Anal. Chem. 2007, 79, 5690-5697
Postcolumn Concentration in Liquid Chromatography. On-Line Eluent Evaporation and Analyte Postconcentration in Ion Chromatography Masaki Takeuchi and Purnendu K. Dasgupta*
Department of Chemistry and Biochemistry, University of Texas at Arlington, Arlington, Texas 76019-0065 Jason V. Dyke
Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409-1061 Kannan Srinivasan
Dionex Corporation, 1228 Titan Way, Sunnyvale, California 94086
On-line sample concentration by evaporation through a narrow-bore membrane tube is described. The device can be deployed just prior to the detector or the sample may be concentrated prior to injection. As solution flows through a solvent-permeable membrane tube, (heated) drying gas (nitrogen/air) flows outside it to remove the solvent. The removal rate increases with increasing sample residence time, drying gas flow rate, and temperature. Various membranes and three concentrator designs (a rectangular maze, a serpentine and a filament-filled helix, the last performing the best) were fabricated and tested for post- and preseparation applications in suppressed anion chromatography. An order of magnitude concentration factors are readily obtained. The present system involves active mass transport radially outward through the walls of a tube. This is a system in which many of the traditional paradigms of flow through a tubular conduit no longer hold true. Because the flow rate continuously varies along the tube, residence time does not scale linearly with residence volume or conduit length. The effects of such mass transport on the parabolic velocity profile of laminar flow remain unknown. Preconcentration is a term common in the parlance of analytical chemistry: it appears in the Web of Science database >800 times in 2006 alone. The term postconcentration in contrast is an unfamiliar one: there are no studies that deal with concentrating an eluite after separation has been accomplished prior to detection (albeit in multidimensional separation techniques the eluite may be “focused” prior to reseparation). However, removal of the solvent is indeed undertaken in some detection approaches in liquid chromatography (LC). The first involves the LC-mass spectrometry interface.1,2 The second is evaporative light scattering * Corresponding author. E-mail:
[email protected]. (1) Cappiello, A.; Famiglini, G.; Palma, P.; Siviero, A. Mass Spectrom. Rev. 2005, 24, 978-989. (2) Zwiener, C.; Frimmel, F. H. Anal. Bioanal. Chem. 2004, 378, 851-861.
5690 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
detection.3 In both approaches, the detector is mass rather than concentration-sensitive and detection is accomplished in the gas phase. There are no studies on concentrating an eluite band by preferential solvent removal and liquid-phase detection with a concentration-sensitive detector. Recently, capillary and intermediate-scale chromatographic equipment has been commercialized. The new generation of low cell volume, concentration-sensitive detectors display the same (or even better) concentration limits of detection compared to their older larger cell volume counterparts. Nevertheless, the majority of LC practice still centers on 4.0-4.6-mm-i.d. columns with eluent flow rates of ∼1 mL/min. If only 10% of the column effluent is sent to a low cell volume detector, there will be no apparent deterioration of performance. Obviously, an order of magnitude gain in performance can result if, rather than discarding 90% of the column effluent, the column effluent is postconcentrated by a factor of 10 prior to detection. If detection is still to be performed in the liquid phase, nebulization, commonly used for solvent evaporation, cannot be used. Also, solvent removal must not lead to excessive band dispersion that will undo the separation. We have previously concentrated analytes with a tubular membrane containing a stationary or flowing liquid while an analyte gas flows around the membrane, resulting in collection of the soluble gas in the liquid as well as evaporative loss of the liquid.4,5 If a column effluent is subjected to selective solvent evaporation as the liquid flows through a small-bore membrane tube prior to detection, a viable postconcentration approach may result. Obviously, such an approach may be equally well used prior to injection; there dispersion is not a consideration. “Membrane desolvation” approaches are used for removal of the solvent prior to nebulization and introduction into a plasma.6 Devices consisting (3) Megoulas, N. C.; Koupparis, M. A. Crit. Rev. Anal. Chem. 2005, 35, 301316. (4) Lindgren, P. F.; Dasgupta, P. Κ. Anal. Chem. 1989, 61, 19-24. (5) Genfa, Z.; Dasgupta, P. K.; Dong, S. Environ. Sci. Technol. 1989, 23, 14671474. (6) Jorabchi, K.; Kahen, K.; Lecchi, P.; Montaser, A. Anal. Chem. 2005, 77, 5402-5406. 10.1021/ac0703799 CCC: $37.00
© 2007 American Chemical Society Published on Web 07/04/2007
Table 1. Specification of Concentratorsa
type
membrane tube
tube i.d., µm
tube o.d., µm
tube wall thickness µm
tube inner surface, mm2
residence volume, µL
inserted filament
inserted filament diam, µm
device designation
maze maze maze maze maze helical helical serpentine
Nafion Nafion ST708Bb ST741Ab ST708Cb Nafion Nafion Nafion
229 229 127 127 127 229 229 229
330 330 330 381 483 330 330 330
51 51 102 127 178 51 51 51
223 223 124 124 124 223 223 223
25.5 17.7 7.9 7.9 7.9 17.7 16.0 17.7
none nylon none none none nylon acrylic nylon
n/ac 127 n/a n/a n/a 127 140 127
I II III V IV VI VII VIII
a The length of all membrane tube was 62 cm. b Radiation-grafted PTFE-based cation-exchange membrane provided by Dionex Corp. c n/a, not available.
