Automated Programmable Preparation of ... - ACS Publications

This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Article pubs.acs.org/ac

Automated Programmable Preparation of Carbonate-Bicarbonate Eluents for Ion Chromatography with Pressurized Carbon Dioxide C. Phillip Shelor,* Kenji Yoshikawa, and Purnendu K. Dasgupta* Department of Chemistry and Biochemistry University of Texas at Arlington Arlington, Texas 76019-0065, United States S Supporting Information *

ABSTRACT: We introduce a novel carbonate−bicarbonate eluent generation system in which CO2 is introduced using programmable CO2 pressures across a membrane into a flowing solution of electrodialytically generated high purity KOH. Many different gradient types are possible, including situations where gradients are run both on the [KOH] and the CO2 pressure. The system is more versatile than current electrodialytic carbonate eluent generators and can easily generate significantly higher eluent concentrations (at least to 40 mM carbonate), paving the way for future higher capacity columns. Demonstrably purer carbonate−bicarbonate eluent systems are possible compared to manually prepared carbonate−bicarbonate eluents and with considerable savings in time. Performance in different modes is examined. The dissolved CO2 is removed by a carbon dioxide removal device prior to detection. Best case noise levels are within a factor of 2−3 of best case suppressed hydroxide eluent operation. The eluent system allows particular latitude in controlling elution order/time of polyprotic acid analytes. Although CO2 introduction is possible prior to hydroxide eluent generation, this configuration causes complications because of electroreduction of CO2 to formate. lthough the first exposition of anion chromatography began with eluents containing NaOH and Na-phenate,1 the first practical eluents were carbonate−bicarbonate based2 and were the first commercially recommended eluents.3,4 Of manually prepared eluents, the carbonate eluent has continued to prevail over four decades. Compared to the electrodialytic generation of hydroxide eluents,5 similar generation of carbonate−bicarbonate eluents is more complex. Two steps are involved: K2CO3 is first generated and part of the K+ is then exchanged for H+ in a controlled fashion to provide a KHCO3− K2CO3 eluent.6 Additionally, the maximum concentration of carbonate/bicarbonate that can be electrogenerated is significantly less than the corresponding hydroxide concentrations. Suppression of carbonate−bicarbonate eluents results in weakly conducting carbonic acid. Increased noise from a carbonatebased eluent, compared to a hydroxide eluent (suppression product is water) has recently been ameliorated by a two-stage suppressor design.7,8 Even before this development, it was possible to remove the volatile carbonic acid by membranebased devices.9−16 A bicarbonate−carbonate gradient was traditionally held as not practical because whereas monovalent HCO3− occupies one anion exchange site, a CO32− ion occupies two and introduction of the latter into the column thus results in the expulsion of an equimolar amount of bicarbonate resulting in a gradient “hump”.17 A CO2 removal device would largely remove such a gradient hump. Such a device was introduced commercially as a “Carbonate Removal Device” (CRD,18 this nomenclature by the manufacturer is obviously a misnomer). But this came long after electro-

A

© 2017 American Chemical Society

generated hydroxide gradients, which, for a variety of reasons are perceived to be superior to carbonate-bicarbonate gradients. As a result, bicarbonate-carbonate gradients have not been explored in detail, taking advantage of the CRD. It is interesting to note that CO2 intrusion into water used in hydroxide eluent generators (HEGs) can also lead to “carbonate humps”. In this case, rather than removing the CO2 postsuppression with a CRD, the preferred approach has justifiably been to remove the carbonate from the eluent with an electrodialytic anion trap column. We are interested in a large range bicarbonate−carbonate gradient capability for the following reason. Given a certain suppression capacity, it will permit a much greater eluent strength span than hydroxide and will thus enable the use of very high capacity columns. The latter will facilitate the analysis of trace components in high ionic strength samples, always the Holy Grail. As noted above, membrane devices for the removal of carbonic acid as CO2 gas, were introduced more than three decades ago and efforts to improve it continues.19 Such CRDs, permit postseparation degassing of suppressed carbonate eluents to lower the conductance background (and the noise). The CRD also greatly improves the linearity of the otherwise nonlinear response observed with carbonate eluents.20,21 Received: July 18, 2017 Accepted: August 25, 2017 Published: August 25, 2017 10063

DOI: 10.1021/acs.analchem.7b02808 Anal. Chem. 2017, 89, 10063−10070

Article

Analytical Chemistry

Figure 1. System setup. The various flow paths are color coded and marked with arrows. The 10-port valve default setting is marked.

