Development of a microfluidic electroosmosis pump on a chip for steady and continuous fluid delivery

Simulation of the micropump

The design and operation parameters for the microfluidic pump are firstly determined in COMSOL Multiphysics^® 5.1, considering the geometry in Figure 2. The simulation of relevant physical principles is done using the computational fluid dynamics (CFD) module and Microfluidics module.

Figure 2:

Geometry and variables considered in the simulations. W is the width in mm, L is the length in mm, V is the position in the bridge where voltage is applied (representing the position of the electrical contact).

The flow is modelled for a Newtonian fluid (water, liquid) of constant density and incompressible considering fluid flow induced by applying an electric field [11]. The charged solution is set to form close to the liquid-electrode interface, known as electric-double layer, which started to move along the pumping channel. The problem to be addressed supposes the fluid flows at very low Reynolds numbers. As such, the flow is solved as stationary creeping flow, also referred to as Stokes flow, using the Creeping Flow (spf) interface. As is characteristic of a Stokes flows, inertial terms of the Navier Stokes equations are neglected in preference to the viscous terms. The Stokes equation is solved to determine the conservation of momentum and the continuity equation is solved for the conservation of mass as.

where p is the pressure in [Pa], I is the identity matrix, μ is the dynamic viscosity in [Pa.s], u is the velocity given in [m s⁻¹] and F is the volume force in [N m⁻³].

And:

where ρ is the volumetric mass density in [kg m⁻³].

A laminar inflow boundary condition with average velocity is set at the inlet as:

where L _entr is the length of the fictitious entrance given in [m], ∇ _t is the tangential gradient, p _entr is the pressure at the fictitious entrance in [Pa] and n is the boundary normal pointing out of the domain.

The outlet is set with a pressure boundary condition:

where p ˆ 0 is the pressure at the outlet with suppressed backflow, in [Pa].

The fluid electric properties are uniform and the walls are set with electroosmotic velocity, tangential electric field and stationary.

where u _EO is the electroosmotic mobility [m² V⁻¹ s⁻¹] and E _t is the tangential component of the electric field, [V m⁻¹] given by

Here, E is the electric field vector in [V m⁻¹], obtained as

where V is the electric potential [V].

The flow that balances viscous stresses with electrical forcing (electroosmotic flow) is given by the velocity term u _EO which can be deduced from.

where ϵ _r is the relative permittivity [F m⁻¹], ϵ ₀ is the permittivity of vacuum ≈1/36π10⁻⁹ [F m⁻¹] and ζ is the Zeta potential [V].

Physics interface of electric currents (ec) is set with current conservation.

where Q _j is the current source in [A m⁻³] and J is the current density (volume) [A m⁻²] defined as

Here σ is the electrical conductivity [S m⁻¹] and J _e is the current density generated externally [A m⁻²].

The electric displacement, D [C m⁻²] is related to the fluid’s relative permittivity ϵ _r as.

The electric insulation follows

The mesh is set to a predefined size as Fine with maximum element size of 1.22 × 10⁻⁴ m and minimum element size of 2.3 × 10⁻⁵ m. The solver used is PARDISO. The geometry resulting from simulations, giving the results within the defined volumetric flow rate range, was fabricated.

Microfabrication

The procedure for microfabrication has 6 main tasks (A–F) and is followed according to Figure 3.

Figure 3:

Schematics of the fabrication steps.

Individualization of cellulose membrane

Because gases cannot be pumped in an electroosmotic pump, such a pump cannot fill itself with liquid from an initially empty state [13]. To overcome this limitation, the pump developed integrates a cellulose filter (pore size: 0.2 µm, diameter: 47 mm, thickness: 120 µm, GE Healthcare WhatmanTM) to drive the fluid to the main channel and bridges where the electrical contacts are.

The paper shape is designed in AutoCAD and imported to Silhouette Studio software (Figure 4A) to be defined by xurography (blade nr 1, thickness 15, cutting velocity 10 cm s⁻¹, single cut, 17 s). The individualized paper filters are depicted in Figure 4B.

Figure 4:

Shape of the cellulose membrane to be integrated in the EO pump. (A) Design imported to the silhouette studio software and (B) individualized cellulose filter.

