Anal. Chem. 1989, 6 1 , 939-945
Table 111. Recovery of Triazines at 50 ng per Gram of Various Vegetables lettuce simazine simetryn atrazine prometon ametryn propazine prometryn terbutryn
98.4 97.3 97.8 99.0 96.3 97.8 95.0 95.3
recovery,” % spinach chicory 97.6 97.0 98.3 98.8 97.2 96.9 95.7 96.2
98.4 99.6 98.0 98.3 96.9 97.3 96.5 95.1
939
LITERATURE CITED Ambrus, A.; Lantos, J.; Visi, E.; Csatlos, I.; Sarvari, L. J . Assoc. Off. Anal. Chem. 1981, 6 4 , 733-742. Khan, S. U.; Marriage, P. B. J . Agric. Food Chem. 1877, 2 5 ,
endive
kale
97.9 98.3 97.5 99.1 98.1 98.0 96.3 96.6
97.5 99.4 97.0 97.6 98.5 97.4 95.8 95.9
Mean values calculate from four determinations.
1408-141 2. Muir, D. C. G.; Baker, B. F. J . Agric. Food Chem. 1878, 18, 111-116. Popl, M.; Voznakova, 2 . ; Tatar, V. J . Chromatogr. Sci. 1983, 2 , 39-42. Mangani, F.; Bruner, F. Chromatographia 1883, 17, 377-380. Binner, R. Tagungsber.-Akad. Landwirtschaftswiss. D . D . R . 1981, 187-1 92. Xu, Y.; Lorenz, W.; Pfister, G.; Bahadir, M.; Korte, F. Fresenlus’ 2. Anal. Chem. 1986, 3 2 5 , 377-380. Bailey, R.; Lebel, G. L.; Manners, T. G.; Renault, C. M. 8th Eastern Canada Workshop on Pesticide Residue Analysis, Ottawa, May 1976. Lawrence, J. F.; McLecd, H. A. J . Assoc. Off. Anal. Chem. 1877, 60, 979-986. Roseboom, H.; Herbold, H. A. J . Chromatogr. 1980, 202, 431-438. Lee, H. B.; Chau, A. S. Y. J . Assoc. Off. Anal. Chem. 1883, 66, 1322-1326. Di Corcla, A.; Marchetti, M.; Samperi, R. J . Chromatogr. 1887, 405, 357-363. Battista, M.; Di Corcia, A.; Marchetti, M. J . Chromatogr. 1988, 454, 233-242. Weber, J. B. Spectrochlm. Acta, Part A 1867, 23A, 456-462. Gordon, J. E. J . Chromatogr. 1865, 18, 542-555. Funasaka, W.; Hanai, T.; Fujimura, K.; Ando, T. J . Chromatogr. 1872, 72, 187-191. Funasaka, W.; Hanai, T.; Matsumoto, T.; Fujimura, K.; Ando, T. J . Chromatogr. 1974, 88, 87-97. Nielen, M. W. F.; Frei, A. W.; Brinkman, U. A. Th. J . Chromatogr. 1984, 317, 557-567. Kaczvinsky, J. R.; Saitoh, K.; Fritz, J. S. Anal. Chern. 1883, 5 5 , 1210-1 215. Green, D. R.; Stull, J. K.; Heesen, T. C. Mar. Pollut. Bull. 1988, 17, 324-329. Green, D. R.; Le Pape, D. Anal. Chem. 1987, 5 9 , 699-703. Di Corcia, A.; Liberatori, A.; Marchetti, M.; Samperi, R. Proceedings of the Workshop “Organic Micropollutants in the Aquatic Environment”, heid in Berlin, Oct 1986; pp 103-116.
than 10 parts per trillion can be measured, while the limit of sensitivity of this method for triazine residues in vegetables can be set a t about 10 ng/g of vegetable. The reusability of both the Carbopack and SCX cartridges WBS estimated by carrying out repeated extractions of the eight triazines from aliquots of water and vegetable extracts. After each extraction, the Carbopack bed was restored with 3 mL of methylene chloride, followed by 2 mL of methanol and 2 mL of water, while 4 mL of 0.12 mol/L HC1 in methanol, 2 mL of methanol, and 1 mL of acetonitrile were passed sequentially through the SCX bed to restore it. After six such water extractions, recovery of the analytes considered did not decrease significantly. Vice versa, after three runs with vegetable extracts, the Carbopack B cartridge partially failed to quantitatively extract triazines. Registry No. H20,7732-18-5; simazine, 122-34-9; simetryn, 1014-70-6; atrazine, 1912-24-9; prometon, 1610-18-0; ametryn, 834-12-8; propazine, 139-40-2; prometryn, 7287-19-6; terbutryn,
RECEIVED for review July 19, 1988. Accepted December
886-50-0.
1988.
