Effect of channel geometry on ion-concentration polarization-based preconcentration and desalination

Abstract

Polarization of the ion-selective systems results in the formation of ion-depleted and ion-concentrated zones in the electrolyte layers adjacent to the system. One can employ ion-concentration polarization for the removal of charged large molecules and small ions from the flowing liquid. Removal of large molecules from the flowing solution and their local accumulation is often referred to as preconcentration, removal of small ions as desalination. Here, we study the effect of the channel geometry on the removal of charged species from their water solutions experimentally. Straight, converging, and diverging channels equipped with a pair of heterogeneous cation-exchange membranes are compared in terms of their effect on preconcentration of an observable fluorescein dye and on desalination of water solution of potassium chloride. Our results show that preconcentration of the dye is not significantly affected by the channel geometry. The distance of the preconcentration band from one of the membranes was approximately the same in all tested channel geometries. The major difference was in the location of the band within the channel, when the conical channels localized the band at one of the channel walls. The straight channel showed a slightly broader range of applicable flow rates. The semibatch desalination of 0.01M KCl solution turned out to be more efficient in conical channels, which was associated with a larger volume of the channel available for the accumulation of the concentrated solution. Our results suggest that conical channels can be advantageously used in transforming the ion-concentration-polarization-based semibatch desalination into a fully continuous one.

I. INTRODUCTION

Ion-concentration polarization (ICP)^1–3 is a phenomenon that has been intensively studied in the last two decades specifically in the view of its potential to preconcentrate biomolecules (such as proteins^4,5 or DNA^6,7) for their downstream processing.^8,9 Ion-concentration polarization occurs at systems exhibiting so-called ion selectivity.^10–12 Ion selectivity is based on electrostatic interactions between mobile ionic species in the electrolyte solution and a so-called fixed charge present in the ion-selective system. Fixed charges are dissociated functional groups covalently bound (immobile) on the solid support. The effective characteristic dimensions over which the electrostatic forces act in water solutions are on the order of nanometers. The scale of electrostatic interactions predetermines the internal characteristic dimension of ion-selective systems that has to be on the same order. Ion-selective nanochannels,^13,14 ion-selective particles (ion-exchange resin particles),^15,16 or ion-selective membranes (ion-exchange membranes in the field of desalination)^17,18 are examples of the aforementioned systems.

One can use the ion-selective systems in a passive way for simple ion-exchange;¹⁹ however, the connection of DC electric field on these systems brings a lot of scientifically interesting phenomena and opens up the door for many unprecedented applications such as ICP-based preconcentration and desalination. One of the major consequences of the polarization in a DC electric field is a formation of ion-depleted and ion-concentrated zones on the sides of an ion-selective system.³ The ion-depleted zone is used in ICP-based applications to control the transport of ions, and its occurrence is accompanied by the formation of electroconvection localized in the ion-depleted zone.^20,21 Electroconvection refers to the bulk motion of the electrolyte caused by the electric field, and it is a result of the action of the electric field on a spatial charge formed by mobile ions. The mechanisms underlying the electroconvective motion, which is common to all ion-selective systems,^22–24 depend on the experimental conditions and the geometry of the ion-selective system.^25–27 These mechanisms have been nicely reviewed in several recent articles.^28,24 A water-splitting reaction proceeding at the interface between the ion-selective system and the electrolyte might be another phenomenon occurring at ion-selective systems on the formation of the ion-depleted zone.^29–32 A water-splitting reaction generating H⁺ and OH⁻ ions is characteristic for ion-exchange membranes and ion-exchange resin particles. Its occurrence has a negative effect on ICP-based applications. For that reason, cation-exchange membranes (such as Nafion) are preferred to anion-exchange membranes because anion-exchange membranes show higher susceptibility to the water-splitting reaction.

