High sensitivity optical detection methods in hydroxide eluent suppressed anion chromatography via postsuppression ion exchange

Anal. Chem. 1087, 59, 1963-1969

1963

High Sensitivity Optical Detection Methods in Hydroxide Eluent Suppressed Anion Chromatography via Postsuppression Ion Exchange Hideharu Shintani’ and Purnendu K. Dasgupta* Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409-4260

Detectlon by posteuppreselon Ion exchange, as an adjunct to suppressed hydroxlde eluent Ion chromatography, Involves exchange of elutlng sample anlon or concurrently present Ht for a fluorescenthptkally absorbing anlon or catlon such as anthranllate or Cea+. Membrane-based postsuppresslon devices permlt rdlabk contlrmowr operattan, and detecttan limlts can be better than those attainable by conductlvity detwtlon. The anlon exchange membrane-anthranllate system also ylelds excellent detectlon llmlts by UV abrorptbn except that membrane degradatlon le a problem. The Ion exchange at the membrane Interface Is not an equlllbrlum process. The dependence of the membrane response on an external hlgh frequency electrlcal field suggests that the transport of the Ions through the membrane Is not a rate-determlnlng factor.

Hydroxide eluents, along with new large-capacity ion-exchange suppressors have opened new vistas in suppressed anion chromatography, permitting highly sensitive detection and facile gradient elution (1). Hydroxide eluents are clearly those of choice, as far as detectability is concerned, in nonsuppressed anion chromatography as well (2). In suppressed anion chromatography with a hydroxide eluent, virtually nonconducting water is produced as the background, permitting particularly sensitive conductometric detection. Because the elution of sample ions causes the appearance of protic acids (largely ionized a t low concentrations,for all but extremely weak acids) in a nonionic background, other, potentially more sensitive, detection modes become viable. All eluting ions, for instance, may be exchanged partially or quantitatively after the suppressor by an intensely absorbing, fluorescing, or electroactive ion followed by absorptiometric, fluorometric, or electrochemical detection. There are a number of ways in which such postsuppression ion exchange may be carried out. The sample anion emerges from the suppressor as the corresponding acid; it is possible, in principle, to exchange the proton for some other cation by cation exchange or the anion for some other anion by anion exchange. In practice, however, the extent of this exchange can be too small to be of analytical utility, depending on the selectivity coefficient (3) of the ion exchanger phase for the desired ion vs. the proton or the analyte anion, as appropriate. However, the anion exchange mode represents a significant difference from the cation exchange mode that should be recognized. This is that in the cation exchange mode, one is always attempting to exchange H+, the second most poorly retained cation, for some other cation. The prospect of doing this with high efficiency for any cation other than Li’ (the most poorly retained cation on most exchangers) is not good. Good exchange for Li+ is viable, however, and sensitive del Permanent address: National Institute of Hygienic Sciences, Department of Medical Devices, Tokyo, Japan.

0003-2700/87/0359-1963$01.50/0

tection of Li+ by a suitable method may be an attractive way to conduct “replacement ion Chromatography” as originally reported by Downey and Hieftje (4). On the other hand, in the anion exchange mode it is the sample anion that is being exchanged and the attainable exchange efficiency should go up with the retention time of the analyte ion. Detectability in isocratidy performed chromatography generally decreases with increasing retention time, but a large retention time also signifies a large affinity of the separator exchanger resin for the late eluting ion. If the ion exchange selectivity of the packed-column postsuppressor device parallels that of the separator, the increased exchange efficiencies for analyte anions with increasing retention times should compensate, in terms of detectable concentrations, for the increasing band volume of late eluting ions. There is yet another way in which postsuppression ion exchange can be used to perform useful optical detection. If a strong acid form cation exchanger is treated with a weak acid such as anthranilic acid (0-aminobenzoic acid, HAn), the uptake of this compound by the resin may take place by protonation of the -NH2 group or, likely to a lesser extent, as the uncharged molecule by ion exclusion. When such a column is used as the postsuppressor, a dynamic equilibrium involving various forms of HAn, i.e., the protonated cation, the carboxylate anion, the zwitterion, and the uncharged molecule, is present. Consequently, with the suppressed hydroxide eluent (i.e., essentially pure water) flowing through it, a measurable amount of HAn bleeds off the column, representing a detector background. Because the uptake capacity is relatively large compared to the bleedoff rate, the latter can remain reasonably constant over a long period of time. The existing equilibrium is perturbed whenever the pH of the postsuppressor influent is altered, as when an acidic band due to an eluting sample enters the column. The decreased pH shifts the equilibrium such that either mechanism of retention (cationic retention or ion exclusion) is promoted, the HAn concentration in the effluent decreases, and thus a negative absorbance/fluorescence signal is expected to result due to sample elution. The results are expected to be the same if a cation exchange membrane postsuppressor is used with an external solution of HAn. Presumably, one may perform postsuppression ion exchange using either packed-column or membrane-based postsuppressors. The selectivities of different types of ion exchange resins vary considerably and characterization of ion exchange membranes, with the possible exception of the perfluorosulfonate cation exchanger Nafion, is still in its infancy. We have therefore investigated the utility of postsuppression ion exchange and UV/fluorescence detection by using a wide variety of ion exchange resins and membrane postsuppressors with primary emphasis on the following three modes: (a) cation exchange of the proton for Ce3+,(b) anion exchange of the sample anion for An-, and (c) indirect detection based on the controlled release of protonated HAn from a cation exchange resin-membrane. 0 1987 American Chemical Society