20,
Electrodialytic Membrane Suppressor for Ion Chromatography Douglas L. Strong and Purnendu K. Dasgupta* Department of Chemistry and Biochemistry, Texas Tech Uniuersity, Lubbock, Texas 79409-1061 A dual membrane helical eiectrodialytlc suppressor Is described. A platinum-wire-filled tube made of Naflon perfluorosulfonate membrane, inserted in another perfluorosulfonate membrane tube, is colied into a helix. The helical assembly Is inserted wlthln an outer jacket packed with granular conductlve carbon. An alkaline eluent, e.g., NaOH or Na,CO,, flows in the annular channel between the two membranes and pure water flows through the Inner membrane and the outer jacket, countercurrent to the eluent flow. A dc voltage (typlcally 3-8 V) Is applied across the carbon bed and the platinum wire. Na’ in the eluent mlgrates to the cathode compartment resulting In water as the suppressed effluent and NaOH as the catholyte effluent. The dual membrane design prevents direct electrode contact with the eluent; bubble-Induced noise In the suppressed eluent due to any residual gas Is ellmlnated or mlnimlzed with a microporous gas-permeabie membrane tube or by applying sufflcient back pressure to the detector exit. Up to 500 pequiv of NaOH/mln can be quantitatively suppressed with a membrane length of 50 cm and a band dispersion of 106 pL (20-pL sample). With typical eluents, the system permits detectlon llmlts In the low-parts-per-blliion level for most common anIons.
The advent of ion chromatography (IC) irrevocably changed the way anionic analysis is performed (1). Today conductometric IC flourishes both in the chemically suppressed form as originally introduced and in a single column version (2). Research in this laboratory has largely centered on suppressed IC, in particular on the use of membrane devices (3-10). In suppressed IC, the introduction of membrane-based suppressors to replace packed-column devices was a major event (11), permitting continuous and temporally invariant performance. For hydrodynamically well designed membrane suppressors, the attainable exchange capacity depends on the rate of ion transport through the membrane (3, 12, 13). Further, quantitatively exchanging high eluent concentrations requires proportionately high regenerant concentrations. The transmembrane passage of an ion similarly charged to the matrix of the ion exchange membrane (the “forbidden” ion) is prevented solely by the Donnan potential. This barrier is hardly absolute; with the thin membranes used in present suppressors, undesirable penetration of the forbidden regenerant counterion (e.g., sulfate from a dilute H2S04regenerant) occurs significantly at practical regenerant concentrations (14). To minimize regenerant penetration, lower regenerant concentrations may be used at higher flow rates. However, such a practice consumes too much liquid with a proportional
0003-2700/89/0361-0939$01.50/00 1989 American Chemical Society
940
ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989
amount of waste production. In fact, regeneration of the spent regenerant by a large packed-bed suppressor is recommended for many situations (15). Electrodialysis, charge-selective transport through a n ion exchange membrane and electrolysis, provides in principle a potentially important solution to the above dilemma. The best known application of electrodialysis with ion exchange membranes is related to chlor-alkali cells wherein it is desirable to produce salt-free NaOH from the electrolysis of brine (16). In the simplest possible embodiment of an electrodialytic membrane suppressor (EMS),an alkaline eluent, e.g., NaOH, flows on one side of a cation-exchange membrane while pure water flows on the other side of the membrane. With suitably inert electrodes placed in each flow channel, a potential is applied across the membrane with the eluent side held positive. The applied potential causes the eluent cations (Nat) to migrate across the membrane, thus resulting in eluent suppression; Ht and OH- are respectively produced at the anode and cathode to maintain electroneutrality (unfortunately, electrolytic gases are also simultaneously produced). The concept of an EMS for IC is not new (13); There are two U S . patents (17,18) and a recent paper by Chinese workers (19) which describe such a device. None of these devices are strictly electrochemically regenerated; all require acids to function. In this paper, we describe a low dispersion wateroperated EMS and demonstrate its chromatographic performance over a range of conditions.
EXPERIMENTAL SECTION Reagents and Material. Eluents were prepared by dilution of stock "carbonate-free" 50% NaOH solutions (J. T. Baker). Water used was distilled and deionized with a specific resistivity 214 MRScm. Dodecylbenzenesulfonic acid (DBSA) was commercial detergent grade (Bio-soft S-100 or Stepantan H-100, Stepan Chemical Co., Northfield, IL). Platinum and platinumcoated tungsten/molybdenum wires were obtained from Aesar/Johnson-Matthey (Seabrook, NH). Perfluorosulfonate cation exchanger (Nafion) membrane tubing was obtained in three sizes, 020X (-400 pm i.d., -50 ym wall), 8 1 1 X (-625 pm i.d., -125 pm wall), and 815X (- 1000 pm i.d., 125 pm wall) from Perma-Pure Products (Toms River, NJ). Conductive carbon granules were prepared by grinding spectroscopic grade carbon rods and manually sieving through screens of appropriate mesh size. All other reagents used were of analytical reagent