The flow of a water solution through the developed ion-depleted zone results in the removal of ionic species from the flowing fluid. The degree of the ion removal depends primarily on the velocity of the flow and the applied voltage sustaining the ion-depleted zone.⁷ The ionic species, especially the large ones being the coions with respect to the fixed charge in the ion-selective system, are not allowed to enter the ion-depleted zone due to strong electric force acting upon these molecules in the opposite direction. Also, small coions are mostly removed from this zone for the same reason. However, a low concentration of these coions is found in the ion-depleted zone in which they function as a bridge for the transport of counterions. Small counterions are the major electric current carriers, and their flux sustains the ion-depleted zone. At the same time, a certain amount of counterions remain with removed coions because of the electroneutrality condition. The overall result of the simultaneous action of the ion-depleted zone and forced convection is the accumulation of ionic species at the ion-depleted zone and desalination of the passing liquid. This phenomenon has found many interesting applications related to preconcentration,^33,34 biosensing,^8,35 separation,^9,36,37 ionic control,^38,39 and desalination.^40–42

The pioneering work related to the use of ion-depleted zones for removal of proteins from a given solution was published by Astorga-Wells et al.⁴³ The authors used a simple capillary equipped with two cation-exchange membranes to realize the ion-concentration polarization preconcentration. By tuning the flow rate and voltage, they were able to selectively preconcentrate certain proteins and release them into a mass spectrometer for downstream analysis. Later, Han and co-workers exploited a selectivity of a nanochannel connected to a microchannel for preconcentration of fluorescently labeled proteins.⁴ The application of the voltage on the system drove both electro-osmotic flow inside the microchannel, which brought the molecules of interest toward the nanochannel-microchannel junction, and sustained the ion-depleted zone. The preconcentration factor of the used protein reached 10⁶. Since then, many publications on the use of ion-concentration polarization preconcentration, often in association with improved biosensing, have appeared.^44,45 This type of preconcentration works for charged species in general. One exception to this rule might be large counterions in association with ion-selective membranes. These ions may adsorb on the surface of the membranes and cause their fouling accompanied by deleterious effects such as water splitting.²⁹

Besides preconcentration, the ion-depleted zone can be used for desalination of flowing solutions. One of the major advantages of this approach is the fact that the membranes are prevented from a certain type of fouling by the very presence of the ion-depleted zone. The ion-depleted zone limits possible interactions of membranes with foulants. ICP-based desalination, unlike electrodialysis,⁴⁶ also allows one to use only one type of the membranes, e.g., cation-exchange ones.^41,47 Two cation-exchange membranes creating two opposite walls of a microchannel formed depletion and concentration zone within the same channel and allowed continuous withdrawal of partially de-ionized water and a concentrated solution. One of the major problems of this type of desalination was control of the position of the boundary between the ion-concentrated and ion-depleted zone. Later, such a desalination unit was equipped with porous membranes that controlled the flow of water solution through the main channel and divided the solution into brine, desalinated solution, and a solution of medium salinity.⁴⁸

In this work, we study the preconcentration and desalination experimentally in two cation-exchange membrane systems, which vary in the geometry of the channel. Namely, we use a straight channel, converging [wide-to-narrow (WN) conical channel], and diverging channel [narrow-to-wide (NW) conical channel]. The geometry of the channel affects the local values of velocities and electric fields, which in turn can influence IPC-based phenomena. We perform two larger experimental studies, one aimed at direct observation of the preconcentration of a fluorescent dye and the other one on the degree of desalination one can achieve in this semibatch desalination. Both studies are performed in the dependence on the applied voltage and flow rate.

II. EXPERIMENTAL SECTION

A. System fabrication

All systems used in this work were fabricated from PMMA plates and heterogeneous cation-exchange membranes (Ralex CEM) kindly provided by MemBrain a. s., Czech Republic. First, two PMMA plates with a size of 70 × 30 mm² were structured by mechanical milling. The bottom plate contained a channel of given dimensions (the depth of 300 μm, the width depending on the geometry, see Fig. 2), the top plate was provided with two openings for insertion of small pieces of the heterogeneous cation-exchange membrane. Two pieces of the CEM were thoroughly sealed by Acrifix 192 UV curable glue (2 cm apart) at the designated locations. The sealed membranes were let swell in a KCl solution for 48 h, and then both sides of the PMMA plate were polished on a polisher. Polishing of the top plate removed parts of the membranes that stack out of the PMMA plate. In the next step, the two PMMA plates were aligned and tightly bonded by our solvent assisted thermal bonding technique.⁴⁹ The top PMMA plate containing the membranes was provided with reservoirs for placing source electrodes, the inlet and outlet channels with short pieces of Tygon tubing that served as fittings for Teflon tubing. The schematics of the used chips can be seen in Figs. 1(a) and 1(b).

