Vitamin C-reduced graphene oxide improves the performance and stability of multimodal neural microelectrodes.

Nanocarbons are often employed as coatings for neural electrodes to enhance surface area. However, processing and integrating them into microfabrication flows requires complex and harmful chemical and heating conditions. This article presents a safe, scalable, cost-effective method to produce reduced graphene oxide (rGO) coatings using vitamin C (VC) as the reducing agent. We spray coat GO + VC mixtures onto target substrates, and then heat samples for 15 min at 150°C. The resulting rGO films have conductivities of ∼44 S cm-1, and are easily integrated into an ad hoc microfabrication flow. The rGO/Au microelectrodes show ∼8x lower impedance and ∼400x higher capacitance than bare Au, resulting in significantly enhanced charge storage and injection capacity. We subsequently use rGO/Au arrays to detect dopamine in vitro, and to map cortical activity intraoperatively over rat whisker barrel cortex, demonstrating that conductive VC-rGO coatings improve the performance and stability of multimodal microelectrodes for different applications.

Introduction

Our understanding of the circuitry underlying brain function and disease is largely informed by the ability to record and modulate neural activity with high-fidelity electrodes. Typically, these are composed of inorganic materials such as doped silicon (Si), platinum (Pt), gold (Au), and iridium oxide (IrOx), as they are adequately conductive and biocompatible, electrochemically stable, and easy to pattern into a variety of structures using standard lithographic techniques. However, resolving the spatial (10–100 μm) and temporal (1–10 ms) scales governing cellular-level activity requires miniaturizing the electrode contacts (Kellis et al., 2016). Microscale electrodes based on these materials typically exhibit high impedance (>1 MΩ) and modest charge injection capacities, which degrade the signal-to-noise ratio (SNR) of recordings and limit safe neuromodulation (Cogan, 2008; Ludwig et al., 2011; Boehler et al., 2020). Furthermore, monitoring the dynamics of biochemical species is crucial to investigate mechanisms of neural function and disease, but electrodes with limited surface area and appropriate surface chemistry suffer from low sensitivities and sub-optimal detection limits above physiological levels (Roychaudhuri, 2015).

Common strategies to overcome these limitations are nanoscale roughening and coatings on the surface of metallic or Si electrodes (Aregueta-Robles et al., 2014), which result in a higher effective surface area and additional adsorption sites for analytic detection (Cogan, 2008). The overall effect is an increase in the electrochemical double-layer capacitance, leading to impedance reduction and enhanced capacity to store and safely deliver stimulation charge. Common materials used for electrode surface enhancements include: titanium nitride (TiN) (Zhang et al., 2017; Khan et al., 2019), nanoscale carbons like carbon nanotubes (CNTs) (Keefer et al., 2008; Shoval et al., 2009) and nanodiamonds (Turcheniuk and Mochalin, 2017; Puthongkham and Venton, 2019), conductive polymers (CPs) such as polypyrrole (PPy) (Harris et al., 2013) and poly(3,4-ethylenedioxythiophene) doped with polystyrene sulphonate (PEDOT:PSS) (Ludwig et al., 2011; Castagnola et al., 2015; Ganji et al., 2017a; Ganji et al., 2017b; Aqrawe et al., 2019), and hybrids of these materials (Behler et al., 2009; Castagnola et al., 2009; Luo et al., 2011; Sriprachuabwong et al., 2012; Kolarcik et al., 2014; Zhang et al., 2017). However, TiN is an extremely stiff ceramic that requires high-temperature processing and specialized nitridation treatments to be effective (Zhang et al., 2017; Feng et al., 2019). Nanocarbon coatings also require complex chemical modifications to promote adhesion to most electrode surfaces (Keefer et al., 2008; Kostarelos, Bianco and Prato, 2009; Aregueta-Robles et al., 2014). Similarly, CPs are often electrochemically polymerized onto the target substrate, followed by additional post-processing to selectively coat only the exposed contact sites (Harris et al., 2013; Aregueta-Robles et al., 2014; Castagnola et al., 2015). Even if directly cast as thin films onto the target substrate—as is possible with solution-processable materials, such as PEDOT:PSS (Khodagholy et al., 2015; Ganji et al., 2017a; Ganji et al., 2017b)—many CP coatings still suffer from degradation, mechanical failure, and delamination, due to multiple factors including poor adhesion (Green et al., 2012) and excessive thickness of the coating layer (Cui and Zhou, 2007).

