ON-CHIP FIELD AMPLIFIED SAMPLE STACKING UNDER SUPPRESSED ELECTROSOMOTIC FLOW CONDITIONS R. Bharadwajl and J. G. Santiago* ‘Department of Chemical Engineering, Stanford Univemity ‘Depmtment ofMechanical Engineering, Stcanford Universi@ ABSTMCT We present theoretical and experimental results for concentration enhancement using field amplified sample stacking (FASS). We use an acidified poly(ethylene oxide) (PEO) coating to minimize dispersion due to EOF. We model the FASS process as a onedimensional electromigration and dispersion of two background electrolyte ions and one sample ion across an initial concentration gradient. Regular perturbation methods are used to solve for the concentration fields. Also, we use CC&based full-field, quantitative, epifluorescence imaging to experimentally measure the unsteady concentration fields and validate the model. KEYWORDS: Sample stacking, dispersion, electroosmotic flow, polymer coating 1. INTRODUCTION Field amplified sample stacking (FASS) is a sample preconcentration technique that leverages conductivity gradients between a sample solution and background buffer as shown in Figure 1 [I]. The rate of concentration enhancement in FASS is limited by molecular diffusion and dispersion. Advective dispersion is a result of internallygenerated pressure-driven flow resulting from a mismatch in electroosmotic velocity of the high and low conductivity regions of the channel [2]. The aim of this work is to investigate FASS dynamics under reduced electroosmotic flow @OF) conditions. High Conductivity buffer Low Conductivity Sample E OQO Qg &Q uw w, +---a / Stacked Analyte 7th lnternat~onal O-974361 I-0.O/~TAS2003/$15.0002003TRF Conference October Figure 1: Schematic of FASS process. A gradient in background buffer conductivity is used to create a gradient in electric field strength. Sample ions that drift from a region of high velocity to one of low velocity, accumulate or “stack” at the interface between these regions. This process increases sample concentration and can be used to increase signal-to-noise ratio in on-chip CE. on Miniaturized Chemical and Blochemlcal 5-9, 2003, Squaw Valley, Callfornla USA Analysts Systems 2. THEORY We model FASS as a one-dimensional electromigration and diffusion of two background electrolyte ions (A+, B-) and one sample ion (C) across an initial concentration gradient. Dispersion effects associated with mismatched EOF are approximated using an effective dispersion coefficient, Deffi which is equal to the molecular diffusivity in the absence of convective effects. The dimensionless governing equations are then: where Eo, v,, D, are respectively the characteristic scales for electric field, mobility, and dispersion coefficient. The initial conductivity gradient is modelled as an error function with a characteristic length scale s. The dimensionless parameters governing this system of equations are the background-buffer-to-sample conductivity ratio, 1/, and a Peclet number, Pe, (electrophoretic-to-diffusive flux ratio). Since the concentration of sample ions is much smaller than the buffer ions (Cc/Ca - O.Ol-O.OOl), we employ regular perturbation technique to decouple the buffer and sample ion concentration fields. At zeroth order, buffer ions follow binary electrolyte dynamics and the buffer-buffer interface is described by a purely diffusive error function. The sample ion distribution (the first-order problem) develops as a diffusive wave and the peak concentration increases until the theoretical maximum enhancement is reached ( Figure 3). We have also used this method to predict FASS dynamics with other initial conditions including a top-hat distribution for the initial sample concentration profile. We will present this work in a future paper. 3. EXPERIMENTAL Experiments were conducted in a borosilicate microchip (Micralyne, Alberta, Canada) with a staggered, double-T injection region. An Olympus epifluorescent microscope, an ICCD camera (Roper Scientific), and 10x objective were used to obtain images of concentration fields. Channels were flushed with an acidified poly(ethylene oxide) (PEO) solution [3] to suppress EOF and minimize flow-induced dispersion. 7th lnternat~onal 596 Conference October on Miniaturized Chemical and Blochemlcal 5-9, 2003, Squaw Valley, Callfornla USA Analysts Systems Figure 2: (a) Schematic of microchip for single interface stacking. Width and centreline depth of channels were 50 and 20 microns, respectively. (b) Epifluorescence, CCD images showing establishment of initial condition for conductivity gradient and subsequent stacking across the interface. The sample was negatively charged 17 pM bodipy dye and the buffer was HEPES at pH = 7. 4. RESULTS AND DISCUSSION Figure 3 shows the measured sample concentration distribution. The sample concentration develops in a wave-like manner with a peak concentration approaching the 4 (b) (4 Figure 3: (a) Measured intensity profiles for fluorescent sample ions. The raw data was normalized using brightfield and darkfield images. To reduce the image noise, the raw data was low-pass filtered using a Gaussian kernel with a standard deviation width equal to 10% of the FWHM at each time step. The electric field in the stacking region was 619 V/cm and y = 4. Time between individual profiles is 0.15 s. (b) Model prediction of the sample concentration field for the conditions of the experiments shown in (a). 7th lnternat~onal Conference October on Miniaturized Chemical and Blochemlcal 5-9, 2003, Squaw Valley, Callfornla USA Analysts Systems 597 theoretical maximum, equal to y. The temporal increase in peak concentration is roughly exponential. Figure 4 suggests that, for the same Pe, the timescale to achieve the theoretical maximum increases with increasing y . For efficient FASS, therefore, high y in combination with high Pe is essential. The model curves in Figures 3 and 4 were generated by varying the dispersion coefficient, Defl, to provide the best fit with the peak concentration data. The parameter s was set equal to the width of the measured initial concentration profile of the sample ion. To account for finite EOF, a uniform advective velocity (equal to the axial average of the product of electric field and mobility) was added as suggested by the work of Anderson and Idol [4]. Both Figures 3 and 4 show that there is good qualitative agreement between the measured and model predictions of the spatial and temporal concentration distribution. The model slightly overpredicts the width of sample ion profiles, and this might be due to three-dimensional effects near the channel interesection. Further investigation of this discrepancy and quantitative comparison of model predictions and experimental data is currently underway. Figure 4: Comparison of experimental and theoretical rates of dimensionless concentration increase. The experimental data were obtained from the concentration profiles shown in figure 3. For the 1 f 0.8 i=x 7 0.6 z E. co.4 y = 4, Experiment Model, Pe =I 11 y = 7.6, Experiment Model, Pe = 14 0.2 1 0.5 1.5 time (s) y = 7.6 case, it to reach the maximum in higher Pe. In applied electric sample region V/cm. takes longer theoretical spite of a both cases field in the was 6 19 REFERENCES 1. 2. 3. 4. Quirino, J.P., and Terabe, S., J. Chromatogr. A, 2000, 902, pp. 119-135. Burgi, D.S., and Chien, R.L., 1991, Anal. Chem., 63, pp. 2042-2047. Preisler, J., and Yeung, E.S., Anal. Chem., 1996, 68, pp. 2885-2889. Anderson, J.L., and Idol, W.K., Chem. Eng. Commun., 1985,38, pp. 93-106. 7th lnternat~onal 598 Conference October on Miniaturized Chemical and Blochemlcal 5-9, 2003, Squaw Valley, Callfornla USA Analysts Systems