1964

ANALYTICAL CHEMISTRY, VOL. 59, NO. 15, AUGUST 1, 1987

Table 1, Postsuppressor Membrane Ion Exchangers membrane type 1 2

3 4

material Nafion 020 Nafion 811X nAfion 815X polyethylened

id., pm

wall, pm

Cation Exchangers 60 675 75 1000 125 1100 50 360

mean surface area per unit length," cm

EWb

device typeC

0.130 0.236 0.351 0.361

1100 1100 1100 1250

AB,C C D

Anion Exchangers 50 0.333 2800 D PTFE' 1010 PTFE' 1100 50 0.361 2400 D PTFE' 500 300 0.232 1350 A 8 unknownf 500 50 0.172 1400 C 9 PEVAg 300 240 0.152 1700 A 10 microporous polypropyleneh 400 30 0.135 8900 A nBased on log mean diameter of dry tubes, dimensional changes are significant upon wetting. *Equivalent weight, determined by procedures previously described (3). 'A, hollow; B, filament-filled, linear; C, filament filled knotted; D, packed with glass beads, see text. Chemically bonded cation exchange sites produced by proprietary techniques. e Poly(tetrafluoroethylene), radiation grafted with vinylbenzyl chloride and then quaternized, see ref 3. 'Reportedly some form of fluorocarbon, ion exchanger preparation is proprietary. 8Poly(ethylvinyl acetate). hThe base material is Celgard X-20 hollow fibers (manufactured by Celanese Corp., Charlotte, NC); these are then radiation grafted like the PTFE tubes. 5 6 7

EXPERIMENTAL SECTION A Model 4000i ion chromatograph (Dionex Corp., Sunnyvale, CA) was used in all chromatographic experiments along with a Dionex anion micromembrane suppressor (AMMS). The instrument came to our laboratory in a prototype version; aside from modifications made to the instrument as described in a previous paper (I), a short column serving as a pulse dampenerfmixer supplied by the manufacturer was installed between the pump outlet and the injector. Preliminary work was conducted with the conductivity detector connected serially ahead of the optical detector, with a membrane-based or a short packed column ion exchanger in between. In later work, as represented by all chromatograms presented herein, the conductivity detector was omitted to avoid unnecessary dispersion. All work was conducted with an eluent flow rate of 1.0 m l l m i n and a regenerant (12.5 mM H,SO,) pressure of 15 psi. The optical detectors used were a Model SF 757 W-vis detector (KratosISchoeffel, Inc., Ramsey, NJ), a Model RF 430 fluorescence detector (Shimadzu Scientific, Columbia, MD), or a Model FS 970 fluorescence detector (Kratos f Schoeffel). For cerium(III), anthranilate, fluorescein and fluoresceinsulfonate, magdala red, rhodamine 6G, and salicylate, the optimum respective excitationlemission wavelengths as determined by a Model LS-5 spectrofluorometer (Perkin-Elmer Corp., Norwalk, CT) were 256f 350,327f 420,278/513,540/570, 5251550, and 3271377 nm, respectively. These wavelength combinations were used with the Shimadzu instrument which utilizes monochromators on both excitation and emission sides. The Kratos instrument utilizes a high pass filter on the emission side. Experiments were conducted with Ce3+and anthranilate (An-) with this detector with the excitation wavelengths as stated above and with emission filters with respective 50% cutoff wavelengths being 320 and 389 nm. The absorptiometric detector (Kratos 757) normally utilizes a heat exchanger a t the flow cell exit which generates significant back pressure. The back pressure due to this and the postsuppressor ion exchanger was sufficiently large to exceed the pressure rating of the AMMS, causing it to malfunction. The heat exchanger was therefore removed with an unavoidable increase in detector noise level. Reagents. Water used in this work was distilled and then passed through activated carbon and two sequential deionizing tanks. It met or exceeded the specification of ASTM type I reference reagent water. Anthranilic acid (Eastman Kodak) was recrystallized from hot water. Lithium anthranilate or salicylate a t a desired pH was prepared by adding the necessary amount of LiOH solution to a solution of the free acid of known concentration and diluting to volume. Cerous sulfate (Alfa Products) had a stated purity of 99.9%. Fluorescein (sodium salt, Uranine, Aldrich), rhodamine 6G, and magdala red (both from Pfaltz and Bauer) were soluble dye grade and likely contained significant levels of impurities. Fluoresceinsulfonic acid, sodium salt, was

