Stability Assessment of Ultramicroelectrode Arrays in Neural Stimulation: an Electrochemical Impedance Spectroscopy Analysis

Abstract

The overall focus of this research program is to investigate the viability and safety of microelectrode arrays (MEAs) and ultramicroelectrode arrays (UMEAs) for microstimulation (μStim) in the brain. We aim to assess their potential to deliver adequate electrical stimulation for neural activation without causing electrode or tissue damage. To this end, here we conducted galvanostatic Electrochemical Impedance Spectroscopy (EIS) to assess the behavior of electrodes of various diameters both pre- and post- stimulation in a rat model. Our first-generation electrode arrays (G1) were designed with electrode diameters spanning 3 orders of magnitude and including subcellular dimensions. Using these devices, we delivered 0.4 nC/phase stimulation in vivo. We conducted in vitro EIS measurements before and after the current pulsing. Our findings show that these electrode arrays, including UMEAs, exhibit stable electrochemical behavior following a 0.4 nC/phase current pulse. To validate the EIS frequency range used for the G1 measurements, we designed a second-generation electrode array (G2) that included a terminating lead and a looping lead. This array provided the ability to assess the accuracy contour plot of our EIS system and electrode array platform. This exploratory research contributes to the ongoing knowledge of UMEAs in μStim applications. These developments may result in high-fidelity, non-damaging multichannel μStim, and improved neuromodulation technology for the treatment of neural diseases and injuries.

I. Introduction

Invasive neural interfaces are rapidly advancing with the introduction of novel neural interfaces. High-density microelectrode arrays, such as the Neuropixels array [1], have significantly enhanced our ability to capture action potentials from neurons, providing unprecedented insight into neural activity. Concurrently, the development of ultramicroelectrode arrays (UMEAs) has shown promise in extending the functional lifespan of neural recordings and minimizing the biological tissue response due to their reduced scale [2][3][4], while also offering a finer resolution for interfacing with the nervous system. Previously our group has demonstrated that device design affects the sensitivity and stability of neuromodulation [5][6][7]. In this work our efforts are towards optimization of site size for suitable charge delivery and stability. The outcomes of these efforts could enable arrays of UMEAs for stable, high-resolution neuromodulation [8][9].

Primary somatosensory cortex (S1) is an important site for the study of intracortical microstimulation (ICMS), capable of restoring sensory perceptions when stimulated [10][11][12]. Layer-specific reactions within the S1 highlight the need for precise electrode placement and stimulation parameters for optimal neuroprosthetic outcomes [6]. These stimulation studies utilize the primary somatosensory cortex to investigate the longevity and functionality of ICMS. The emerging use of subcellular electrodes, which elicit less immune response and offer higher resolution stimulation, is promising for the next generation of these applications. These electrodes, however, require careful charge management to prevent damage, a challenge that can be addressed by spreading the stimulation charge across multiple channels [12][13].

There is a critical knowledge gap in the design of UMEAs for microstimulation. Minimized site-size is desired to allow for ultra-thin substrates as well as high-density electrode spacing. Ultra-thin substrates have been shown to improve long-term viability and biocompatibility. High-density electrode spacing will allow spatial and temporal current field delivery. However, decreasing site size results in increased charge density, which could result in damage to the electrode structure or damage to neural tissue. These issues highlight the necessity for rigorous evaluation of UMEAs during microstimulation to ensure safety and effectiveness—a challenge not yet fully addressed by the current body of research.

Addressing this pivotal challenge, our study seeks to investigate the operational stability of electrode arrays of various diameter electrodes after in vivo electrical stimulation. We employ Electrochemical Impedance Spectroscopy (EIS) to assess the integrity of the electrodes both before and after delivering controlled electrical stimulation to S1 in the rat model. Through EIS analysis, we measure the stability of the impedance of the electrodes, where stability may indicate minimal damage to the electrode. This approach is designed to test the hypothesis that UMEAs can deliver stimulation without compromising the electrode’s functionality in an in vivo model.