grade. Equipment. Single-piston reciprocating pumps (Models E120-5and B-94, Eldex, Inc., Menlo Park, CA) with packed columns serving as pulse dampeners or peristaltic pumps (model Minipuls 2, Gilson Medical Electronics, Middleton, WI) were used for regenerant pumping. Chromatography was conducted on a Model 4000i ion chromatograph (Dionex Corp., Sunnyvale, CA) with a 50-pL loop and a AS4A or a AS5A column. The conductivity detector (CDM-I) was calibrated with a standard KCl solution. dc potentials (0-10V) were provided by filtered and regulated power supplies operable in either constant current or constant voltage mode. Procedure. Suppression-related experiments were performed by pumping the desired eluent and "regenerants" through the designated channels of the EMS. Bubble-induced noise in the suppressed eluent due to any residual gas was minimized either by placing a back pressure regulator set to 45 psi (Optimize Technologies, Bend, OR) at the detector exit or by placing a gas permeable membrane (GPM) device before detector. Of the many types of membranes tried, the best performing GPM consisted of a 30 cm length of a 400 pm i.d. tube made of Accurel microporous polypropylene (0.2 pm pores, 70% surface porosity, Enka AG, Wuppertal, FRG) filled with a sandpaper-roughened 260 pm diameter nylon monofilament (6 lb. strength fishing line). The membrane was put inside a poly(tetrafluoroethy1ene) (PTFE) jacket tube provided with entrance/exit T-connections; C02-free air was drawn through the PTFE jacket by a suction pump. Flow rates were measured frequently, by timed volumetric collection. Band dispersion was measured by injecting a 20-pL sample of NaN03 and measuring the peak width at half-height
-
b
-a
L
J
1
OYLer
y""'
Figure 1. Cross section of the (a)single membrane EMS and (b) dual membrane EMS. (c) Side view of the helical dual membrane EMS. See text for details. ( Wl,J with (a) the injector connected directly to the detector and
(b) the injector connected to the detector via the EMS. Band dispersion was calculated as the square root of the difference of the W12 values ( 4 ) . EMd Designs. Device T y p e 1. This design, shown in cross section in Figure la, utilizes a single 50 cm length of Nafion 020X (B), containing a 254 pm diameter Pt-coated tungsten wire (A), inserted inside an 18 gauge type 304 stainless steel hypodermic needle tubing (C, 840 pm i.d., Small Parts, Inc., Miami, FL). Inlet/outlet liquid connections were made with subminiature polypropylene tees (Ark-Plas, Inc., Flippin, AR). Eluent flows in the inner channel while regenerant flows countercurrent in the outer channel. The steel shell is held negative with respect to the central conductor. Deuice T y p e 2. This involves (cross-section shown in Figure lb) a Pt wire (D)filled 50 cm length of Ndion 020X (E), inserted inside a Nafion 8 1 1 X tubing (F), and the whole inserted inside a stainless steel shell (G). The internal diameter of the shell ranged from 6.3 mm (referred to as device type 2a) to 1.8mm (13 gauge needle tubing, device type 2b). Efforts were then made to reduce the outer shell diameter to the minimum possible (840 pm i.d.), by utilizing a prestretched 811X membrane for E (device type 2c). Electrode connections were made as for device 1. Eluent flows in the annular channel between E and F while regenerant flows in the innermost and outermost channel. Fluid connections are similar to those for device type 3 which is described in detail. Deuice T y p e 3. This design is similar to that for type 2 except that the dual membrane wire-filled assembly is coiled into a helix to promote better mass transfer to the membrane (6). Specifically, a 40 cm length of Ndion 811X is soaked in methanol and stretched to 250% of its original length and allowed to completely dry while under tension. A 368 pm diameter Pt wire or a 381 pm diameter Pt-coated molybdenum wire (H) is then inserted in a -50 cm length of the stretched membrane (I) and this is inserted into a -50 cm length of an unstretched 811X membrane tube (J). The coiling of the assembly is carried out on a 1.25 mm diameter support (18 gauge needle tubing); the support is then removed and the coil is slightly stretched to prevent successive turns from touching each other. A platinum-coated tungsten/molybdenum wire (K) is deployed through the center of the coil. The electrodes (H, K) are accessible only through one end of the EMS. The entire assembly is then inserted into a PTFE jacket tube (L, 23 cm X 4 mm i.d.). With a glass wool plug (M) serving as a retainer, 20-28 mesh conductive carbon (N) is packed in the jacket space with a vibrating tool. The other end is then also secured with a glass wool plug. One end of the complete EMS device (device type 3a) is shown in Figure ICwith exaggerated detail to reveal the connection means. The jacket tube (L) terminates in a '/a in. and a in. polypropylene tee (0,P)with the help of a PTFE tube spacer (Q).The outer membrane (J) is sealed to the tee (P) by an inserted stainless steel tube segment (R). The inner membrane
ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989
(I) is sealed to the tee (P) by inserted stainless steel (S) and PTFE tube (T) segments. All seals are reinforced by a pair of nichrome wire crimps (U). The fluid flow channels are indicated. A variation within this basic design was also experimented with a helix containing Nafion 815X as the outer membrane and 811X as the inner membrane (device type 3b).