introduced) followed by a microfabricated metallic gradient mixer (V100, www.agilent.com), a high pressure conductivity cell, injector (25 μL loop), an analytical column (IonPac AS9HC, 4 × 250 mm), a suppressor (ESRS 500, 4 mm), a modified 2 mm CRD 200 carbon dioxide removal device, and a CD25 conductivity detector (cell maintained at 35 °C). The highpressure conductivity cell used to monitor the eluent consisted of two stainless steel HPLC tubes electrically isolated but fluidically joined by a PEEK union. The tubes functioned as electrodes for a Dionex CDM-1 conductivity detector. The high-pressure conductivity detector, injector, and column were all housed in a column oven (40 °C) that was part of a Surveyor AS-AP autosampler. The 10-port valve allowed facile reversal of the flow path through the HEG and engasser; the default condition follows the path described above. Note that the AS9-HC column is not intended for long-term use with hydroxide eluents. We are using it here because of its high capacity and therefore utility with carbonate gradients. Although some experiments with pure hydroxide eluents have been described, this is purely for comparative purposes, it is not being suggested that the column be so used. Our experiments with hydroxide were limited enough that there was no significant change in capacity or selectivity due to such use. Engasser Pressure Control. The pressure in the engasser jacket was controlled using an electronic pressure regulator (EPR, Proportion-Air GX1ANKKZP900PSG1). The EPR can control gauge pressures up to 6.2 MPa; maximum attainable CO2 pressure at ambient laboratory temperature (22 °C) is ∼6 MPa. A PEEK column blank (9 mm i.d. 6 cm long) was used as a ballast and connected between the EPR and one tee of the engasser; the second tee was plugged during operation after purging the system with CO2. The permeation rate of CO2 is dependent upon the partial pressure of CO2; even at zero gauge

In the present paper we propose a new paradigm in automated programmable generation of carbonate−bicarbonate eluents. We suggest precisely the opposite of a CRD: introducing variable amounts of CO2 through a membranebased engasser into an electrodialytically generated hydroxide or manually prepared carbonate solution to prepare carbonatebicarbonate eluents for IC. Eluent gradients can be generated both with programmed changes in the hydroxide/carbonate concentration and the CO2 pressure. The strategy can also generate carbonic acid gradients, recently demonstrated to be useful for ion exclusion chromatography.22

■

EXPERIMENTAL SECTION Engasser. The engasser has been described previously22 and is shown in detail in Figure S1 of the electronic Supporting Information (SI). Briefly, the engasser is a length of gas permeable tube (Teflon AF tube, 0.18 mm i.d., 0.74 mm o.d., taken from degasser units integral to electrodialytic eluent generators available from www.thermofisher.com) enclosed within an impermeable jacket tube that terminates at tees at each end, permitting access both to the annular space as well as the lumen of the Teflon AF tube. The engasser used throughout had an active length of 5.5 cm. Laser Pure grade Carbon Dioxide (www.airgas.com P/N CD LZ200) was used throughout. Chromatographic System. Figure 1 schematically shows the chromatographic system. Other than the engasser, all other components are commercial off-the-shelf components (www. thermofisher.com, except as stated). A GP40 gradient pump pumped water through a 10-port stainless steel injector valve (www.vici.com, plumbed as shown), an EG40 hydroxide eluent generator (HEG) equipped with KOH EGC-III cartridge and degasser, the engasser (via which desired amounts of CO2 are 10064


Article

Analytical Chemistry pressure the engasser annular space has ∼1 atm of CO2 available to permeate. Pressures henceforth are reported as absolute values. Conversion of gauge pressure readings was performed by adding the ambient laboratory barometric pressure (typically 98.5 kPa during these experiments) to the EPR reported gauge pressure. A minilab 1008 data acquisition card (www.mccdaq.com) allowed 10 bit analog EPR control and 12 bit recording of the high pressure conductivity data and pressure registered by the EPR with software written in LabView. Engasser Permeation Rate Determination. The permeation rate of the engasser was determined at 25 °C by conductometric titration with 49.8 ± 0.1 mM NaOH solution (as determined by primary standard potassium hydrogen phthalate) prepared from 50% w/w NaOH (www.fishersci. com). The solution was pumped through the engasser at a constant flow rate while varying the external CO2 pressure (pCO2,ext) from 110 to 3450 kPa over 4 h; the experiment was repeated twice. Subsequent shorter-duration (1 h) titrations were performed periodically using 40 mM HEG KOH (a concentration relevant to the generated eluents) with pCO2 centered at 690 kPa (near the equivalence point for the eluent flow rate used) using both ascending and descending pCO2 ramps. In all titrations, 10 min isobaric zones were included before and after the pressure ramp to facilitate pressure and conductivity data alignment (see Figure S2). The KOH titrations were performed at chromatographic temperatures of 40 °C with the column in place to provide enough back pressure to prevent H2 bubble formation at the HEG.