obtained from Molecular Probes (Eugene, OR) as a pure compound. Test sample anions were prepared from sodium or potassium salts. Standard carbonate-free potassium hydroxide was used as eluent; details have been described previously ( I ) . Postsuppressor Ion Exchanger. The large majority of the work was conducted with tubular membrane-based ion exchangers (5). Pertinent characteristics of the membrane tubes used in this work are listed in Table I. Membrane tube types 1-4 and 8 were obtained from Perma-Pure Products, Toms River, NJ. Type 8 is produced by Toyo Soda Co., Japan, from proprietary material and by an undisclosed process. Type 9 was obtained from Dionex Corp., the preparation details of this membrane are also proprietary. Types 5-7 and 10 were custom made by RAI Research Corp., Hauppage, NY (type 10 is available commercially from RAI). Designs of individual devices were dictated by the tube diameters. Unlike the situation with use as suppressors, only a relatively small length of membrane is necessary for use as the postsuppressor exchanger; the pressure drops in these devices are therefore small, regardless of the specific device type. For membranes with internal diameters 5400 pm (types 1,6,9, lo), typically a hollow fiber was used; this is designated as device type A. Stainless steel syringe needle segments of appropriate size were inserted a t each end for making connections; various approaches to making connections to membrane tubes have been described (3,5-7). Device type B utilizes the same small bore (5400 pm) fibers except with a filament inserted inside. The device is used in a linear configuration. For membrane tubes having diameters between 500 and 800 ym (types 2 and a), we utilized a filament-filled knotted construction (designated device type C). This configuration has not previously been described. It has been shown for conventional tubes that dispersion is greatly decreased upon knotting the tube (8), and such a tube has become the preferred mixing line for flow injection applications in our laboratory (9-11). Tying knots into membrane tubes cannot be directly practiced; such attempts shut off the flow cross section. However, the cross section cannot be completely shut off if an inner member is present within the fiber lumen. This was practiced therefore for membrane tube types 1 , 2 and 8, with 250-, 630-, and 430-pm nylon monofilaments (6,15, and 12 lb. strength fishing line, respectively) inserted within the lumen and tying knots (of diameter ca. 4 mm) with the filament filled fiber, such that successive knots touch each other. For membrane tubes of diameters 21000 pm, we packed a length of the fiber with glass beads, using glass wool plugs a t each end to retain the beads (device type D). Inlet/outlet tubes were inserted into the fiber and wire-crimped externally. Glass bead size was typically 62 pm (Ace Glass, Inc., Vineland, NJ); 300 and 500 ym size beads (Thomas Scientific, Swedesboro, NJ) were also used as indicated. The membrane devices were connected between the suppressor and the detector, typically immersed in a solution of the fluor-

ANALYTICAL CHEMISTRY, VOL. 59, NO. 15, AUGUST 1, 1987

escent compound under test. Only with a very dilute external solution was it deemed necessary to renew the external solution continuously. This was accomplished by putting the membrane device within an external jacket in a tubular suppressor-like configuration (3) and allowing the solution of the fluorescent compound to flow continuously through the external jacket by gravity. The effect of an external ac field was tested on membrane devices of the externally jacketed type, constructed from membrane tube types 1 and 6 by using a stainless steel tube for the outer jacket. A short length of 500 pm i.d. Tygon tube was used as the connecting line between the membrane device and the detector and a micro-puncture was made in the wall of this tube with a 27-gauge hypodermic needle. A 100 pm diameter gold wire was then inserted from the outside through the needle into the fiber lumen until it extended through the entire length of the membrane tube. A short length of gold wire was left protruding out of the Tygon tube wall. The puncture was sealed over with polyurethane foam sealant. The field (10 kHz, sine wave, 10 V (peak to peak)) was generated by a Tektronix FG 504 function generator and applied across the gold wire and the stainless steel jacket. Penetration of the fluorescent compound external to the membrane into the internal flow stream was measured under normal operating conditions, i.e., with suppressed eluent flowing through the device, which was immersed in the solution of the fluorescent compound. The concentration of the fluorescent compound in the membrane effluent was measured by measuring the fluorecence/absorbance of the effluent under optimal emission/absorption conditions and comparing this value with those of bracketing standards under the same conditions. The percentage of a suppressed analyte exchanged by the postsuppressor device for the fluorescent ion of interest was calculated by comparing, on an ionic equivalent basis, the peak heights obtained from a sample containing known amounts of chloride and sulfate ( 5 nmol each) injected on a AS4A column (30 mM KOH, 1 mL/min) followed by the suppressor and the postsuppressor with that obtained by direct injection of a standard solution of the fluorophore into a water carrier flowing directly into the detector. The effect of dispersion on the results was taken into account via consideration of the peak width ( I ) . Ion exchange resins in short packed columns (4.6 X 50 mm) were also used as postsuppressor ion exchangers. Strong acid type cation exchanger (H+form Rexyn-101, Fisher Scientific), weak acid type cation exchanger (H+ form Amberlite CG-50), strong base type anion exchanger (OH- form Dowex-BX@,and several weak base type ion echangers IRA-45, IRA-47, IRA-68, IRA-93, and IRA-94 (all from Sigma Chemical; IRA-47 was obtained as the hydrochloride, the others as the free base) were typically ground and sieved to 200-mesh size before use. The Dowex 2x8 column was first washed with a small or large volume of 10 mM anthranilic acid (HAn) solution to convert the column partly or wholly to the An- form and then it was washed with 2100 mL of water. The cation exchange columns were loaded with HAn by passing 2 mL of saturated HAn through the column. The cation exchange columns were converted to the Ce3+ form by passing several milliliters of concentrated Ce2(S0J3 solution through the columns and washing with larger volumes of water. An- was loaded on the weak base type ion exchanger columns by passing 10 mL of 10 mM LiAn through the respective columns. The reproducibilityof the results presented for the membrane devices were found to be S 5 % , while those for packed column devices changed with the time.

