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
Admittance Scanning for Whole Column Detection Brian N. Stamos,† Purnendu K Dasgupta,*,† and Shin-Ichi Ohira‡ †
Department of Chemistry and Biochemistry, University of Texas, Arlington, Texas 76019, United States Department of Chemistry, Kumamoto University, 2-39-1 Kurokami, Kumamoto 860-8555, Japan
‡
S Supporting Information *
ABSTRACT: Whole column detection (WCD) is as old as chromatography itself. WCD requires an ability to interrogate column contents from the outside. Other than the obvious case of optical detection through a transparent column, admittance (often termed contactless conductance) measurements can also sense changes in the column contents (especially ionic content) from the outside without galvanic contact with the solution. We propose here electromechanically scanned admittance imaging and apply this to open tubular (OT) chromatography. The detector scans across the column; the length resolution depends on the scanning velocity and the data acquisition frequency, ultimately limited by the physical step resolution (40 μm in the present setup). Precision equal to this step resolution was observed for locating an interface between two immiscible liquids inside a 21 μm capillary. Mechanically, the maximum scanning speed was 100 mm/s, but at 1 kHz sampling rate and a time constant of 25 ms, the highest practical scan speed (no peak distortion) was 28 mm/s. At scanning speeds of 0, 4, and 28 mm/s, the S/N for 180 pL (zone length of 1.9 mm in a 11 μm i.d. column) of 500 μM KCl injected into water was 6450, 3850, and 1500, respectively. To facilitate constant and reproducible contact with the column regardless of minor variations in outer diameter, a double quadrupole electrode system was developed. Columns of significant length (>1 m) can be readily scanned. We demonstrate its applicability with both OT and commercial packed columns and explore uniformity of retention along a column, increasing S/N by stopped-flow repeat scans, etc. as unique applications. n Tsvet’s first experiment on chromatography of plant pigments, he necessarily carried out visual whole column detection (WCD). Although his ability for quantitation may have been limited by his visual prowess, he saw not only what eluted from the column but also what did not, not feasible in modern HPLC. Indeed, if such a capability existed, premature column deaths from recalcitrant components in the injected samples could have been well prevented. In principle, if bands could be quantitated on-column, an end-column detector would be superfluous. The overwhelming use of metallic column enclosures in the current practice of liquid chromatography inhibits any type of electromagnetic imaging, however. Virtually all WCD work in the last 50 years has centered on optical detection; the first attempt involved limited motorized movement of a glass column through a spectrophotometer to determine partition constants.1 Birks and co-workers studied WCD first theoretically2 and later experimentally with a single long illuminator and 14 stationary photodiodes using then prevalent glass columns.3 Most other WCD studies are credited to Pawliszyn and co-workers in capillary electrophoresis (CE) or, more commonly, capillary isoelectric focusing (CIEF) systems.4 One commercial WCD instrument5 measures absorbance at 280 nm over a 5 cm long imaging zone of a CIEF system using a fiber-optic array for illumination. Other commercial capillary scale imaging detectors (www.deltadot. com; www.paraytec.com) image substantially smaller lengths, for purposes not related to WCD.
I
© 2017 American Chemical Society
Related pioneering work has involved axial optical detection through the entire column where differentiation yields a conventional chromatogram;6 late-eluting strong absorbers can make this impractical. Other WCD studies have relied on laser-induced fluorescence which can be ultrasensitive but generally requires prior derivatization. As is evident from chromatographic simulations,7 the separation between a given analyte pair is often complete oncolumn long before they are seen by end-column detectors. WCD would have saved time and allowed measurement immediately after separation and before further dispersion, possibly improving the LOD. In gradient elution, a gradient that is too steep or too early or both can cause a pair of just separated analytes to remerge, even change elution order.8 One rarely relies on models to formulate a gradient; it is typically an iterative trial and error process. An ability to see the separation develop will permit changing conditions in near real time. The only intrinsic chromatographic analyte identifier is the “retention factor”. In gradient elution, the time variant speed can be measured by WCD; this can uniquely identify an analyte. As for optical WCD, instantaneous imaging works well for small lengths. If a long zone is illuminated with the same Received: April 15, 2017 Accepted: June 1, 2017 Published: June 1, 2017 7203
DOI: 10.1021/acs.analchem.7b01412 Anal. Chem. 2017, 89, 7203−7209
Article
Analytical Chemistry
reagent grade and used as received (www.fishersci.com). House distilled water, passed through an ARIES water purification system (www.ariesfilterworks.com), provided 18.2 MΩ·cm−1 water. All solutions were filtered through a 0.2 μm filter before use. Linear Stage Setup. Referring to Figure 1, a CRK Series linear stage (www.boschrexroth.com) with 1360 mm of stage