of a porous poly(tetrafluoroethylene) (PTFE) tubular membrane, kept in an oven, and bathed externally by a counterflow of argon, are commercially available.7 However, there are analyte loss problems.8-10 Bishop and Mitra11 achieved significant solvent removal/sample preconcentration by passing the sample through either polar (ion exchange) or nonpolar hollow fibers (when hexane was the sample solvent), initially reporting collection in autosampler vials and later12 directly into the loop of a loop injector. However, the authors do not mention that any ions present in the sample accumulate on the ion exchange fiber and permanently reduce its ability of solvent removal. We report here the results of such studies as applied to suppressed anion chromatography, a technique we are most familiar with. It represents both an opportunity and a challenges removing the matrix solvent while retaining solute ions is not a major problem; however, water has the highest boiling point, specific heat, and latent heat of evaporation compared to other solvent components commonly used in LC. EXPERIMENTAL SECTION Initial Studies on Membrane Tube-Based Concentrators. Initial studies on the performance of different membrane devices were carried out with 10-cm lengths of each membrane. Figure S1 in Supporting Information (SI) shows the schematic of such concentrators. Details of the design are given in SI; The test system is schematically shown in Figure S2. In essence, water was pumped through the membrane tube (kept in a heated enclosure) at 5-25 µL/min while 0.1-2 L/min preheated air or N2 flowed outside the membrane. The response to injected 200 nL of 1 mM NaNO3 was monitored with a capillary-scale UV or conductivity detector. The membrane tubes used in this experiment and their characteristics are given in SI. Maze Concentrator. Figure S3 in SI schematically shows the maze-type concentrator. These devices were built with the Nafion tubes with an active length of 62 cm. Details of construction are (7) Cetac Technologies Inc., Omaha, NE. U6000AT+ Ultrasonic Nebulizer with Membrane Desolvator. http://cetac.com/prods/usn/u6000/extras/ u-6000at_brochure.pdf. (8) Kahen, K.; Jorabchi, K.; Montaser, A. J. Anal. At. Spectrom. 2006, 21, 588591. (9) Βendahl, L.; Gammelgaard, Β. J. Anal. At. Spectrom. 2005, 20, 410-416. (10) Juresa, D.; Kuehnelt, D.; Francesconi, K. A. Anal. Chem. 2006, 78, 85698574. (11) Bishop, E. J.; Mitra, S. J. Chromatogr. A 2004, 1046, 11-17. (12) Bishop, E. J.; Mitra, S. J. Pharm. Biomed. Anal. 2005, 37, 81-86.