RESULTS AND DISCUSSION Mechanism of Operation. Ion Exchange Column Postsuppressors. The selectivity coefficient of Rexyn 101, a poly(styrene-divinylbenzene) strong acid type cation exchanger for Ce3+ vs. H+ was found to be very high (log S = 3.5; see eq 8 in ref 3 for a definition of the selectivity coefficient). This is so high that good exchange efficiency is not likely to result at low levels of H+. Only 7% and 3% exchange were observed for the chloride and the sulfate peaks, respectively (see Experimental Section for details of this determination), for the Ce3+-loadedRexyn 101 postsuppressor.

1965

The exchange efficiencies were marginally better for the Ce3+-loaded CG-50 (weak acid cation exchanger) postsuppressor, 14% and 9%, respectively, but still far smaller than desired. The dual requirements of maintaining low back pressures after the suppressor and of low band dispersion preclude the use of longer columns to obtain higher exchange efficiencies. Attempts at anion exchange, using An--loaded packed anion exchange columns, were even less successful. Although selectivity coefficients against specific anions were not measured; T-T interactions between the aromatic rings are expected to lead to strong retention of An- on poly(styrene-divinylbenzene)-based anion exchangers. Accordingly, no measurable detector response was observed with the An--loaded strong base type anion exchange column and only two of the five weak base type columns, IRA-45 and IRA-68, produced any detectable response a t all (exchange efficiency 5 5 % ) when used as postsuppressor devices. With the packed column postsuppressors, the HAn-loaded cation exchanger mode, which produces negative signals, led to the best detectability. Approximately 40% of the chloride and sulfate equivalents were observed to be translated into the fluorophore signal with the Rexyn-101 column. The same figure for the weak acid type column is much lower, 5%. However, the fundamental limitation of this mode of operation is the high background bleedoff rate of the HAn which deteriorates the signal to noise ratio (S/N), as well as limits the operational l i e before the column must be reloaded with HAn. Membrane Postsuppressors. Other than the facility of continuous rejuvenation and elimination of certain other operational difficulties encountered with a packed column suppressor, there is little fundamental functional difference between a packed column and a membrane suppressor. With the postsuppressors used here for ion exchange, however, this is not the case. For all but extremely low concentrations of the fluorescent ion put outside the membrane, a finite measurable transport of the ion occurs through the membrane, even if there are hardly any ions to be exchanged on the inside of the membrane (Le., internal stream is pure water). The exact extent of this transport depends on membrane parameters such as total available surface area, membrane thickness, etc. The fact that with any given Nafion membrane device immersed in a Ce2(S04)3solution, the effluent fluorescence with pure water flowing through the membrane is dependent on the external Ce3+concentration shows that this transport is governed primarily by the concentration differential and not by an ion exchange process at the membrane interface. If a t this point exchangeable concentration of cations, including H+ ions, in the inner flowstream is momentarily increased, as from an eluting sample, H+ ions will be exchanged for Ce3+ at the membrane interface. We believe that the response of the Ce3+-loadedmembrane system to H+impulse is not governed by the laws followed by an ion exchange process a t equilibrium. In the present situation, exchangeability of H+ for Ce3+seems to partly involve a pseudocatalytic effect of promoting the concentration differential governed transport of the Ce3+. The interior wall of the membrane is essentially saturated with Ce3+;the release of Ce3+from this surface due to the transient passage of the H+-rich eluite band represents an impulse-response system rather than a system at equilibrium or one a t steady state as in a suppressor. For trace amounts of either H2S04or HC1, directly injected into the postsuppressor as described in the Experimental Section, we have observed for example an exchange efficiency of 75435% of H+ for Ce3+on an equivalent basis for device type 3D(62)-10 (Table 11) with 100 pM external Ce3+. Not only is this far too high to be dictated by the laws of equilibrium ion exchange in view of the measured selectivity coefficient

1966

ANALYTICAL CHEMISTRY, VOL. 59, NO. 15, AUGUST 1, 1987

Table 11. Response of Membrane-Based Postsuppressorso

device typeb

external fluorophore

% exchanged'

lC(250)-15 2C(600)-1 2C(600)-20 3D(62)-10

A. Cation Exchange Membrane, Cationic Fluorophorese 86 (78) 0.5 mM Ce(II1) 0.5 mM Ce(II1) 11 (6) 0.5 mM Ce(II1) 37 (26) 58 (57) 0.01 mM Ce(II1)

4C(630)-15

0.05 mM Ce(II1) 0.1 mM Ce(II1) (0.15 pM)f 0.25 mM Ce(II1) 0.5 mM Ce(II1) 1.25 mM Ce(II1) 2.5 mM Ce(II1) 5.0 mM Ce(II1) 0.5 mM Ce(II1)

1C (250)-15

0.01 mM HAn

2B(600)-5 2B(600-5) 3D(62)-10

5D(300)-20 6D(300)-15

6D(62)-15 7C(430)-15 8A-20 9A-60 10A-20

84 (71) 84 (74) 86 (71) 94 (85) 94 (85) 99 (78) 99 (88) 6 (4)

SINd

350 (110) 70 (14) 120 (30) 270 (93) 50 (15), UV 256 nm 340 (100) 340 (105) 350 (100) 380 (120) 380 (120) 270 (70) 80 (25) 17 (4)