Figure 2. Conductivity titration of 49.75 mM NaOH with CO2 delivered through a 5.5 cm engasser at absolute pressures of 110− 3450 kPa. The two equivalence points can be determined by the intersection of the corresponding two linear segments.

conductance shows no further increase. The two equispaced end-points, indicated by arrows in Figure 2, occur respectively at 1396 and 2807 kPa. Stoichiometry dictates at the first equivalence point:

■

RESULTS AND DISCUSSION Determination of CO2 Permeation Rate of the Engasser. CO2 introduction in the present utilization of the engasser is kinetically rather than thermodynamically governed. Above a solution pH of ∼8.3, pCO2 is negligible; the pK1 for carbonic acid (this term hereinafter includes hydrated CO2) being ∼6.3. For any permeative transfer system, the permeation rate is proportional to the differential pressure of the gas across the barrier. If the receptor solution has a pH throughout high enough to have a negligible internal pCO2, the permeation rate R (in nmol/min) is constant across the length of the tube and is linearly proportional to pCO2,ext. From analogous heat transfer equations for a cylindrical pipe, we derive that R = (kL pCO2,ext )/ln(do/d i)

(2)

FC = 2R

where C mM NaOH is flowing at F mL/min. The combination of eqs 1 and 2 lead to k = [FC × ln(do/d i)]/[2 × L pCO2,ext,equiv,1]

(3)

where pCO2,ext,equiv,1 is the external CO2 pressure at the first equivalence point. Likewise k may be computed using the second equivalence point (pCO2,ext,equiv,2). The perceptive reader will appreciate that the effect of any carbonate contamination of the NaOH used can be eliminated by a difference equation: k = [FC × ln(do/d i)]/[2 × L(pCO2,ext,equiv,2

(1)

− pCO2,ext,equiv,1)]

where R is the permeation rate in nmol/min, L is the length of the tube of o.d. and i.d. do and di, respectively (all in cm), and k is the permeability constant (nmol/(min·cm·kPa)), independent of pCO2,ext (kPa). Figure 2 shows the results of a single conductometric titration experiment to determine the permeation rate. The presently used EPR was designed for a greater level of gas consumption. Set point maintenance involved rather frequent automatic pressurization/depressurization (Figure S2), a moving average filter spanning 20 s was applied for the pressure data in Figure S2. The Teflon AF membrane, however, behaves as a large capacitor, the modest gas pressure oscillations were not observed in the conductivity signal, which displayed a noise level equal to the DAQ bit noise. During the titration, NaOH is converted stepwise to Na2CO3 and then NaHCO3. Further addition of CO2 does not result in perceptible ionization of the added CO2 and solution

(4)

For F = 1 mL/min, k was computed from eqs 3 and 4 for three titrations to be 4.50 ± 0.12 and 4.53 ± 0.08 nmol/(min·cm· kPa), respectively, statistically indistinguishable (combined: 4.515 ± 0.88 nmol/(min·cm·kPa)), also indicating negligible carbonate contamination of the NaOH. For HEG KOH, we relied on the nominal concentration that the manufacturer states it is generating and only the first equivalence point was determined. Based on three titrations spanning as many weeks, k consistently ranged from 3.98 to 4.08 nmol/(min·cm·kPa), ∼10% lower than that obtained with the secondary standard NaOH titration. Further investigation revealed that this particular pump-HEG combination is producing an eluent concentration slightly greater than the nominally programmed value, not exhibited by other HEG setups in our laboratory. In any case, we refer to this nominal manufacturer-stated HEG KOH value throughout as the 10065


Article

Analytical Chemistry

Figure 3. (a) Computed species distribution as CO2 is added to a constant (40 mM) concentration of KOH. (b) Computed species distribution as a 0−50 mM KOH gradient is imposed on a background of constant total dissolved carbonate species concentration of 20 mM. For carbonic acid (this includes dissolved CO2 aside from true carbonic acid), pK1 and pK2 were assumed to be 6.30 and 10.32, respectively. No activity corrections were performed.