in SI; in essence, the Nafion tube is strung through a narrow maze in an Al block with 1/8-in.-wide and 1/4-in.-deep channels. A cover plate provides closure, and the block is heated with planar siliconecoated heaters placed on the top and bottom of the Al block. Liquid flows through the tube, gas flows through the maze countercurrent, and block temperature is controlled by a temperature sensor placed in the gas exit point. In order to evaluate water evaporation rate from the membrane tube, 10 µM sulfuric acid was pumped through the concentrator. Water evaporation rate, We (µL/min), was calculated as
We ) Fi(1 - Ci/Ce)
(1)
where Fi is the influent flow rate (100, 200, 300, 400, and 500 µL/ min), Ci is the influent sulfuric acid concentration, and Ce is the effluent sulfuric acid concentration calculated from the measured conductance. To ensure accuracy, sulfuric acid and pure water were alternately pumped using a two-syringe system. Effect of the drying gas fIow was evaluated using air or N2 at a constant 5 SLPM flow with the flow accomplished both by pressure and by vacuum. Helix Concentrator. Figure S6a shows the schematic of helical membrane tube-based concentrator. The membrane is made into a filament-filled helix, largely following the senior author’s early work.13,14 The helix is jacketed with a glass tube (through which drying gas flows) and wrapped externally with a heating filament. Details of construction are given in SI; a photograph is shown in Figure S7. Serpentine Concentrator. Figure S6b shows the schematic of the serpentine membrane tube-based concentrator. A nylon monofilament (127-µm diameter, 62 cm long) filled Nafion membrane tube (229-µm i.d., 330-µm o.d., 62 cm long) as above was woven on a polypropylene mesh with 0.5-mm grids in a 3-D serpentine pattern15-17 as shown. See SI for construction details. Chromatographic Test Arrangements. An ion chromatograph (IC) was operated with the maze-, helix-, and serpentinetype concentrators. Table 1 shows the specific details for each (13) Dasgupta, P. K. Anal. Chem. 1984, 56, 96-103. (14) Dasgupta, P. K. Anal. Chem. 1984, 56, 103-105. (15) Curtis, M. A.; Shahwan, G. J. LC-GC. Mag. 1988, 6, 158-164. (16) Dasgupta, P. Κ. J. Chromatogr. Sci. 1989, 27, 422-448. (17) Waiz, S.; Cedillo, Β. Μ.; Jambunathan, S.; Ηοhnholt, S. G.; Dasgupta, P. Κ.; Wolcott, D. Κ. Anal. Chim. Acta 2001, 428, 163-171.
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
5691
Figure 1. Setup for running chromatograph with maze/helical/ serpentine-type concentrator
concentrator tested in this manner. The IC system was configured for use as a 2-mm system (GP40 gradient pump, 150 µL/min; EG40 eluent generator, 21.5 mM potassium hydroxide; LC30 chromatography oven, 30 °C; AG11-HC 2 mm guard column + AS11-HC 2 mm separation column; ASRS ULTA-II 2-mm anion suppressor, suppressor current 50 mA, external water mode (200 µL/min); CD20 conductivity detector, all from Dionex Corp.). A sample consisting of a common anion test standard (0.1 ppm chloride, 0.1 ppm nitrite, 0.2 ppm sulfate, and 0.2 ppm nitrate, 25 µL) was used for this system. The concentrator was placed between the suppressor and conductivity detector as shown in Figure 1. Water evaporation rate was calculated as difference between influent and effluent flow rate (both were gravimetrically measured). Concentration factor, CF, based on peak height/area was calculated as
CF ) Cw/Cwo
(2)
where Cw and Cwo are the analyte peak height or area (area computed in terms of signal intensity‚volume, rather than signal intensity‚time, as the flow rates vary) with/without concentrator, respectively. Preinjection evaporative concentration of common anions was also studied. A maze-type concentrator was exclusively used. The sample (the same aforementioned common anion standard) was continuously pumped through the concentrator by a syringe pump at an influent flow rate of 135-200 µL/min, to fill a 25-µL injection loop, which was periodically injected in to the IC system described above. RESULTS AND DISCUSSION Effect of Membrane Type and Operating Variables. The performance of different membrane tubes and the effect of operating parameters such as influent flow rate, drying gas flow rate, and temperature was evaluated using the short (10 cm) straight tubular membrane concentrators (Figure S1). Results for six different membrane devices are shown in Figure 2a under otherwise comparable conditions. One radiation-grafted cationexchanger membrane tube (A, ST730), showed negligible evaporation; however, the other (B, ST728) clearly showed evaporative concentration. Note that some dispersion is unavoidable when an evaporation device is connected; as such, if there is little or no 5692
