the differential potentiometric flow cell with twoidentical tubular chloride selective electrodes. 1, sensing tubularchloride ion selective electrode; 2, tubular reference electrode; 3,connections to ion-meter; 4, stainless-steel grounding electrodes;5, polyvinylidenedifluoride body

occurance of two extraneous signals viz. the injectionpeak and a so-called system peak [2] were found to be themajor limitation finding appropriate conditions. In orderto avoid the system peak the eluent pH was raised to 6.8where phthalate is totally ionized [Ref. 2, p. 123]. At thatpH the capacity factors for all anions are lower leading toreduced residence times but concommitantly the chloridesignal overlapped with the (negative) injection peak. Fur-ther optimization of the phthalate concentration and theeluent flow rate finally resulted in a reasonable compro-mise between resolution, absence of disturbances fromextraneous peaks and analysis time as exemplified by thechromatogram depicted in Fig.3A. With an injection vol-ume of 20 µl chloride, nitrate and sulfate can be deter-mined in the concentration range 0.5–20 mg/l which wasregarded appropriate for the intended application. Theanalysis of extract solutions containing 7% ethanol gaverise to a large spurious signal preventing useful evaluationof anion signals. This problem can be circumvented by

precisely matching the chromatographic eluent and theextractant with respect to their ethanol content. It is note-worthy that this measure was feasable only with theHamilton PRP X-100 since all other columns tested [21]exhibited a dramatically increased backpressure whenethanol was used as eluent modifier. Test runs on realsamples revealed yet another problem. Owing to the rela-tively high amount of total cations present the injectionpeak became positive, adversely affecting the evaluationof the chloride signal (see Fig. 3B). The simultaneous po-tentiometric detection with a chloride selective electrode[6] was thought to solve this problem without changingthe optimized chromatographic conditions and was henceinvestigated.

Performance of the potentiometric flow-through chloride detector

The use of a chloride ISE in flow-through measurements(and in particular in flow injection analysis (FIA)) is asubject of numerous papers published [e.g. 22–24]. Manydifferent cell configurations have been developed andtheir respective performance behaviours have been stud-ied. Tubular silver chloride electrodes of the second kindhave been demonstrated not only to be a particularly sim-ple version of a flow-through detector but also to offer ex-cellent features in terms of detectability, precision, dy-namic range, response time and long-term stability [22,24, 25]. The experimental conditions of typical FIA mea-surements have much in common with IC detection in thatin both cases a discrete sample is injected into a continu-ously flowing solution which during transportation under-goes dispersion (generally the dispersing element in FIpotentiometry is a piece of coiled tubing whereas in IC itis the separation column) and eventually reaches the de-tector as a transient concentration zone. When addition-ally the carrier solution in FIA is chosen identical to thechromatographic eluent the only differences left are thedegree of dispersion and the fact that in IC chloride ionsare separated from other anions present in the sample. Theion-exchange separation causes that chloride – irrespectiveof the initial composition of the sample – is always ac-companied by the cation used for preparation of the eluent(i.e. potassium in the present case) during the passagethrough the detector. Therefore, an initial examination ofthe potentiometric chloride detector was carried out by FI-measurements using a dual line set-up with the phthalateeluent serving as the carrier stream and potassium nitratesolution as reference stream. The dimensions of the con-nection tubing between injection valve and potentiometricflow-cell were chosen to match those of the IC systemwithout column. Chloride standards of variable concentra-tion were prepared in phthalate solution precisely match-ing the composition of the chromatographic eluent. In ini-tial experiments the low conductivity of the eluent wasfound to create severe problems with electrostatic noiseand streaming potentials. In order to minimize this threemeasures were undertaken, i.e. (i) the distance betweenthe sensing and the reference electrode was reduced tominimum to keep the impedance of the circuit low, (ii)grounding electrodes were inserted in front of the two sil-

125

Fig.3. A Typical chromatogram obtained after injection of an ionstandard containing 10 mg/l chloride, 10 mg/l nitrate and 20 mg/lsulfate using optimized fast non-suppressed ion chromatography(see Experimental). B Chromatogram of a real filter extract dem-onstrating the strong interference of the extraneous positive cationpeak on the chloride signal with conductivity detection (lower line)and absence of interference when potentiometric detection withchloride detection (upper reversed line) is employed