amount of light, S/N will suffer, however. Using all the light at one point at a time gives good S/N, but scanning a long region does incur a time penalty. Mechanically scanned imaging is not inferior: it is the present norm for optical scanners rather than the original invention of at-once imaging.9 Regardless of the method, WCD provides a sequence of temporal images that can potentially reveal surprising details of the chromatographic process. In gradient elution, the retention factor changes in a nonlinear fashion and often very differently for different analytes. As a result, it is not uncommon for a trailing analyte band to go past another initially ahead during an eluent gradient,8 but this will never be seen by an end-column detector. A compact, easily movable sensing head is a necessity for scanning. Conductivity detection is the mainstay of ion chromatography (IC), of interest to us. We recently described admittance (incorrectly, but more commonly, termed contactless conductance) detectors that can work with very small capillaries even at low specific conductance levels.10−12 Small capillaries and low specific conductance make for a high impedance system (pure water in a 10 μm capillary with electrodes 1 mm apart represents an ohmic resistance of 2.3 TΩ) and very low current levels. Nevertheless, with high-gain, ultralow bias-current transimpedance amplifiers to immediately amplify the signal in a properly shielded enclosure, a compact translatable sensing head is possible. Admittance detection (AD) requires electrically nonconducting column enclosures as is used in IC anyway. Ions in solution (rather than stationary phase-bound ions) contribute the majority of the admittance signal due to their much greater mobilities.13
Figure 1. Scanning instrument setup. AD: Pass-through admittance detector (details in Figure 2); C: OT capillary column; H: aluminum capillary column holders. The bottom part attaches to the rails of the linear stage and can be easily moved and then anchored as required by the length of the column. Platforms hang over the linear stage with adjustable height to allow different detectors to be used. A PEEK tee T is used to hold the head of the column and perform injections with the aid of solenoid valve SV and injection inlet I. A drilled out union at the other end fixes C in place using a ferruled nut. The linear stage has 1.36 m of usable travel and is driven in 40 μm steps by stepper motor SM.
■
travel was used to sweep the detector across the capillary column. One revolution of the screw translates to 16 mm of travel. The stage has a lower and upper T-rail along each side. Aluminum blocks were machined to fit in the rails at each end of the stage to make X,Z-adjustable platforms H to which a 10− 32 threaded tee T and a 10−32 threaded union (with 0.5 mm through passages) are mounted to hold the head and the exit of the capillary column C. The tee T shown in the inset serves as part of the injection system that includes a solenoid valve SV and a sample injection entrance I. The manner in which the injection system operates has been described elsewhere.13 We used two detectors: A TraceDec AD (www.istech.at) with a high sensitivity sensor head (HS3059) was used with 200% gain and a voltage output of −12 dB. The other detector used the same electronics that we have previously described11 except the primary amplifier was an OPA606 (www.ti.com). This device is now commercially available from Analytical Foundry (AF) (www.analyticalfoundry.com). Either sensor head was attached to the moving stage via custom 3D-printed holders. The end platforms, adjustable along the stage travel, allow any column length up to the stage maximum. Once the platforms are fixed in place, the capillary is held taut and ferruled male nuts at each end fix the capillary so that it does not change its position by the drag of the sweeping sensor head. The linear stage is moved by a high torque (425 oz·in., 3.00 N·m) NEMA 24 stepper motor SM (1/4 in. shaft) and a stepper motor driver (QJ8060, www.buildyourcnc.com) powered by a 24 V DC 6.5 A supply (RS-150−24, www.meanwellusa.com) deliberately currentlimited to 4.1 A. Column and Injection Arrangement. For the commercial packed capillary column (Dionex AS11-HC, 0.4 × 250 mm), a GP-40 pump (Dionex, split-flow configuration) provided 10 μL/min flow through the column. Sample (1.5
MOVABLE ADMITTANCE DETECTION In its present form, AD was first conceived in 1998.14,15 It was quickly realized that the sensor can be moved along the tube; this ability was first exploited by Unterholzner et al.16 in capillary electrophoresis (CE). They showed, in a nonuniform electric field, the apparent migration order of analytes may change with the on-column detector location; the detector can be advantageously located at a position where the maximum resolution between a difficult to resolve analyte pair occurs. With enhanced placement ability of a smaller sensor head, Macka et al.17 measured apparent selectivities for various ions at 11 separate detector locations. Gillespie et al.18 then used a commercial detector (TraceDec) and manually positioned it in different places along a monolithic capillary column, to characterize progress of the surfactant coating being applied. Gillespie et al.19 also manually moved an AD during a chromatographic separation from one stationary location to a second to optimally capture separations of both early and late eluters. Paull20 recorded 20 successive chromatograms with the sensor head manually placed at 20 equispaced stationary locations beginning at the head of the column and demonstrated that successive display of these images can depict the separation as a virtual movie. Here, we explore the potential of automated repeated electromechanically scanned admittance imaging of both open tubular (OT) and packed capillary columns.