(i) Use of an electrogenerated hydroxide eluent, which is converted variously to carbonate and bicarbonate by the CO2 engasser. Gradients are possible: (a) hydroxide gradient at constant pCO2,ext, (b) varying pCO2,ext at constant hydroxide concentration, and (c) both are varied. (ii) Use of a pumped Na2CO3 eluent (with a gradient pump, the concentration may be varied if desired), which is converted to bicarbonate by the engasser to the desired extent. (iii) Use of a pure carbonic acid eluent with low capacity anion or cation exchangers. Because the pH remains acidic, this mode is also compatible with silica-based ion exchange columns. This application is different from the central focus of the present paper; this will be presented elsewhere. The use of pure carbonic acid as eluent in ion exclusion chromatography has already been reported.22 Modes of Deployment. Several modes of deployment are apparent: (a) The HEG precedes the engasser, the engasser runs a pCO2,ext gradient; (b) The engasser precedes the HEG; the HEG runs a gradient; (c) The engasser precedes the HEG; a pCO2,ext gradient is run on the engasser; and (d) The HEG precedes the engasser; a gradient is run on the HEG. Each of these results in different consequences, not to speak of the fact that within each deployment domain, a monotonic gradient in the indicated parameter may not result in a monotonic change in eluent strength. In addition, in all the above deployments, gradients can be run both on the engasser and the HEG. Species Distribution in Different Gradient Modes, Chromatographic Consequences. The consequences of operational mode (a) above are shown in Figure 3a. An increasing amount of CO 2 (0−50 mM equivalent) is introduced into a constant concentration of 40 mM KOH. Up to the point that 20 mM CO2 has been introduced, essentially hydroxide is converted into carbonate, and while the

concentration. Also, unlike the pumped NaOH case, results of titrations using descending vs ascending pressure ramps consistently differed by 3% indicative of some minor hysteresis in the system. Values reported are the average of the ascending and descending pressure ramp titrations. The permeation rate can also be estimated from the slope of the titration plots. The accuracy of this estimation depends on the calibration of the cell constant as well as the listed equivalent conductance values of NaOH, Na2CO3, and NaHCO3 at the concentration level the titration is carried out. In Figure 2, the NaOH to Na2CO3 (step 1) and Na2CO3 to NaHCO3 (step 2) conversion steps display respective descending slopes of 4.25 and 0.515 μS/(cm·kPa), r2 = 0.9995 and 0.9970. The equivalent conductance (λ) of 50 mequiv/L NaOH, Na2CO3, and NaHCO3 were, respectively, reported to be 227, 93.2, and 80.6 μS/cm per meq/L,23 the conversion factors for the difference thus being 133.8 and 12.6 μS/cm per meq/L. Converting the above slopes to concentration units one thus gets 31.8 and 40.9 μeq/(L·kPa) for steps 1 and 2, respectively. Multiplying by the flow rate and the natural log of the outer to inner diameter and dividing by the length of the engasser and stoichiometric correction, k for all three trials is computed to be 4.00 ± 0.11 and 5.26 ± 0.01 nmol/(min·cm· kPa) for steps 1 and 2, respectively. The former was determined to be 11.4% lower on average than the value computed from the equivalence points while the latter was 16.4% higher. This apparent discrepancy likely originates from errors in the reported λ values. Our data on 50 mM NaOH (based on the ordinate intercept of the step 1 fits) indicates a λ of 230.9 μS/ cm per meq/L, close to the reported value of 227.23 However, at ∼50 mequiv/L, we determined λ values of Na2CO3 and NaHCO3 to be 112 and 97.5 μS/cm per meq/L, significantly higher than that in the Dow report.23 Chromatographic Application. The following basic modes of anion exchange chromatography are possible with eluents made with a CO2 engasser: 10066


Article

Analytical Chemistry Table 1. Gradient Elution Conditions; Linear Ramps Connect the Successive KOH and pCO2,ext Pointsa gradient G1, [KOH] = 40 mM

a

gradient G2, [KOH] = 40 mM

gradient G2, [Na2CO3] = 40 mM

gradient G3/G4, pCO2,ext = 1380 kPa

dual gradient

time min

PCO2,ext kPa

time min

PCO2,ext kPa

time min

PCO2,ext kPa

time min

[KOH] mM

time min

PCO2,ext kPa

[KOH] mM

0 4 12 14 17 18 30

276 276 827 1380 1380 276 276

0 4 16 24 26 29 30 40

2650 2650 2340 1790 1380 1380 2650 2650

0 4 16 24 26 29 30 40

1330 1330 1170 896 689 689 1330 1330

0 5 20 23 26 31.5 32 37

21 21 25 30 45 45 21 21

0 8 18 20 26 29 30 40

172 172 552 552 1380 1380 172 172

7.5 7.5 24 24 60 60 7.5 7.5

The same data are presented in graphical form in Figures S11−S15 in the Supporting Information.

