Gradient Elution Moving Boundary Electrophoresis
(GEMBE), a Brief Users Guide

David Ross
National Institute of Standards and Technology


Gradient elution moving boundary electrophoresis is a recently described technique for electrophoretic separations using microfluidic channels or capillaries [1]. The microfluidic device required for GEMBE is simple, consisting of two reservoirs connected by a relatively short, straight microchannel or capillary (see Figure 1). The channel and one of the reservoirs is filled with a run buffer and the other reservoir is filled with a sample to be analyzed. As with more conventional methods for electrophoretic separation, a high voltage is applied to the ends of the channel to drive electrophoretic separation of the analytes. However, with GEMBE there is no injection. A controlled solution counterflow is used to selectively allow analytes to enter the channel, one at a time, where they are detected as step-wise increases in the detector signal. The counterflow is typically controlled through application of a pressure to the head space of the run buffer reservoir. At the beginning of a separation, the applied pressure is high, so that none of the analytes of interest will enter the channel. Over time, the pressure is gradually reduced so that each analyte can enter the channel for detection and quantitation.

Figure 1. Schematic of GEMBE apparatus. Detailed discussion of each of the components is included in the text.

GEMBE has a number of advantages over conventional methods of electrophoresis: 1) The simple channel structure is much easier to implement in a multiplexed format. 2) High resolution separations can be performed with relatively short channels, making device fabrication simpler and less expensive. 3) Because there is no injection, there is less variability in quantitative measurements of analyte concentrations. 4) The trade-off between separation time and resolution in GEMBE can be controlled by the rate at which the pressure is varied rather than by the length of the separation channel and the electroosmotic mobility, so method development and optimization is much faster and easier. 5) The counterflow can be used to exclude proteins, particulates or other matrix interferents from entering the separation channel, so that GEMBE can be used to analyze complex samples with very little sample preparation [2].

Because GEMBE is qualitatively different in a number of ways from the more conventional electrophoretic or chromatographic separation methods, the construction and mode of operation of a GEMBE apparatus is not intuitively obvious for many who have worked with other analytical separation techniques. Therefore, in order to facilitate its wider use, the goal of this article is to provide a description of the hardware required for a GEMBE separation, and a brief guide to the selection of optimized GEMBE separation parameters.

Equipment Required*

The equipment required for GEMBE consists of a separation channel, reservoirs for buffer and sample, a high voltage source, a pressure controller, and a detector. The basic layout is shown in Figure 1.

Separation Channel. The separation channel for GEMBE can be any of the types of microchannels or capillaries commonly used for capillary electrophoresis (CE) separations. Generally the best channel for GEMBE will be shorter in length and have a smaller inner diameter than the typical channel or capillary used for CE. The smaller inner diameter is to minimize Taylor-Aris dispersion [3, 4] that results from the pressure driven flow used with GEMBE.

The most commonly used GEMBE separation channels in our lab are short lengths of fused silica capillary, with an inner diameter from 5 μm to 30 μm and a length from 3 to 5 cm. We have also used simple microfluidic channels machined in plastic, wet etched in glass, or bought from a commercial vendor.

Reservoirs. For a GEMBE separation, there are two reservoirs: the sample reservoir and the run buffer reservoir. Generally, the run buffer reservoir is larger and is sealed to allow for pressure control of the head space for regulation of the buffer counterflow. The fluid volume in the run buffer reservoir is generally left large enough so that the flow of analytes through the capillary during a GEMBE separation will be sufficiently diluted in the run buffer reservoir so that it will not interfere with subsequent separations. In our lab, we typically use run buffer reservoirs with a fluid volume of 1 mL or 2 mL.

The sample reservoir is generally left at ambient pressure, but is often covered to reduce evaporation, particularly for small samples and/or long analysis times. It is important that the sample reservoir can be easily rinsed with fresh buffer or sample solution between separations. The volume of sample reservoirs used in our lab range from 20 μL to 1 mL, though 200 μL is the preferred volume.

When working with fused silica capillaries, the connection to the sample reservoir does not have to be pressure tight, so it is typically made simply by drilling a 360 μm diameter hole (to match the o.d. of the capillary) through the wall of the sample reservoir and carefully sliding the end of the capillary through the hole. To make a reliably pressure tight seal between the capillary and the run buffer reservoir, we typically use a miniature capillary compression fitting (of the type sold by Upchurch Scientific or LabSmith). This allows the run buffer reservoir to be reused many times without leaks. Connection to the pressure controller and high voltage source is best done through the lid of the run buffer reservoir.