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
evaporation, peak heights will actually decrease. This is indeed observed for the Nafion-coated Celgard tube (D). The graft level (and hence the ion-exchange capacity) of ST730 was less than half of ST728. Further studies with radiation-grafted cation exchangers (vide infra) were limited to ones with higher ionexchange capacities, and all showed substantial evaporative concentration. Evaporative concentration with the COOH-functionalized polypropylene tube is discernible but small. This coating technique results in relatively thick coatings and little evaporation occurs. In contrast, both Nafion tubes show significant evaporative concentration. The extent of evaporative concentration was much greater for the larger bore tube. This is to be expected inasmuch as compared to the smaller membrane tube (E, 89-µm-i.d., 57-µm wall), the larger tube (F, 229-µm-i.d., 51-µm wall) had about the same membrane thickness, 2.6× greater inner surface area, and 6.6× greater residence volume. In fact, the data for the larger tube are presented with a smaller drying gas flow rate because with the same gas flow rate of 5 SLPM, liquid evaporation was so extensive that at best the liquid reached the detector intermittently, resulting in severe noise. It is important to note here that many of the paradigms that we hold as axiomatic regarding flow through tubular conduits break down when evaporative loss occur through the tube walls. This is because flow rate through the tube changes along the length of the tube and residence time is no longer linearly related to the residence volume. For the larger tube, volumetric evaporation rate per unit length is larger than that for the smaller tube. As a result, flow rate decreases at a steeper rate than that in the smaller tube, and under otherwise identical conditions, residence time in the larger tube is greater than what the ratio of the internal volumes will indicate. In a similar vein, for tubes of identical bore, residence time is more than linearly proportional to the length of the tube. The departure from the linear model is obviously greater as a greater fraction of the influent liquid is subjected to loss, for a situation where the influent flow is large and only a small portion of this liquid evaporates, the system will behave close to the conventional model. Thus, in the low influent flow rate range of 5-25 µL/min, with the extent of evaporation of the influent flow being significant, the ratio of peak height with/without the concentrator increased not linearly but exponentially as the influent flow rate was decreased. This is readily visible in the data of Figure 2b. Note that, while conductance was monitored in Figure 2a, the other panels in Figure 2 used absorbance detection; any concentrationsensitive detector will similarly respond to the concentration gains attained through evaporative concentration. The primary purpose of gas flow is to remove the water vapor that emanates from the membrane tube. At 30 °C, each liter of water vapor-saturated N2 holds ∼8.6 µL of liquid water. As such, flow rates higher than ∼1.2 SLPM aids primarily to diffuse away the water vapor emanating from the membrane surface. As such, this behavior should be governed by a relationship akin to the Gormley-Kennedy equation for flow through a cylindrical tube18 where a plateau is exponentially approached. This behavior is indeed seen in Figure 2c over a drying gas flow rate range of 0.1-2 SLPM. (18) Gormley, P. G.; Kennedy, M. Proc. R. Ir. Acad. Sci. Sect. A 1949, 52, 163169.
Figure 2. Comparison of sodium nitrate peak with (red) and without (black) 10-cm membrane-based concentrator. Sodium nitrate was injected 3 times for each experimental condition. All experiments: unless otherwise noted, 10 µL/min influent flow, t ) 30 °C, 5 SLPM N2 flow, Nafion, 89-µm i.d. (a) Effect of membrane tube material: (A) cation exchanger Cex ST730, (B) cation exchanger Cex ST728, (C) COOH-functionalized Accurel PP, (D) Nafion-coated Celgard, (E) Nafion, 89-µm i.d., (F) Nafion, 229-µm i.d. operated with 3 SLPM N2. (b) Effect of influent flow rate. 5-25 µL/min influent flow rate as indicated, 1 SLPM N2 flow throughout. (c) Effect of N2 flow rate. 0.1-2 SLPM N2 flow rate, as indicated. (d) Effect of temperature, 24-80 °C, as indicated; 1 SLPM N2 flow.