pCO2,ext the principal difference is the maximum CT level attainable. Especially for short engassers, with water flowing through the engasser, the effluent CO2 concentration is far from the equilibrium value (equilibrium C T equals HCO2*pCO2,ext, HCO2 being 0.365 mM/kPa). At pCO2,ext = 500 kPa, for example, the equilibrium CT is 183 mM. Previous experiments with Teflon AF tubes approximately half as permeable as the ones presently used indicated that at F = 0.5 mL/min, 5.5 and 160 cm long engassers reached only 4.3 and 63.1% of equilibrium.22 The present 5.5 cm engasser in deployment mode c (water influent @ 0.5 mL/min) had an effluent CT of 7.85 mM at pCO2,ext of 500 kPa. In contrast, in mode (a), CT can go up to the permeation rate (2.34 nmol/ (cm·min·kPa)) controlled value (as long as the effluent pH remains alkaline), which for that 5.5 cm engasser at F = 0.5 mL/min and pCO2,ext = 500 kPa translates to CT = 9.67 mM. Mode (d) shares with mode (b) constant pCO2,ext operation. This can be considered the simplest hardware option: the CO2 engasser can be operated simply connected to a CO2 cylinder and operated at a fixed pressure. As before for mode (a) versus (c), mode (d) allows greater CT than mode (b), the difference being significant for long engassers. Also, when the HEG is placed after the engasser, the degasser following the HEG (used to remove electrolytic gases) removes excess CO2 at eluent pH values < ∼8. Isocratic Separation. We show results for the separation of 19 anions ranging from fluoride to chromate with the HEG (several isocratic KOH levels: 18, 20, 22, 24, 30, 35 mM KOH) preceding the engasser (operated at 1380 kPa pCO2,ext, leading to ∼24 mM CT) in Figure S3. In general, higher KOH levels result in shorter analysis times but at the expense of resolution (data for 45 mM KOH were obtained but are not shown for this reason). In this isocratic mode, more than 1 h is needed to achieve a separation of nearly all the ions in this mix (bromide and chlorate are not resolved). This highlights the general problem of isocratic eluent systems, resolution comes at the expense of analysis time. Figures S4−S6 depict the relationship between the KOH concentration and the retention of various ions. Polyprotic acid derived anions selenite, arsenate, and phosphate behave differently from the others. As the pH increases, the effective charge on these anions increase at greater HEG KOH, resulting in greater retention despite increased eluent strength. In pure hydroxide eluent IC (whether isocratic or gradient), these anions are usually near completely dissociated; in the HCO3−/CO32− eluent systems there is considerable latitude for choosing where these ions will elute. Phosphate and arsenate, for example, may be separated as the mono-, di-, or trivalent species. The large difference in

extent may vary among analytes, there will be a general increase in eluent strength. This gradient region and direction is indicated as G1. Extant instrumentation will permit the generation of mixed (heretofore unexplored) hydroxide− carbonate eluents by pumping a carbonate solution through a hydroxide generator. The carbonate concentration could even be varied if a gradient pump was used. Note that there is relatively little change in pH in this eluent system. Past the point that the maximum eluent strength is reached, further addition of CO2 will cause carbonate to be converted to bicarbonate, and the eluent strength will decrease. So it is not meaningful to go across the entire width of the plot (G1 + G2). To run an effective gradient in the right half, one would start at some point in the right quartile of the plot and move leftward. This gradient G2 is the same as that previously mentioned for configuration (ii), where a carbonate eluent is pumped through the CO2 engasser that starts with a high pCO2,ext, which decreases during the run. This increase in carbonate/ bicarbonate ratio during the gradient results in a much greater pH change (and, thus, may substantially affect retention of polyprotic acid anions) than the G1 gradient. Note that toward the right end of the G2 zone, especially past the bicarbonate end point, a linear gradient in pCO2,ext does not, in practice, result in a linear increase of total carbonate species (CT) because the internal pCO2 level can no longer be neglected; and hence, the introduction rate is no longer linearly related to pCO2,ext. This is particularly true when long engasser lengths or low flow rates are to be used.22 The present experiments used very short (5.5 cm) engassers, however; the CO2 introduction rate decreased only by ∼2% when raised above the bicarbonate end point. The consequences of deployment mode (b) are shown in Figure 3b. The first part of the gradient (G3) results essentially in a linear increase in the bicarbonate concentration. Bicarbonate is a rather weak eluent and G3 thus represents a very weak gradient system, possibly useful only for very weakly retained ions. Continuing to the next gradient zone G4 results in the conversion of bicarbonate to carbonate with an increasing carbonate to bicarbonate ratio and rapid rise in eluent strength (and pH). When essentially all of the carbonate species are converted to CO32−, although additional hydroxide is added, most analytes are unlikely to experience a further increase in eluent strength with the additional increase in hydroxide concentration. The only difference in deployment modes (a) and (c) is the order of the devices: either hydroxide solution flows into the engasser or an aqueous CO2 solution flows into the HEG. For the same flow rate and the same engasser and maximum 10067


