Article pubs.acs.org/ac
Electrodialytic Ion Isolation for Matrix Removal Shin-Ichi Ohira,* Kenta Kuhara, Mayu Kudo, Yuko Kodama, Purnendu K. Dasgupta, and Kei Toda Department of Chemistry, Kumamoto University, 2-39-1 Kurokami 860-8555, Japan Department of Chemistry and Biochemistry, The University of Texas at Arlington, Arlington, Texas 76019-0065, United States S Supporting Information *
ABSTRACT: We report a fully automated online sample pretreatment system for ionic analytes that extracts the ionic analytes from the sample and largely removes the nonionic sample matrix and can preconcentrate the analyte. Sample pretreatment is a key analytical process; conventional pretreatment is conducted in a difficult to automate batchwise manner. The present system relies on the transport of ions induced by an electric field to a water acceptor. Cations and anions are simultaneously and separately collected into individual acceptor streams which can be directly introduced to a chemical analyzer. Common inorganic ions (≤10 meq/L) are quantitatively transferred from samples within a few seconds. Small nonionic molecules are transferred by 0.5−10%, and proteins are not transferred at all. The method has been successfully applied to drinking water, urine, and cow’s milk with 3.7 ± 2.5, 3.8 ± 2.6, and 4.6 ± 2.6%, respectively, in variance (n = 10). Present results agreed well with those from conventional pretreatment methods. Interestingly, when calcium in milk is measured by the present method, the results correspond to the total calcium by conventional methods; i.e., it can extract calcium from its protein-bound form in milk.
I
described. Sample (donor) and acceptor solutions are separated, e.g., with a regenerated cellulose dialysis membrane. The target analyte ions (and small molecules) are transferred from the donor to the acceptor; the analyte concentration gradient provides the motive force. The acceptor analyte concentration exponentially approaches that in the donor, resulting in a long time to equilibrium. Donnan dialysis uses an entirely different principle. A high ionic strength acceptor is separated from a larger (donor) sample volume by, e.g., a cation exchange membrane. Because of the large ionic strength gradient, acceptor cations come through to the donor side. To maintain electroneutrality, donor cations go to the acceptor.12 Because of the smaller acceptor volume, large concentration factors can be attained. Transport of uncharged molecules to the acceptor is typically much lower than in conventional dialysis. By adding an appropriate complexing agent for the analyte to the receiver, selective preconcentration becomes possible.13 However, attainment of equilibrium is not any faster than in conventional dialysis. Trace metal preconcentration for IC14 and atomic spectrometry15 were demonstrated, but it has had little practical use; a high ionic strength sample matrix is hardly attractive. The above processes use concentration gradient as the sole motive force for transport. In electrodialysis, an electric field provides the primary motive force to move ionic analytes from donor to acceptor. Electrodialysis is widely used for desalin-
on analysis is important. In the present context, by ion, we mean small inorganic and organic anions and cations but not charged colloidal particles or large molecules, e.g., proteins. Atomic spectrometry is mostly used to determine metals; alkali and alkaline earths are also measured by ion chromatography (IC) and ion selective electrodes. Anion analysis is dominated by suppressed conductometric IC;1 mass spectrometry (MS) and/or tandem MS are often used for trace and ultratrace analysis.2 Regardless of the specific technique, many real samples can only be analyzed after some pretreatment. Such pretreatment may have the objective(s) of removing (a) particulate matter, (b) substances that would irreversibly bind to a separation column, (c) interferences, and/or (d) analyte enrichment.3,4 (Ultra) filtration, (ultra) centrifugation, solid phase extraction, and liquid−liquid extractions are the most common in ionic analysis.5 Sample preparation is often the bottleneck; a dedicated conference is now in its 14th year.6 Also, while most instruments allow automated sequential analysis, most sample pretreatment methods cannot be carried out online. Batch pretreatment is not easily integrated with an automated analysis system, without a complex, e.g., robotic, interface. Dialysis in Sample Pretreatment. Dialysis approaches are particularly relevant to the present paper. Dialysis membranes have been used to remove particles, cells, cellular debris, and large molecules, especially from biological samples. Automated dialysis-based pretreatment systems for flow analysis have been reviewed.7 Online dialysis for measuring Na/K in vegetables by flame photometry8 and anions in wastewater,9,10 milk,5 engine coolant,5 tablets,5 and organic solvents,11 all by IC, have been © 2012 American Chemical Society