When working with microfluidic chips the choice of reservoir is dependent on the chip interface. For flat plastic or glass chips, we have used clamp-on reservoirs with o-rings or PDMS gaskets to seal to the chip surface. Another approach we have used is high performance double-sided tape to attach both sample and run buffer reservoir to a flat glass or plastic chip. When microfluidic chips are available with Luer lock ports, we have taken advantage of this by using a cut-off plastic syringe as the buffer reservoir, and a modified syringe plunger for the lid.

High Voltage Source. Any of the high voltage power supplies commonly used for capillary or microchip electrophoresis would also work for GEMBE. However, lower maximum voltage is required than for conventional capillary electrophoresis (because of the shorter channel length used with GEMBE), and only a single high voltage channel is required, since there is no need to define an injection as with microchip electrophoresis.

Pressure Controller. We have found that control of the counterflow in GEMBE is most easily accomplished by controlling the pressure applied to the head space of the run buffer reservoir. We have also tried using a liquid-filled syringe to directly control the volume flow rate through the separation channel, but found that method to be irreproducible because of the long time constant associated with device mechanical compliance, and uncontrolled volume changes due to electrolysis at the electrodes. With control of the head space pressure, the response to changes in pressure is almost instantaneous and the flow rate is insensitive to electrolysis.

The desirable features of the pressure controller are: 1) It should have a wide enough range to drive fluid flow at velocities comparable to the electrokinetic velocity of the analytes to be measured (10 kPa or 100 kPa is sufficient for most microfluidic geometries). 2) It should be precise and stable enough to drive a smooth, steady flow through the separation channel. 3) It should be controllable via a computer for application of a pressure ramp and synchronization with the electrophoresis voltage and detector signal. In addition, for the greatest flexibility, the pressure controller should be bi-directional so that it can apply both positive and negative pressures (relative to ambient). There are very few commercially available products that have all the desired features, and they are typically high precision pressure control and calibration systems. They are fairly expensive ($10,000 to $15,000 in 2006), though not much more so that the multi-channel high voltage power supplies often used for microchip electrophoresis. In our lab we have used the APC600 from Mensor, typically with a p.s.i. ( kPa) range.

Detector. In principle, any detector compatible with conventional CE could also be used with GEMBE. In our lab, we have used fluorescence detection [1] (both arc lamp/microscope and laser/photomultiplier tube systems), contactless conductivity detection [2, 5] (TraceDec, Groton Biosystems, Boxborough, MA), and channel current detection [6, 7] (in which the electrophoresis current through a very short separation channel is used as the detector signal).

Operation and optimization of a GEMBE separation.

Once the hardware configuration is set, the important operational parameters for a GEMBE separation are the starting pressure, the voltage, the pressure ramp rate, and the duration of the pressure ramp. In this section, we will discuss the operation of the GEMBE apparatus and the selection of the optimal values for those parameters.

After the GEMBE instrument is assembled, the run buffer reservoir is filled with buffer and sealed. A positive pressure is then applied to fill the separation channel and rinse it with run buffer. For a 5 cm long, 15 μm i.d. capillary, we typically use a 30 kPa rinse pressure (flow velocity approximately 4.7 mm/s, volume flow rate approximately 0.8 nL/s). After rinsing for approximately 2 minutes, the sample reservoir is also filled with run buffer. The rinse pressure is typically applied continuously between measurements and during sample loading to prevent bubbles from entering the separation channel.

The procedure for starting a GEMBE separation is as follows: 1) With the high voltage off and the pressure set to the rinse pressure, a new sample is introduced into the sample reservoir. 2) The ground electrode is inserted into the sample reservoir (see Figure 1). 3) The high voltage is turned on and the pressure is reduced to the starting pressure for the separation. The pressure is typically held at this value for 10 s to 20 s. 4) The pressure is then ramped downward until all of the analytes of interest are detected. 5) The pressure is then increased to the rinse pressure, and the voltage is turned off.