■
EXPERIMENTAL SECTION Chemicals. Sodium salicylate (as eluent) and chloride, chlorate, perchlorate, iodide, thiocyanate, sulfate, and thiosulfate, as sodium or potassium salts as test anions, were all 7204
DOI: 10.1021/acs.analchem.7b01412 Anal. Chem. 2017, 89, 7203−7209
Article
Analytical Chemistry μL) was manually injected (injector: MXP9900-000, www.idexhs.com). For the OT system, 360 μm o.d. fused silica (11−21 μm i.d.) and cyclic olefin polymer (COP, 19 μm i.d.) capillaries were used. For chromatographic experiments, capillaries were coated with AS18 latex (COP capillaries were sulfonated before coating); details of COP column preparation appear elsewhere.21 The OT column injection system used for the capillary columns has been previously described.13 Figure 1 inset indicates how this injection system is built around the capillary holding tee. The volume injected can be controlled with both pressure and injection time. Except as stated, all OT columns were connected to a 92 cm long, 11 μm ϕ silica restriction capillary at the exit. Injections, controlled through a LabVIEW program were performed at 80 psi with a range of 2− 7.5 s (injected volumes of 0.17−0.65 nL; 1.8−6.8 mm initial zone length on a 11 μm ϕ column). Scanning Cell Electrode Arrangement. As sleeve type detection electrodes cannot be slipped over the sizable endfittings, we developed a double-quadrupole electrode design. That dimensioned for the packed capillary column is described below. Two identical top and bottom halves are each built on a flat nonconducting board I glued to a grounded 3 mm thick metallic sheet M (Figure 2a). Two pairs of stainless steel
each quadrupole are connected with a short wire W (inset, Figure 2b), and final assembly is completed. Spring-equipped screws S apply pressure from the bottom so that differences in outer diameter along the length of the scanned column do not cause loss of contact (Figure 2b). Silicone lubricants were lightly sprayed on the columns to reduce friction before scanning. Linear Stage Control. The stepper motor was controlled by a CY8CKIT-059 prototyping kit (www.cypress.com) containing a programmable system on chip (PSoC5LP). In this work, 1/2 stepping (equal to 0.9° rotation, 40 μm of linear movement/microstep) was used. The system keeps track of the stage location at any time. The velocity Vs is the product of the microstep length and the stepping rate fs (Hz), namely, 0.04 fs mm/s. Detector Data Acquisition and Positional Synchronization. With the TraceDec, the internal ADC uses a sampling frequency (f) of 34.14 Hz, the maximum. The AF AD uses a ΔΣ ADC onboard the PSoC5LP, acquiring 18-bit data at f = 3.6 kHz. Hereinbelow, the initiation point of each scan is designated “home”; this is actually ∼30 mm from the head of the column because of the width of the detection head and column portion inside the T-holder. The length resolution of the acquired data is Vs/f. The length resolution can be improved (with penalties) by (a) reducing Vs (increases scan time) or (b) increasing f (increases noise). The PSoC synchronizes the movement of the stage with the data acquisition, allowing each detection point to be taken at precisely the same location on the column through a LabVIEW routine that allows Vs, number of steps to travel, wait time (if any, between successive scans), total number of scans to be made, and f as input parameters. Presently, scans were unidirectional, meaning the stage returned home to begin the next scan. In addition, at each terminal end, the stage stopped for 2 s to avoid inertial glitches. In this work, Vs ranged from 0.4 to 4.0 cm/s and was 2.8 cm/s unless otherwise stated (Vs,max was 10 cm/s; return speed could be higher for unidirectional scans). It is possible in principle to use bidirectional scanning.
Figure 2. Double quadrupole admittance detector. (a) View of bottom part. Two pairs of steel cylinders Q are separated by a rectangular cutout CO and silver-epoxied to insulating board I, glued to a metal bottom M that is electrically grounded. The shield SH (MuMETAL inside, polyimide outside) is shown in detail on the left, and the column C is slipped through the slit to rest in the snug-fitting hole. An identical top half is then put in assembled. (b) Assembled view from the side with inset showing view from the column end. SMB connector pin SP connects to each top electrode set; wire W connects each top and bottom pair within a quadrupole. Spring-tensioned screws S firmly hold the top and bottom together.