Received: April 22, 2012 Accepted: May 30, 2012 Published: May 30, 2012 5421
dx.doi.org/10.1021/ac300982k | Anal. Chem. 2012, 84, 5421−5426
Analytical Chemistry
Article
ization16 and acid/base/metal recovery from wastewater.17 Martin first demonstrated electrodialytic desalting;18 a subsequent “improved micro desalter”19 could remove salt from a single saline drop of amino acids on a filter strip using a mercury cathode and a cellulose covered anode. The tubular dual membrane electrodialytic membrane suppressor,20 subsequently commercialized in a planar format,21 is the first widely used flow-through electrodialytic device.22 Although not depicted as such, the suppressor is an electrodialytic pretreatment device: It removes, e.g., the base from sample anions in a strong base matrix before presenting them to the detector. The dual tubular off-line electrodialyzer of Okamoto,23 devised to remove the NaOH matrix from NaOH-fused rock sample extracts, was configurationally identical to the suppressor in ref 20. In the same year, others described the off-line use of a commercial suppressor24 or a home-built planar “electrodialysis” system25 for removing base prior to determining trace anions in strongly alkaline samples. Conversely, microfluidic suppressors that functions by liquid− liquid extraction have been described.26−28 The use of two dissimilar membranes (cation exchange and anion exchange membranes (CEM, AEM)) was discussed early29 but explored only recently with cation/anion exchanger resin beads and utilized to make different electrolytes.30,31 With polarities reversed, the device can act as a flow-through deionizer/desalter that removes salts from a saline protein solution.32 Others have previously studied similar removal of NaCl from sulfanilic acid with limited success.33 Operated in the constant voltage mode, the total charge passing through the desalter is a measure of the total charge on the ions removed; it behaves as a charge detector.34,35 Operated with a central ground electrode and one electrode each exterior to the CEM/AEM, cations/anions can be added/subtracted from the central channel through the CEM/AEM permitting, e.g., the generation of salts, acids, bases, or buffers of any desired pH or concentration.36 If Na+ and Cl− can be completely removed from a mixture containing a large protein molecule by the desalter, then it is logical to believe that the respective CEM/AEM effluent solutions will contain the cations/anions originally in the sample. However, there are problems: chloride will oxidize to chlorine at the anode and attack the AEM. Even if the desired cation/anion is immune to electrochemical reduction/oxidation, automated sample processing is difficult with uncontrolled amounts of electrolytic gas present. Herein, we rely on the basic principle outlined above but describe devices that overcome these problems by appropriate design and demonstrate practical applications.
construction are given in the Supporting Information (Figure S1).
Figure 1. The ion transfer sample pretreatment devices. (a) Threelayer device (3L) and (b) five-layer device (5L). PP, plastic plate; SG, solution layer gasket; BP, bipolar membrane; M, ion transfer membrane; D, dialysis membrane. A photo of the 5L-40 device is shown in Figure S1 in the Supporting Information.
Automated Ion Isolation Interface to IC Systems. The ion isolation interface to the independent cation chromatography (CIC) and anion chromatography (AIC) systems consisted of the electrodialytic device, a syringe pump with 8port selector valve, three peristaltic pumps. and two solenoid valves. The details are in the Supporting Information (Figure S2). Briefly, the syringe pump handled the sample(s) and the other streams were pumped peristaltically. The acceptor streams flowed through the IC sample loops. The system alternated between ion isolation/transfer and cleaning/preparefor-new-sample modes, with a 15 min cycle time, largely controlled by the time for chromatography (Supporting Information, Figure S3).