RESULTS AND DISCUSSION Device Designs, Current Efficiency, and Gas Evolution. With NaOH flowing through the eluent channel and water flowing through the “regenerant” channel(s), Na+ migrates through the membrane to the cathode under the influence of the applied potential. Electroneutrality of the catholyte is maintained by the electrolytic production of OH-
2H20 + 2e-
-
20H-
+ H2
Similarly at the anode, a corresponding amount of H+ is generated
H20 - 2e-
-
2H+
+ Y2O2
For a single membrane device (type 1)the anodic H+ directly neutralizes the eluent OH-; for a dual membrane device, the generated H+ migrates across the anode membrane to the eluent solution. Since H+ migrates through the membrane faster than Na+, the rate-limiting process for the suppression of the eluent remains the transport of Na+ through the membrane for both single and dual membrane devices. The minimum amount of current necessary to achieve quantitative suppression is obviously dependent on the total quantity of Na+ to be transported. For a C molar solution of NaOH flowing through the EMS at a volumetric flow rate of V mL/min, the total flux of Na+ in mequiv/min, Q, is given by
Q = C V mequiv/min
(5)
Multiplying Q by the Faraday constant F (coulombs/equiv) and appropriately changing the time unit then directly yields the minimum necessary current imin
imin= FCV/GO mA
(4)
Putting in numerical values, it is readily computed, for example, that &,, is 80 mA for a 0.05 M NaOH solution flowing at 1 mL/min. In practice, some electrolytic breakdown of water, as represented by eq 1and 2, unavoidably occurs without concomitant Na+ transport through the membrane. The degree to which this occurs depends on the electrical resistance of the EMS, specifically the spatial resistance profile along the EMS. Unlike other more familiar examples of electrodialysis, suppression in IC requires quantitative removal of the eluent cation. Development of high electrical resistance at some point in the eluent channel of the EMS device is thus unavoidable. The EMS device can be visualized as a number of smaller serially connected devices, each attempting to exchange successively smaller amounts of influent Na+, while they are electrically connected in parallel. It is possible, in principle, to have a succession of physically separate EMS devices such that individually optimal voltages can be applied to each unit. With a single EMS device required by practical considerations, a current in significant excess in imin,Le., less than 100% current efficiency, is necessary to achieve quantitative suppression. Device type 1, with only a single membrane interposed between the two electrodes and no separate eluent channel, predictably produces the best current efficiency. Quantitative suppression of 0.05 M NaOH flowing at 1 mL/min was achieved for example at a current of 1.07ik; at imh, the EMS effluent was 99.6+% exchanged. Another benefit of this simple design is the low hydraulic resistance in all fluid channels. Unfortunately, the electrolytic production of oxygen
Q41
in the eluent channel is so abundant (ca. 0.36 mL/min for the suppression example given above) that it is impractical to apply sufficient back pressure or devise a GPM device of sufficiently small dispersion to produce an acceptably stable base line. The dual membrane design was therefore explored. In such a device, the anode is separated from the eluent channel by a second membrane. The occurrence of gas in the suppressed effluent is thus drastically reduced, albeit not totally eliminated. As to the continued presence of small amounts of gas in the eluent channel in the dual membrane EMS, the electrolysis of water at the membrane surface appears to be responsible. We draw this conclusion based on the following observations. Operating the EMS under normal fluid pressures but without applied voltage and deliberately introducing pressurized gas (H2/02)into the electrode channels do not result in any gas bubble formation in the eluent channel; i.e., the Nafion membranes are not particularly permeable to H2 or 02.The analysis of the gas formed in the eluent channel of the dual membrane EMS shows it to be a mixture of H2 and O2 in a 2:l volumetric ratio. Further, the amount of gas in the eluent channel increases with increasing applied voltage. All of these suggest that if the applied voltage is sufficiently high, the potential difference between the membrane surface constituting the analyte channel can be high enough to cause electrolytic breakdown of water. In practice, applying back pressure to the detector exit or using a GPM are both effective in dealing with any residual gas in the eluent channel of the dual membrane EMS. The former is much preferred because no additional sources of dispersion are introduced; the GPM also tends to leak eventually over extended periods of operation. Tubular Nafionmembrane-based EMS devices are sufficiently robust to tolerate detector back pressures at least up to 60 psi without problems. Applying back pressure not only reduces the size of any gas bubbles in solution and thus diminishes the magnitude of the noise but also reduces the actual amount of gas produced. Increasing the pressure increases the activity of H2/Oz present in the eluent channel and this in turn inhibits the progress of reactions 1 and 2 (electrolytic breakdown of water) at the same applied voltage. Indeed, a t least part of the effectiveness of a GPM may well be due to the back pressure it presents to the EMS. In any case, without a deliberate attempt to deal with the electrolytically generated gases, no useful EMS device is possible. In principle, it is possible to operate an EMS device with electron donor/acceptors that are Donnan-forbidden toward transmembrane passage, e.g., ferrocyanide/ferricyanidein the anode/cathode channels, respectively, thus eliminating gas formation in the electrode channels. It may further be possible to recycle such anolyte/catholyte solutions from one compartment to the other. At the present stage of development, however, the added complexity of such a system largely negates the advantages of a water-regenerant EMS. Operating Characteristics. For all designs, the extent of suppression attained was relatively insensitive to the anolyte/catholyte flow rate within the range of 0.3-1.0 mL/min. However, since at higher currents (2300 mA) warming of the EMS device was perceptible, the anolyte/catholyte flow rates were maintained at 20.5 mL/min. Band dispersion of device 3a was measured to be 106 f 5 pL for a 20-pL injected sample. The relationship between current and conductance of the suppressed effluent is shown in Figure 2a for four dual membrane EMS devices with 0.05 M NaOH as eluent flowing at 0.5 mL/min. The devices are operated in constant current mode. Complete suppression is never achieved with the cathode at a large distance from the outer membrane (device 2a) even with a current 7.5 times imin.Device 2c is designed
942
ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989
1-
10
40 I
400
a
C
8
300
35
200
30
100
25
I!