Article

Analytical Chemistry selectivity between HCO3− and CO32− also allows substantial increases in eluent strength with only modest changes in the KOH concentration. For example, the retention factor of the monovalent ions decreases on average 50% when HEG KOH is increased from 20 to 25 mM. Gradient Separations. All four possible gradients (G1− G4, Figure 3) were explored using deployment modes (a)−(d) as well as combined KOH and pCO2,ext gradients. In addition, gradient G2 was performed using manually prepared Na2CO3 as the feed instead of HEG KOH because no electrodialytic eluent generator is needed. Although gradients G3 and G4 were not particularly attractive individually, a gradient transitioning from G3 to G4 was much better. In total 9 different gradients were studied. Optimization was performed using deployments (a) and (d); the same gradient settings were used for (b) and (c), respectively, after switching the 10-port valve in Figure 1 to reverse the flow path through the HEG and engasser. Table 1 provides the gradients studied. Figure 4 compares several gradient separation modes; configuration (a) or (d) was used. With isocratic 40 mM

multiprotic anions like arsenate and phosphate to elute earlier than in G1. However, if the separation of only phosphate and arsenate is of interest, the difference in selectivity appears to be the greatest when they are triply charged (Figure S6) and they separate further apart in G1. The pK2 values for H3AsO4, H3PO4, and H2SeO3 are, respectively, 6.8, 7.2, and 8.3. A high pCO2,ext level maintains selenite singly charged, causing it to elute as early as chlorite (not shown in Figure 4). Running G2 with manually made Na2CO3 rather than HEG KOH and appropriate reduction in pCO2,ext achieves nearly the same results. Interestingly, due to the presence of impurities, baseline shift associated with manually made Na2CO3 is much greater. Due to the transition from a HCO3− form column to a CO32− form in G2, there is a tendency for ions to focus on the transition zone. The narrower peak of arsenate compared to phosphate in the G2 Na2CO3 gradient is a direct result of this focusing. Because of this transition, iodide and sulfate are difficult to resolve in G2 using HEG KOH. Gradient G3/G4 was therefore operated beginning with conditions as close to the transition as possible, with a small amount of CO32− already in the eluent. This resulted in selenite now eluting between phosphate and arsenate. Gradient G3/G4 provides separation abilities similar to G2 but the latter was simpler to optimize and more forgiving with respect to the HCO3−−CO32− transition. There is a larger concentration of HCO3− present in the G2 gradient when the transition to CO32− begins; relative to HCO3−, the dependence of eluting power of CO32− on its concentration is related to √C. A combination of G3 and G2 gradients may provide the best eluent program for the separation of a broad range of analytes; in the present system, the same ends are achieved by running dual gradients as detailed below. The best operating conditions in the dual gradient systems involved maintaining a constant ratio of KOH to pCO2,ext with the ratio high enough so the eluent is alkaline during the entire separation. Although [CO32−] concentrations at the beginning of the run are similar to those in G1, better separations of the early eluting ions were observed here. In G1, initial [OH−] is sufficiently high such that it is the dominant eluent ion. In the dual gradient run as conducted, selenite, phosphate, and arsenate are all completely ionized and elute later than in G2 or G3/G4. Comparison of the background conductance traces (Figure S7) of the various methods with respect to noise and shift is provided in Table 2. Noise was measured in consecutive 30 s long segments in the isocratic regions (both at low and high eluent strength ends). While G1 and the dual KOH/pCO2,ext gradients had the lowest starting noise, it increases with the increasing CO2. Gradient G2 had the lowest and most consistent background noise at 7 nS/cm, while the gradient baseline shift between the start and the end of the gradient was the lowest for G3/G4 (0.71 μS/cm). The gradient baseline shift for G2 was nearly as good but negative (−0.94 μS/cm). Except when manually prepared Na2CO3 was used, baseline noise was comparable to suppressed HEG gradient systems (typ. specifications 1−3 nS/cm at modest OH− concentrations, isocratic). In the present system, the noise primarily originates from the incomplete removal of CO2; we have observed lower backgrounds with even greater concentrations of CO 2 previously with a similar CRD device.22 Why the HEG Should Precede the Engasser. As would be evident from Figure 1, the 10-port valve makes it simple to switch the order in which the fluid goes through the HEG and