■
EXPERIMENTAL SECTION Reagents. All chemicals were from www.nacalai.co.jp. Working standards were prepared as needed daily from stock solutions of analytical grade reagents. IC eluents were prepared from methanesulfonic acid, sodium carbonate, and sodium bicarbonate. Distilled−deionized water was used throughout. Electrodialytic Devices. The initial electrodialytic cation isolation device contained three solution layers (a 3L device) and an active length of 80 mm; the combined cation−anion isolation device had five solution layers (active lengths: 40/120 mm 5L-40/5L-120). The planar cascading structure of these devices is similar to presently available commercial suppressors22 as shown in Figure 1, except for a greater number of layers and addressable flow channels. Details of device
■
RESULTS AND DISCUSSION Electrophoretic Ion Transfer. In its simplest form, electrophoretic ion transfer from a sample to an acceptor can be achieved with a single barrier therebetween to prevent diffusive mixing. Initially, we investigated cation transfer in an annular tubular geometry with a Pt-wire inserted inside a dialysis hollow fiber surrounded by a stainless steel jacket (structural details in Supporting Information, Figure S4a). Without applied voltage (Vapp = 0), the maximum transfer to the acceptor was predictably ∼50%, but with Vapp = 3.5 V and a 5422
dx.doi.org/10.1021/ac300982k | Anal. Chem. 2012, 84, 5421−5426
Analytical Chemistry
Article
metal analyte ions on the sample side cannot occur but this may not be true for all analytes. Regardless of whether one is interested in one or both charge types as analytes, the 5L device will thus be more generally useful and we focused on that design. Simultaneous Cation and Anion Transfer from Sample Matrix with a Five-Layer Device. The sample channel is separated by individual ion transfer membranes on both sides from the acceptor channels which in turn are separated from the electrodes by the isolator membranes (Figure 1b). Figure 2 shows that from a illustrative sample
residence time (tR) of 5.4 s, nearly quantitative transfer (the equilibrium between electrophoretic transport and back diffusion) is reached. The electrophoretic mobility and the diffusion coefficient are related by the Nernst−Einstein equation.37 Interestingly, while the most electrophoretically mobile Ca2+ moves to the acceptor the most rapidly at the start, at equilibrium, it is the slowest diffusing Li+ that is most quantitatively transported. This arrangement, however, is not practical: electrolytic gases and redox conversions at the electrodes pose potential problems. The 3L device (Figure 1a) solves this problem by incorporating an isolator membrane between the acceptor channel and the cathode. We initially used a bipolar membrane (BPM) consisting of a CEM and an AEM bonded together.38 Upon application of a reverse voltage (CEM side negative), essentially, all the potential drops across the highly resistive interface. Theoretically, at Vapp ∼>2.1 V, the field strength becomes high enough to enhance dissociation of water.39,40 The H+ and OH− thus produced migrate through the CEM/ AEM to the corresponding electrodes. The ions generated by water dissociation can provide the necessary counterions for the transferred sample ions. The gasket screens in the solution flow path break up laminar flow and assist mass transfer to the membrane;41 this is not possible in the tubular device. The utility of the overall design for cation transfer was studied with different ion transfer membranes. Choice of Ion Transfer Membrane. The electrical field is the driving force for ion transport. Prima facie, the ion transfer membrane should have: (a) low electrical resistance, (b) high ion permeability, and (c) high charge selectivity. Only ion exchange membranes (IEMs) meet (c) and have been used for electrodialysis.42 However, a cation transfer experiment with a CEM in Figure 1a produces no analyte in the acceptor. The ion exchange capacity of the membrane alone is orders of magnitude larger than the analyte amount expected to be “transported” across. Most analyte ions have greater affinity for the ion exchange sites than H+/OH−, the membrane basically becomes a sink for the analytes. The analyte ions may be successfully transferred from the sample but little or nothing is detected in the acceptor. Even if the ion exchange capacity is deliberately reduced, the residual capacity will translate to sample carryover. Logically, the capacity will need to be zero; i.e., unlike industrial electrodialysis, an IEM cannot be used as the ion transfer membrane for analytical scale ion transfer. Data are presented in Figure S5 (Supporting Information) for cation transport experiments with 50 μM test cation solutions with (a) a CEM and (b) a porous hydrophilic membrane of 0.45 μm pore size. In (a), the analytes were barely detectable in the acceptor except at high Vapp. In (b), ∼≤75% was transported even at Vapp = 12 V; back diffusion through the large pores likely governs this. In contrast, a small pore size regenerated cellulose dialysis membrane (molecular weight cutoff (MWCO) 8 kDa) showed very good transport, with ∼50% transport by Vapp = 3 V and essentially quantitative transport by Vapp ≥ 4.5 V (Supporting Information, Figure S6). While we did experiment with membranes of both larger and smaller MWCO, none provided better overall performance than the 8 kDa MWCO membranes. The larger pore size membranes could not accomplish near-quantitative analyte transfer from the sample while the smaller MWCO membranes required a higher Vapp to perform at the same level. All further experiments were conducted with the 8 kDa MWCO membrane. In this particular experiment, oxidation of the alkali
Figure 2. Simultaneous cation and anion transfer from standard solutions with 5L-120. Isolator membranes were bipolar membranes (BPMs) or ion exchange membranes (IEMs). All solution flow rates: 300 μL min−1. Test solution: Mixture containing 100 μM each of all cations and anions indicated. The solutions were collected at the outlet of solution channels followed by IC analysis. Left and right panels show cation and anion in the solutions, respectively. Top, middle, and bottom panels show solution passed through acceptor (+), sample, and acceptor (−), respectively.