B. Cation Exchange Membranes, Neutral/Anionic Fluorophoresg 27 (33) 95 (60) 0.05 mM HAn 63 (72) 110 (60) 220 (110), UV 210 nm 55 (60) 140 (75) 0.05 mM HAn, pH 4.3" 0.05 mM HAn, pH 6.0" 58 (60) 155 (85) 55 (60) 140 (75) 0.05 mM HAn, pH 8.1" 78 (82) 110 (60) 0.1 mM HAn 0.5 mM HAn >loo (>loo)' 100 (50) 0.01 mM HAn (1.24 pM)f 9 (12) 43 (30) 11 (14) 40 (27) 0.05 mM HAn 25 (29) 73 (44) 0.1 mM HAn 78 (68) 92 (42) 0.5 mM HAn 0.01 mM HAn 26 (33) 90 (60) 43 (55) 100 (67) 0.05 mM HAn 0.1 mM HAn 54 (61) 95 (56) 220 (120), UV 210 nm 90 (100) 95 (56) 0.5 mM HAn 23 (30) 80 (55) 0.05 mM HAn, pH 6" 0.1 mM HAn, pH 6" 31 (33) 110 (60) 7 (5) 50 (20) 0.5 mM HAn, pH 6" 3 (3) 20 (10) 1.0 mM HAn, pH 6" C. Anion Exchange Membranes, Anionic Fluorophores' 0.2 5 mM HAn, pH 6" 0.01 mM HAn, pH 6" 17 (27) 31 (35) 0.05 mM HAn, pH 6h 0.1 mM HAn, pH 6" 62 (60) 0.5 mM HAn, pH 6h 63 (60) 1.0 mM HAn, pH 6" 63 (65) 60 (60)' 5.0 mM HAn, pH 6" (7 pM)/ 10 mM HAn, pH 6" 57 (60)' 10 mM HAn, pH 6" 0.05 mM HAn, pH 6" 5 mM HAn, pH 6" 10 mM HAn, pH 6" 5 mM HAn, pH 6" 5 mM HAn, pH 6 5mM HAn, pH 6

7 (3) 13 (13) 14 (13) 4 (4)

0.3 (0.3)

3 120 (76) 220 (130) 220 (110) 220 (110) 220 (120) 210 (110) 200 (110) >500 (>500), UV 210 nm 600 (3201, UV 210 nm 50 (11)

90 (45) 90 (45) 40 (20) 2 (1)

k

Unless otherwise noted, results are for fluorescence detection with the Shimadzu detector. The Kratos detector yields a S/N typically -3 times better for anthranilate and -5 times better for Ce(II1). bThe device code involves the membrane tube type as the first numeral(s) (see Table I), the configuration is denoted by letters A-D (Table I), the value in parentheses is the diameter of the filament in fim inserted in the device (type B, C) or the diameter of beads using as packing (type D), the number following the hyphen is the length of the membrane in centimeters. 'Estimated values based on the signal magnitude, see Experimental Section. dSignal to noise ratio for 5 nmol of injected chloride (value for 5 nmol of sulfate in parentheses), chromatographic conditions BS in Table 11. 'All signals in this mode are positive. /For a few cases, the concentration of the fluorophore that appears in the effluent stream under background conditions (only suppressed eluent passing through postsuppressor) is given in parentheses. #All signals in this mode are negative. "pH adjusted with requisite amount of LiOH. 'Response appears to be superstoichiometric, see text. 'Actual exchange efficiencies are likely higher, see text. No response. of Nafion for Ce3+ vs. H+(log S = 3.53) but the exchange efficiency increases further with increasing external Ce3+ concentration. It is to be noted that the selectivity of the membrane for Ce3+is sufficiently high so that exchange sites are saturated with Ce3+,regardless of the external Ce3+concentration. Indeed, there is every indication that the response can be superstoichiometric when one considers that 100% of the H+ in the eluite band probably does not contact the membrane wall in a 10 cm long device.

The relative responses obtained with the various membrane postsuppressors are shown in Table 11. Similar to the high exchange efficiency (H+for Ce3+) observed for the Nafion device above, the exchange efficiencies for C1- and SO4'- for An- with the anion exchange membrane device 6D(300)-15 (5 mM external H h ,pH 6) were high and the same for both anions, 60%. This once again points to the fact that this type of membrane-based exchange is not governed by the laws of an equilibrium process. While the exchange of H+ for Ce3+