6
n 4
2 V
0
100
200
300
0
1 ' '
,
'
3
0
0
1 20 0
Flgure 2. (a) Eluate conductance as a function of current for different EMS designs: 50 mM NaOH eluent, water regenerant, all flows 0.5 mL/min. (b) Current-voltage relationships of the different EMS designs; conditions as in part a . (c) Current required (left ordinate) for complete suppression (detection limit noise limited) and current efficiency (right ordinate) as a function of eluent concentration.
with minimum possible distance between the cathode and the outer membrane and exhibits much better current efficiency than device 2a, achieving quantitative suppression at 5imh. The hydraulic resistance to catholyte flow in this device is quite large and catholyte effluent leaves the EMS as a uniformly gas bubble interspersed stream. Device 2b (not shown) utilizes a cathode-outer membrane separation intermediate between 2a and 2c and exhibits an intermediate performance. As these experiments clearly suggested the importance of the proximity of the cathode to the outer membrane, the helical devices were designed with granular conductive carbon in intimate contact with the outer membrane (device type 3). Such a design exhibits little hydraulic resistance in the outer channel. As Figure 2a shows, device 3a performs significantly better than device type 2. The replacement of a steel shell with a carbon cathode also led to the possibility of reversing the polarity of the applied voltage, i.e., with the carbon functioning as the anode. (Note that when Pt-coated Mo wire is used as the central electrode/helix support in device type 3, the Pt-coating tends to peel off as a result of coiling. The bare Mo is rapidly attacked if held anodic. Such a device can, however, be used with the polarity reversed.) Interestingly, this performance, shown as device 3ar, is somewhat better than the normal electrode polarity. Detailed experiments showed that the carbon electrode was being slowly oxidized under these conditions, resulting in the production of acidic material. With pure water flowing through all three flow channels, the outer jacket effluent pH dropped from near-neutral to -4.5 as current was switched on and returned to neutral slowly over 1.5 h following the cessation of current flow. For a glassy carbon electrode anodically activated by cycling for three weeks to a potential of 1.8 V (vs an Ag quasi-reference electrode in 0.1 M H,SO,), Kepley and Bard (20) found that the activated surface material incorporated -26% 0 by weight. In our system, with an applied potential difference as high as 8 V, oxidation doubtless proceeds to organic acids, potentially even to COP. Although the EMS device obviously cannot be operated indefinitely with such a reverse polarity, the improvement in this operating mode led us to conduct experiments involving deliberate doping of the catholyte/anolyte solutions with small amounts of electrolyte. Solutions of 1 mM DBSA, 1 mM DBSA-Na salt, 0.5 mM HZSO,, and Na2S04were each used as (a) catholyte only, (b) anolyte only, and (c) both catholyte and anolyte. Improvements in terms of current efficiency and/or attainable background conductances over the use of pure water for the same test conditions as in Figure 2a were discernible only with DBSA or DBSA-Na
-
salt and if used as the catholyte only. The improvements were marginal however, and judged unworthy of further pursuit. In terms of current efficiency, even the performance of device 3ar falls far short of what can be achieved with a single membrane device: the current necessary to achieve quantitative suppression was nearly 2.5i- Device 3b, with a larger surface area and greater membrane thickness than 3a, exhibited a current-conductance relationship (and also current-voltage behavior, vide infra) nearly identical to device 3a. The surface area and wall thickness factors appear to essentially nulllify each other; however, total gas production in the suppressed eluate channel was perceptibly lower. The hydraulic radii of the electrode channels are significantly larger for this device and thus much larger flow rates are possible through these channels at modest pressures. This is useful for suppressing high concentration eluents (e.g., >200 mM NaOH) as the attendant requirements for current and thus for heat dissipation increase. The current-voltage relationships (operated in a constant voltage mode) for devices 2a, 2c, 3a, and 3ar are shown in Figure 2b under the same test conditions as in Figure 2a. The i-E behavior, except for the finite intercept on the E axis, is nearly ohmic for all devices. The intercept is approximately the voltage a t which significant electrolysis begins. This is also a measure of the static resistance of the EMS device prior to suppression because the iR drop in the solution controls the actual potential a t the electrodes and thus the onset of significant electrolysis. Predictably device type 3ar, with a slightly acidic anolyte resulting from oxidation of the carbon, has the lowest extrapolated intercept; all others contain water as the anolyte and a larger V intercept. Similarly, among the type 2 devices, the effect of the cathode-outer membrane separation is clearly evident in the differences in the V intercepts. The incremental increase in current with voltage is a function of the efficiency of mass transport within the membrane device. The slope of the lines in Figure 2b is approximately the same for each device type and the superior mass transport efficiency of the helical configuration is reflected in a -60% greater slope of device type 3 over device type 2. The low resistance of the outermost channel obtained with conductive carbon packing is also likely an important contributing factor toward the superior performance exhibited by type 3 devices. Current Efficiency] Suppression Capacity, and Background Conductance. The current necessary to achieve quantitative suppression and current efficiency relative to iminare shown in Figure 2c. (Quantitative suppression is
ANALYTICAL CHEMISTRY, VOL. 61,NO. 9,MAY 1, 1989