■
RESULTS AND DISCUSSION Detector Suitability: A Tale of Two Detectors. The response time of a detector is obviously critical to scanning. Although admittance detector related publications now number more than 1000, there are not many commercial detectors. We started with the leading commercial detector (TraceDec) and of necessity (see below) used our own design.11 The differences in detector characteristics are compared in the following sections. Effect of Scanning Speed on Signal Fidelity. TraceDec response to a 0.95 nL injection of 500 μM KCl is shown in Figure S1 at scan speeds ranging from 0.4 to 4 cm/s (initial zone length, 2.7 mm in a 21 μm ϕ silica capillary; water carrier, 0.82 mm/s). Scan was begun at different times for different scan speeds after injection, so in all cases, the detector reached the sample apex at the same location (∼5.3 cm from home). The perceived signal height (from initial baseline) decreased with increasing scanning speed as 1:0.85:0.62:0.42 with no significant change in the peak area. At the two higher scanning speeds, the peak is also obviously misshapen. This is not due to an inadequate f; even at the fastest scanning speed, 8 data points are being acquired per cm. Rather, this is due to the imbedded signal processing algorithm for noise reduction.
cylinders Q (9 mm long, 1.59 mm ϕ ea.) serve as electrodes on each half. Each pair is silver-epoxied (SE, Figure 2b inset) to I; the central pin SP of a female SMB connector also connects to the electrode pair/silver epoxy patch through a hole on the other side of I/M. The electrode pairs are 1 mm apart and are separated by a 10 mm long, 0.5 mm wide cutout CO therebetween on I (Figure 2a). The ground shield is a foldedover 100 μm thick sheet of MuMETAL (www.magneticshield. com) foil, to which a 25 μm thick adhesive-backed polyimide tape (www.kaptontape.com) is attached on one side. A 16 × 5 mm long piece is folded in half, with the polyimide facing outward, the effective thickness being ∼250 μm. A 1.6 mm hole is drilled through the center of the 8 × 5 mm piece and a slit cut from the short edge to the hole (see right inset SH in Figure 2a) to allow the shield SH to be slipped around the column C (this is not needed for the OT version, the shield can simply be inserted from one end). With the shield positioned on the slot CO, the column is then placed on the bottom set of electrodes. The top half is now put in, forming the two quadrupole electrodes Q; the top and bottom electrodes in 7205
DOI: 10.1021/acs.analchem.7b01412 Anal. Chem. 2017, 89, 7203−7209
Article
Analytical Chemistry Signal processing with filters of an unknown algorithm can lead to unanticipated problems.22 Due to overfiltration, this detector never shows any short-term noise; from a finite signal at some concentration, the signal abruptly disappears on the next dilution. The AF detector time constant (τ) is controlled by an integrating capacitor in the RMS-DC converter (AD637a, www. analogdevices.com). With τ = 250 ms, the response behavior was better than the TraceDec but qualitatively similar in that the peak height decreased and the tailing increased with scanning speed (Figure S2). With τ = 25 ms, the signal became independent of scanning speeds up to 2.8 cm/s, and some effect was perceptible at still higher speeds (Figure S3). There was far less tailing at all speeds relative to τ = 250 ms, a direct comparison is shown in Figure S4. Figures S2−S4 used f = 1 kHz. As shown in Figure S5, f = 500 Hz may be sufficient. For the TraceDec, the highest available f (34 Hz) was already demonstrably insufficient; even lower f (16.6 and 5.6 Hz) made things predictably worse (Figure S6). Admittedly, the detector was designed for stationary applications. Positional and Interval Reproducibility. These parameters were measured by applying three ∼3 mm wide marks with a metallic paint pen ∼5 cm apart on the outside of a silica capillary. Both detectors registered large spikes when passing over the metallic ink marks. Multiple scans were performed at 1.6 cm/s for both detectors; for the AF AD, a slower Vs of 0.4 cm/s was also used. For the TraceDec, the reproducibility in locating the marks in all cases (four scans, three marks) were within one data point (Figure S7); unfortunately, at 1.6 cm/s and f = 34 Hz, this resolution is rather large (0.47 mm). The detector did not separately see the two edges of each mark (Figure S8). For the AF AD, with f = 1 kHz and Vs = 1.6 cm/s, the data resolution is theoretically 16 μm, but it cannot really be lower than the step resolution of 40 μm; any given location is oversampled. The AF detector discerned each edge of the mark (Figure S9). In 5 scans over the three marks, the standard deviation of the location of each edge (6 edges) varied from 40 to 190 μm (average ± sd: 100 ± 60 μm). However, the electrodes fit snug on the capillary, and each pass rubs off some of the paint (at an even higher scanning speed of 2.8 cm/s, the first pass is enough to remove all paint) resulting in diminishing registration of the marks with successive scans. As such smearing may affect the results, a freshly marked capillary was scanned at 0.4 cm/s (Figure 3). The location of each edge exhibited a standard deviation (SD) of 15−37 μm. The precision (as SD) in measuring the width of each of the 3 marks was 26−67 μm compared to 100−180 μm at Vs = 1.6 cm/s. At Vs = 0.4 and 1.6 cm/s, the SD of the measured intervals between the three marks was 115−140 and 120−150 μm, respectively. For the TraceDec at Vs = 0.4 cm/s, this was 0.47 mm (limited by data resolution). To put matters in perspective, the typical initial length occupied by the sample in the column is 1500 μm. To measure locational precision for a liquid edge inside the capillary, we injected a 500 μM KCl solution into a capillary filled with hexane, sealed the ends of the capillary, and repeatedly scanned the interface region (the interface produces an easy to recognize sharp spike). On a 21 μm i.d. silica capillary with a scan speed of 1.9 cm/s and f = 1 kHz, the locational precision was 35 μm in terms of standard deviation (n = 5), essentially the same as the step resolution. See ac7b01412_si_002.mpg to look at the reproducibility.
