Programmable Electrochemical Stimulation on a Large-Scale CMOS Microelectrode Array

A. Pixel and Electrode Array Design

The microelectrode array is integrated on a CMOS sensor with 131,072 pixels arranged in a 256 × 512 array. An image of the array architecture and a simplified pixel schematic are shown in Fig. 1. Each pixel contains SRAM cells which determine whether it operates in sensing mode or stimulation mode (Fig. 1(b)), allowing the array to be configured for stimulation with any arbitrary spatial pattern. In stimulation mode (Fig. 1(c)), the switches can be biased in the linear region for bipolar voltage stimulation, or biased as cascoded current sources for bipolar current stimulation. Stimulation polarity and timing are determined by external digital controls. Here we focus only on the voltage stimulation mode, while the other operating modes of the chip are described in [30].

B. Layout and Packaging

The circuits are implemented in 180 nm CMOS technology, and individual pixels are 10μm × 10μm with an exposed top metal electrode. The overall chip is 25 mm², with the electrode array occupying 13.1 mm² [30]. The chip is wirebonded to a small printed circuit board module, and the bondwires are encapsulated with epoxy, leaving the active array area exposed [31]. The module plugs into a custom data acquisition board hosting ADCs, DACs, power management, and an FPGA module (Opal Kelly XEM7310) with a USB interface.

A. Electroplating of Gold

The pixels are initially fabricated using the aluminum top metal from the foundry, which is prone to oxidation and corrosion at moderate voltages in common aqueous buffers. We construct a small open fluid chamber around the sensor array with a polypropylene tube sealed with silicone adhesive. We then remove the aluminum layer in acid (Aluminum Etchant Type A, Transene), exposing the underlying tungsten vias and titanium nitride diffusion barrier layer. These materials are conductive and stable, and can be used as-is, but it can also be valuable to introduce other metals to the surface. We filled the chamber with 2 mL gold plating solution (SG-10, Transene), plus a small amount of surfactant (2 μL Triton-X100) to ensure complete wetting of the microscale pixels. A small vibrating dental plaque remover tip was used to agitate the solution and improve mixing, and the chip was positioned on a hotplate set to 75°C. The experimental setup is shown in Fig. 2.

The pixels to be electroplated are programmed with STIMU_EN=1 (as illustrated in Fig. 1(c)), and these pixels are pulsed between 1.8 V and 0 V using the on-chip switches. The pulse duration was 1 sec with 50% duty cycle. A stainless steel counter electrode is biased at 1.8 V. For the intervals of time when the stimulated pixels are at 0 V, they act as cathodes and gold ions in the plating solution are deposited on the pixels. The remaining pixels have STIMU_EN=0 and are held at 1.2 V, and do not receive any gold deposition.

In Fig. 3, we show electron microscope images of a 16 × 16 array of pixels after varying amounts of plating time (Zeiss LEO 1530 SEM). The electrodeposited gold at each pixel has a large surface area, with sharp dendrites and multiple nucleation sites. Dendritic structures are expected for electroplating in cyanide baths carried out at high deposition potential and low pulse duration [33], [34]. The increased surface area would be expected to decrease the interfacial impedance and provide higher total stimulation current. The deposition potential used here is 1.8 V and t_ON is 0.5 sec. Based on these experiments, we selected t = 120 sec for gold electrodeposition. The in-pixel SRAM means that these electroplating protocols can be applied to arbitrary spatial patterns within the electrode array. For example, Fig. 4 shows gold selectively electroplated in a Sierpinski triangle fractal pattern.

B. Programmable Generation of Hydrogen Bubbles

Spatial control of gas evolution can be valuable for patterning cells on a microelectrode array surface [13]. In aqueous solutions, there is a relatively small window (≈ 1.23 V) of electrochemical stability, beyond which water molecules may become unstable. The electrolysis reaction at the cathode produces hydrogen gas as well as hydroxide ions:

Due to the low solubility and low mass density of hydrogen gas in an aqueous solution, the hydrogen gas tends to coalesce and produce bubbles. To study the bubble formation on the CMOS array, we employed a PBS buffer (pH 7) with 50 mM KCl salt as the electrolyte. The gold-coated pixel surface was stimulated to act as a cathode, while a silver wire held at 1.8 V was used as the counter electrode. We observe rapid nucleation and growth of hydrogen gas over the stimulated area (Fig. 5). The growth rate of the bubbles depends on the spatial distribution of hydrogen, and its diffusion rates. The analytical solution to the unsteady growth of hydrogen bubble diameter (d) is expected to scale approximately with $\sqrt{t}$ as

where d₀ is the initial diameter, d_t is the diameter at time t, and D_H is the diffusivity of hydrogen in an aqueous solution (≈ 10 × 10⁻⁹ m²/s). The parameter β represents the degree of hydrogen supersaturation. Monitoring the sizes of several bubbles (Fig. 6), we see that their growth fits well to Eq. 2. The fit value for the supersaturation parameter is β = 0.13. Since this experiment was performed using the fractal electoplated array, it is reasonable that β is slightly lower than typical values from the literature (0.22 – 0.63) [35], where electrolysis is usually carried out over a large uniform electrode surface. The experiment suggests that stimulation time < 1s could be enough to induce local cell detachment.

C. Spatially Controlled pH Modulation

Electrochemical modulation of local pH has been used to influence the viability of biofilms [36] as well as to support programmable DNA synthesis [16]. There are multiple parameters which affect generated pH gradients, including the stimulation mode (potentiostat versus galvanostat), redox chemistry, reservoir geometry, and stimulation time. Here, we utilize the [OH] radicals produced as a result of the electrolysis of water to modulate pH. The hydroxyl radicals will tend to react and disperse over time, and the extent of pH modulation will depend on the number of excess [OH] radicals introduced, as well as their spatial confinement. Mass transfer at the pixel is described by the faradaic equation:

where z is valence (the number of electrons consumed per [OH] produced), F is Faraday’s constant, I is the stimulation current at each pixel, and t is the effective duration of stimulation. I is a function of the applied stimulation voltage, but in these experiments we generally do not measure the current directly. To monitor the spatio-temporal pH changes, we prepared an electrolyte consisting of DI water mixed with pH indicator dye (0.04% bromothymol blue plus ≈ 50 mM potassium chloride). The dye was orange-yellow at the initial pH (≈ 5.5), and it turns blue at basic pH (> 7.4). Fig. 7 shows that within 1 second the pH begins to change near stimulated pixels. The dispersion of hydroxyl radicals beyond the stimulated pattern (Fig. 7 (e)) appears to initially follow a diffusion model ($L_D=2\sqrt{Dt}$), but it soon transitions to an advection regime (L_D = vt), where L_D is the dispersion distance from the stimulated region, D is the diffusivity of hydroxyl ions in an aqueous solution (≈ 5 × 10⁻⁹ m²/s), v is the advection velocity, and t is the time elapsed. We suspect that bubble generation creates recirculation currents of the hydroxyl radicals and speeds up the horizontal pH dispersion [37]. The typical vertical velocity of hydrogen bubbles due to electrolysis is in the range of 40,000 – 100,000 μm/sec [35]. We observe lateral dispersion which is faster than would be expected from diffusion alone, but which is still approximately two orders of magnitude slower than the vertical motion of microscale bubbles which detach from the surface. Slower dispersion is preferable when the aim is to stimulate a meaningful spatial pH gradient. Introducing redox couples such as hydroquionine [16], [17] could help to influence pH without bubble formation, although adding redox couples could also affect cell metabolism.