Figure 4. Comparison of all gradient modes using KHCO3/K2CO3. The various gradient conditions are presented in Table 1. Letters in parentheses correspond to the deployment modes. The analytes (with concentrations in mg/L in parentheses) are identified as follows: 1, fluoride (3); 2, acetate (20); 3, formate (6); 4, chlorite (10); 5, bromate (20); 6, chloride (3); 7, nitrite (15); 8, bromide (25); 9, chlorate (25); 10, nitrate (12); 11, selenite (40); 12, iodide (35); 13, phosphate (30); 14, arsenate (40); 15, sulfate (10); 16, selenate (20); 17, oxalate (20); 18, thiosulfate (20); 19, chromate (35).

KOH and running a pCO2,ext gradient, gradient G1 resulted in poor resolution of early eluting ions even with the lowest CO2 pressure possible being used in the beginning of the run. However, using the same KOH concentration and a high pCO2,ext in the beginning and slowly reducing pCO2,ext to create gradient G2 resolved the early eluting ions and allowed 10068


Article

Analytical Chemistry Table 2. Blank Chromatogram Parameters of Various Chromatographic Modesa low strength:

relevant interval, min noise, nS/cm high strength: noise, nS/cm interval (min) starting background, μS/cm shift, μS/cm

gradient G1

gradient G2, KOH

gradient G2, Na2CO3

gradient G3/G4

dual gradient

0.2−2.5 2.6 ± 1.4 20.0−22.5 6.1 ± 2.6 1.87 1.56

0.2−2.5 7.0 ± 2.5 31.2−33.5 7.2 ± 3.7 3.63 −0.94

0.2−2.5 4.0 ± 1.2

0.2−2.5 12 ± 3.6 31.5−34.1 7.5 ± 3.8 2.76 0.71

0.2−2.5 1.53 ± 0.58 30.0−32.5 8.7 ± 4.7 1.32 1.80

b

10.2 8.80

a Noise measurements were taken where applicable in isocratic regions at low and high eluent strength. Noise is reported as ±1 std dev. bBackground changing too steeply to measure noise.

formate can be optimized, it will provide a route to formic acid−formate based eluents popular with electrospray ionization mass spectrometric detection.

the engasser. Separations using G1, G3/G4, and the combined KOH/pCO2,ext gradients with the engasser ahead (modes (b) and (c)) were nearly identical to those with the HEG ahead (modes (a) and (d), respectively, see Figures S8−S10). While nearly the same separation was achieved in each case, modes (b) and (c) consistently exhibited higher background and greater baseline shift than (a) or (d), respectively. Gradient G2 showed the largest change upon putting the engasser upstream (Figure 5). The background increased by ∼30 μS/cm, and the

■

CONCLUSIONS A gas permeable membrane was used for controlled introduction of CO2 into a flowing alkaline stream for the production of HCO3−/CO32− eluents. Low noise gradients with minimal gradient baseline shifts can be attained using multiple approaches that offer particular advantages to tailor the separation of polyprotic acid anions.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.7b02808. Engasser construction details, conductance and pressure data with pCO2,ext sweep at a fixed [NaOH], chromatograms: fixed KOH- varying pCO2,ext, retention factors as a function of [KOH] for above; baseline for different gradient systems, comparison of HEG ahead vs engasser ahead; see text for details (PDF).

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. Fax: 817-272-3808.

Figure 5. Comparison of gradient G2 wherein (a) the HEG precedes the engasser and (c) the engasser precedes the HEG. The gradient conditions are presented in Table 1. The analytes (with concentrations in mg/L in parentheses) are identified as follows: 1, fluoride (3); 2, acetate (20); 3, formate (6); 4, chlorite (10); 5, bromate (20); 6, chloride (3); 7, nitrite (15); 8, bromide (25); 9, chlorate (25); 10, nitrate (12); 11, selenite (40); 12, iodide (35); 13, phosphate (30); 14, arsenate (40); 15, sulfate (10); 16, selenate (20); 17, oxalate (20); 18, thiosulfate (20); 19, chromate (35).