Fabrication and individualization of PDMS membrane

SYLGARD™ 184 Silicone Elastomer Kit containing 3-dimethyl syloxane (DMS) and 184 silicone elastomer (cross-linking agent) is weighted and manually mixed. The mixtures are deaerated in a vacuum dessicator (1-800-4Bel-Art, Bel-Art Products) for 1 h and absolute pressure 0.22 ± 0.09 atm (R5 rotary vane vacuum pump, Busch). To avoid bubbles, each mixture is gently poured on top of an auxiliary surface. A spin coater was used to set the PDMS (polydimethyl siloxane) thickness (120 μm, step 1:500 rpm, 5 s, 500 rpm s⁻¹, step 2: 965 rpm, 20 s, 100 rpm s⁻¹, modular spin coater ws-650-23NPP, Laurell Technologies Inc). The PDMSis then cured at 70 °C for 1 h 40 min (digitally controlled Memmert GmbH + Co. KG 100–800 oven). All steps of the PDMS preparation are done inside a laminar flow hood (Faster-BSC-EN) to avoid surface contamination.

After curing, the PDMS membrane covers the silicon substrate except the inlet and outlet reservoirs, main channel and bridges, as presented in Figure 5A in blue. The definition of the structures onto the PDMS is conducted by xurography using the Silhouette Curio™ cutting printer and blade (Figure 5B). The auxiliary surface for spin coating is chosen because the PDMS cannot be directly spin coated onto the microfabricated Si substrate as the cut would damage the electrodes (see section C). Instead, the PDMS membrane is fabricated on another Si wafer or on a polystyrene plate (PS, 3 × 100 × 100 mm³, CRYLUX^®), ensuring an adequate adhesion during the spin coating process without compromising any future step of peeling, and later transferred to the microfabricated Si substrate. Tests are performed at various combinations of cutting depths and speeds to access the best PDMS cut finishing (Figure 5C).

Figure 5:

Rapid prototyping of the microfluidic channel by Xurography. (A) Schematics of the coverage of PDMS membrane on silicon/graphene substrate; (B) Setup for the xurography fabrication. On top of the stage, two platforms are placed to secure the plate in place. A blade with size varying from 1 to 3 is used in the tests; (C) Xurography parameters for the fabrication of 120 µm PDMS membrane with silhouette curio and parameters for transfer procedure.

Fabrication of chip with electrical contacts

The surface elements of the microchip flow pump follow typical cleanroom procedures at INESC MN (cleanroom class 100/10). The chip (Figure 6) is fabricated in a standard 6-inch silicon wafer (single side polished Si, mechanical grade, 0.65 mm thick, <100>, University wafer). The electrical components are defined by laser photolithography (Direct Laser Lithography System DWL 2.0, 6″, HeCd laser 422 nm, critical dimensions 0.8 µm, Heidelberg Instruments) of Al_98.5Si_1.0Cu_0.5 (in percentage) deposited by magnetron sputtering (DC Magnetron Sputtering, 2 kW, 3 mTorr, Nordiko 7,000, 6″). Wet etch using TechniEtch Al 80 MOS etchant from Microchemicals is followed by a 2nd photolithography and the deposition of TiW (N) and Cu by ion beam deposition (Nordiko 8,000 coupled to Nordiko 3,600, 8″). Liftoff, 3rd lithography, RF sputtering of Al₂O₃ (UHVII, base pressure of 5 × 10⁻⁷ Torr, applied power of 200 W_rf, 543 Hz, argon flux of 43.9 sccm, deposition pressure of 4.6 mTorr, 6″) for electrode protection and electric passivation, 2nd liftoff and chip individualization (DISCO DAD 321 Dicing Saw, 6″) follows.

Figure 6:

Electroosmotic pump: from the design to the chip. (A) schematics of the chip with electrical contacts. green: electrical contacts Al_98.5Si_1.0Cu_0.5 (3,000 Å); red: cellulose membrane; light blue: passivation layer (Al₂O₃ 3,000 Å); dark blue: TiW (300 Å) + Cu (300 Å) and (B) chip with electrical contacts after microfabrication.