II. Methods

A. Electrochemical Impedance Spectroscopy

In this study, the primary objective of Electrochemical Impedance Spectroscopy (EIS) was to assess the stability of our novel G1 device before and after stimulation sessions in a rat brain model. The G1 devices, manufactured by NeuroNexus Technologies, Inc. (Ann Arbor, MI), represent the first generation in our series, and were fabricated on silicon substrate with gold contacts that were coated with Sputtered Iridium Oxide (SIROF). Each device comprised four shanks, 3 mm in length and 15 μm in thickness (Fig. 1(a)). These shanks supported 32 circular electrode sites, eight per shank, with diameters of 5, 10, 15, 20, 25, 30, 40, 50 μm. The detailed arrangement of these electrodes is depicted in Figure 1(a), showcasing the specific layout of sites on each shank, with inter-site gaps of 10–40 μm, inter-shank distances of 37 μm (body) and 100 μm (tip), and a shank width of 63 μm.

Fig. 1.

Overview of the Experimental Setup for Electrochemical Impedance Spectroscopy and Neural Stimulation. (a) G1 device, the silicon-based design with a detailed view of the electrode sites, ranging from 5 to 50 μm, on the four shanks. (b) EIS setup, a potentiostat-galvanostat with connections to a working electrode (WE), sense (S), reference electrode (RE), counter electrode (CE), and ground (GND). WE and S are connected to a multiplexer for channel selection, further connected to the UMEA immersed in a phosphate-buffered saline (PBS) solution for testing. (c) Somatosensory cortex of a rat brain, the target area for neural interfacing. (d) Signal processing system setup, including the stimulator connected to the signal amplifier, and the ground (GND) reference. The ZIF-Clip^® headstage is connected to the implanted UMEA, with an additional RE placed on the skin. (e) G2 device, the silicon-based design with a detailed view of the electrode sites on the four shanks, and the highlighted terminal sites and loop sites.

For EIS measurements, we utilized the Autolab PGSTAT12 (Metrohm, Utrecht, The Netherlands) equipped with a three-electrode setup (Fig. 1(b)). The working and sense electrodes were connected to two 16-channel multiplexers (MUX) for automated EIS across 32 channels. A platinum-wire electrode (6.0301.100; Metrohm, Utrecht, The Netherlands), served as the counter electrode. A Silver/Silver Chloride electrode (6.0726.100; Metrohm, Utrecht, The Netherlands) was used as the reference electrode. The electrode assembly, alongside the MUX, ZIF-Clip^® headstage, and other components, was housed within a copper Faraday cage to minimize electrical interference. The ZIF-Clip^®, attached to a micromanipulator, ensured only the electrode tips were immersed in the phosphate-buffered saline (PBS) solution within the testing beaker. The PBS solution contains phosphate-buffer 10 mM, KCl 2.7 mM, and NaCl 137 mM, with a pH of 7.4, at room temperature. The PBS was prepared by diluting from a commercial 10X stock.

The EIS data was acquired using the NOVA software in conjunction with the Autolab system, over a frequency range of 1–100 kHz with a resolution of 10 data points per decade. A sine wave potential modulation of 0.01 V RMS was applied at each frequency to each electrode site, and each site’s data was recorded thrice to ensure consistency and reliability of the impedance readings.

To obtain an accuracy contour plot for the Autolab system, we utilized G2 devices (Fig. 1(e)). The G2 device, despite its differences from the G1 model, shares a similar foundational design, providing a relevant reference for G1 device performance assessment. The G2 has specialized terminal and loop designs enabling accuracy measurements of the instrumentation setup. The terminal channels terminate beneath the dielectric and thus allow an estimate of the open-circuit impedance of the fully instrumented system (i.e., the Autolab connected to the MUX, then through a ZIF-Clip^® connector, and finally through the G2 device). The loop design is conductive trace material that runs the entirety of the shank length and returns to a different connector pin; however, it is insulated by the dielectric layers over the entirety of the trace, allowing an estimate of the short-circuit impedance of the fully instrumented system. To augment the fully instrumented EIS, we also measured open-circuit and short-circuit configurations for the Autolab alone, the Autolab+MUX, and the Autolab+MUX+ZIF-Clip^®. For all equipment, data from four iterations were averaged.