Reduced graphene oxide (rGO) is an attractive coating material for neural electrodes due to its predominantly capacitive nature, electrochemical stability, ease of processing and tunability, and lower impedance and higher charge delivery characteristics compared to metals and Si (Apollo et al., 2015; Wang et al., 2019; Hejazi et al., 2020a, 2020b). In the context of electrode technologies, rGO, like many other hybrid carbon-based nanomaterials (Laurila et al., 2017), has been primarily explored as a surface area enhancing material for biochemical sensors (Robinson et al., 2008; Wang et al., 2009; Shao et al., 2010; Wang et al., 2014; Yang et al., 2014; Ng et al., 2015; Fang et al., 2017), while only a few works have used rGO for neural recording and stimulation (Apollo et al., 2015; Liu et al., 2015b; Dalrymple et al., 2019; Wang et al., 2019). For most biosensing applications, rGO films have been produced by direct reduction of dry films of its hydrophilic precursor, graphene oxide (GO) (Alazmi et al., 2016). Many of these post-assembly reduction steps typically involve exposure to hydrazine vapor (Gilje et al., 2007; Stankovich et al., 2007; Becerril et al., 2008; Eda et al., 2008; Wallace et al., 2008) or high temperature annealing at temperatures T > 1,000°C (Chen et al., 2016; Wang et al., 2018). However, these processes pose serious health and safety concerns (Zhang et al., 2010; Ding et al., 2015), and may also be incompatible with other electrode components, such as encapsulating polymers. For materials such as parylene-C, polyimide (PI), SU-8, and polyethylene terephthalate (PET), the glass transition temperature (T) lies in the ∼120°C–300°C range (Yokota et al., 2001; Harder et al., 2002; Lin et al., 2011; Chung and Park, 2013; Kahouli et al., 2014), which precludes high temperature annealing. Furthermore, it has been shown that hydrazine causes long-term degradation of a variety of polymeric films (Dine-Hart et al., 1971; Goel et al., 2009).

Alternative reduction methods consist of liquid-phase reduction via strong acids, such as hydrochloric (Ning et al., 2014) and hydriodic (Lin et al., 2013) acids, or ammonia-rich mixtures (Wallace et al., 2008; Pham et al., 2010). In these processes, GO and the reducing agent are mixed at moderate temperatures (T > 60°C) for a few minutes, followed by vacuum filtration to form the rGO films (Wallace et al., 2008; Fernandez-Merino et al., 2010). However, rGO solutions suffer from colloidal instability (Xu et al., 2014; Liu et al., 2015a) and surfactants like N-methyl-2-pyrrolidone (NMP) or dimethylformamide (DMF) are typically required (Park and Ruoff, 2009). Both NMP and DMF are cytotoxic (Salavagione et al., 2017), and although Cyrene might be a more biocompatible alternative, its long-term stability and safety have not been fully characterized (Salavagione et al., 2017; Pan et al., 2018). Furthermore, surfactants typically lower the electronic conductivity of the final rGO films (Fernández-Merino et al., 2012). Recently, L-ascorbic acid—a.k.a. vitamin C (VC)—has been established as a viable, biocompatible, and safe GO reducing agent (Dua et al., 2010; Zhang et al., 2010; Fernandez-Merino et al., 2010; Ding et al., 2015). In previous studies, VC-mediated GO reduction was conducted in solution, and rGO films were subsequently processed from the reduced suspensions by vacuum filtration. However, vacuum filtration results in large-area films (typically 90–120 μm in thickness), which are difficult to handle, transfer, and pattern, and which are prone to damage during subsequent manufacturing steps. Thus, vacuum filtration is largely incompatible with standard microfabrication processes, and would particularly hinder scale-up of the process to reliably and repeatably fabricate neural electrodes. An alternative, high-through, solutions-based method would therefore be more desirable for realizing rGO-based coatings for neural technologies, as this could be easily integrated into standard fabrication flows for microelectrodes.