Initial Separations and Rough Optimization. With a completely new system, it is best to first run a rapid, low resolution separation to roughly identify the appropriate pressure range for the separation. An example is shown in Figure 2 for the separation of sodium chloride, potassium phosphate, and glycerol-3-phosphate. Initially, a high start pressure and a rapid pressure ramp were used. The resulting plot of detector signal vs. time (Figure 2A) shows steps for the anions chloride and phosphate (glycerol-3-phosphate was not included in the sample for the initial separation), followed by a step corresponding to the reversal of the solution counterflow (similar to the EOF marker seen in conventional CE; when the applied negative pressure was sufficient to overcome the electroosmotic flow in the capillary) and then steps for the cations sodium and potassium. The time derivative of the detector signal (showing a series of peaks as in a conventional chromatogram) is plotted in Figure 2B. For this example, it was desired to measure the concentrations of glycerol-3-phosphate and the phosphate ion. So, the second separation was run with a lower starting pressure (to skip over the long, blank part of the initial separation) and slower pressure ramp rate (to increase the resolution between carbonate and phosphate). Also, it was not necessary to measure the concentrations of the cations, so the pressure ramp was stopped after detection of the phosphate ion step. The third separation (Figures 2B and 2C) was run with the same settings as the second separation, but with a sample containing glycerol-3-phospate in addition to sodium chloride and potassium phosphate. For the final optimization, even higher resolution was desired between glycerol-3-phosphate and the phosphate ion, so the start pressure was reduced again (since quantitative measurement of the chloride concentration was not needed) and the pressure ramp rate was also reduced. The resulting separation (Figures 2D and 2E) shows good resolution between all of the anions (including carbonate that is present due to CO2 in the atmosphere), with a separation time of less than 3 minutes.

Figure 2. Detector Signal (A, C, E) and time derivative of detector signal (B, D, F) for initial optimization of GEMBE separation. All separations were run with a 3.5 cm long 15 μm i.d. capillary and a commercially available contactless conductivity detector (TraceDec) with the detection point located 15 mm from the capillary inlet. Run buffer was 100 mmol/L bistris, 100 mmol/L HEPES, pH 7.1. Step/peak identification: 1, chloride; 2, carbonate; 3, phosphate; 4, glycerol-3-phosphate; 5, electroosmosis/reversal of counterflow; 6, sodium; 7, potassium. Note that the first step/peak in each plot (at approximately 0.2 min) is the result of slight Joule heating as the high voltage power supply is turned on. A, B) Rapid initial separations. Sample: 100 μmol/L sodium chloride, 100 μmol/L potassium phosphate monobasic. Conditions: 2000 V (positive high voltage applied to run buffer reservoir); 68 kPa starting pressure; 500 Pa/s pressure ramp rate. C, D) Slower separation with glycerol-3-phosphate added. Sample: 50 μmol/L sodium chloride, 50 μmol/L potassium phosphate monobasic, 50 μmol/L glycerol-3-phosphate. Conditions: 3500 V (positive high voltage applied to run buffer reservoir); 18 kPa starting pressure; 125 Pa/s pressure ramp rate. E, F) Final optimized separation. Sample: 50 μmol/L sodium chloride, 50 μmol/L potassium phosphate monobasic, 50 μmol/L glycerol-3-phosphate. Conditions: 3500 V (positive high voltage applied to run buffer reservoir); 0 Pa starting pressure; 50 Pa/s pressure ramp rate.

Selection of Starting Pressure. If the starting pressure is too low, it can result in a distortion of the step/peak shape and reduction of the separation resolution (for example, see the results for chloride shown in Figures 2E and 2F). As a general guide, the optimum starting pressure is chosen so that the first analyte step that needs to be quantitative is detected at a time equal to or greater than , where L is the effective channel length (distance between the channel entrance and the detector), a is the acceleration of the pressure driven flow (for a cylindrical capillary, ) and where time zero is defined as the start of the pressure ramp.

Selection of Magnitude and Polarity of Voltage. As with conventional methods for electrophoretic separation, it is best to use the highest voltage that does not result in significant extra dispersion or broadening. This can be determined by running several repeated separations with different voltages (but the same pressure ramp rate) and plotting the step width as a function of the voltage. Generally, the step width should be independent of the voltage unless there is extra dispersion such as Taylor-Aris dispersion [3, 4] or Joule heating dispersion [8] that depends on the voltage. Note that when adjusting the voltage, the starting pressure should also be adjusted and kept proportional to the voltage.