B. Surgery and stimulation

All animal experiments and surgeries were performed under the approval and guidance of the Institutional Animal Care and Use Committee (IACUC) of the University of Florida (Gainesville, FL, United States). In this study, a craniotomy at the right somatosensory cortex (Fig. 1(c)) was performed on anesthetized rats. We implanted the G1 devices such that the tip of the electrode array was 1600 μm below the surface of cortex. and conducted immediate electrical stimulation via an IZ2 Electrical Stimulator, and RZ5D Base Processor (Tucker-Davis Technologies, Alachua, FL) (Fig. 1(d)).

Stimulation pulses were single pulse, cathodic-leading, biphasic, and delivered to the channel sites at amplitudes of 0 and 0.4 nC/phase for all channel sizes for 0.4 ms. For the two stimulation amplitudes, 20 repetitions were applied to each channel. The stimulation sequence was pseudo-random over the 20 repetitions to mitigate non-stationarity. Pulses were delivered at 3 Hz.

Upon completion of the stimulation, the device was carefully removed and cleaned, and the surgical procedure was concluded. EIS of the device was then conducted following the previously described in vitro method.

III. Results

Our results first enabled an analysis of EIS data as a function of electrode diameter. Additionally, we constructed plots of EIS pre- and post-stimulation. Finally, accuracy contour plots were generated allowing for determination of the impedance contribution of each individual instrumentation element.

A. EIS Data as a Function of Electrode Diameter

To assess the electrochemical properties of various diameter electrodes, impedance measurements were taken in vitro across a broad frequency spectrum, ranging from 1 Hz to 100 kHz. Figure 2 provides the EIS measurements for the G1 device before electrical stimulation. The Nyquist plot (Fig. 2(a)) shows that the larger channels correspond to a similar, but overall lower impedance. In the Bode magnitude plot (Fig. 2(b)), we observe a decline in impedance magnitude as frequency increases. For a between-channel comparison, larger channels typically exhibit lower impedance magnitude at lower frequencies. The Bode phase plot (Fig. 2(c)) shows the phase angle between the input signal and the measured response. From low frequencies to high frequencies, the phase angles for channels with smaller sizes (5 μm to 40 μm) showed a trough at around 10 – 100 Hz, and a peak around 10000 Hz. The phase of the largest channel (50 μm) did not show a trough in the lower frequencies, but showed a similar peak in the higher frequency.

Fig. 2.

Electrochemical Impedance Characterization of various electrode diameters. This figure comprises three plots, illustrating the impedance properties of electrode arrays with varying site sizes (5 μm to 50 μm). (a) The Nyquist plot displays the imaginary component of impedance (Z”) vs. the real component (Z’) for each channel size, denoted by different colors. (b) The Bode magnitude plot shows the impedance magnitude across a frequency range from 1 Hz to 100 kHz. (c) The Bode phase plot depicts the phase angle in degrees as a function of frequency.

B. EIS Pre- and Post-Stimulation

The EIS analysis of the G1 devices showed some changes in impedance related to electrical stimulation applied in an in vivo setting (Fig. 3). Nyquist plots for each channel size—ranging from 5 μm to 50 μm—demonstrate shifts in impedance post-stimulation. For the 40 and 50 μm channels, there is a notable increase in impedance after the stimulation. All the impedance data showed large variance, especially at lower frequencies. However, across most electrode sizes, post-surgery impedance values tend to cluster closely to the pre-surgery values, indicating a relative stability in impedance post-procedure.

Fig. 3.

Temporal Impedance Profiles of G1 Devices Across Surgical Interventions. Displayed are Nyquist plots for UMEAs with channel site sizes from 5 μm to 50 μm, measured at three distinct time points: before stimulation, and after stimulation. Each plot showcases the imaginary impedance (Z”) versus the real impedance (Z’) with error bars, with the electrode sizes presented in separate panels for clarity. The data points are color-coded: blue for measurements taken before stimulation, and orange for after the stimulation, illustrating the impedance changes over time.