containing Li+, K+, Ca2+, CH3SO3−, NO3− and SO42− (100 μM each), ≥99% of each ion was removed to the appropriate acceptor channels. We additionally analyzed the anode/cathode acceptor solutions, respectively, for cation/anion analytes; no measurable concentrations were found. More than 99% of the influent anions was recovered in the acceptor with Vapp ≥10 and ≥20 V, respectively, with single IEMs (vide infra) and BPMs as isolator membranes. Predictably, Vapp needed for the same extent of analyte transfer was greater for the 5L compared to the 3L devices. Choice of the Isolator Membrane. The minimum thicknesses of commercially available BPMs (220 μm, Neosepta BP-1E, www.astom-corp.jp) are double that of corresponding IEMs (100−130 μm, CEM: Selemion CMV; AEM: Selemion DSV, www.agec.co.jp), resulting in proportionally greater voltage drops. To achieve near-quantitative ion transfer, significant current is needed. The voltage drop across the 5423
dx.doi.org/10.1021/ac300982k | Anal. Chem. 2012, 84, 5421−5426
Analytical Chemistry
Article
is accurately reflected in the measured acceptor effluent pH as a function of sample concentration (Figure 3); albeit at low
isolator membrane does not contribute to the motive force for analyte transport but only to increased heating that raises undesired diffusive back transport and decreases the net transport. It is not intuitive that applying voltages greater than the optimum can decrease the extent of desired transport: This can be perceived in Figure 2 but more clearly in Figure S7 (Supporting Information). Using CEM/AEM as respective anode/cathode isolators, Vapp needed to attain the same current decreased 2× compared to the use of two BPMs; Vapp needed for the maximum transfer also decreased accordingly. The equilibrium transport data for both BPMs and conventional ion exchange membranes as isolators are shown in Figure 2. Also, compared to a maximum transfer of ∼0.1 meq/L attainable with BPM isolators, the upper limit of quantitatively transferable analyte concentrations increased to 10 meq/L with IEM isolators (Supporting Information, Figure S8). Interestingly, anions require a higher applied voltage than cations to accomplish quantitative transport. At pH > 2.2, a cellulose surface carries a slight negative charge;43 the applied field needs to overcome the Donnan potential. Dimensional and Flow Rate Effects. Current requirements and the resulting Joule heating set the limits of the present device. With 3 mM KNO3 as the sample in a 5L-120 device, the observed current is >100 mA at Vapp = 25 V. 2.5 W dissipated at sub-mL/min flow rates substantially raises the temperature. As seen in i−V relationships at different concentrations of KNO3 as sample (Supporting Information, Figure S9), past Vapp = 2.0 V, the current increases sigmoidally, reaching a plateau at higher voltages where rate of ion transport limits the current flow; the plateau current increases with increasing sample concentration. The 5L-120 device was not tested to determine the maximum allowable sample flow rate. Completion of analyte ion removal depends on the electric field and tR. As Figure S9 (Supporting Information) also indicates, when the plateau current is reached, the electric field can be increased without increasing the current. Other conditions remaining the same, a smaller active length device would exhibit a proportionally smaller current. If it could handle samples of meaningful concentrations at reasonable flow rates, a smaller device would be preferred. We therefore built and tested the smaller 5L-40 device (Supporting Information, Figure S10). Complete transfer of Ca2+ and Li+ were attained by 10 V at a flow rate of 300 μL/min. At the higher sample flow rate of 600 μL/min, the Ca2+ was essentially quantitatively (>98%) transported by Vapp = 25 V and Li+ by Vapp = 30 V. For the anions, predictably the more mobile SO42− was transported faster than the less mobile CH3SO3−. Essentially quantitative transport of both were achieved by Vapp = 20 V at 300 μL/min but not until Vapp = 30 V at 600 μL/min. Vapp needed for quantitative transfer increases not only with flow rate but also with increasing analyte concentration. At a concentration of 50 μeq/L, lithium, potassium, calcium, methanesulfonate, bromide, nitrate, and sulfate were all quantitatively (>98%) transported up to a flow rate of 1000 μL/min at Vapp = 25 V. To put matters in perspective, at flow rates of 300 and 600 μL/min, the estimated sample residence time in the 5L-40 device is 3.7 and 1.8 s, respectively; this is by far faster than Donnan or conventional dialysis methods. Counterions. Cations and anions transferred into the respective acceptor solutions are charge balanced by OH− and H+, respectively, generated by electrolysis/water splitting. This
Figure 3. Acceptor solution effluent pH depends on the sample ion concentration. 5L-40 device; Vapp = 25 V; all flow rates, 300 μL/min.