A

B

ver chloride electrodes and the entire flow-cell wasshielded with aluminium foil. A further improvement ofthe system which was only discovered at a later stage inroutine use was the introduction of an additional flowchannel carrying an ISAB buffer (see Fig. 1) between theoutlet of the conductivity detector and the inlet to the po-tentiomeric flow cell. Modified in this way the baselinenoise (peak-to-peak) was smaller < 50 µV permitting po-tential changes of > 0.2 mV to be detected with a suffi-cient reliability. This value corresponds to a chloride con-centration well below 50 µg/l chloride evidencing excel-lent detectability of the potentiometric detector. The po-tentiometric chloride detector response occurs almost in-stantaneously as evidenced by a comparison of the peak-width at the baseline between the potentiometric and con-ductometric response curves (see Fig.3B). The results ofcalibration performed in the concentration range 0.5–20mg/l chloride using FIA and IC with 20 µl injection vol-ume each are depicted in Fig. 4. For reasons of compari-son the electrode response for continuously supplied cali-bration standards (steady-state signals) is also shown. Thethree curves represent an expected course in that they allexhibit Nernstian response at higher chloride concentra-tion levels and the lower limit of linearity is shifted to-wards higher concentrations as sample dispersion in-creases [26]. Quantitative sample dispersion at the peakmaximum (defined according to Ruzicka and Hansen[27]) amounts to about 3 and 9 in FIA and IC experi-ments, respectively. The precision of measurements was

evaluated by repetitive injections of 20 µl samples intothe FIA and IC systems. The relative standard deviationsin the concentration range 0.5–20 mg/l chloride are typi-cally below 2% in both cases.

Application to the analysis of filter collected airborne particulates

In view of the enormous number of samples that had to beanalyzed in a recent research project [28] sampling fre-quency was a very important factor. In a published workon the application of IC to the analysis of airborne partic-ulates the residence time of sulfate (being generally theanalysis time limiting compound) is often about 10 minand does not fall below 6 min [2, 3, 29–32]. Under our ex-perimental conditions (vide supra) the baseline after sul-fate elution is restored within 3 min permitting a samplingfrequency of 20 per hour to be attained in routine.

A problem repeatedly encountered in the analysis ofaqueous extracts of filter collected particulates using thefast NSIC method with conductivity detection was the in-sufficient resolution between the chloride signal and thepreceding positive cation signal as exemplarily shown inFig.3B. Further examination [21] revealed that the degreeof disturbance was not only dependent on the relative sig-nal heights of the cation and the chloride but also on thekind of cations present in the sample. The presence of di-valent cations (e.g. calcium and magnesium) generally led to a considerably enhanced tailing of the cation peak(an observation also made in previous work [19]) makingproper evaluation of the chloride signal difficult. The si-multaneously performed potentiometric detection withCl-ISE (also shown in Fig.3B) is free of any interferenceof system peaks and hence permits proper evaluation ofthe chloride signal. An alternative approach of conqueringthe cation interference, i.e. sample preparation by the use

126

Fig. 4. Calibration curves obtained using potentiometric detectionwith chloride selective electrode for (g) continuous sample sup-ply, (m) injection of 20 µl sample into an FIA-system, and (p) in-jection of 20 µl into the IC-system, details see text

Fig.5. Comparison of results of the chloride content of filter col-lected airborne particulates using evaluation of the conductometricand potentiometric signals. The linear plot is represented by theequation y = 0.977 x – 0.05, r = 0.995 (n = 38)

of strong cation exchange, was rejected for economic rea-sons and the additional analysis time required.

In the course of application of the proposed methodwith serially arranged detectors in routine it came out thatsome samples could not be evaluated for chloride at all byconductometric detection owing to their low chloride con-tent and the strong interference of the tailing cation signal.In several instances the conductometric response of chlo-ride was severely affected which detracted from reliablemeasurements. In the majority of cases both detectionmethods could be used but the chloride evaluation fromthe conductometric signal was always more difficult toaccomplish than quantitation of the potentiometric signal.The results of analyses of 38 filter extracts collected withdichotomous samplers are condensed in Fig. 5. The goodcorrelation (r = 0.995), a slope value of about 1 and nosignificant intercept demonstrate the reasonable precisionof measurements and the absence of any systematic devi-ations between the two detectors.

Conclusions

The serial arrangement of conductometric and potentio-metric detectors in fast NSIC presented here is capable toovercome limitations of insufficient chromatographic res-olution between the injection (cation) signal and the earlyeluting chloride peak. Due to the high selectivity of thepotentiometric chloride detector it does not respond toother alterations of the eluent composition than thechange of the chloride concentration. The accessibleworking range is similar to that of the conductivity detec-tor so that they supplement each other in an ideal way.