ANALYTICAL CHEMISTRY, VOL. 59, NO. 15, AUGUST 1, 1987

is not expected to be dependent on the identity of the anion (unless the corresponding acid is poorly ionized), equilibrium ion exchange demands that a greater fraction of sulfate, for which the anion exchange membrane shows greater selectivity, be exchanged to a greater extent for anthranilate than chloride. This is not always observed. The effect of an alternating electrical field upon ion transport through ion exchangers has been little explored. Russian scientists have reported faster equilibrium and transport in the presence of such a field (12)and Cox and Twardowski (13)have also observed faster equilibration in the presence of an alternating electrical field in membranebased Donnan dialysis. In unpublished work conducted in this laboratory and elsewhere (Irgum, K., University of Umea, Sweden, personal communication, 1986),significant increases in total exchange capacity of membrane-base suppressors (metal-wire filled, metal jacketed) have been observed upon the application of a 10-kHz electrical field. When such an experiment was conducted with the present cation exchanger-Ce3+postsuppressor, we noted (a) a small decrease in the detector background and (b) a dramatic (nearly an order of magnitude) decrease in response to eluting sample ions. For the anion exchanger-An- postsuppressor, the background remained essentially the same and a small increase (-20%) in response was observed. We believe that transport of nonionic HAn is significantly involved in the latter case. The results indicate that ion exchange as observed in the present postsuppressor is not significantly governed by Donnan dialysis or equilibrium ion transport rates through the membranes. Performance of Membrane Postsuppressors. Table I1 contains selected, but representative, data from the large number of experiments with various membrane devices conducted during this study. Section A is concerned with cation exchange membrane devices operating via cation exchange with fluorescent cations, principally Ce3+. All peaks in this mode are positive. Salient points to be noted include the following: (a) All three Nafion membranes (types 1-3) produce useful devices with good S/N. Very detailed comparison of the performance of different membrane devices is not possible without a very large number of experiments involving indentically made devices. This is because of considerable variation in membrane performance, even among membrane segments from the same manufacturing batch. However, other conditions being equal, the extent of exchange is governed by two factors: total internal surface area available for exchange and the hydrodynamic efficiency with which the influent eluite flux is transported to the membrane surface. Hydrodynamic efficiency is dependent on device design; further, flow is always nonlaminar in the entrance region and leads to better transport to the membrane. Consequently, the exchange efficiencies observed with very short devices can be disporportionately high; even a 1cm long membrane device exhibited sufficient exchange efficiency to show a respectable S / N value. (b) For a given device (3D(62)-10), exchange efficiencies increase monotonically with increasing external [Ce3+],approaching unity for 22.5 mM Ce3+. However, background penetration of Ce3+also increases, deteriorating S/N. Thus, Ce3+concentrations 51.25 mM are regarded as best suited for this work. In contrast, no discernible response could be elicited with cationic dyes (10 mM Magdala Red, 0.1-50 mM Rhodamine 6G) as external fluorophores with any of the Nafion-based devices. The hydrophobic interaction with the membrane matrix is apparently too strong for these compounds to permit facile release. (c) In comparison to the Nafion-based membrane devices, the performance of the polyethylene-based cation exchange

9

1967

membrane is significantly worse, confirming once again the superior characteristics of perfluorosulfonate cation exchanger membranes (14). Among the Nafion devices, membrane thickness does not seem to be an important parameter within the limits of this study, inasmuch as comparable S / N values can be obtained with either the thinnest or thickest membranes by proper choice of external [Ce3+]. (d) As far as fluorescence detection is concerned, the use of Ce3+with Nafion membrane devices consistently produce the best S / N and thus the best detectabilities. Because Ce3+ 800 a t 256 nm), UV detection does absorbs rather poorly (enot produce as good detectabilities. However, because of the close correspondence of the principal mercury emission line with the absorption band of Ce3+,inexpensive but sensitive fixed wavelength (254 nm) absorbance detectors can be used for this purpose. Custom-built dedicated fluorescence detectors which use a mercury pen-lamp as the excitation source are also likely to be particularly attractive. Section B in Table I1 is concerned with experiments that use cation exchange membranes with weak acid type fluorophores, principally anthranilic acid (HAn). The external fluorophore permeates through the membrane, probably the dominant route involving the neutral molecule. A less important route may involve the protonated cation of HAn, a possibility not present for other weak acid fluorophores such as fluorescein. When sample elution causes the membrane influent pH to decrease, background release of HAn is inhibited both via an increase in the fraction of the HAn present as the neutral molecule and via binding of the protonated form to the membrane. Sample elution is therefore accompanied by negative peaks. A quantitative treatment of this response is not possible a t this time. However, it has been observed that a t relatively high external HAn levels, the decrease in concentration of the background effluent HAn from the postsuppressor as a result of sample elution can be higher than the equivalent concentration of the sample, i.e., the response may be superstoichiometric (1C(250)-15, 0.5 mM HAn, and other examples not presented in Table 11). Other salient features of this mode include the following: (a) Exchange efficiencies increase with increased external [HAn] but increases in background fluorescence deteriorate S / N a t high external [HAn]. HAn background penetration rate is an order of magnitude higher than corresponding experiments with Ce3+ as reported in section A, Table 11. (b) At constant and relatively low values of external [HAn], e.g., 0.05 mM, upward adjustment of pH with LiOH has little or no discernible effect (device lC(250)-15). However, if HAn concentration is varied and LiOH is added to maintain the pH constant (e.g., a t pH 6 for device 4C(630)-20),the extent of exchange decreases with increasing [ HAn]. Presumably, increased external availability of Li+ leads to exchange of this species with H+, rather than in the desired manner. (c) Nafion membranes perform better than the polyethylene-based exchanger, similar to the observations made with Ce3+. (d) Unlike Ce3+,HAn absorbs strongly in the UV, the emax being 28 600 a t 210 nm. Consequently, UV detection a t 210 nm typically yields equal to or better detectabilities than fluorescence detection. Section C in Table I1 is concerned with anion exchange membranes in the anion exchange mode. All signals in this mode are positive. In our judgment, this is the mode eventually likely to be the one of choice within the general scope of the proposed technique, although this is not necessarily realized in the present work. This is due to both the lack of commercial membrane tubes which offer performance characteristics comparable to their cation exchanger counterpart, Nafion, and the choice of chemistry. Although results are