Figure 3. Three ∼2 mm wide metallic ink marks made on a silica capillary ∼5 cm apart and scanned five times with the AF AD at 4 mm/s, data acquired at 1 kHz. The top left inset shows a magnified view of the responses to the first mark. The data have been vertically offset slightly to show individual traces and the remarkable reproducibility of the baseline traces.
Baseline Correction. Figure 3 indicates that the undulations in the baseline are not random detector noise. The detector is sensitive enough to pick up minor variations in the coating or tubing dimensions. If the signal variations arose from momentary loss of contact, they would not be reproducible; these must arise from capillary characteristics. Compared to fused silica, COP capillaries, for which manufacturing is subject to much greater tolerances, predictably show much greater shifts in baseline along the length of a column. For this reason, a background subtraction process was always adopted for all chromatographic scans: Four unidirectional scans were initially made, averaged, and stored as the background before any sample injection. This was subtracted from all postinjection scans. It needs, however, to be noted that the background correction is not perfect. There can be a scratch on a column surface or a change in the polyimide coating thickness that is easily recognizable on a scan as a “spike”. Subtraction may diminish the magnitude but will not eliminate it. Column Imaging: Beyond Stationary Detection. In each scan, the observed peak areas more readily provide quantitation. A 20 cm portion of the column was repeatedly scanned (Vs = 2.4 cm/s, f = 1000 Hz) using successive unidirectional scans 22.3 s apart. The results are shown in Figure 4. With each scan, the analytes further progress down the length of the column. As we are looking at the column with injection/exit ends respectively at left/right, appearance of the bands are mirror images of that displayed by an end-column detector. In scan 33, the void dip has just moved into our view window. In scan 35, peaks 3 and 4 just appear in the observation zone, well resolved with 3 trailing 4. It will be observed that, as scans progress, 3 catches up with 4 and, by the time the pair exits the view window, 3 is ahead of 4. A similar situation occurs with the peaks 6 and 7; 6 is far ahead of 7 when they first appear. They completely merge by scan 67, and although it is not possible to see because of the very limited ordinate scale for each trace, 7 eventually elutes before 6. In 7206
DOI: 10.1021/acs.analchem.7b01412 Anal. Chem. 2017, 89, 7203−7209
Article
Analytical Chemistry
Formats of Presenting a Dynamic Separation. The present system generates three-dimensional data; we are not as yet sure of the best mode(s) of presentation. Figures 4 and S12 show only one presentation mode. Obviously, successive images can be stitched together to make a movie (Movies ac7b01412_si_004.mpg and ac7b01412_si_003.mpg, respectively, comprise 51 and 23 scans, successive scans begin 11 s apart) and can reveal features on a dynamic basis that is not possible to depict, at least as clearly, in a temporally static fashion. For example, one often observes unusually retentive spots on the column, where analytes “slow down”, and this can be more pronounced for some analytes than others. Another static mode of temporal presentation of the chromatographic process is simply the visualization of the raw numerical data in Microsoft Excel using conditional formatting which allows one to highlight cells in color schemes that rely on their numerical values. One example is shown as the Table of Contents graphic, and another color scheme is shown in Figure S13. A “heat map” depiction is also possible (Figure S14). Retention Factor. Although we traditionally define the retention factor k as (t − t0)/t0 where these terms have their usual meaning, k is more simply defined presently in terms of the distance the mobile phase (indicated by the void dip/peak) and the analyte have, respectively, traveled at a certain time t (dm,t and da,t, respectively):
Figure 4. Repeated scans on 20 cm length of a 11 μm id, 360 μm od AS18 latex coated silica capillary of total separation column length 60 cm. Injection: 430 pL (4.5 mm initial zone length) of 500 μM each of Cl−, ClO3−, SO42−, I−, SCN−, ClO4−, and S2O32−(numbered analytes 1−7 in that order). Eluent, 6 mM sodium salicylate; flow rate, 5.2 nL/ min (mobile phase velocity 0.1 cm/s), Vs 2.4 cm/s. Each scan 8.3 s; interval between unidirectional scans 22.6 s, Analytical Foundry detector. The scan numbers appear (odd numbered scans imply away from home) below each trace. Peak 0 is the void. The small ordinate scaling makes the bands appear as very broad. See Figure S12 inset for a different view.