In the present work, we propose a novel method to produce rGO coatings for neural microelectrode arrays that is safe, high-throughput, compatible with common electrode materials, and easy to integrate into standard microelectrode fabrication processes. Taking advantage of the slow reaction kinetics of VC as a reducing agent at room temperature (Zhang et al., 2010; Fernandez-Merino et al., 2010; Abulizi et al., 2014), we demonstrate the direct fabrication of rGO-coated neural microelectrode arrays, using GO + VC mixtures spray cast onto bare Au contacts micropatterned on Si wafers. To maximize the electronic and electrochemical properties of the coatings, we have optimized processing parameters such as VC concentration and heating time. We also demonstrate the fabrication and one-step coating process of planar rGO/Au microelectrode arrays encapsulated in parylene-C for microelectrocorticography (μECoG) recordings, cortical stimulation, and neurochemical sensing. In vitro characterization of the electrochemical properties of rGO/Au microelectrodes indicates that rGO coatings are stable and significantly enhance the electrode surface area, leading to impedance reduction, as well as charge storage and charge injection capacity increases compared to same-size Au electrodes and other coating materials from literature. The enhanced surface area afforded by the rGO coatings also results in an increase in adsorption site count, enabling the in vitro detection of the neurotransmitter dopamine (DA) with a high sensitivity and low detection limit. Finally, to demonstrate the feasibility of rGO/Au μECoG arrays for monitoring microscale neural circuits at high resolution, we show high-density mapping of evoked cortical responses in rat somatosensory cortex from whisker stimulation. Altogether, we demonstrate the potential of VC-reduced rGO coatings as a unique strategy to improve standard electrode technologies, in order to advance future multimodal neural interface applications.

Results and discussion

Optimization of the VC reduction method and rGO film characterization

An overview of the VC reduction process is illustrated in Figure 1. The reaction kinetics of VC in water is well established to be of the first-order, and are thus exponentially proportional to the solution temperature (Wang et al., 2017). This implies that VC solutions at room temperature may be used for many days without concerns over material degradation or colloidal instability (De Silva et al., 2017; Wang et al., 2017). Furthermore, when VC is mixed with GO dispersions at room temperature, at least two days of continuous stirring are required before complete reduction of the solution is achieved (Zhang et al., 2010; Abulizi et al., 2014; De Silva et al., 2017). Altogether, this means that it is possible to mix GO and VC, and subsequently form films by directly spray coating GO + VC mixtures onto the target substrate (Figure 1B). The resulting films appear in the characteristic light brown color of GO when dried, and they can then be reduced at any subsequent point in time through a simple heating step. Once the films are heated, they progressively become darker, as a result of the reduction process (Figure 1C and Video S1).

Figure 1

VC-rGO thin film fabrication flow

(A) The substrate is preheated to 150°C for 5 min, followed by (B) spray coating of the GO + VC mixed solution.

(C) After an additional 15 min of baking at 150°C, the GO + VC films are reduced into conductive VC-rGO films. Bottom panels show the reduction process on glass slides and parylene-C-encapsulated μECoG arrays pre-patterned in Au on a Si wafer. The transition in color of the GO + VC films from a light brown in panel (B), to a dark black in panel (C), is indicative of GO reduction. Scale bars: 25 mm.

We varied VC concentrations and heating time, and measured the sheet resistance of the resulting film (Figures 2A and 2B). The sheet resistance shows dependence on VC concentration up to 50 mM, and on heating time only up to 15 min at 150°C, beyond which no improvement is observed. Accordingly, at a VC concentration of 50 mM and a heating time of 15 min at 150°C, the optimal sheet resistance of the VC-rGO thin films is 981.50 ± 125.96 Ω sq−1. From stylus profilometry, the average film thickness is 236.17 ± 26.37 nm, yielding an average DC conductivity of 43.86 ± 4.08 S cm−1. This value is comparable to previous rGO films from VC reduction in solution (Zhang et al., 2010; Fernandez-Merino et al., 2010), and is also higher than some hydrazine-reduced films (Gilje et al., 2007; Wallace et al., 2008). Figure S1 and Table S1 summarize the DC conductivity of rGO films obtained from different published reduction methods (Domingues et al., 2013, Huang et al., 2011, Konios et al., 2015, Li et al., 2008).

Figure 2

Optimization and characterization of the VC-rGO films

(A and B) Dependence of the dry VC-rGO film sheet resistance on (A) VC concentration and (B) heating time at 150°C (mean ± std. dev., N = 3 samples).

(C and D) AFM of VC-rGO films made using 50 mM VC on glass (C) before and (D) after 15 min of baking at 150°C, showing flake morphology and film roughness.

(E) SEM of a VC-rGO film on glass.

(F) Raman spectra of GO + VC and VC-rGO films on glass, showing the characteristic graphitic peaks in the D and G bands. The ID/IG ratio is also provided for each film type.