Fabrication of PMMA support

Two PMMA (polymethyl methacrylate) substrates (150 × 150 mm, 4 mm-thick) are fabricated in a high precision 3 axis CNC micromilling machine (SuperTech) with specific designs for top and bottom casings (e.g., pocket, inlet, and outlet structures, Figure 7). These encase the silicon chip with electrical contacts, cellulose filter and PDMS structures.

Figure 7:

Schematics of the (A) top plate and (B) bottom plate of the PMMA casing in AutoCAD for the encapsulation of the silicon substrate. The top casing presents inlets to contact the main reservoir and the electrodes of the system, an outlet, and perforations to unite the two PMMA plates. The bottom casing includes a pocket structure and perforations aligned with the ones present in the top casing.

Bonding

The PDMS is peeled from the substrate, is permanently bonded to the microflabricated Si chip (300 s, 133 Pa O₂ atmosphere, 30 W, 25 °C, RF-power, benchtop 6″ Plasma Cleaner PDC-002CE, Harrick Plasma) and encloses the paper filter. A second PDMS layer is permanently bonded to the first to close the microfluidic system. An example of PDMS bonded to a dummy silicon substrate is depicted in Figure 8.

Figure 8:

PDMS permanently bonded to a Si substrate. In this example, the Si is not microfabricated.

Assemblage

All the elements are assembled to obtain the microfluidic chip pump as shown in Figure 9.

Figure 9:

Microfluidic chip flow pump.

Power supply module

The power supply module for this setup is built around the Bellnix OHV12-1.0K1500P, an ultracompact, ultralow ripple noise DC-DC converter. This component accepts a nominal input voltage of 12 V and outputs a fully adjustable voltage from 0 to 1,000 V, with a maximum output power of 1.5 W. The relevant electric specs of this converter are shown in Table 1.

In the complete power supply module, additional components are added both to the input and output sides of the OHV12 converter to enhance its operating characteristics. Moreover, a resistor network is employed to change the control voltage that defines the output voltage of the converter.

An electric field is then imposed to the electrical contacts of the microfluidic pump. The device is tested with de-ionized (DI) water colored with red food coloring die (non-alcoholic, Globo) (Figure 10). Snapshot images of the resulting flow field at the outlet tubbing are captured during flow delivery experiments using a Levenhuk DTX-50 Digital Microscope and analyzed with Image J software.

Figure 10:

Schematics of the experimental setup.

Simulation of the micropump

Simulations start by comparing results from 2D and 3D representations of the flow field. The channel is defined by walls in the x-direction and the electric field is applied along the bridges y-direction. It is assumed that the electric potential is steady and constant between the bridges.

The simulated flow velocity depicted in Figure 11 represents the velocity values u _x over the line x=x _out (black squares) or at z=½ channel height for cut plane x=x _out (red circles) considering the flow is fully developed and the system is represented in 2D or in 3D, respectively. The electroosmotic velocity profile is expected to be plug-like, as shown by the red circles in Figure 11. The 2D results (black squares) are common of a Poiseuille flow instead of an electroosmotic (EO) flow. In the EO flow, the maximum velocity is expected near the electrode surface and bottom wall and will decrease linearly to the non-slip channel top. Because the flow velocity in the 2D representation of the channel is failing to express this effect, the 2D results are discarded from the study. Following studies are performed considering 3D channel representations.

Figure 11:

Results of simulated velocity profiles in 2D vs. 3D channels. 3D simulations were performed for a channel height of 120 μm.

Figure 12 shows the influence of the dimensions and spatial arrangement of the electrodes in the channel in the simulated pumping capacity.

Figure 12:

Simulated volumetric flow rate dependence of geometrical parameters: (A) W _bridge =W _bridge,1 =W _bridge,2 , (B) L _bridge =L _bridge,1 =L _bridge,2 and (c) W _pumping. Other input parameters are E=2.6 V mm⁻¹, x _bridge,1 =4.3 mm, x _bridge,2 =20.3 mm, x _out=23 mm, L _bridge,1 =L _bridge,2 =11 mm, L _pumping=23 mm, W _bridge,1 =W _bridge,2 =0.6 mm and W _pumping=2 mm.