k=
dm,t − da,t da,t
(1)
While we can estimate k by simply taking the positions of the injection void and the analyte peak obtained during a scan, the imaging is not instantaneous: if the analyte peak location da,t is imaged at time t, the mobile phase front is not imaged until a slightly later time t′ (i.e., what is measured is dm,t′); the difference t′ − t is the interval for the sensor head to travel from the analyte peak to the void. The error committed by taking dm,t′ as dm,t and the resulting error in k can be significant if the mobile phase velocity (Vm) cannot be neglected relative to Vs. In such a case, the appropriate expression for k is given by (see the Supporting Information for derivation):
both cases, a monovalent ion was initially ahead of a divalent ion but the latter eventually goes ahead. The reason is discussed later, but presently, we just note that it would have been impossible to see this behavior with any stationary detector. Second, the present column is 60 cm long. It would be apparent that any individual ion was completely separated from the others at some point within our view window of 20 cm. With an end-column detector, we will only see greatly broadened bands under isocratic conditions. With a programmable-position detector and the knowledge as to when the desired separation is complete, quantitation can be advantageously performed with much greater sensitivity as soon as the band(s) of interest is(are) separated from others and the detector moved to intercept the peak. An example is shown in Figure S10. After all desired quantitation is complete, the column can be flushed out. In such programmable stationary operation, even a slow responding detector can be used, whereas in scanning operation any peak will likely be distorted by such a detector (Figure S11). Packed vs OT columns. In the present application, conductance is the major contributor to the admittance signal. An ion displays much greater mobility in solution than in stationary phase-bound form.14 As such, OT ion chromatography, where the phase ratio (defined in the present case as the ratio of the ion exchange equivalents in the mobile phase to that in the stationary phase22) is much larger than for a packed column, provides correspondingly greater sensitivity. Although the sensitivity is poorer for a packed column, it is still possible to follow a separation. Figure S12 shows serial scans on a commercial 0.4 mm i.d. anion exchange column depicting the separation of chloride and bromate.
⎛ dm,t ′ ⎞⎛ V ⎞ k = ⎜⎜ − 1⎟⎟⎜1 − m ⎟ Vs ⎠ ⎝ da,t ⎠⎝
(2)
Figure 5 shows the plots of k values for a mixture of 7 anions dissolved in water injected into a Na−salicylate eluent. Similar data for the analytes dissolved in the eluent appear in Figure S15, and comparisons for selected anions are shown in Figure S16. Predictably, the k values begin at a higher level when the sample is in water but eventually converge to the same values. In both situations, however, the analytes initially display much higher retention. In an open tube, retention is dependent on mass transfer to the wall. Steady state flow in such small capillaries are highly laminar, the least efficient vector for transport to the wall. Sample injection in this system is followed by a period to flush out the excess sample just outside the column head. During this interval, there is essentially no flow through the column. When chromatography begins, column flow starts abruptly from near zero. During the finite time and length needed to establish laminar flow, mass transfer is more efficient than at steady state and leads to the observed higher k values. A second factor is more important with the sample in water. The eluent strength initially experienced is much less than that 7207
DOI: 10.1021/acs.analchem.7b01412 Anal. Chem. 2017, 89, 7203−7209
Article
Analytical Chemistry
accommodated, saving time. Figure S20 shows that signal averaging a 2.5 cm zone centered on a low concentration thiosulfate peak (n = 10, taking 51 s) improved the SNR by a factor of 3.14, almost exactly the theoretical gain of 3.16. Further averaging and SNR increase is possible because the signal is oversampled (see the Supporting Information for details). Imaging Column Homogeneity. Admittance scanning is well suited for ascertaining the uniformity of the bed structure of a packed column. Figure 6 shows the data for two
Figure 5. Retention factors calculated according to eq 2 for the data in Figure 4. At the beginning of the scan window, the retention factors for all analytes are much higher than at the end. The two factors contributing to this, the self-development of an eluent gradient and the finite length needed for the full development for laminar flow from noflow conditions, are discussed in more detail in the text.