AFM of the dried films before (Figure 2C) and after (Figure 2D) the optimized reduction reveals similar flake sizes and distributions, demonstrating that the VC reduction method preserves the structure of the individual graphene flakes. SEM images (Figure 2E) show that individual rGO sheets endow the film with a wrinkled and folded morphology, which contributes to increasing the effective surface area. Finally, Raman spectroscopy (Figure 2F) reveals clear distinctions between the lattice structures of the GO + VC and VC-rGO films. The D band, which is related to the breathing modes of the six-atom carbon lattice, is especially sensitive to edge defects and domain sizes (Stankovich et al., 2007), and thus, the ID/IG ratio is a useful measure for distinguishing between GO and rGO films (Kudin et al., 2008; Kaniyoor and Ramaprabhu, 2012; Ferrari and Basko, 2013). Prior to reduction, overlapping graphitic flakes appear as defects in the Raman spectra, and thus there is an initially pronounced D peak, resulting in an ID/IG ratio of 0.854 for the GO + VC films. Upon exposure to heat, though, there is extensive reduction of the carbon lattice and removal of oxygen moieties from the GO sheets, resulting in smaller sized domains in which sp2 vibrations may occur (Stankovich et al., 2007). This, in turn, causes an increase in the intensity of the D band and in the ID/IG ratio (1.013) for the VC-rGO films. Given the limited range of the Raman system (180–2,580 cm−1), the 2D and (D + D′) bands do not appear in the spectra, although they are expected in the range of 2,600–3,000 cm−1. However, these peaks show only minor changes in shape or location in response to the reduction step, as has been discussed in previous reports (Kaniyoor and Ramaprabhu, 2012; Muhammad Hafiz et al., 2014).

In addition to the changes in the properties described above, GO + VC films also become more hydrophobic following reduction (Figure S2). For dry GO + VC films, the water contact angle is comparable to bare glass (28.1° ± 0.2° for GO + VC vs. 22.9° ± 1.7° for glass, N = 3 samples). After reduction, the contact angle of the VC-rGO films nearly doubles (45.6° ± 0.7°, N = 3), in agreement with previous studies (Xu et al., 2014; Liu et al., 2015a). The increased hydrophobicity of VC-rGO may result in longer lifetimes for these films in aqueous environments, such as might be found in living systems. Indeed, a few reports have already highlighted the advantages of hydrophobic carbon nanomaterial coatings to prevent moisture ingress from causing delamination, corrosion, or other failures in both implantable (Zhao et al., 2016) and wearable (Chun et al., 2019) biomedical designs. To further explore this potential benefit in our own work, we evaluated the shelf-life and electrolytic stability of VC-rGO films aged in (i) air, and (ii) 1X PBS at room temperature, over 30 days. Conductivity measurements at the initial and final timepoints (Figure S3) reveal that VC-rGO is highly stable against oxidative degradation in both environments, with the films retaining (93.5 ± 0.8)% and (70.3 ± 3.7)% of their starting conductivity in air and in PBS, respectively (N = 3 samples). This is a significant advantage over alternative coatings—such as PEDOT:PSS, for which conductivity has been shown to decay exponentially with time (Stepien et al., 2017; Kee et al., 2019)—and could be especially advantageous for long-term implants in biological media.

Characterization of rGO/Au and Au μECoG properties in vitro

Optical microscopy of the microfabricated Au and rGO/Au μECoG arrays reveals the rougher surface topology following deposition of the VC-rGO coating (Figures 3A and 3B). We observed for all devices that the VC-rGO coating adhered well to the underlying Au layer, and in every instance appeared to be a continuous coating at the exposed contact sites, as well.

Figure 3

In vitro characterization of the rGO/Au μECoG arrays

(A and B) Optical microscopy images of (A) a bare Au and (B) an rGO/Au μECoG array. Insets show an individual channel from each array, including the 50 × 50 μm contact.

(C and D) Bode plots of the impedance (C) modulus and (D) phase for the Au and rGO/Au μECoG arrays. Points are means, shaded regions are std. dev. (N = 3 devices, 48 individual channels total for each electrode type).

(E) Example cyclic voltammograms (CVs) for the Au and rGO/Au μECoG electrodes in the safe operating voltage window of Au (−0.6 V ∼ +0.8 V) and in the full water window of rGO (−1.4 V ∼ +1.6 V). Scan rate, 200 mV s−1. Values of CSCc are included for each material.