Here the averaged velocity u _{average3D ,out} is obtained by averaging the velocity over the cut plane x=x _out and the volumetric flow rate at the outlet, Q _out , is calculated from the averaged velocity

A _cs is the cross-section area of the channel. For E=2.6 V mm⁻¹, x _bridge,1 =4.3 mm, x _bridge,2 =20.3 mm, x _out=23 mm, L _bridge,1 =L _bridge,2 =11 mm, L _pumping=23 mm, W_bridge,1 =W _bridge,2 =0.3 mm, L_pumping=23 mm, W_pumping=2 mm, L _V→out=2.7 mm, Q _out is found to be 18.2 nL min⁻¹.

The increase of the fluid flow rate is seen to increase with the surface of the electrodes, W _bridge. This increase will increase the electric double layer surface and consequently the DC electroosmotic pumping. While there is a linear proportionality between the width of the bridges and the resulting volumetric flow rate (Figure 12A), this linear dependence is not observed for other dimensions. For the same applied electric field, the increase of the bridge length leads to a decrease of Q _out (Figure 12B) due to the decrease in field intensity and consequently a decrease of the Coloumb force. The increase in the width of the pumping channel increases Q _out (Figure 12C) due to a compromise between the electroosmotic force and the back pressure exerted in the pumping area. When W _pumping becomes smaller the back pressure dominates over the electroosmotic force resulting in lower flow rates. The width of the pumping channel is found to be the dimension that has the biggest influence in the final value obtained for the volumetric flow rate (Figure 12C).

Microfabrication

The pump was fabricated at INESC MN cleanroom class 10/100 facilities using silicon as substrate and typical semiconductor methods to define the actuation strategy. This minimized contamination during the fabrication process, which is lead at controlled temperature and humidity. The use of direct write laser lithography with minimum feature size of 0.8 μm provides the necessary resolution to precisely define the dimensions and position of the electrode structures.

Apart from the electrode structures, the supporting materials chosen to integrate the microfluidic pumpdefine its capabilities. Paper substrates are often used in microfluidics as they achieve simple and cost-effective fabrication. Their porosity and hydrophilic nature are ideal to this end as they guarantee the fluid is perfectly distributed to the electrodes before and during actuation. Additionally, paper can be used to eliminate pressure-driven flow and air bubbles in the microchannel [14]. Poly (methyl methacrylate) (PMMA), was chosen for the pump encapsulation as it is extensively used in microfluidics nowadays due to its many advantages such as optical transparency, chemical stability, biocompatibility, low electrical conductivity and wide range of design possibilities [15]. This choice allowed the visualization of the process while attributing a robust character to microfluidic devices. Two PMMA substrates (4 mm thick) were fabricated with specific designs for top and bottom casings (e.g., pocket, inlet, and outlet structures) to encase the silicon substrate. Plastic screws and nuts (M3) were used to compress the various parts of the microfluidic pump. A critical requirement of the channel material is that it needs to be easily fabricated and peeled. PDMS has been chosen as the channel material as it is one of the most used polymers in the microfabrication of microfluidic devices in the laboratory, mainly because of its optical transparency, gas permeability, biocompatibility, and ease of fabrication and molding [16]. The physical characteristics of PDMS are known to be highly dependent on the monomer/curing agent ratio and curing temperature chosen, increasing the possibility of fabrication of structures and range of applications [17]. This tuning ability is used to achieve the DMS:cross-linking agent ratio with mechanical properties that presented the best finishing after xurography (Figure 13). The diferences are presumed to be due the Young Modulus of the material to be dependent on the amount of curing agent used in the formulation. For a given curing temperature, as the amount of curing agent increases, the PDMS has a more elastic behaviour, reflected by the decrease of the value of the Young modulus from 10:1 ratio up to 6:1 in [17]. After establishing the 3:1 ratio, the minimum cutting velocity is found to be 3 cm s⁻¹, as it is the one at which the definition of the walls is smoother (blade nr 1, thickness 15, single cut). The corner of the channels has round shape due to the working mode of the blade. Nevertheless, xurography provides the ability to obtain channels in a very expedite way when comparing to more traditional fabrication techniques such as soft lithography. Another important contribution to the good definition of the PDMS structures is the adhesion of the PDMS to the auxiliary surface. Silicon wafers are seen to be unsuited for processing the PDMS layer. The cured PDMS layer ruptures when peeling because of the strong bonding it creates with the Si. Polystyrene plates are also tested and result in appropriate peeling, without tearing of the PDMS membrane. The groove formed on the PS after cutting the PDMS is not seen to influence finishing of the patterned PDMS layer.