further down the column because of the dilution of the eluent by the sample matrix. At the beginning of the column, k’s are much higher as the sample ions are largely surrounded by water and reach a stable value as the steady state eluent concentration is reached (Figure 5). In effect, the initial part of the column experiences gradient elution. This “inlet effect” is smaller when injected analytes are dissolved in the eluent (see Figure S16 for a comparison): the equilibration is more rapid (see, e.g., by 8 cm from home, kchlorate is stable when the sample matrix is the eluent but continues to fall when the sample is in water, eventually reaching the same value). In a larger bore column, the inlet effect zone is even longer (compare Figure S17 with Figure 5). The order of elution of divalent−monovalent analyte pairs often changes from the time they are first seen by the detector. Both thiosulfate and sulfate bands are initially seen behind perchlorate and iodide, respectively. Salicylate is a monovalent eluent. The logarithm of the retention factor of a monovalent analyte decreases linearly with log[salicylate], whereas that of a divalent ion decreases with the square of log[salicylate]. As a result, as the analytes experience an increasing eluent concentration moving forward from the original water matrix, the divalent anions speed up, merge, and then move ahead (Figures S16a,c and S18; this is dynamically seen in ac7b01412_si_003.mpg). The observed k for strongly retained analyte ions like perchlorate is also significantly affected by the injected concentration due to self-elution, especially initially (Figure S19). Improving Signal to Noise Ratio (SNR) through Stopped-Flow Multiple Scans and Signal Averaging. Being able to use the column as a separation canvas permits stopped-flow signal averaging to improve detection and quantification for analyte peaks with poor S/N, thanks to very slow liquid phase diffusion. Only the peak(s) of interest and not the whole column need(s) to be scanned. Since the liquid is stationary, bidirectional scan data are easily
Figure 6. Packed column scans to determine packing uniformity. The bottom four scans (left ordinate) are of the same column filled with KCl solutions of different concentrations. Sufficient solution was passed through each time to have a new medium each time. The top scan is that of a larger i.d. packed capillary anion exchange column.
commercial packed capillary columns, one a 100 μm i.d. silica capillary packed with 5 μm C18-derivatized silica and the other a 400 μm i.d. PEEK capillary filled with an anion exchange packing. The first was filled with 0.75, 1.25, and 1.50 mM KCl and the second with 4 mM Na-benzoate. As noted before, ions in the interstitial liquid contribute more to the admittance signal than do adsorbed/bound ions, the scanned image across the column provides a picture of the liquid−particle distribution homogeneity; a greater admittance indicates a larger void fraction. While admittance detectors have previously been used to explore column packing homogeneity by manually moving the detector,23 automated scanning makes this far more practical and with far greater resolution. The spatial features are highly reproducible (Figures S21 and S22). In numerous column scans, we are yet to encounter the ideal perfectly packed column. In general, most columns appear to be better packed in the bottom half than the top half. The present system allows one to probe nonuniformity even for an OT column by measuring the local changes in k for an analyte (Figure S23).
■
CONCLUSIONS An admittance detector with a new double quadrupole electrode design was adapted to a motorized linear stage enabling imaging admittance detection for both OT and packed capillary columns and permitting visualization of IC separations in real time. The detector can also be operated in punctuated 7208
DOI: 10.1021/acs.analchem.7b01412 Anal. Chem. 2017, 89, 7203−7209
Article
Analytical Chemistry Notes
stationary mode, positioning it to quantitate an analyte after it has just separated from others, combining the merits of both stationary and scanning detection. For an analysis that may be legally important, a scanned column can be archived. Desired regions of a column can also be repeatedly scanned to improve S/N. There is a penalty paid in noise in going from a stationary to a scanned configuration that can be significant at high scanning speeds but is modest at low scanning speeds. Tallarek, Guiochon, and others have used NMR imaging to look into how an analyte distributes three-dimensionally in the entrance region of a packed column.24−26 For an ionic analyte, similar insights, but in a 2-D manner, can be obtained presently, in arguably a much simpler manner. Similarly, it allows one to check the uniformity of a packed or OT column. By measuring variances in the local retention factor, even the presence of unusually retentive sites in a column can be identified. As with any new technique, the full potential of admittance imaging or other scanned imaging approaches will only be known in time. However, the following are already apparent: An end-column detector only sees what elutes while the present approach can also see what perhaps has not eluted and take appropriate action. A periodic admittance image of a column is its ongoing fitness report: a shift in the packing structure or irreversible analyte binding will change the image, of significance to quality control. Scanning WCD generates serial evolving images of a column, each scan adding to the available data. In repetitive analysis situations (e.g., in quality control), known analytes with known retention/migration behavior are involved. In such cases, with a temporally expanding database, rapid, mathematically deconvoluted quantitative analysis27,28 should be possible on-the-fly, long before full physical separation has been attained; computing power is increasingly inexpensive. Absorbance/ fluorescence scanning of a transparent column with a fiberoptic based detector should be straightforward, but sensitivity in the OT case may be a problem.
■
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work was supported by National Science Foundation (CHE-1506572), NASA (Grant Nos. NNX11AO66G and 15AM76G), and Thermo Fisher Scientific. We acknowledge the gift of packed capillary columns from Column Scientific (Xiamen, China). We thank C. Phillip Shelor and Akinde F. Kadjo, respectively, for artistic and mathematical assistance and Sajad Tasharofi for the liquid interface scanning data.