FIG. 2.

The microfluidic systems used in the experimental study with (a) converging channel, (b) straight channel, and (c) diverging channel. The upper row of pictures shows the top view of the systems; the bottom row of pictures shows the bottom view of the systems. The channels were filled with a concentrated trypan blue solution for easy orientation.

FIG. 1.

Schematics of the chips used in the experiments: (a) side view and (b) 3D view. The schematics show the chip with a straight channel. The chips with converging and diverging channels differed only in the geometry of the channel.

All fabricated systems studied in this work are depicted in Fig. 2. These systems are (i) a system with a converging channel referred to as a wide-to-narrow conical channel [Fig. 2(a)], (ii) a system with a straight channel [Fig. 2(b)], and (iii) a system with a diverging channel referred to as a narrow-to-wide conical channel [Fig. 2(c)]. This figure also contains all dimensions of each microfluidic system along with the direction of flow and connection of the electric voltage.

B. Experimental setup

The studied experimental systems were connected to a programmable double syringe pump model 4000 (from New Era Pump Systems, Inc.) by Teflon tubing with an inner diameter of 0.8 mm. The syringe pump was used for the delivery of fresh solutions at a given flow rate. Two gold electrodes placed in the membrane reservoirs were used as the source electrodes and were connected to an electrophoresis power supply Consort EV232. The positively biased electrode was placed on the concentrating membrane (CM) and the negatively biased electrode (ground) on the depleting membrane (DM). We purposely inserted a resistor with a resistance of 1000 Ω in the electric circuit (in series with respect to the fluidic system) and measured its voltage in the course of the experiment. This voltage was measured on a multimeter Agilent 34410A. The ratio of the voltage and the resistance of the resistor give the electric current that passes through the circuit. The chosen connection protects the equipment from electric shortcuts, which may occur in the case of the electrolyte solution leaks. The system was placed over a camera Pixelink PL-D722CU. A dark reader transilluminator from Clare Chemical Research was used in experiments with fluorescein for its exposure. The measurement of the voltage and the imaging of preconcentration in the chip were controlled by a script written in a Matlab environment.

C. Preconcentration and desalination experiments

We performed two different experimental studies. In the first one, we observed preconcentration of a negatively charged fluorescein dye and the dependence of its preconcentration in the three systems on the applied voltage and flow rate. In the second experimental study, we focused on the degree of desalination of a 0.01M KCl solution. The experiments related to the preconcentration of the fluorescein dye were carried out in the following manner. The channel was filled with the prepared solution of the fluorescein dye, and the chambers for placing the source electrodes were filled with a 10× TAE buffer. The measurement of the voltage and the imaging of the system were initiated in Matlab. The pictures of the system were taken with a frequency of 1 per 10 s. A chosen voltage was connected to the system in which no flow of the electrolyte solution was applied yet. The syringe pump set at a given flow rate was turned on approximately 200 s after the connection of the voltage. The whole experiment took around 1800 s, after which the experiment was terminated, and all measured data saved. The experimental study was conducted for the applied voltages of 50, 100, 150, 200, 250, and 300 V and the flow rates of 1, 2, 5, 10, and 20 μl/min. The voltages yielded average electric fields of 1.7, 3.3, 5, 6.7, 8.3, and 10 kV/m. The average electric fields were calculated by dividing the applied voltage by the distance of the source electrodes. This distance was approximately equal to 3 cm. The flow rates calculated for a flow-through area of 0.6 mm² gave linear velocities of 1.7, 3.3, 8.3, 16.7, and 33.3 mm/min. This flow-through area corresponds to that one of the straight channels and the smallest flow-through areas of the converging and diverging channels. We also use this flow-through area in Sec. III. The last picture taken in each experiment was then analyzed with respect to the efficiency of preconcentration and the location of the front meniscus of the fluorescein band measured from the concentrating membrane. The concentration of fluorescein in the original solution was 10⁻⁵ M.