It is certainly a debatable point whether the benefits offast analysis with cheap separation columns are out-weighted by the necessity of using an additional detectorand also whether it is worth to do so when high-perfor-mance columns can be applied for a similar purpose with-out any of the problems mentioned. Irrespective of the an-swer given it can be stated that the present system consti-tutes a viable means of reliable anion determination beingpotentially capable to detect also other halides andpseudohalides [2, 6, 33]. The low cost of the extension ofthe NSIC system with an additional (readily self-made)potentiometric detector is probably attractive when IC isapplied only occasionally and for institutions where bud-gets are low. This also holds for other combinations of de-tectors (e.g. amperometry and conductometry or electro-chemical and spectrophotometric detectors) which offerrelatively simple means of selectivity improvement.

References

1. Small H, Stevens TS, Baumann WC (1975) Anal Chem 47 :1801

2. Haddad PR, Jackson PE (1990) Ion chromatography – Princi-ples and applications. Elsevier, Amsterdam

3. Weiß J (1991) Ionenchromatographie, 2nd edn. Verlag Chemie,Weinheim

4. Heckenberg AL, Haddad PR (1984) J Chromatogr 299 :3015. Tarter JG (1984) J Liq Chromatogr 7 :15596. Lockridge JE, Fortier NE, Schmuckler G, Fritz JS (1987) Anal

Chim Acta 192 :417. Han SH, Lee KS, Cha GS, Liu D, Trojanowicz M (1993)

J Chromatogr 648 :2838. Rocklin RD, Johnson EL (1983) Anal Chem 55 :49. Han K, Koch WF (1987) Anal Chem 59 :1016

10.Butler ECV, Gershey RM (1984) Anal Chim Acta 164 :15311.Slanina J, Bakker FP, Jongejan PAC, Van Lamoen L, Möls JJ

(1981) Anal Chim Acta 130 :112.Wheals BB (1983) J Chromatogr 262 :6113.Williams RJ (1983) Anal Chem 55 :85114. Jackson PE, Haddad PR (1988) J Chromatogr 439 :415.Talasek RT (1989) J Chromatogr 465 :116.Slais K (1991) J Chromatogr 540 :4117. Jupille T, Burge D, Togami D (1982) Chromatographia 16 :31218.Behrend P, Kipplinger A, Behnert J (1985) Labor Praxis 9 :32419.Frenzel W, Rauterberg A (1992) Mikrochim Acta 106 :17520.Frenzel W (1991) Fresenius J Anal Chem 340 :52521.Rauterberg A (1989) Diploma Thesis, Technical University of

Berlin 22.Frenzel W (1989) Fresenius Z Anal Chem 335 :93123.Trojanowicz M, Matuszewski W (1983) Anal Chim Acta 151 :

7724.Van Staden J (1986) Anal Chim Acta 179 :40725.Frenzel W (1994) Poster Presentation, Flow Analysis Confer-

ence VI, Toledo, Spain26.Trojanowicz M, Matuszewski W (1982) Anal Chim Acta 183 :

7127.Ruzicka J, Hansen EH (1988) Flow injection analysis, 2nd edn.

Wiley, New York28. Israel GW, Erdmann A, Shen J, Frenzel W, Ulrich E (1992)

Analyse der Herkunft und Zusammensetzung der Schweb-staubimmission. Fortschrittberichte VDI, Reihe 15: Umwelt-technik, Nr. 92. VDI Verlag, Düsseldorf

29.Mulik JD, Sawicki E (eds) (1979) Ion chromatographic analy-sis of environmental pollutants, Vol. 2. Ann Arbor Science,Ann Arbor

30.Sawicki E, Mulik JD, Wittgenstein E (eds) (1978) Ion chro-matographic analysis of environmental pollutants. Ann ArborScience, Ann Arbor

31.Miller DR Byrd JE, Perona MJ (1986) Water Air Soil Pollut 32 :329

32.Verein Deutscher Ingenieure (1988) Messen partikelgebun-dener Anionen in der Außenluft, VDI 3497, Blatt 3 and 4. VDI-Handbuch Reinhaltung der Luft, Vol 4. VDI Verlag, Düs-seldorf

33.Trojanowicz M, Probozy E, Meyerhoff ME (1989) Anal ChimActa 222 :109

127