ORCID

Purnendu K. Dasgupta: 0000-0002-8831-7920 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the U.S. National Science F o undation (CHE- 1506 572 ), NASA (Grant N o . NNX15AM76G), Dionex Corporation (now part of Thermo Fisher Scientific) and the Hamish Small Chair endowment at the University of Texas at Arlington.

retention time of all ions decreased. It is well-known that dissolved CO2 can be electroreduced to formate.24 To confirm, the system was operated without the column, and the suppressed effluent collected after CO2 removal. With 690− 2760 kPa CO2 and 10−40 mM KOH, the column effluent was found to contain 0.3−4.2 μM formic acid by standard IC analysis. The concentration increased with either pCO2 or the HEG current. The values measured would be consistent with the behavior observed in Figures S8−S10, but much more formic acid/formate is likely produced under the conditions in Figure 5. Operational modes (b) and (c) are therefore not recommended, and practical implementation does not require the 10-port valve. On the other hand, if the production of

■

REFERENCES

(1) Small, H.; Stevens, T. S.; Baumann, W. C. Anal. Chem. 1975, 47, 1801−1809. (2) Small, H.; Solc, J. Proceedings, Conference on The Theory and Practice of Ion Exchange; Society of Chemical Industry: London, 1976. (3) Mulik, J.; Puckett, R.; Williams, D.; Sawicki, E. Anal. Lett. 1976, 9, 653−663. (4) Anderson, C. Clin. Chem. 1976, 22, 1424−1426. (5) Strong, D. L.; Dasgupta, P. K.; Friedman, K.; Stillian, J. R. Anal. Chem. 1991, 63, 480−486.

10069


Article

Analytical Chemistry (6) Shen, G.; Lu, Y.; Chen, F.; Zhang, F.; Yang, B. Talanta 2016, 159, 143−147. (7) Dionex ERS 500 Suppressor 031956, Revision 10, August 2015; https://tools.thermofisher.com/content/sfs/manuals/Man-031956IC-Dionex-ERS-Suppressor-Man031956-EN.pdf. (8) Pohl, C. A.; Srinivasan, K.; Omphroy, B. U.S. Patent 9,400,268, July 26, 2016. (9) Sunden, T.; Cedergren, A.; Siemer, D. D. Anal. Chem. 1984, 56, 1085−1089. (10) Siemer, D. D.; Johnson, V. J. Anal. Chem. 1984, 56, 1033−1034. (11) Shintani, H.; Dasgupta, P. K. Anal. Chem. 1987, 59, 802−808. (12) Saari-Nordhaus, R.; Anderson, J. M. J. Chromatogr. A 2002, 956, 15−22. (13) Douglas, D. R.; Saari-Nordhaus, R.; Despres, P.; Anderson, J. M. J. Chromatogr A 2002, 956, 47−51. (14) Anderson, J. M.; Saari-Nordhaus, R.; Benedict, B. C.; Sims, C. W.; Gurner, Y.; Pham, H. A. U.S. Patent 6,444,475. September 3, 2002. (15) Bose, R.; Saari-Nordhaus, R.; Anderson, J. M. J. Chromatogr. A 2002, 956, 71−76. (16) Bose, R.; Saari-Nordhaus, R.; Sonaike, A.; Sethi, D. S. J. Chromatogr A 2004, 1039, 45−49. (17) Rocklin, R. D.; Pohl, C. A.; Schibler, J. J. Chromatogr. A 1987, 411, 107−119. (18) Dionex CRD Carbonate Removal Device; https://tools. thermofisher.com/content/sfs/manuals/Man-065068-CRDCarbonate-Removal-Device-Man065068-EN.pdf. (19) Masunaga, H.; Higo, Y.; Ishii, M.; Maruyama, N.; Yamazaki, S. J. Chromatogr. A 2015, 1392, 69−73. (20) Doury-Berthod, M.; Giampaoli, P.; Pitsch, H.; Sella, C.; Poitrenaud, C. Anal. Chem. 1985, 57, 2257−2263. (21) Midgley, D.; Parker, R. L. Talanta 1989, 36, 1277−1283. (22) Shelor, C. P.; Dasgupta, P. K. J. Chromatogr. A. 2017, na. (23) The Dow Chemical Company, Filmtec Membranes, Tech Manual Excerpt; http://msdssearch.dow.com/ PublishedLiteratureDOWCOM/dh_003c/0901b8038003ccb2. pdf?filepath=liquidseps/pdfs/noreg/609-02127.pdf&fromPage= GetDoc. (24) Kapusta, S.; Hackerman, N. J. Electrochem. Soc. 1983, 130, 607− 613.

10070