The desalination experiments proceeded in a very similar way. First, the Matlab script initiated the measurement of the voltage, and then a given voltage was connected to the system along with the flow of the chosen flow rate. We started to collect the samples of the desalted electrolyte after 20 min from the commencement of the experiment. The sample collection also took 20 min. The electrolytic conductivity of the sampled electrolyte was measured on conductometer LAQUAtwin EC-33. This conductivity (σ_output) was compared with the initial conductivity of the 0.01M KCl (σ₀), and a degree of desalination was calculated as the ratio of (σ₀ − σ_output)/σ₀. The initial electrolytic conductivity of the 0.01M KCl was 1.17 mS/cm.

III. RESULTS AND DISCUSSION

A. Preconcentration of the fluorescein dye

Our experimental systems do not rely on the electro-osmotic delivery of the fresh electrolyte solution to the preconcentration region of the system, and thus, the applied voltage driving the depletion under the depleting membrane and the flow rate of the electrolyte solution can be controlled independently. There is an opposite effect of these two factors on the preconcentration of the fluorescein dye. While larger applied voltages will develop longer depletion region and thus will provide for better conditions to preconcentrate the dye in the space between the depleting and concentrating membrane, higher flow rates will push the fresh electrolyte into the depletion region and thus will decrease the effect of the preconcentration.⁷ In the first experimental study, we carried out a set of experiments that studied the effect of the flow rate and the applied voltage on the quality of preconcentration, i.e., whether the preconcentration occurs under the given condition and what shape the preconcentration band takes. Then, we analyzed the position of the preconcentration band in the preconcentration channel and found the limiting (or critical) condition under which the preconcentration works with 100% efficiency. This 100% efficiency is defined as preconcentration in which we cannot detect any fluorescein leaks from the preconcentration channel; i.e., no fluorescent signal is detected in the outlet channel. The major aim of the study is to find out whether the channel geometry affects the critical parameters of the preconcentration, which could be potentially tuned for a given application.

Figures 3 and 4 show the experimental results obtained from this experimental study. Figure 3 shows the effect of the applied voltage on the formation of the fluorescein preconcentration band under the constant flow rate of 5 μl/min for a (a) wide-to-narrow conical channel (WN channel), (b) straight channel, and (c) narrow-to-wide conical channel (NW channel). The flow rate of 5 μl/min corresponds to the linear velocities of 8.3 mm/min for the straight channel, 8.3 decreasing to 2.8 mm/min for the NW channel, and 2.8 increasing to 8.3 mm/min for the WN channel. The applied voltage has a strong effect on the position of the preconcentration band. The fluorescent signal moves further away from the depleting membrane with the increase in the applied voltage. Qualitatively, this dependence is approximately the same for all systems studied. At the voltage of 50 V, the signal is seen in the whole preconcentration channel indicating that a small amount of the fluorescein leaks through this depletion region. However, starting with a voltage of 100 V, no fluorescent leaks are observed for both WN and NW channels in which a clear dark zone develops between the depleting membrane and the preconcentration band. In the case of the straight channel, one may still observe small leaks around the upper wall under this voltage. With further increase in the voltage, the signal for fluorescein occupies a smaller space within the channel; in other words, the fluorescein becomes more concentrated. The preconcentration, thus, works well under these conditions. The shape of the fluorescent bands depends on the geometry of the channel. The preconcentration band in the WN channel is mostly localized along the upper wall (as seen in the images) with faint fluorescence spread across the center of the channel. In the case of the straight channel, the fluorescence of the preconcentration band is across the whole channel with slightly higher fluorescence coming from the upper wall. The signal for the NW channel is voltage-dependent. While for the low voltages the highest fluorescence comes from the bottom part of the channel, this strong fluorescence moves to the top part of the channel for higher source voltages. In all cases, the preconcentration band tends to be located at one of the channel walls. This observation has its probable origin in the asymmetry of the channel (with respect to the centerline of the channel) and more importantly in strong electroconvection developing under the depleting membrane. Our recent results show that strong electroconvection is intrinsic to cation-exchange membranes used in this work when a sufficiently large voltage is applied to the system.¹⁶ At the same time, the appearance of the electroconvection is dependent on the voltage, and it takes a form of a vortex array at smaller voltages and one large vortex for large voltages. Although the values of the voltages are significantly higher in this work, the shape of the vortex significantly affects the shape of the preconcentration band, which is very dynamic in its nature. The vortex influences local velocity fields and eventually pushes the preconcentration band toward one of the walls. While this effect does not have to be beneficial for biomolecule preconcentration, it can be significant in continuous desalination by ion-concentration polarization. Here, the separation of the ion-concentrated solution and ion-depleted solution introduces significant challenges. From this perspective, the narrow-to-wide conical channel will provide the best geometry.