assumed to be indicated by a background conductance less than or equal to that obtained with a packed suppressor column. In practice,, it is generally indicated by a major reduction in the base-line noise.) Although there is an increasing current requirement with increasing demand on the suppression capacity, it is interesting to note that the current efficiency actually increases with increasing eluent concentration. We believe that this behavior is partly traceable to the increased conductance of the suppressed eluent with increasing eluent concentration. The conductance of an NaOH eluent after suppression should be, in principle, independent of the eluent concentration because it is suppressed to water. In practice, it has not been possible to purify NaOH solutions to the exclusion of all other anions but hydroxide. In this work, we did not make any attempts to purify the 50% stock NaOH solution. EMS device type 3 was found capable of quantitatively transporting over 500 Kequiv of Na+/min (0.4 M at 1.3 mL/min), given enough current. This is likely not the upper limit of the device capability. However, past this point heat dissipation becomes increasingly a major problem and base-line noise, in our judgment, becomes unacceptably high. The background conductance of the suppressed eluent increases monototonically with the eluent concentration; with 400 mM NaOH, for example, the suppressed conductance is 20 pS/cm. It is easily computed that the presence of either -3 mol % (relative to NaOH) carbonate or -0.01% mol % chloride in the eluent is sufficient to account for this background conductance; the increase in background conductance is not due to the inability of the EMS to quantitatively exchange the Na+. Simultaneous Electrodialysis and Chemical Regeneration. A continuing problem in IC analysis is the determination of trace components in very high ionic strength samples, e.g., brine or 50% NaOH. Presently, such samples can only be analyzed after some form of pretreatment. One potential solution to this problem is separation columns of high exchange capacity, operated in conjunction with correspondingly high ionic strength eluents and thus necessitating low dispersion membrane suppressors of very high dynamic exchange capacity. The lure of increasing the maximum exchange capacity of a chemically regenerated membrane suppressor by using electrodialysis as a supplement is a tempting one. Indeed, it has been suggested (17)that application of a potential allows the desired suppression process “to proceed with enhanced efficiency” and thus can increase exchange capacities compared to those obtained by chemical suppression alone. Experiments, however, show otherwise. Device 3a was operated in the completely chemical mode (6);with 0.25 M DBSA flowing at 1mL/min through each regenerant channel, the device was able to exchange 94% of 200 pequivfmin influent Na+ (0.2 M NaOH at 1mL/min). Application of potential produced only minor changes. Up to a current of 100 mA, the background conductance actually increased. Upon further increase in the current, a modest increase in the extent of suppression, to 97%, was observed and did not show any further change upon increasing the current to 400 mA. This behavior is understandable in that with purely chemical suppression, exchange of Na+ for H+occurs at both the inner and the outer membranes. With application of potential, the anode membrane essentially becomes unavailable for Na+ transport while the Na+ transport through the cathode membrane increases because of accelerated migration rates in the direction of the electrical field. The overall result of these opposing factors is little net change; certainly an additive increase in exchange capacity is not obtained in this manner. Performance. All data are presented for device type 3. The base-line stability is shown in Figure 3 with 0.05-0.2 M
943
200 m M 150 mM
E 100 mM
-
9
“;I%
el
50 m M
E
J 2 i
I
0
2
4
Time (min) Flgure 3. Baseline noise levels as a function of eluent concentration suppressed, eluent flow rate 1 mL/min. The noise levels for the two lowest concentrations are shown 1OX magnified as well. See Figures 4 and 5 for base-line noise at more typically used eluent concentrations.
I
t
0
2
4
6
Minutes Flgure 4. Typical chromatogram with a hydroxide eluent, 28 mM NaOH, 0.7 mL/min; column HPIC AS5A. I n order of elution: (upper) F- (l),CI- (l),NO2- (3), NO3- (3), SO,’- (3), concentrations in ppm in parentheses, i= 100 mA; (lower) CI- (3), NOz- (3), NO,- (lo),SO -: (lo),concentrations in ppb, i = 120 mA.
NaOH flowing at 1 mL/min, quantitatively suppressed. Although base-line noise is somewhat higher than that obtained with a packed column suppressor, the base-line noise decreases rapidly with decreasing eluent concentration. Operation at typical eluent concentrations around 20-30 mM NaOH exhibits respectable performance with typical detection limits in the low-parts-per-billion level. (Figure 4). The lower chromatogram in Figure 4 was obtained with a 0.05 Hz frequency cutoff digital filter;this improves signal t~ noise (S/N) by a factor -2. The utility of the EMS is not limited to hydroxide eluents, attractive performance can be obtained with a carbonate eluent as well (Figure 5). Ideally, gradient anion chromatography with an EMS should not only be conducted with a highly pure NaOH eluent but also involve current programming the EMS. The potential of the EMS for gradient IC is evident in Figure 6a even with constant current operation and eluents made from 50%
944
ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989
b
a
ii 0
8
Minutes
16
l
0
,
*
J
4
,
8
Minutes
U
+
1
#
0
Figure 6. (a) Gradient hydroxide elution; time (min), YO E: 0, 20% E l + 80% E2; 3, 100% E2; 3.1,90% E2 10% E3; 10, 100% E3. (El = H,O, E2 = 14 mM NaOH, E3 = 150 mM NaOH). Column HPIC AS4A, current 300 mA. I n order of elution (ppm): F- (lo),Ci- (lo), NO3- (30),Cr0:(401, BF, (30),SCN- (30). (b) Isocratic elution with
2
Minutes
4
6
Flgure 5.
Typical chromatogram with a carbonate eluent, 2 mM Na2C03 0.15 mM NaHCO,, 1.5 mL/mln, column HPIC AS4A, i = 40 mA: (upper) F- (l),CI- (l),NO2- (3),NO3- (3),Sod2-(3).concentrations in ppm in parentheses; (lower) F-, CI- (20 ppb each, both hidden in water dip), NO3-, SO,2- (30 ppb each).
