Online Supplementary Material Forces acting on a particle in a concentration gradient under an externally applied oscillating electric field Yuan Luo, and Levent Yobasa) Department of Electronic and Computer Engineering, Hong Kong University of Science and Technology, Hong Kong, P. R. of China Contents: I. Fitting parameters S-2 II. Governing Equations S-2 III. Boundary Conditions S-4 IV. Numerical Simulations S-5 V. Simulation Results S-5 VI. References S-7 I. Fitting parameters β π | = π1 π 3 + π2 π to describe the curves in Fig. 3(a) The least-square fitting parameters, π1 and π2 , used in the relation |π are as listed in Table S-I. β π | = π1 π 3 + π2 π. TABLE S-I. Least-square fitting parameters |π π1 × 1010 , mm/s (V-rms)3 π2 × 105 , mm/s V-rms 10 3.29 2.54 25 4.37 1.54 50 3.59 0.94 100 1.55 0.86 π, μS/cm II. Governing Equations Both the immersion medium and the electrolytes are based on the phosphate buffered saline (PBS) solution and thus predominantly contain NaCl, a symmetric binary electrolyte (π + = −π − ≡ π0 ; π + and π − are the valance of the ionic species Na+ and Cl−, respectively), for which the flux of ions is given by the Nernst-Planck equations: π±± = −π· ± ∇π ± β π0 π ± ± π· π ∇π + π ± βπ ππ΅ π (S-1) where π· + = 1.4 × 10−9 m2/s and π· − = 2.0 × 10−9 m2/s are the diffusion coefficients of Na+ and Cl– ions, π ± are their molar concentrations, π is the elementary charge, ππ΅ is the Boltzmann constant, π is the absolute temperature, βπ is the solution velocity vector, ∇ is the gradient operator, and π is the electrical potential governed by the Poisson equation: π» 2π = − ππ΄ ππ0 + (π − π − ) ππ π0 (S-2) where ππ΄ is the Avogadro constant, ππ = 79 is the relative permittivity of water, and π0 = 8.85 × 10−12 F/m is the vacuum dielectric constant. The bold letters denote vectors. In Eq. (S-1), the three terms on the right hand side refer to the contributions due to diffusion, electromigration and convection of ions. In a typical microfluidic system, both the convective and diffusive contributions to the total current are β , where βπ¬ is the much smaller than that from electromigration.1 In view of these considerations, we can simply state π± = ππ¬ electric field vector given by −∇π and π is the electrolyte conductivity by π0 π ππ΅ π (π· + π + − π· − π − ). Furthermore, the charge conservation equation in the stationary state, π» β π± = 0, for an oscillating electric field can be written as: S-2 β)=0 π» β ((π + ππππ π0 )π¬ (S-3) β , and π = √−1. where π = 2ππ is the angular frequency with π being the temporal frequency of the applied electric field π¬ β is a complex vector expressed as π¬ β = Re[∇π sin(πππ‘)] whereas π is the electrical potential phasor. Note that π¬ Under the quasi-electroneutrality assumption, the conductivity of a binary symmetric electrolyte (e.g. NaCl) can be described by the convection–diffusion equation:2 π·πππ π» 2 π = ππ β β π»)π + (π ππ‘ (S-4) where π·πππ = 2π· + π· − ⁄π· + + π· − is an effective diffusivity ( π·πππ = 1.65 × 10−9 m2/s). In the stationary state, the time derivative in Eq. (S-4) is zero. For the fluid motion in stationary state, Stokes equation is used due to low Reynolds number (π π~10−3 , for a typical flow velocity of ~1 mm/s and ~50 µm channel dimension): βπ ππ» 2βπ − π»π = −π (S-5) π» β βπ = 0 (S-6) β π is the electrical body force, given by: where π = 10−3 Pa s is the dynamic viscosity of water, π is the pressure and π 1 2 1 π(π π ) 2 β π = ππ π¬ β − |π¬ β | π»(ππ π0 ) + π»(ππ€ ( π 0 ) π |π¬ β| ) π 2 2 πππ€ (S-7) where ππ and ππ€ are the free electrical charge density and the mass density, respectively. We assume that the permittivity of the electrolyte does not experience a noticeable change during the experiment and hence consider βππ depends solely on the Columbic contribution ππ βπ¬. The free charge density under an oscillating electric field in the bulk can be written from (S-2) and (S-3) as: ππ = − ππ π0 π»π β βπ¬ π + ππππ π0 (S-8) and the time-average electrical body force is βπ= π 1 β ∗] Re[ππ π¬ 2 where βπ¬∗ is the complex conjugate of the electric field. S-3 (S-9) III. Boundary Conditions Fig. S-1 shows a 3D illustration and a 2D schematic layout of the computational domain whereby each labeled segment (e.g. ππ′ , ππ) in the latter denotes a particular surface for which the boundary conditions were defined according to Table SII. The potential at the boundary ππ′ was assigned to the applied rms voltage (71 V-rms, 500 kHz) with respect to the ground potential set at the boundary ππ ′ . The walls facing the microcapillaries and the channel top and bottom were assumed electrically insulating. The boundaries ππ, ππ and ππ were maintained at the prescribed ionic concentration and normal inflow velocity while those π′ π ′ , π ′ π ′ and π ′ π ′ were given by the convective flux and atmospheric pressure. No normal flux (insulation) and no slip were assumed for all the remaining boundaries. FIG. S-1. The computational domain: (a) 3D illustration and (b) 2D schematic layout (not true to scale). TABLE S-II. Boundary conditionsa set for the computational domain in Fig. S-1 Boundaries ππ′ Eq S-3 π = 71 π=0 Eq S-4 β)=0 β β (−π·πππ ∇π + ππ π β)=0 β β (−π·πππ ∇π + ππ π Eqs S-5 and S-6 βπ = 0 βπ = 0 ππ ′ ππ, ππ β βπ±=0 π π =15 mS/m π£π₯ = 4 ππ β βπ±=0 π π = 100, 50, 25 or 10 µS/m π£π₯ = 1 π′π′, π′π′, π′π′ β βπ±=0 π β β (−π·πππ ∇π) = 0 π πatm Others β βπ±=0 π β)=0 β β (−π·πππ ∇π + ππ π βπ = 0 a β ) = 0; Convective Surface normal vector: βπ; Electric insulation: βπ β π± = 0; Insulation: βπ β (−π·πππ ∇π + ππ flux: βπ β (−π·πππ ∇π) = 0; The velocity component in the x direction (mm/s): π£π₯ ; Applied voltage in V-rms: π; Atmospheric pressure: πatm = 1.01 × 105 Pa S-4 IV. Numerical Simulations Governing equations were numerically solved for the 3D geometry (Fig. S-1a) through the finite element method (FEM) analyses in COMSOL Multiphysics Software 3.5 (Comsol Inc., MA). The electric field and convection–diffusion equations were solved first to obtain the ionic concentration profile, which was subsequently used as an initial condition to the coupled electric field, convection–diffusion, and flow equations. In the simulations, the electrolytes were set at an injection conductivity of 15 mS/cm, while the particle immersion was assigned at 10, 25, 50 or 100 µS/cm. A tetrahedral mesh was used with a maximum element size of 0.8 µm for the microcapillary domains and of 10 µm for the channel domains. The chosen mesh was sufficiently fine. A finer mesh produced no significant change, indicating that the solutions are convergent and grid independent. V. Simulation Results Fig. S-2 shows the steady-state distribution of ionic concentration (conductivity) at a horizontal plane situated 10 µm above the channel floor (electrolyte streams: 15 mS/cm; particle stream: 50 µS/cm; excitation: 71 V-rms 500 kHz). A considerable influx of ions can be seen through the microcapillaries into the particle stream, which is in agreement with the experimental observations (Fig. 2 of the main text). 15 Conductivity (mS/cm) 0.05 Electrolyte y Particle x Electrolyte FIG. S-2. Steady-state concentration (conductivity) distribution across the horizontal plane situated 10 µm above the channel floor. The dark dashed lines on the partitions represent the sidewalls of the microcapillaries located out of the horizontal plane. The electrolyte and particle streams were set at an injection conductivity of 15 mS/cm and 50 µS/cm, respectively. Excitation: 71 V-rms, 500 kHz. S-5 Fig. S-3a describes the electric field strength variation along a representative trajectory of 10 μm particles immersed in distinct conductivities (legend) undergoing a trapping event (71 V-rms 500 kHz). The trajectory follows a diagonal path originating at π₯ = 30, π¦ = 25, π§ = 30 μm and terminating at π₯ = 10, π¦ = 5, π§ = 10 μm in reference to the center point of a microcapillary opening. As shown, the field intensity toward the microcapillary opening (origin) continues to monolithically increase despite the immersion conductivity rise due to the electrolyte (15 mS/cm) influx under Maxwell-Wagner polarization. We also evaluated the ∇ log(π) averaged along the same trajectory in response to the increased applied rms voltage and found a linear correlation (π 2 = 0.944) between the two (Fig. S-4). Fig. S-3b shows the fluid velocity component at distinct immersion conductivity levels (legend) along the line that represents the projection of the longitudinal symmetry axis of the microcapillary pair on the horizontal plane situated 10 μm above the channel floor (dashed line in Fig. S-2). The flow velocity switches sign (direction) as the particle stream midpoint (origin) is crossed, and it vanishes once the applied voltage is removed as shown for the simulated case corresponding to the lowest immersion conductivity (10 μS/cm). FIG. S-3. Line plots: (a) electric field strength along a representative trajectory of 10 μm particles undergoing a trapping event with the applied oscillating electric field (following a diagonal line that originates at π₯ = 30, π¦ = 25, π§ = 30, and terminates at π₯ = 10, π¦ = 5, π§ = 10, all in μm in reference to the center point of a microcapillary opening located at the origin of the plot); (b) the fluid velocity component along the line that represents the projection of the longitudinal symmetry axis of the microcapillary pair on the horizontal plane situated 10 μm above the channel floor (dashed line in Fig. S-2). The electrolyte streams were set at an injection conductivity of 15 mS/cm while the particle stream was assigned at 10, 25, 50, and 100 μS/cm (legend). Excitation: 71 V-rms 500 kHz. S-6 log (c) / (m–1) 10 ×104 8 6 4 2 0 0 20 40 60 Voltage/(V-rms) 80 FIG. S-4. The concentration gradient (∇ log π) averaged along the representative trajectory (same as the one defined for Fig. S-3a) as a function of the applied voltage rms magnitude. The solid line represents the best least-square fit (π 2 = 0.944). The electrolyte streams were set at an injection conductivity of 15 mS/cm while the particle stream was assigned at 50 μS/cm. Excitation: 500 kHz. References 1 A. Castellanos, A. Ramos, A. Gonzalez, N. G. Green, and H. Morgan, J. Phys. D: Appl. Phys. 36, 2584 (2003). P. García-Sánchez, M. Ferney, Y. Ren and A. Ramos, Microfluid. Nanofluid. 13, 441 (2012). 2 S-7