FIG. 3.

The effect of the applied voltage on the formation of the preconcentration band of fluorescein at the flow rate of 5 μl/min for a (a) wide-to-narrow conical channel, (b) straight channel, and (c) narrow-to-wide conical channel. The red dashed lines indicate the position of the depleting (DM) and concentrating (CM) membrane and the blue dashed lines the boundary of the preconcentration channel. The velocities in the figure are local linear velocities.

FIG. 4.

The effect of the flow rate on the formation of the preconcentration band of fluorescein at the voltage of 150 V for a wide-to-narrow conical channel (a), straight channel (b), and narrow-to-wide conical channel (c). The red dashed lines indicate the position of the depleting (DM) and concentrating (CM) membrane and the blue dashed lines the boundary of the preconcentration channel. The velocities in the figure are linear velocities.

Figure 4 shows the dependence of the preconcentration band formation on the flow rate applied to the system at a constant voltage of 150 V. The displayed images indicate the opposite role of the flow rate. The front boundary of the preconcentration band moves toward the depleting membrane with increasing flow rate. Qualitatively, the shift is very similar in all three channels. Interestingly, the straight channel starts to leak significantly at the flow rate of 10 μl/min (linear velocity of 16.8 mm/min) and the other two systems at 20 μl/min (linear velocity of 33.3 mm/min). The shape of the preconcentration band corresponds to those observed in Fig. 3. In the case of the WN channel, the fluorescein localizes at the top part of the channel with detectable fluorescence across the whole channel. The fluorescence signal for the straight channel is evenly distributed across the channel at the flow rates of 1 and 2 μl/min (linear velocity of 1.7 and 3.3 mm/min), and then it concentrates at the top wall for 5 μl/min (linear velocity of 8.3 mm/min). The NW channel develops a fluorescent band rather in the bottom part of the channel with a high-intensity fluorescent spot at its front.

When compared to the shape of bands displayed in Fig. 3, one can see that the qualitative appearance of the fluorescence signal is the same for the WN channel and the straight channel (mostly the position of the fluorescent zone changes). There are, however, differences observed in the NW channel. Increasing voltage tends to drive the preconcentration band toward one of the walls and clearing most of the preconcentration channel. Unlike that, the increasing flow rate tends to develop highly fluorescent zone close to the centerline of the channel. We believe that this behavior is given mainly by the mutual interaction of pressure-driven flow and electroconvection developing under the depleting membrane.