(F) Voltage transients at the Au and rGO/Au μECoG arrays in response to 7.5 μA biphasic current pulses (duration, 650 μs; inter-pulse delay, 20 μs). Black dashed lines denote the water window limits of the Au electrodes.

The higher degree of surface roughness imparted by the VC-rGO coating results in a reduction of the overall electrode impedance across the 1–103 Hz frequency range (Figure 3C). In the higher frequency range (i.e., >103 Hz), the Au and rGO/Au microelectrodes behave similarly, which is expected given that this region is dominated by the solution resistance (R) and electrode geometric surface area, rather than by material properties (Franks et al., 2005; Boehler et al., 2020). At the reference frequency of 1 kHz, the average impedance of the Au and rGO/Au electrodes are 177.00 ± 8.37 kΩ and 21.12 ± 18.29 kΩ, respectively. With such a lower impedance, it is expected that the thermal noise in the case of the rGO/Au microelectrodes will be lower than for bare Au electrodes, suggesting an improvement in overall recording quality (Lempka et al., 2011; Scalia et al., 2012). Fitting the impedance to an equivalent circuit model (Table S2 and Figure S4) further reveals that the double-layer capacitance, C, of the bare Au electrodes significantly increases by ∼400x after rGO coating (Au: C = 11.32 ± 0.76 μF cm−2 vs. rGO/Au: C = 4.40 ± 1.20 mF cm−2, p < 0.05). Such increase in C can be ascribed to the higher effective surface area, but also to the intrinsically higher capacitance of rGO compared to Au (Piela and Wrona, 1995; Lee et al., 2013). The significantly lower values of R observed for the rGO/Au electrodes (Au: 0.180 ± 0. kΩ cm2 vs. rGO/Au: 0070.035 ± 0.022 kΩ cm2, p < 0.05) suggest that the VC reduction method sufficiently removes excess oxygen terminations from the GO, further improving the charge-transfer characteristics at the VC-rGO layer (Yang and Gunasekaran, 2013).

The increased C and enhanced electrochemical surface area of the VC-rGO coatings should also result in a larger area voltammogram for the rGO/Au electrodes, which is indicative of the increased ability to collect and store more charge at the electrode-electrolyte interface (Figures 3E and S5). Indeed, in the water window of Au (−0.6 V ∼ +0.8 V) (Ganji et al., 2017b), the CSCc of uncoated Au electrodes is 1.41 mC cm−2, while for rGO/Au electrodes in the same potential window, the CSCc is 9.73 mC cm−2. However, when extending the analysis to the full operating window of rGO (−1.4 V ∼ +1.6 V), the CSCc of the rGO/Au electrodes becomes 45.58 mC cm−2, indicating that the VC-rGO coatings considerably increase the charge storage capabilities of bare Au contacts.

Given the greater charge storage capacity of the VC-rGO coatings, it may also be expected that the rGO-coated electrodes should be capable of injecting more charge for stimulation studies compared to uncoated electrodes. From CV, it is possible to calculate the total safe charge injection limit, Q, using Equation 1, which represents the total theoretical amount of charge that an electrode can inject, based on the double-layer capacitance and water window limits of the material. The average Q of the Au electrodes is 0.016 mC cm−2, compared to 13.20 mC cm−2 for the rGO/Au μECoG arrays (Table S3). This ∼1000-fold increase in the total safe injection limit is due to the higher C of rGO compared to Au, but also follows from the broader rGO potential window compared to most metals. In terms of Q, too, the VC-rGO coatings reported here outperform many other standard neural electrode materials, such as IrOx and TiN (Weiland and Humayun, 2014; Kang et al., 2016; Yamagiwa et al., 2015), and approach the higher limits of nanocarbon-based fiber microelectrodes (Apollo et al., 2015; Hejazi et al., 2020a, 2020b) (Table S3).

We have also characterized the electrode charge injection properties using stimulus waveforms more relevant to neural stimulation experiments (Figure 3F). Figure S6 provides an overview of the components that contribute to the polarization potential response in a typical voltage transient experiment (Ganji et al., 2017b; Cisnal et al., 2018; Driscoll et al., 2021), while Figure S7 shows complete voltage transients and CIC evaluation for the Au and rGO/Au μECoG electrodes. In response to 7.5 μA of input current, the bare Au electrodes exceed their safe operating window, reaching polarization potentials greater than ±1 V depending on the current pulse polarity (Figure 3F). In contrast, the rGO/Au electrodes do not exceed ±0.5 V for the same levels of current injection (Figure 3F). Indeed, under the same biphasic pulsing protocol, rGO/Au electrodes can be subjected to currents of up to 15 μA before they exceed their operating window (Figures S7A and S7B). Furthermore, for the bare Au electrodes, the charge injection capacity is CIC = 0.36 ± 0.02 mC cm−2, while for rGO/Au electrodes, CIC = 1.09 ± 0.13 mC cm−2 (Table S3 and Figure S7C). The rGO/Au CIC values are comparable to PEDOT:PSS coatings on bare metallic electrodes (Ganji et al., 2017a; Ganji et al., 2017b), as well as hybrid PEDOT:PSS/CNT coatings on Pt (Luo et al., 2011).