Figure 13:

Cut finishing of 0.6 mm-wide bridges on PDMS of (A) 10:1 ratio and (B) 3:1 ratio formulations, spin coated on PS plates. The darkest region is the result of the cut onto the PS plate right below the PDMS.

The power supply has been characterized prior to flow experiments. Considering the relative voltage difference of the power supply to be RVD=(x ₂ −x ₁ )/x ₁ , where x ₁ is the set value and x ₂ is the read value, the power supply stabilizes around 1,000–2,000 s, and its derivative over time dRVD/dt after that period is found to be lower than 5 × 10⁻⁶ (Figure 14). The higher the voltage set, the faster the power supply achieves a stable voltage. This aspect is taken into consideration when applying a DC voltage to the microfluidic pump.

Figure 14:

Stabilization times for [100:100:500] V imposed to the power supply.

A preliminary study of the flow has been conducted to evaluate the pump performance. Before the measurements start, the outlet reservoir is filled with DI water died red while the inlet reservoir is filled with DI water alone, as represented schematically in Figure 10. The pressure at both inlet and outlet is left to equalize by having the system open to air. This way, the influence of pressure-driven flows is limited when the measurements take place. At time t=0 s a DC voltage is applied to the electric contacts microfabricated onto the Si chip.

The liquid starts to move towards the outlet and the distance travelled by the fluid in a vertical position, from bottom to top, is recorded. Figure 15 shows an example of the flow of DI water with coloring die at the outlet tubing when applying an electric field of 13.2 V mm⁻¹ to the electrodes. The distance travelled by the fluid in the tubing at the outlet is obtained subtracting the position of the meniscus at time t + Δt from that at time t.

Figure 15:

Pumping results obtained with the microfluidic pump while imposing an electric field of 13.2 V mm⁻¹ during 1,800 s. The figure shows the advancement of the fluid consequence of the electroosmosis. The temperature of the room is 22 °C.

From this example, assuming that the inner diameter of the tubbing is that given by the manufacturer, of 0.86 mm, it is possible to roughly estimate the volumetric flow rate of Q _out =(161 ± 20) nL min⁻¹. The error is given by the standard deviation of results based on 24 data points for the averaged volumetric flow rate. The value of Q _out obtained experimentally is found to be in good agreement with the simulated results under ideal flow conditions (Figure 16). These simulated volumetric flow rate results, which have a maximum difference ratio of just 7.3% to the experimental results, have a considerable impact on minimizing trial and error in the construction of the pump and on achieving the targeted conditions. Moreover, the results demonstrate the microfluidic actuation capability to drive very small volumetric flow rates down to 45 nL min⁻¹, which is also useful in organ-on-a-chip contexts. Here, the fluid shear stress and mass transport regime of nutrients are both dependent on the flow rate, and these will greatly influence the cellular response [18].

Figure 16:

Simulated and experimental volumetric flow rates as a function of the applied electric field, considering x _bridge,1 =4.3 mm, x _bridge,2 =20.3 mm, x _out=23 mm, L _bridge,1 =L _bridge,2 =11 mm, L _pumping=23 mm, W _bridge,1 =W _bridge,2 =0.6 mm, W _pumping=2 mm and L _V→out=2.7 mm.

It is necessary to correct the contributions of inner tube diameter, non-zero pressure load at the outlet, surface wettability on the flow path, evaporation, etc., to the measurement error [19] and to assess the deviations of the experimental values to the simulated data. Nonetheless, the preliminary results are very encouraging.