Temperature plays of course perhaps the most vital role in determining the amount of the water lost through evaporation. The rate of loss of water should be linearly related to the water vapor pressure, and the vapor pressure of water increases rapidly with temperature. With increasing temperature, the extent of evaporative concentration increased dramatically as shown in Figure 2d. The ratio of the sodium nitrate peak height with/ without concentrator with the vapor pressure of water exhibited an excellent linear correlation (r2 ) 0.9920, t ) 22-80 °C). Choice of a Membrane. Several membranes other than those already discussed above were investigated for evaporative preconcentration for which data are not presented in detail. This included (a) a radiation-grafted PTFE anion exchanger (Aex 88, 72-µm i.d., 216-µm o.d., Dionex Corp.), (b) a porous Celgard polypropylene membrane (200-µm i.d., 250-µm o.d., 40% surface porosity), and (c, d) similar polypropylene membranes (209-µm i.d., 263-µm o.d. and 400-µm i.d., 460-µm o.d., coated with a very thin layer (∼1 µm) of plasma polymerized poly(dimethylsiloxane), similar to silicone-coated Celgard membrane tubes described in ref 19. At 30 °C and 5 SLPM N2 flow, the water loss rates were respectively (a) 0.22, (b) 0.23, (c) 0.083, and (d) 0.19 µL/min/ cm, all of which are substantially lower, for example, than 0.93 µL min-1 cm-1 observed for the hollow Nafion tube (229-µm i.d., 330-µm o.d.). Of these, only the anex membrane tube has a much smaller bore than the comparison standard, and this may produce (19) Ullah, S. M. R.; Adams, R. L.; Srinivasan, K.; Dasgupta, P. K. Anal. Chem. 2004, 76, 7084-7093.
comparable evaporation if used in a comparable dimension. Anionexchange membranes cannot of course be used for evaporative concentration in suppressed anion chromatography as analyte anions will be captured. However, such a membrane will be of interest in dealing with cationic (as well as neutral) analytes. For the cation exchangers, it appears that the extensive solvation of the hydrogen ion present in such an exchanger plays a paramount role; this will be examined again in a later section. Strong acids in ionized form cannot pass through a negatively charged cationexchange membrane, although weak acids in molecular form are not subject to the Donnan barrier. Thus, cation-exchange membranes would be the basis of preferred concentration devices in anion analysis, anyway. Porous membranes can be used of course for evaporative concentration as discussed in the introduction;4,5,7 the extent of transport to and from the membrane is expected to be a function of fractional surface porosity, pore tortuosity, and membrane thickness.20 Interestingly, the evaporation rate through the porous Celgard membrane was not as high as that of the better cation-exchanger membranes. This type of membrane also requires maintenance in that pores can be gradually blocked by evaporating solids;18 coating with a thin asymmetric layer, e.g., of siloxane, may obviate pore blockage but it did not improve water transport rates. Henceforth we used cation-exchange membranes. Water Evaporation as a Function of Influent Flow Rate. One would expect that, in the absence of significant dissolved solute concentrations, as long as the influent rate is sufficiently (20) Dasgupta, P. K. ACS Adv. Chem. Ser. 1993, 232, 41-90.