1968

ANALYTICAL CHEMISTRY, VOL. 59, NO. 15, AUGUST 1, 1987

reported in Table I1 only for anthranilate, we also studied fluorophores such as fluorescein, salicylate, and fluoesceinsulfonate. Fluorescein produced essentially no response, although penetration into the inner flow stream was measurable. Fluoresceinsulfonate was only marginally better. Salicylate, especially at elevated pH (-9.5), produced easily quantifiable response; the S / N values, however, were an order of magnitude worse than those attainable with anthranilate. Best detectabilities were obtained with anthranilic acid and pH 6 was judged to be the minimum suitable pH (data not shown) for effective operation. Operation is possible a t higher pH but oxidative degradation of anthranilate is exacerbated. There are large differences in the performance of different types of membranes. Even among the radiation grafted PTFE membranes, the performance difference is very large. A substantial amount of grafted activated aromatic rings may remain unconverted during this process of membrane preparation and can profoundly affect the hydrophobic character of the membrane. This is not revealed by measurements of equivalent weight. The background penetration rate of HAn is more than an order of magnitude lower under the given conditions than with corresponding experiments with Ndion membranes. This allows much higher external anthranilate concentrations with stable base lines and allows excellent detectabilities by UV. With device 6D(300)-15 a t external anthranilate concentrations 2 5 m M ,the background concentration is, however, high enough to show nonlinearity in the fluorescence intensity-concentration relationship. The exchange efficiencies calculated on the basis of a linear relationship for these conditions should therefore be regarded as lower limits; it is unlikely that the actual values are lower than those with 0.5-1 mM anthranilate. Excellent detectabilities, rivaling or surpassing those obtained by the Nafion-Ce3+-fluorescence detection system may be obtained by UV detection a t 210 nm with anion exchange membrane-An- systems, as shown by the data for device 6D(62)-15. The principal shortcoming of the anthranilate anion exchange membrane system is, however, rapid poisoning of the membranes, presumably via oxidative degradation of adsorbed anthranilate. Deterioration of membrane performance is accompanied by visible discoloration of the membrane. When not in use, storage of the membrane in dilute HCl solutions retards, but does not eliminate, degradation. Unfortunately, none of the other fluorescent substances, less prone to such degradation, yielded comparably attractive results. Experiments were also conducted with dilutionless passive membrane ammonia introduction reactors (IO)following the postsuppressor device for raising the pH in attempts to improve detection limits for the anthranilate system. With either cation or anion exchanger postsuppressors, detectabilities did not improve. Chromatographic Performance. In terms of chromatographic performance, acceptable dispersion was observed with all devices but those of large diameters packed with larger (1300Wm) beads. It should be noted that no special emphasis was placed in this work to minimize the dispersion in the postsuppressors. Isocratic UV absorption chromatograms are shown in Figure 1 for cation exchange membrane postsuppressor-Ce3+ and anion exchange membrane postsuppressor-An- combinations. The significantly superior S / N ratio for the latter combination, in spite of the obvious band broadening introduced by the large bead packing (nitratebromide resolution is completely lost), is noteworthy. Better detection limits are attained for the former combination, however, with fluorescence detection. Under isocratic elution with hydroxide eluents, this combination is capable of routinely attaining detection limits 2- to 5-fold better for

3

0 2 4 6 8 1 0

0 2 4 6 8 1 0

TIME, MIN.-

Figure 1. Postsuppression ion exchange and UV detection with a six-aniorr standard: (A) cation exchange postsuppressor 3D(62)10 with

0.5 mM external Ce3+,256 nm; (B) anion exchange postsuppressor 6D(300)15 with 10 mM external HAn, pH 6, 210 nm. Conditions were as follows: AS4A column, 25 mM KOH, 1 mL/min, 5 nmol of each ion; (1) iodate, (2) chloride, (3) &rite, (4) nitrate, (5) bromide, (6)sulfate. the more common (nonchelating) anions with the Kratos FS 970 fluorescencedetector compared to the Dionex conductivity detector, a bipolar pulse device that produces very respectable limits of detection itself. For example, by use of the chromatographic conditions listed in Figure 1, the signal/noise ratio for a 5-pmol sulfate peak (retention time 8.6 min) was observed to be 6 for the conductivity detector and 15 for Ce3+-postsuppression exchange and fluorescence detection with the Kratos detector operated a t a P M T voltage of 550 V. Response linearity in fluorescence detection was checked for chloride and sulfate samples with the above two postsuppressor-external fluorophore combinations under isocratic elution conditions and was found to be linear from essentially the detection limit to 50 nmol, for each injected anion. Nonlinearity was pronounced with injections containing 500 nmol of each anion. In addition, it should be noted that the membrane behaves like a capacitor. Discharges of the fluorophore as a result of sample injection require a finite recharging time to maintain constant response. For example, with injections of high concentration levels of samples repeated rapidly, the device response decreases continuously. While these behavior are elicited under chromatographically unrealistic injection frequencies, it should be clear that with this detection method, quantitation of minor constituents in a sample containing a large amount of an early eluting ion must necessarily be conducted by means of standard addition. In terms of practical utility it is not convenient to install a postsuppression device if it must be constantly cared for. To this end, we have investigated the utility of a hybrid device. A device of the type 1A-10 was taken and an external jacket tube (3 mm i.d.) was put around it. The jacket was then completely filled with 400-mesh cation exchange resin in Ce3+ form and the jacket ends were closed off. Fifty-microliter samples of 1 mM HC1 were then repeatedly injected with water flowing through the device, and the fluorescence due to liberated Ce3+was monitored. The response stabilized after the fiist few injections and then remained uniform for the next 100 injections, at which point the experiment was terminated. A device of this type, used possibly in a disposable fashion, may provide a convenient way to practice postsuppression ion exchange. The detection modes described above are compatible to a degree, with gradient elution. Identical gradient chromato-