analyte concentrations, the measured pH of these unbuffered solutions are understandably less acidic/basic than calculated values. It is potentially problematic that the cation/anion acceptor solution becomes basic/acidic, respectively. Especially if nonalkali/alkaline earth metals are of interest, these will precipitate in the hydroxide form, as will higher concentrations of Mg2+. Some anions, e.g., NO2−, S2O32−, ClO2−, are also unstable in acid solution and/or are highly reactive. Using a dilute acid or base (e.g., HCl/KOH), respectively, in the cation/anion acceptor channel is not only ineffective but also detrimental. No appreciable change in the effluent pH, compared to the use of DIW alone, is seen. All pH modifiers represent highly mobile ions, at low concentrations not only do they quickly disappear from where they are initially put but also contamination of the anion/cation acceptors with Cl−/K+ is immediate. If only one charge type of ions is to be analyzed, appropriate isolator compositions can be devised, e.g., for transition metal isolation; we will discuss these aspects elsewhere. Presently, we noted that for high concentrations of Mg2+, if an acid is added online for pH adjustment immediately to the cation acceptor effluent, complete recovery is observed. Volumetric dilution from acid/base introductions can be readily prevented by passive membrane-based acid/base addition, where the reagent can be in the gas phase.44−46 The device residence time is so short that any “precipitated” particles are still too small to settle and immediately redissolve in acid. Decomposition of susceptible anions, such as nitrite, sulfite, thiosulfate, etc., will also likely be prevented if base is immediately added.47 Behavior of Nonelectrolytes. Nonelectrolyte transport will occur by diffusion but should not be affected by an electric field. We studied the behavior of bovine serum albumin (BSA), dichloromethane, methyl isobutyl ketone (MIBK), benzene, and ethylbenzene. The effluents were analyzed by reversed phase HPLC-UV (or UV spectrophotometry alone for BSA), and the results were compared with those of small test cations and anions (Figure 4). Predictably, with a MW of 66.5 kDa, BSA shows zero transport across an 8 kDa MWCO membrane. Under an electric field, the ions were transported quantitatively to a specific acceptor while the nonelectrolytes were transported equally (within the precision of the present measurements) to both acceptors, to an extent of ∼0.5−10% in each, depending on the specific compound. Diffusivity of these nonelectrolytes, other than BSA, in water is almost equal to the 5424
dx.doi.org/10.1021/ac300982k | Anal. Chem. 2012, 84, 5421−5426
Analytical Chemistry
Article
Figure 4. Transfer efficiency of ionic vs nonionic components; 5L-40 device; all flow rates, 300 μL/min. All test compound concentrations: 100 μM, except BSA (100 μg/L).
ions tested.48 While the ions were almost quantitatively transferred at Vapp = 25 V, the transported nonelectrolyte concentrations showed no dependence on applied voltage. Reproducibility, Sample Carryover, Preconcentration Potential. Reproducibility was examined at three concentration levels, containing 10, 100, and 1000 μM of each ion, respectively. The standard deviation (Supporting Information, Figure S11) ranged from 4.2−7.1, 1.3−3.8, and 1.8−3.3%, respectively, at the 10, 100, and 1000 μM levels (n = 10 each). The experiments were carried out in the low to high concentration order; carryover was measured by running blanks immediately after 1000 μM samples to be 0.1 ± 0.08% (n = 10). We have reported here only on the matrix isolation capability of the present system and not on its preconcentration abilities. The astute reader will realize that preconcentration is readily performed using an acceptor/sample flow rate ratio