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
5693
high, the absolute amount of water evaporated would be the same, regardless of the precise influent flow rate. This was largely observed for a 62-cm-long Nafion tube (229-µm i.d., 330-µm o.d.) in the maze concentrator for an influent flow rate range of 100300 µL/min. Detailed results are presented in SI in Figure S8. There was no consistent pattern of evaporation rate change with the influent flow rate, but it increased monotonically with temperature as in Figure 2d. Over the flow rate range of 100-500 µL/min, the average ( sd (n ) 3) water evaporation rates (in µL/min) were as follows, respectively: 23.5 ( 9.2, 39.6 ( 6.7, 68.9 ( 5.6, 97.6 ( 9.6, 136.7 ( 6.6, 159.6 ( 2.7, and 187.4 ( 4.6 at 22, 30, 40, 50, 60, 70, and 80 °C (the latter four data points do not include the 100 µL/min influent flow rate, for obvious reasons). These data fit a linear pattern:
evaporation rate (µL/min) ) 2.92 ( 0.08 (t °C) - (44.94 ( 4.44),
r2 ) 0.9960 (3)
Stripping Gas Flow. Even if the mass flow rate and temperature of the drying gas is maintained the same, it is interesting that the exact arrangement for the gas flow makes a difference whether the drying gas flow is supplied by a positive pressure or by applying a vacuum. In the latter case, outside the membrane, the pressure is negative relative to ambient, increasing the pressure differential across the membrane and thus enhancing evaporation. Stripping gas flow arrangement and detailed data are respectively presented in SI in Figure S9 and Table S1. The difference is easily perceptible but it is not large. For the same maze concentrator described in the paragraph above, operating with 100 µL/min influent flow and t ) 40 °C, 5 SLPM N2 flowing under pressure and under vacuum resulted in respective evaporation rates of 64.2 and 72.9 µL/min. However, all other data reported here were obtained under pressurized gas flow, as the necessary arrangement is obviously simpler. Ionic Form of the Membrane Makes a Large Difference. In a suppressed IC system, the concentrator device can be placed after the column, either before or after the suppressor, prior to the detector. The advantage of placing it upstream of the suppressor is that one is dealing with an alkaline eluent stream with analyte bands that are present as salts and there is no possibility whatsoever of any analyte loss. Needless to say, the precise original ionic form the cation-exchange membrane of the concentrator will be immaterial as it will rapidly be converted to the ionic form of the eluent cation. Interestingly, no increase in peak heights was observed at all when the maze-type concentrator described above was used in a chromatographic system and placed before the suppressor, even with an 80 °C device temperature. Other experiments confirmed that the membrane must be in H+ form for any significant evaporative concentration to be observed. This suggests that the high affinity of the proton for water and the ability of the positive charge to easily tunnel from one water cluster to another21 must substantially account for the high evaporation rates observed for these membranes, which are indeed greater than those observed for comparable porous membranes. Chromatographic Application. Postcolumn Deployment of a Maze/Helix/ Serpentine Concentrator. The water evapora(21) Yeager, H. L. ACS Symp. Ser. 1982, 180, 49-64.
5694 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
Figure 3. Comparison of water evaporation rate between maze-, helical-, or serpentine-type concentrator at 150 µL/min, 5 SLPM of nitrogen. Water evaporation rate was calculated as (influent flow rate) - (effluent flow rate). See Table 1 for device specifications.
tion rates for maze, helix, and serpentine concentrators are shown in Figure 3 as a function of device temperature with a constant membrane length of 62 cm, a drying gas flow of 5 SLPM N2, and an influent flow of 150 µL/min. In all cases, if simply operated at room temperature without active thermal control, the exit gas temperature was several degrees cooler than the entry (room) temperature due to adiabatic evaporative cooling. The water evaporation rate increased in the order serpentine, maze, and helix concentrators. In the first design, a good portion of the membrane is in contact with the support grid and evaporation from this area is inhibited. The serpentine design leads to good mass transfer in the flowing liquid, leading to good band dispersion characteristics,15 but previous applications did not have any need for good mass transfer outside the tube, which, obviously is a limitation in the present application. The filament-filled helix also leads to leads to good mass transfer in the flowing liquid and hence good band dispersion characteristics,11,12 but externally it is suspended in space and mass transfer away from the membrane is not inhibited. Figure 4 shows the chromatogram with nylon monofilamentfilled Nafion helix, which demonstrated the highest water evaporation rate. (Note however, the radiation grafted cation exchangers like ST708B have a much smaller bore and if we were to compare on the basis of induced dispersion, these will outperform the Nafion tubes.) The peak heights of chloride, nitrite, sulfate, and nitrate increased dramatically with temperature. However, the peak widths (in terms of time) also increased and there was deterioration of resolution. Part of this is due to the relatively longand large-bore (30 cm, 0.254 mm) transit conduit between the evaporator exit and the detector. It is also useful to note that the carbonate peak decreases in the membrane concentration process and essentially disappeared at the higher temperatures, as shown in the inset of Figure 4. It is well-known that the broad carbonate peak, often present in a large concentration in samples exposed to ambient air, poses a problem in the trace determination of
Figure 4. Chromatogram with postcolumn concentration with nylon filament-filled Nafion helix concentrator (Table 1, device VI), chromatographic flow rate 150 µL/min, and 5 SLPM N2 drying gas. (a) Without concentrator, (b) gas exit temperature of 14 , (c) 20 , (d) 25 , (e) 30 , (f) 35 , (g) 40 , (h) 45 , and (i) 50 °C. (1) Chloride, (2) nitrite, (3) carbonate, (4) sulfate, and (5) nitrate. The inset graph shows the response around the original carbonate peak.