To quantify the obtained results, we evaluated the distance of the front meniscus of the preconcentration band from the concentrating membrane. The distance equal to 20 mm (distance between the concentrating and depleting membrane) means that the fluorescence spans the whole preconcentration channel and most probably some fluorescein leaks through the depleting region; the distance equal to 0 mm would correspond to a situation in which all fluorescence would be detected on the concentrating membrane. The obtained data are plotted in Fig. S1 of the supplementary material. Figure S1 in the supplementary material shows the dependence of the distance on the applied voltage for the WN channel [Fig. S1(a) in the supplementary material], straight channel [Fig. S1(b) in the supplementary material], and NW channel [Fig. S1(c) in the supplementary material]. Each curve in these graphs is for a constant value of the flow rate. All the graphs show that the distance of the preconcentration band from the concentrating membrane decreases with increasing voltage. The increasing flow rate has the opposite effect; i.e., the distance increases with increasing flow rate. The shape of the curves is qualitatively very similar for all three channels. Since it is essentially impossible to directly compare the data from all three channels in a single plot, we carried out another experiment in which we determined a critical flow rate and critical linear velocity at a given connected voltage. The critical flow rate and critical linear velocity are defined as the flow rate and linear velocity at which the fluorescein starts leaking from the preconcentration zone under the given voltage. The obtained results are plotted in Figs. 5(a) and 5(b) for the critical flow rate and critical linear velocity, respectively.

FIG. 5.

(a) Dependence of the critical flow rate on the connected voltage for the NW channel (black line), straight channel (red line), and WN channel (blue line) and (b) the dependence of the critical linear velocity on the average electric field for the NW channel (black asterisks), straight channel (red asterisks), and WN channel (blue asterisks). Straight dashed lines are linear fits of the experimental data. The fitting equations are given in the main text.

The critical flow rate [Fig. 5(a)] monotonously increases with increasing voltage for all tested channels. Interestingly, the results are almost identical for the two conical channels. Unlike that, the critical flow rate is approximately 15% higher for the straight channel than the conical ones. However, from the point of both desalination and preconcentration, this difference does not play any significant role. The critical linear velocity, which was calculated as the ratio of the found critical flow rate and the flow-through area of 0.6 mm², is linearly proportional to the average electric field strength [Fig. 5(b)]. This area is the flow-through area of the straight channel and the smallest area of the converging and diverging channels. The experimental data were fitted with linear functions. These equations allow one to obtain a critical value of the linear velocity based on the applied average electric field. Both values are theretically transferable to other systems. The equations are (i) v_NW = 3.53 × 10⁴ for the NW channel, (ii) v_S = 3.94 × 10^7.3 for the straight channel, and (iii) v_WN = 3.34 × 10^4.4 for the WN channel. The electric field is substituted in kV/m, and the resulting linear velocity is in mm/min. As documented below, these numbers correlate well with the values of the average electric fields and linear velocities used in the work of others.^43,50 For example, the linear velocities used for the preconcentration of proteins in a system made of a PEEK capillary equipped with two pieces of Nafion membranes⁵⁰ ranged from 16 to 80 mm/min under the applied voltages between 300 and 1000 V. In another work coming from the same authors, the same range of linear velocities was used when the average electric fields were between 6.3 kV/m and 63 kV/m.⁴³

The analysis of the experimental data shows that the channel geometry does not significantly affect either position, more precisely the distance from the concentrating membrane, or critical parameters for preconcentration of the fluorescein. On the other hand, it significantly influences the shape of the preconcentration band. While the band is localized across the channel or very close to the centerline in the case of the straight or NW channel at low flow rates, most of the fluorescence comes from the channel wall in the case of the WN channel. One can, thus, use different channel geometries depending upon the application.

B. Desalination of potassium chloride solution

In the second parametrical study, we carried out a simple desalination of 0.01M KCl solution by using the ion-depleted region as a nonmechanical barrier for small ions. These experiments were performed in a semibatch mode in which the desalted solution was continuously withdrawn from the chip and the concentrated solution accumulated in the preconcentration channel. We measured the conductivity of the collected desalted solutions and evaluated the degree of desalination as $D=\frac{{\sigma }_0-{\sigma }_{\mathrm{output}}}{{\sigma }_0}$. The obtained results are plotted in Figs. 6(a)–6(f) as the dependencies on the flow rate and the linear velocity, respectively. Specifically, Figs. 6(a) and 6(d) show the degree of desalination obtained for the WN channel, Figs. 6(b) and 6(e) for the straight channel, and Figs. 6(c) and 6(f) for the NW channel.