C. Accuracy Contour Plots

We constructed accuracy contour plots to delineate the impedance contributions of the components of our electrochemical system (Fig. 4). This analysis enabled determination of the range within which precise and accurate impedance measurements were obtained when utilizing our system. The Autolab, represented by the blue line, provides the baseline measurement, illustrating the system’s capabilities without additional components from 2 Ω to 3×10¹² Ω at low frequencies around 0.01 Hz, to 1.6×10⁵ Ω to 4.3×10⁵ Ω at high frequencies around 5×10⁵ Hz. With the integration of the MUX, depicted in green, we observe an expected increase in the lowest measurable impedance to 53 Ω to 2.5×10¹¹ Ω at around 0.01 Hz, and 47 Ω to 6064 Ω at 5×10⁵ Hz, suggesting additional circuitry introduces additional base level of impedance. The inclusion of the ZIF-Clip^® (ZC32, Tucker-Davis Technologies, Alachua, U.S.A.), shown in red, slightly reduces the highest measurable impedance to 2.1 × 10¹⁰ Hz, while not largely decreasing the lowest measurable impedance. However, the most insightful data comes from the G2 device measurements, marked in purple, where the impedance range is most restricted to 1266 Ω to 1.4×10⁹ Hz at around 0.01 Hz, and 1074 Ω to 2485 Ω at 5 × 10⁵ Hz, confirming the hypothesis that adding components to the system incrementally affects the impedance limits.

Fig. 4.

An accuracy contour plot of impedance with different components of an EIS measurement system over a wide frequency range from 1×10⁻² Hz to 5×10⁵ Hz. The impedance values for individual system components are plotted to assess their impact on the overall measurement capability. The Autolab potentiostat, indicated by the blue lines, serves as the foundational measurement device, followed by the MUX (green lines) which represents the addition of a multiplexer for channel selection. The ZIF-Clip^®, represented by the red lines, includes the impedance contributions from the Autolab, the MUX, and the ZIF-Clip^® connector. Finally, the G2 device curve (purple lines) incorporates the entire system assembly, reflecting the cumulative impedance of the Autolab, MUX, ZIF-Clip^®, and the electrode itself. Each bottom line represents the lowest impedance possibly accurately recorded by the system at each frequency, and the top line represents the highest impedance possibly accurately recorded by the system.

The results from the G2 device included in the plot provided a reference of operational window for the G1 device in terms of impedance, offering assurance that the measurements are within the system’s accurate detection capabilities. This result is an essential reference for validating the G1 device’s potential application in neural tissue.

IV. Discussion

Our results show some one measure of electrode stability as a function of diameter. These data enable potential optimization of electrode size enabling high-density, ultramicroelectrode array development for neuroprostheses.

The EIS data offer a understanding of the stability and performance of different electrode diameters before and after neural stimulation. Notably, the post-stimulation data reveals that the impedance levels across nearly all electrode sizes exhibit a consistent pattern, maintaining a similar trajectory as pre-surgical readings. This consistency is important, suggesting that UMEAs can function within the neural environment with minimal alterations to their electrical characteristics under safe stimulation parameters, an essential requirement for reliable neural interfaces.

The accuracy contour plots play a crucial role in validating the precision of our impedance measurements from the sites of G1 and G2 devices. Since the impedance measurements for the accuracy contour plot includes the electrode itself, we can have a benchmark for the accuracy of our impedance data. This allows us to use these impedance data to further look into questions like whether electrodes can function well in stimulation and recording tasks with such impedance values. Additional questions regarding this could be a thorough analysis of the accuracy of our recording results on different site sizes, especially the small ones like the 5 μm ones.

Together, these observations enable future investigation of UMEA neuroprosthesis development. EIS before and after stimulation is the first step for us to further investigate the stability of G1 and G2 devices in a more intense and realistic environment, such as larger stimulation amplitudes, or different stimulation waveforms. One could also further investigate the EIS in vivo longitudinally to determine electrode stability over time post-implantation. On the other hand, analysis of the accuracy contour plot provides us with a detailed understanding of our current system capabilities and guides us in strategizing future parameter selections for EIS analysis. In general, the current analyses serve as a basis for formulating hypotheses that enable further investigation, particularly on how these impedance changes may affect long-term functionality and integration of neural interfaces.