A GEMBE separation can be run for the analysis of anions, cations, or both, and with the high voltage set to either polarity (positive voltage or negative voltage at the run buffer reservoir). Typically, the polarity is set so that the electrophoretic motion of the analytes is in the direction from the sample reservoir toward the run buffer reservoir (positive high voltage for anions, negative high voltage for cations). The direction of the buffer counterflow is then from the run buffer reservoir toward the sample reservoir. For some applications, however, it is desirable to reverse the peak order, and with GEMBE this is simply accomplished by reversing the polarity so that the direction of electrophoresis is from the run buffer reservoir toward the sample reservoir and the counterflow is from the sample reservoir toward the run buffer reservoir. For the analysis of both anions and cations in a single run, the direction of the counterflow is reversed in the middle of the separation as shown in ref[5] and in Figures 2A and 2B.

Selection of Pressure Ramp Rate. The pressure ramp rate is the primary parameter to be varied in the trade-off between resolution and separation time. Slower pressure ramp rates result in separations with higher resolution but longer separation times, while faster pressure ramp rates result in faster, lower-resolution separations. Theoretically, the resolution scales as the -¼ power of the pressure ramp rate if the step width is determined predominantly by diffusion or dispersion in the separation channel; and it scales as the -½ power of the pressure ramp rate if the step width is determined by the width of the detector. The dependence of the separation time on pressure ramp rate is more complicated, but in most cases, it scales roughly as the -½ power of the pressure ramp rate. The separations shown in Figure 2 provide a good example of the optimization of a separation by changing the pressure ramp rate.

For many separations, a constant pressure ramp rate is appropriate. As with gradient elution chromatography, however, a variable pressure ramp can be used with slow ramp rates used to provide good separation of the most difficult to resolve analytes, while faster ramp rates are used for the remainder of the separation. Good examples of variable pressure ramps can be found in Refs[1, 5].

Selection of Pressure Ramp Duration. The final (and simplest) parameter for optimization is the pressure ramp duration. The duration of the pressure ramp should be set so that the pressure ramp ends just after the detection of the final analyte of interest, allowing enough baseline after the final step for quantitative measurement. Taking the separation shown in Figure 2E and 2F as an example, the pressure ramp could have been stopped at 2.5 min, without loss of any useful information.


The information provided here is meant as a guide for the design and operation of GEMBE separation devices. It includes a description of the hardware requirements and of the best mode of operation as it is currently understood by the author. Although the information here is presented as specific to GEMBE, parts of it (the hardware descriptions in particular) are also relevant for the related techniques of temperature gradient focusing (TGF) [9] and gradient elution isotachophoresis (GEITP) [10].

* Certain commercial equipment, instruments, or materials are identified in this report to specify adequately the experimental procedure. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.

  1. Shackman, J. G., Munson, M. S., Ross, D., Anal. Chem. 2007, 79, 565-571.
  2. Strychalski, E. A., Henry, A. C., Ross, D., Anal. Chem. 2009, 81, 10201-10207; Strychalski, E. A., Henry, A. C., Ross, D., Anal. Chem. 2011, 83, 6316-6322.
  3. Taylor, G. I., Proc. R. Soc. Lond. 1953, 219, 186-203.
  4. Aris, R., Proceedings of the Royal Society a-Mathematical Physical and Engineering Sciences 1956, 235, 69-77.
  5. Flanigan, P. M., Ross, D., Shackman, J. G., Electrophoresis 2010, 31, 3466-3474.
  6. Ross, D.,Romantseva, E. F., Anal. Chem. 2009, 81, 7326-7335.
  7. Ross, D.,Kralj, J. G., Anal. Chem. 2008, 80, 9467-9474.
  8. Virtanen, R., Acta Polytechnica Scaninavica 1974, 123, 1-67.
  9. Ross, D.,Locascio, L. E., Anal. Chem. 2002, 74, 2556-2564.
  10. Shackman, J. G.,Ross, D., Anal. Chem. 2007, 79, 6641-6649.
  11. Ross, D, Electrophoresis 2010, 31, 3650-3657.
  12. Ross, D, Electrophoresis 2010, 31, 3658-3664.