FIG. 6.

The dependence of the degree of desalination on the applied flow rate under selected source voltages for a (a) wide-to-narrow conical channel, (b) straight channel, and (c) narrow-to-wide conical channel. The dependence of the degree of desalination on the linear velocity under selected applied voltages for a (d) wide-to-narrow conical channel, (e) straight channel, and (f) narrow-to-wide conical channel.

The most efficient desalination occurs at low flow rates and high source voltages. Interestingly, the flow rate of 5 μl/min (a linear velocity of 8.3 mm/min) provides desalination degree close to 1 for all systems. This degree of desalination is independent of the applied voltage. These experiments evidence the fact that ICP with forced flow removes both coions and counterions efficiently from the flowing solution and that any preconcentration of large biomolecules is accompanied by a significant increase in the ionic strength due to the increase in the concentration of small ions. The degree of desalination then decreases with increasing flow rate and decreasing applied voltage for the straight channel. It reaches less than 50% under the least favorable conditions corresponding to the flow rate of 20 μl/min (a linear velocity of 33.3 mm/min) and the voltage of 100 V. By comparing this degree of desalination with the position of the fluorescent band within the preconcentration channel [see Fig. S1(b) in the supplementary material], one can see that the fluorescein dye leaks through the depletion region under these conditions. This observation is consistent with the low degree of desalination obtained under the same conditions. Interestingly, the degree of desalination is better for both conical channels tested in this work. The degree of desalination is very close to one for the flow rates up to 15 μl/min (linear velocity of 25 mm/min) when higher voltages are applied (200 or 300 V). The degree of desalination for 100 V decreases with increasing flow rates and reaches very small values for the highest flow rate (20 μl/min). The reason why the conical channels are better in the semibatch desalination of the KCl solution might be in larger volumes of the preconcentration channels, which allow more ions to be accumulated in that part of the systems.

The results obtained from both the preconcentration and the desalination experiments indicate that the conical channels could be advantageously used in the desalination of flowing solutions. They develop the preconcentration band at one of the preconcentration channel walls and offer higher capacity for accumulating small ions. Both facts can play a significant role in the transformation of the semibatch desalination into a fully continuous mode (both desalted and concentrated solutions are withdrawn continuously). Although this type of desalination cannot, at this moment, compete with large desalination units in terms of the throughput and overall performance, its realization offers an advantage in limited interactions of potential foulants with the membrane and can thus prolong the desalination cycles without the need to clean the membranes or replace them.

IV. CONCLUSION

We studied the effect of the channel geometry on the ion-concentration-polarization-based preconcentration and desalination experimentally. ICP was realized with the use of two heterogeneous cation-exchange membranes, which allowed independent control of the applied voltage sustaining the depleting region in the preconcentration channel and the flow rate of the tested water solutions. We tested two conical channels, one converging, one diverging, and one straight channel. Our preconcentration experiments realized with a fluorescent dye showed that there is a minimal effect of the channel geometry on the operating conditions under which successful removal of the fluorescent dye from the flowing solution occurs. The band of the preconcentrated fluorescein dye occurred at similar distances in all three chips when the same conditions of the preconcentrations were applied. The straight channel had a larger range of applicable flow rates. The differences were in the shape of the preconcentration band. Both conical channels showed a tendency to accumulate the fluorescein dye at one of the channel walls and the straight channel across the whole channel. Unlike preconcentration, the semibatch desalination showed better efficiency in the conical channels. The applicable range of flow rates was larger, which was probably associated with the fact that the conical channels have larger volumes and may accumulate more ions. The obtained results indicate that the ICP desalination of flowing water solutions will benefit from the conical geometry of the channels.

SUPPLEMENTARY MATERIAL

See the supplementary material for a complete set of experimental data related to preconcentration of fluorescein.