+
NaOH. Similarly, although the ultimate in detectability is not possible under this condition, the facility with which very strongly retained anions are eluted rapidly with symmetric peaks is illustrated in Figure 6b by the ability of the EMS to suppress large eluent concentrations. Extent of Ion Exchange and Detectability. The issues of base-line noise and limits of detection are more complex with a water-operated EMS than with an acid-regenerated membrane suppressor. With the latter device, limits of detection are not predicated, per se, by the extent of ion exchange of Na+ for H+. Since some penetration of the regenerant acid is unavoidable (14), complete acid-base neutralization of the suppressor effluent occurs without the need for stoichiometric ion exchange. A typical situation, a 30 mM NaOH eluent flowing a t 1 mL/min being suppressed by 12.5 mM H2S04flowing a t - 5 mL/min, may result in a penetration of -5 nmol of HzSO,/min. If 99.99% of the influent Na+ is exchanged, the suppressor effluent will consist of 1.5 pM Na2SO4,3.5 pM HzS04. Indeed, under typical conditions of chemically regenerated membrane suppression, the suppressor effluent is very slightly acidic (8). Neutralization of all of the alkaline equivalents represented by the eluent, whether by ion exchange or regenerant penetration, is necessary for the limit of detection to be solely dictated by base-line noise. In the above example, there is no theoretical or practical barrier toward the detection of a low concentraton analyte; an analyte X- with a concentration of 1 pM at the peak apex represents a net addition of 1 pM HX over the background composition of 1.5 pM Na2S04,3.5 WMH2S04. This should be easily detectable. On the other hand, if ion exchange took place without the possibility of regenerant penetration as in the case of the water-operated EMS, the same 99.99% ion exchange will leave a concentration of 3 pM NaOH in the suppressed effluent. It is obvious that a positive peak cannot result until the effluent analyte concentration is larger than this; consideration of equivalent conductance values in fact indicate that it has to be very substantially larger
400 mM NaOH eluent, 1.3 mL/min; HPIC AS4A column, i I n order of elution: I-, SCN-, C104- (30 ppm each).
= 780 mA.
to produce an observable positive response. The effects of remnant alkaline eluent concentrations on analyte response have been recently calculated (21). The authors intended the calculations to represent the case of a chemically regenerated membrane suppressor. Because regenerant penetration was ignored, the calculations do not really apply to the intended case; in fact they are much more appropriate for the present water-operated EMS. Presence of other anionic impurities in a real eluent do not significantly complicate the situation; the requirement that all alkaline equivalents must be neutralized before a positive peak results is still valid. In this light, consider the detection of the sulfate peak in the lower chromatogram of Figure 4. With 10 pequiv of sulfate injected and appearing in a peak volume of an estimated 250 KL,the concentration a t the peak apex is 40 nequiv/L. For this to be detected as a positive peak, >99.9999% of the 28 mM alkaline eluent must be exchanged by the EMS! EMS Current Requirements. Noise and Detectability. At or near stoichiometric ion exchange levels, the observed noise levels with an EMS appear to be related to the presence of gases in the suppressed effluent. Although there are no visible bubbles which flow to the detector cell (bubble noise is indicated typically by sharp spikes), presence of microbubbles adhering to the cell electrodes or variations in degree of supersaturation of electrolytic gases in the eluate is possible. No data are available on the dependence of the conductance of pure water upon the content of dissolved oxygen or hydrogen. We were unable to experimentally find a significant change in the conductance of water upon continued electrolysis with platinum electrodes. Current or temperature monitoring equipment with sufficient resolution was unavailable to measure the current through the EMS or the temperature of the suppressor effluent; it was therefore not possible to overrule the possibility that the noise seen with the EMS is associated with small variations in these factors. An experiment was performed with pure water flowing through all channels; the noise significantly increased upon applying current through the EMS, without any significant change in the mean background conductance.
ANALYTICAL CHEMISTRY, VOL. 61, NO. 9, MAY 1, 1989
After most of the influent Na+ is removed, increasingly more energy has to be expended to remove a unit amount of the remnant Na+. The current-conductance relationship in Figure 2a illustrates this. Detectability improves as the remnant Na+ is continually removed by application of a greater voltage (causing more current to flow) because the neutralization of the suppressor effluent, as discussed in the previous section, more nearly reaches stoichiometry. The greater applied voltage however makes it likely that gas evolution noise in the suppressed channel will increase, with an attendant effect on the base-line noise. Consequently, there is no significant correlation of the base-line noise with the mean conductance of the suppressed effluent with the EMS operating at nearstoichiometric neutralization levels. In Figure 4, the upper chromatogram, obtained a t 100 mA (3.2imh),may reveal a slightly less noisy base line if expanded to the same scale as the lower chromatogram; however, under these conditions, 10 ppb injected sulfate cannot actually be detected. While an EMS can always be operated under conditions where the detection limit becomes noise-limited, very low detection limits are not always necessary. If it is of value to operate with a lower EMS current, the choice can be made by repeatedly injecting an analyte of the lowest concentration of interest with increasing EMS currents until no significant change in the peak height is observed. Analyte Loss. The potential loss of analyk anions through the membrane to the anode compartment was investigated by repeatedly injecting tap water and 1Ox diluted tap water through an EMS device and comparing the resulting peak area values with those obtained with a packed column suppressor. Within the precision of the measurements, there was no significant difference between the area values. Additionally, although the local tap water contains several millimolar Ca and smaller concentrations of Mg, no evidence of membrane poisoning by alkaline-earth metals was observed. Previous EMS Designs. An EMS device is first of all a membrane suppressor,it is therefore always possible to operate it in the conventional chemically regenerated mode. It is perhaps noteworthy that the considerable commercial success of suppressed IC owes a large debt to the patented concept of chemical suppression. Incentives to bypass this are therefore substantial. A true “acid test“ of an EMS for anion chromatography should therefore be carried out under conditions in which no suppression is possible without applied voltage, Le., an acid cannot be used in the regenerantlelectrode channels. None of the devices described (17-19)meets this criterion. The most recent (19)uses acid concentrations as high as 0.2 N in both the static reservoirs comprising the electrode compartments. Gas evolution and problems arising therefrom are mentioned in only one previous report (17). This study found it possible to operate only below an applied potential difference of 1.23 V. In our experience, such an applied voltage is quite inadequate to completely suppress any meaningful eluent concentration. Not surprisingly, the actual claims in the patent (17)involve only chemical and not electrodialytic suppression. Another device (18), generally similar in design to our device type 2 except for the use of planar, sheet membranes, supposedly operates with 200 V typically (ranging up to 550 V) applied. The total power consumption (with currents as high as 1.2 A) is large enough to convert the few mL/min of total influent flow to super-
945
heated steam; it is superfluous to criticize this report for not addressing gas evolution problems. While it is impossible to duplicate the design of the most recently reported static reservoir EMS (19)due to insufficient detail given, we find it inexplicable how such a device, with closed electrode compartments, can safely operate with applied voltages substantially above that needed for the electrolytic breakdown of water. This report is the first one, however, to show actual chromatographic performance (5 mM Na2C03 eluent, 2 mL/min), presumably under operative electrodialytic conditions (4 V, 50 mA). Note that the device should be a workable chemical suppressor if no voltage is applied.