certain inorganic anions with IC. In particular, when atmospheric trace gases are measured with IC, the broad tailing of carbonate peak covers the response of other ions that elute in this region.22 The concentrator therefore is useful also for reducing carbonate response. Clearly, further efforts need to be made to make smaller dispersion concentrators and also to avoid postconcentrator dispersion by locating the detector as close as possible thereafter. Inexpensive computing power and adaptive logic-based algorithms increasingly rule the practice of analytical sciences. Because the drying gas stream can very readily cooled or heated, we do not think it far-fetched that a particular separation can be tuned by programming the postconcentrator temperature. While Figure 4 has depicted the chromatograms with the traditional time units as abscissa, this does not give one the whole picture when the individual flow rates through the detector from the top to the bottom chromatogram change by a factor of 6. Figure 5 shows the same chromatograms with a volumetric abscissa scaling. Interestingly, if one computes the peak resolution of sulfate and nitrate between 14 and 30 °C, the resolution in fact remains unchanged. Band dispersion originates in the parabolic velocity profile of laminar flow where the velocity at the wall boundary layer is zero. How and to what extent the velocity profile in a tube, which has significant mass flux to the wall, is altered from traditional laminar flow is an interesting question; to our knowledge, such a system has never been studied. Concentration Factor Based on Peak Height and Area Ratios versus Influent/Effluent Flow Rate Ratios: Loss of Analyte. The concentration factor CF (eq 2) can be calculated in (22) Boring, C. B.; Al-Horr, R.; Genfa, Z.; Dasgupta, P. K. Anal. Chem. 2002, 74, 1256-1268.
Figure 5. Chromatogram of Figure 4, plotted with volume abscissa.
Figure 6. Relationship between the peak height and area ratios vs the influent to effluent flow rate ratio.
terms of either height or area ratios. Figure 6 shows that while the height-ratio-based CF for the nitrate peak rapidly departs from the gravimetrically measured influent/effluent flow rate ratio due to dispersion, the area ratio (where the area is based on volume; see eq 2) remains linear with the influent/effluent flow rate ratio, indicating that no loss of nitrate occurred during evaporative concentration. This is clearly not the case for carbonate, which produces the weak volatile carbonic acid in the suppressor; it can be totally lost during such concentration. Area ratio data are also plotted in Figure 6 for an analyte derived from an acid from intermediate pKa, nitrite; these data do not indicant any significant loss of nitrous acid. It remains to be established Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
5695
Figure 8. Time variation of concentration factor based on nitrate peak height with maze-type concentrator. Figure 7. Chromatogram of precolumn concentration with Nafion membrane tube-based maze-type concentrator at 5 SLPM of nitrogen, 60 °C. (a) Without concentrator and (b) 200 , (c) 170 , (d) 150 , (e) 140 , and (f) 135 µL/min. (1) Chloride, (2) nitrite, (3) carbonate, (4) sulfate, and (5) nitrate. The inset graph shows extended figure around peak 3.