ANALYTICAL CHEMISTRY, VOL. 59, NO. 15, AUGUST 1, 1987

i

2

far the most prominant for case B. The most notable aspect of the chromatogram in (C) is that the oxalate/tartrate and citrate peaks are negative-these anions complex the background Ce3+and reduce the fluorescence. This suggests an application of the Ce3+chemistry for the detection of chelating anions. Perhaps more importantly, such behavior represents additional information regarding qualitative identification that can be available uniquely in this detection mode. The relative sensitivities in the chromatograms in Figure 2 may be discerned from the following data. Chromatogram A was obtained with 30 pS full scale (the last peak is -10 KS in height), and chromatogram B was obtained a t a detector setting 20 times less sensitive than in chromatogram C. Gradient program and peak identification are as follows: time (min) 0 8 13 23 40; 70 B (0.25 M KOH) 0 2 8 16 43 (eluent A is water). Peak identities are as follows (in the order of elution): (1) iodate, (2) acetate, (3) formate, (4) bromate, (5) chloride, (6) nitrite (7) bromide, (8) nitrate, (9) malate, (10) sulfate, (11) oxalate (unresolved from tartrate), (12) phthalate, (13) chromate, (14) phosphate, (15) citrate, 5 nmol each. We hope to report in the future similar work on anion exchange membrane-I03- systems. The facile displacement of IO3- from anion exchange sites, its sensitive detection by electrochemical means, or optical detection of iodine in the ultraviolet (e8 X lo4)after an amplification reaction with acidic iodide are powerful incentives.

lol!ll

I

15

- -

15

I

10

1969

I U

+

+

-

-

4

ACKNOWLEDGMENT We thank J. Lee (RAI Research Corp., Hauppage, NY) for the custom-made radiation grafted ion exchange membrane tubes, J. Kertzman (Perma-Pure Products, Toms River, NJ) for the other various ion exchange membrane tubes, and Dionex Corp. for the use of their equipment and columns.

LITERATURE CITED

'B

I

I

L

0

5

10

(1) (2) (3) (4) (5)

'14

I

I

I

I

I

I

15

20

25

30

35

40

TIME, MIN.

Fisure2.

(4-

ity detection. (B) Fluorescence detection, anion postsuppressor 8D(300)-15with 10 mM external HAn, pH '

exchanger 6. (C) Fluorescence detection, cation exchanger postsuppressor 3D(62)-10 with 0.5 m M external Ce3+. AS5A column, 1 m l l m i n , with Shimadzu RF 430 detector.

grams for three different detection modes, (A) conductivity, (B) anion exchanger-An-, and (C) cation exchanger-Ce3+(the latter two by fluorescence detection) are shown in Figure 2 for a 16-anion standard sample. The background shift is by

Shintani, H.; Dasgupta, P. K. Anal. Chem. 1987, 59, 802-808. Okada, T.; Kuwamoto, T. Anal. Chem. 1985, 57, 829-833. Dasgupta, P. K. Anal. Chem. 1984, 56, 96-103. Downey, S. W.; Hieftje, G. M. Anal. Chim. Acta 1983, 753, 1-13. DaSQUDta, P. K.; Bliqh, R. Q.: Lee, J.; D'Aqostino, V. Anal. Chem. 1986, '57, 253-257.(6) Dasgupta, P. K.; Bligh, R. Q.; Mercurio, M.A. Anal. Chem. 1985, 57, ABA-AR9 .- . .- - . (7) Mercurio-Cason, M. A.; Dasgupta, P. K.; Blakeley, D. W.; Johnson, R. L J. Membr. Sci. 1986, 27, 31-40. (8) Engelhardt, H.; Neue, U. D. Chromatographia 1982, 15, 403-408. (9) Hwang, H.; Dasgupta, P. K. Anal. Chim. Acta 1985, 57, 1009-1012. (IO) Hwang, H.; Dasgupta, P. K. Anal. Chem. 1980, 58, 1521-1524. (11) Dasgupta, P. K.; Yang, H A . Anal. Chem. 1986, 58, 2839-2844. (12) Mal'tsev, G. I.; Kurin, M. N.; Tikhomirov, I. A. Sov. Electrochem. (Engl. Trans/.) 1970, 6, 1834-1837. (13) Cox, J. A.; Twardowski, 2 . Anal. Lett. 1980, 13, 1283-1291. (14) PerfluorinatedIonomer Membranes, ACS Symp. Ser.. 180 Eisenberg, A,, Yeager, H. L. Eds.; American Chemical Society: Washington, DC, 1982.

RECEIVED for review January 16, 1987. Accepted April 22, 1987. This research was supported by the Office of The Basic Energy Sciences, Division of Chemical Sciences, U.S. Department of Energy, through Grant No. DE-FG 05-84-ER13281. However, this article has not been subjected to review by the DOE and no endorsements should be inferred.