CONCLUSIONS The present work demonstrates the viability of a purely electrically regenerated membrane suppressor for IC. In a variety of situations, frequent regenerant replacement is inconvenient and a source of pressurized deionized water can be provided. The water-operated EMS is ideal for such cases. The electrodialytic suppressor operated with pure water as the catholyte does more than replace a chemical regenerant with electricity, however. Its use in a reverse mode, to generate different concentrationsof pure NaOH for use as the IC eluent in a current-programmed manner, may prove more intriguing. ACKNOWLEDGMENT We thank John R. Stillian for much valuable exchange of ideas and information. The gift of the membrane tubes used in this work by Jack Kertzman, Perma-Pure Products, is gratefully acknowledged. LITERATURE CITED (1) Small, H.; Stevens, T. S.; Bauman, W. C. Anal. Chem. 1975, 4 7 , 1801- 1809. (2) Gjerde, D. T.; Fritz, J. S. Ion Chromatography, 2nd ed.;Alfred Hijthig: New York, 1987. (3) Dasgupta, P. K. Anal. Chem. 1984, 56. 96-103. (4) Dasgupta, P. K. Anal. Chem. 1984, 5 6 , 103-105. (5) Dasgupta, P. K. Anal. Chem. 1984, 5 6 , 769-772. (6) Dasgupta, P. K.; Bligh, R. Q.; Mercurio, M. A. Anal. Chem. 1985, 5 7 , 484-489. (7) Mercurio-Cason, M. A.; Dasgupta, P. K.; Blakeley, D. W.; Johnson, R. L. J. Memb. Sci. (1986), 2 7 , 31-40. (8) Shintani, H.; Dasgupta, P. K. Anal. Chem. 1987, 5 9 , 802-808. (9) Shlntani, H.; Dasgupta, P. K. Anal. Chem. 1987, 5 9 , 1963-1969. (IO) Gupta, S.; Dasgupta, P. K. J. Chromatogr. Sci. 1988, 26, 34-38. (11) Stevens, T. S.;Davis, J. C.; Small, H. Anal. Chem. 1981, 5 3 , 1488- 1492. (12) Dasgupta, P. K. I n Ion Chromatography; Tarter, J. G. Ed.; Marcel Dekker: New York, 1987. (13) Stillian, J. R. LC L19. Chromafogr. HPLC Mag. 1985, 3 , 802-808. (14) Dasgupta, P. K.; Bligh, R. Q.; Lee, J.; DAgostino, V. Anal. Chem. 1985, 5 7 , 253-257. (15) Dionex Corporation, Sunnyvale, CA, IC Exchenge 1987. 6(1), 2-3. ( 16) Membrane and Ultrsfikration Technology: Devebpmenfs Since 198 1 ; Torrey, S.,Ed.; Noyes Data Corp.: Park Ridge, NJ, 1984; pp 47-117. (17) Ban, T.; Murayama, T.; Muramoto, S.; Hanaoka, Y. Yokagawa Electric Works Ltd., US. Patent 4,403,039, September 6, 1983. (18) Jansen, K. H.; Fischer, K. H.; Wolf, B. Blotronik Wissenschaftliche Geraete G.m.b.H., US. Patent 4,459,357, July 10. 1984. (19) Tlan, 2 . W.; Hu, R. 2.; Lin, H. S.; J. T. Wu. J. T. J . Chromafogr. 1988, 439, 159-163. (20) Kepley, L. J.; Bard, A. J. Anal. Chem. 1988, 60. 1459-1467. (21) Tlan, 2. W.; Hu, R. 2.; Lin, H. S.; Hu, W. L. J . Chromafogr. 1988, 439, 151-157.
RECEIVED for review December 5, 1988. Accepted January 26,1989. This work was supported by Grant No. DE-FG05MER-13281from the US. Department of Energy. This paper has not, however, been subjected to review by the DOE and no official endorsement should be inferred.
