| Minimizing Dispersion Introduced by Turns on Microchips
Microfluidic devices have been successfully demonstrated for a wide variety of electrically driven separation techniques including capillary electrophoresis (CE), synchronized cyclic electrophoresis, micellar electrokinetic chromatography, open channel electrochromatography and gel electrophoresis. As the advantages of combining sample processing and analysis in a single device are numerous, these separation techniques are being integrated with other physical manipulations and chemical reactions on microchip devices. Increasing the level of integration, however, results in a more complex channel manifold with a greater number of channels and interconnections. In addition to the increased channel manifold complexity, the separations being performed on microchips are becoming challenging. Further improvements, therefore, in separation efficiency and resolving power are required. In CE, separation efficiency and resolving power are generally improved by increasing the electric field strength. There are, however, situations in which field strength increases can degrade resolution. For example, this can be seen with DNA sequencing fragments where the differential mobility of the fragments becomes nearly constant at high field strengths. In such cases separation efficiency and resolving power may be further enhanced by lengthening the separation channel while keeping the field strength constant. Both the increase in separation channel length and the overall increase in the microchip complexity require careful design of the channel manifold so that the microchip retains its overall compact footprint. Often these designs require that several turns be introduced into the various channels on the chip. The introduction of turns adds a geometrical contribution to analyte dispersion as both the migration distance and field strength vary laterally across the width of the channel in the turns (Figure 1A). This results in a "racetrack" effect for the individual molecules in the analyte band where the molecules on the inside wall of the turn migrate faster than those along the outside wall (Figure 1B). A simple, one-dimensional model has been developed to predict the amount of excess dispersion introduced by these turns. This model accounts for migration length differences, field strength differences, and transverse diffusion effects. The geometrical dispersion () generated depends on the ratio of the analyte transverse diffusion time (tD) to the time that the analyte band spends traversing the turn (tt).
Experiments were performed to verify the model developed using the chip design shown in Figure 2 with a variety of channel widths and radii of curvature. A plot of the actual versus the predicted dispersion introduced by turns can be seen in Figure 3. The dashed line on the plot indicates the best fit to the data while the solid line indicates the values predicted by the model. The slope value is well within experimental error of the ideal slope of 1. From these experiments it has been found that the geometrical dispersion can be reduced most effectively by decreasing the channel width and increasing the radii of curvature of the turn.
The model developed also suggests that the geometrical variance may become the dominant source of analyte dispersion for molecules with small diffusion coefficients such as DNA. For separations performed using a Hae III restriction digest of fX174, the excess dispersion introduced by the turns accounts for 60 to 65% of the total variance at detection point D (Figure 2). This leads to a 50% loss in resolution among the peaks between the entry into and the exit from the first turn (Figure 4A and 4B). If, however, a second turn opposite in direction to the first turn is introduced before the analyte molecules have time to diffuse across the separation channel some of the dispersion introduced by the first turn can be removed (Figure 4C and D), thereby, improving the efficiency and resolution of the separation beyond that expected for the distance traveled. Unfortunately, only 51% of the variance introduced by the first turn is removed by the second turn. The added separation length, therefore, leads to no improvement in separation efficiency. Other chip geometries are being investigated, at present, and results using these chip designs will also be reported.