■
(1) Brumbaugh, E. E.; Ackers, G. K. J. Biol. Chem. 1968, 243, 6315− 6324. (2) Gelderloos, G. D.; Rowlen, K. L.; Birks, J. W.; et al. Anal. Chem. 1986, 58, 900−903. (3) Rowlen, K. L.; Duell, K. A.; Avery, J. P.; Birks, J. W. Anal. Chem. 1989, 61, 2624−2630. (4) Wu, X. Z.; Huang, T.; Liu, Z.; Pawliszyn, J. TrAC, Trends Anal. Chem. 2005, 24, 369−382. (5) Convergent Bioscience. iCE280 IEF Analyzer;http://www. proteinsimple.com/documents/iCE280_brochure_revC.pdf. (6) Xi, X.; Yeung, E. S. Anal. Chem. 1990, 62, 1580−1585. (7) Kadjo, A. F.; Dasgupta, P. K. Anal. Chim. Acta 2013, 773, 1−8. (8) Lin, S.-H.; Yu, T.; Sheu, A.; Yang, D.-J.; Pai, S. C. J. Chromatogr. A 2008, 1201, 128−131. (9) Carlson, C. F. Electrophotography. US Patent 2,297,691, 1942. (10) Zhang, M.; Stamos, B. N.; Amornthammarong, N.; Dasgupta, P. K. Anal. Chem. 2014, 86, 11538−11546. (11) Zhang, M.; Stamos, B. N.; Dasgupta, P. K. Anal. Chem. 2014, 86, 11547−11553. (12) Yang, B.; Zhang, M.; Kanyanee, T.; Stamos, B. N.; Dasgupta, P. K. Anal. Chem. 2014, 86, 11554−11561. (13) Small, H. Some Electrochemical Properties of an Ion Exchanger. MSc. Thesis, Queens University, Belfast, Ireland, 1953. (14) Zemann, A. J.; Schnell, E.; Volgger, D.; Bonn, G. K. Anal. Chem. 1998, 70, 563−567. (15) Fracassi da Silva, J. A.; do Lago, C. L. Anal. Chem. 1998, 70, 4339−4343. (16) Unterholzner, V.; Macka, M.; Haddad, P. R.; Zemann, A. Analyst 2002, 127, 715−718. (17) Macka, M.; Hutchinson, J.; Zemann, A.; Shusheng, Z.; Haddad, P. R. Electrophoresis 2003, 24, 2144−2149. (18) Gillespie, E.; Connolly, D.; Macka, M.; Nesterenko, P. N.; Paull, B. Analyst 2007, 132, 1238−1245. (19) Gillespie, E.; Connolly, D.; Macka, M.; Hauser, P.; Paull, B. Analyst 2008, 133, 1104−1110. (20) Paull, B. New Possibilities in Capillary Ion Chromatography. In Pittcon, New Orleans, LA, March, 2008. (21) Huang, W.; Seetasang, S.; Azizi, M.; Dasgupta, P. K. Anal. Chem. 2016, 88, 12013−12020. (22) Wahab, M. F.; Dasgupta, P. K.; Kadjo, A. F.; Armstrong, D. W. Anal. Chim. Acta 2016, 907, 31−44. (23) Connolly, D.; Floris, P.; Nesterenko, P.; Paull, B. TrAC, Trends Anal. Chem. 2010, 29, 870−884. (24) Tallarek, U.; Bayer, E.; Guiochon, G. J. Am. Chem. Soc. 1998, 120, 1494−1505. (25) Tallarek, U.; Baumeister, E.; Albert, K.; Bayer, E.; Guiochon, G. J. Chromatogr. A 1995, 696, 1−8. (26) Baumeister, E.; Klose, U.; Albert, K.; Bayer, E.; Guiochon, G. J. Chromatogr. A 1995, 694, 321−331. (27) Dasgupta, P. K. J. Chromatogr. A 2008, 1213, 50−55. (28) Nikitas, P.; Pappa-Louisi, A.; Papageorgiou, A. J. Chromatogr. 2001, 912, 13−29.
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.7b01412. Response characteristics and spatial resolution/reproducibility of two admittance detectors, images of dispersion along a column, stationary vs moving detection, serial scanned images of separations in different formats, derivation of eq 2, retention factors at various column locations, improvement of SNR by repeat scanning, and uniformity of packed and OT columns along their length (PDF) Four sequential images of the scanner scanned over a 12 cm length (MPG), indicating locational reproducibility Scan containing 23 images (MPG), same data as in Figure 4 Illustrative scan containing 51 images (MPG)
■
REFERENCES
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Fax: 817-272-3808. ORCID
Purnendu K Dasgupta: 0000-0002-8831-7920 7209
DOI: 10.1021/acs.analchem.7b01412 Anal. Chem. 2017, 89, 7203−7209