at what precise combination of volatility and pKa analyte loss, such as that observed for CO2/carbonate, occurs. Other Detectors. To the chromatographic system described in the Experimental Section, after the postconcentrator (device 1 of Table 1, t ) 22-60 °C, 5 SLPM N2) and the conductivity detector, a mass spectrometer was (Thermo-Finnigan, AQA) used to monitor the eluting perchlorate peak (electrospray ionization, 3 kV applied, source temperature 350 °C) at m/z 99. Low microgram per liter sample concentrations were used. The conductivity based peak heights and peak areas (V‚s) showed excellent linear correlation with the reciprocal of the postconcentrator effluent flow rate (r2 ) 0.9977 and 0.9940, respectively). However, the mass spectrometer peak area counts or S/N remained unchanged within experimental uncertainty. Thus, these types of postconcentrators will not benefit mass-sensitive detection, unless advantage is taken of, for example, chip-based nanospray modules (www.advion.com), which work with very high efficiency but only at very low flow rates. The postconcentrators will result in a proportionate gain in sensitivity only with concentrationsensitive detectors. It should be interesting to explore the postconcentration approach with detection methods such as such as amperometry, which is not only concentration sensitive but also mass-transport dependent, the latter being more efficient at a lower flow rate. Preinjection Concentration. The evaporative concentration technique can of course be used in the preinjection mode as well. The technique is ideally suitable for use with continuous process analysis of high-purity water, for example. Figure 7 shows the chromatograms obtained with loop injection of a sample that has passed through an evaporative concentrator (device I, Table 1) at different flow rates prior to loading a sample loop. The peak 5696 Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
heights of chloride, nitrite, sulfate, and nitrate increased dramatically with decreasing sample flow rate, with no effect on chromatographic resolution. As with postsuppressor use, the response from carbonate decreased significantly with the membrane concentrator. The problem with this mode of concentration is that the samples will generally contain non-H+ cations that will be captured by the membrane. As the membrane is converted from the H+ form to some other cationic form, the water transport properties of the membrane will change and the extent of evaporative concentration achieved will decrease. Figure 8 shows the decrease in CF (6.90 at the beginning) as a function of time (6.04 after 24 h of continued injections every 20 min). Similar losses in the CF of other anions were observed as well. In order to prevent the membrane significantly converting to a non-H+ form, one may of course pass the sample through a suppressor before the evaporative concentrator. Alternatively, the same evaporation membrane device can be periodically regenerated after several injections. After every four sample injections, 4 mM HCl was pumped in to the concentrator at 0.7 mL/min for 1 min by a peristaltic pump, followed by water wash at the same flow rate for 19 min to wash out the regenerant. Then, the sample concentration process and injection was recommenced. Residual chloride peak (0.030 µS/ cm) was only marginally higher than the blank chloride response (0.021 ( 0.002 µS, n ) 3). Of course, a strong acid regenerant such as methanesulfonic acid, where the corresponding anion is generally not an analyte of interest, can also be used. In some cases, it may be preferable to alternate between two concentrators. Using the above periodic regeneration protocol, results for nitrate for 24-h continuous operation of the concentrator (influent sample flow rate 135 µL/min, t ) 60 °C) is shown in Figure 8. The CF for all the ions (nitrate, 16.4 ( 0.3; chloride, 16.5 ( 0.3); nitrite, 12.0 ( 0.2; and sulfate, 14.7 ( 0.3) remained constant over 24 h. We have observed that the water transport variation between the H+ form of Nafion and other cationic forms is more pro-
nounced at higher temperatures than at lower temperatures. In this vein, we also operated the concentrator at 30 °C with a sample flow rate of 60 µL/min to still maintain a relatively high CF. The CF loss for nitrate was reduced to only 0.3%/day. The CF loss of other ions was 0.5% for chloride, 0.0% for nitrite, and 1.1% for sulfate and were much better than at 60 °C. However, the precise performance will obviously depend on the cationic composition of the sample, and periodic regeneration or presuppression may constitute more generally applicable approaches. CONCLUSIONS AND FUTURE WORK The evaporative concentration approach provides a simple and easy to automate tool to improve detectability. Pre- and postconcentration modes each have their individual advantages and drawbacks. The general approach may have merits beyond what has been demonstrated here, for example, as an interface between two analysis modes where on-line concentration or removal of a volatile nonaqueous solvent would be desirable, (23) Bao, L.; Dasgupta, P. K. Anal. Chem. 1992, 64, 991-996.
viz. LC-capillary electrophoresis (CE), LC-matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, LC-µLC, etc. In CE systems, porous membranes have been used in the past as a means of sample introduction as an integral part of the separation system.23 It is certainly feasible to carry out electrophoresis in a porous membrane capillary. It is an intriguing question as to how such a system will respond to simultaneous evaporation. ACKNOWLEDGMENT This research was funded by the National Science Foundation Grant CHE- 0518652. Generous support from Dionex Corporation is also acknowledged. We thank Prof. Richard B. Timmons for his help with plasma modification of membranes. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review February 23, 2007. Accepted May 29, 2007. AC0703799
Analytical Chemistry, Vol. 79, No. 15, August 1, 2007
5697