Overall, the higher C, Q, and CIC values, and lower impedance of the rGO/Au electrodes suggest that less power would be consumed by the rGO/Au electrodes in order to maintain the same level of instantaneous electrode current during stimulation experiments (Kelly and Wyatt, 2011; Brunton et al., 2013). VC-rGO coatings also improve the stability of the electrodes under repeated charge injection experiments (Figure S7D). In these tests, we injected continuous biphasic stimulation pulses at 500 μC cm−2 phase−1 at a frequency of 130 Hz over 24 h, and monitored the impedance every 260,000 pulses (11 million cycles total, 2,808C cm−2 phase−1 total charge injection). Such injected charge density exceeds the safety limits for Pt-based cochlear implants set by the American National Standards Institute (ANSI, 216 μC cm−2 phase−1) (ANSI/AAMI, 2017; Dalrymple et al., 2019), thus representing a very rigorous test of the electrode coating stability. At the end of the study, the 1 kHz impedance of the bare Au electrodes increased nearly 5x (113 ± 56 kΩ initial, 537 ± 284 kΩ post-injection), whereas continuous charge injection had no effect on the rGO/Au electrodes (initial impedance: 52 ± 10 kΩ, post-injection impedance: 51 ± 16 kΩ). It may be that long-term exposure to such high charge density forced the Au electrodes to operate at voltages outside their safe operating window, causing degradation over time, whereas the rGO-coated electrodes were not driven outside of their much larger water window by the high charge density. The long-term stability of the rGO/Au impedance further suggests that the VC-rGO coating adheres well to the Au surface, and does not delaminate even under severe pulsing conditions. Furthermore, given the stability of the electrochemical impedance, it is likely that the other advantageous properties of the VC-rGO coating, including its CSCC and CIC, remain largely unchanged, as well.

rGO/Au μECoG arrays for the detection of dopamine

Given the substantial body of literature exploring rGO-coated electrodes for biosensing (Wang et al., 2009; Shao et al., 2010; Wang et al., 2014; Yang et al., 2014; Ng et al., 2015), we also investigated the performance of rGO coatings for DA detection. Figure 4A shows clearly distinguishable oxidation and reduction peaks at the rGO/Au μECoG array for concentrations of DA ([DA]) in the range of 1–100 μM. Note that there is a skewing of the CVs, a drift in the oxidation/reduction peak location, and a change in the peak width depending on [DA]. These phenomena are likely the result of an iR drop across the rGO/Au electrodes, caused by both the non-zero spreading resistance of each separate DA solution, as well as the highly capacitive nature of the VC-rGO coating itself (Pajkossy and Mészáros, 2020). Nevertheless, following background subtraction, a calibration curve can be constructed, relating peak oxidation currents to [DA], and showing a linear trend for [DA] < 50 μM (Figure 4B). Above this limit, the linear correlation breaks down due to an increase in DA molecules adsorbed to the electrode surface, leading to a transition from an adsorption-controlled to a diffusion-controlled regime (Venton and Cao, 2020). The fitting equation in the linear range is i = 0.029C + 1.044 (R = 0.9965), where i is the peak oxidation current in μA, and C is the concentration in μM. The slope of this fitting curve, m = 29 nA μM−1, is the sensitivity of the rGO/Au microelectrodes, while the limit of detection (LOD) is 62.53 ± 16.59 nM (calculated from Equation 2, with a SD of the background signal from the scan in 1X PBS of s = 0.60 ± 0.16 nA). It is important to highlight that the LOD here is purely theoretical; the lowest concentration experimentally resolved with the rGO/Au devices was [DA] = 1 μM. Future work will be necessary to explore the experimental LOD of the rGO/Au devices, and confirm that it agrees with the value determined theoretically.

Figure 4

In vitro detection of dopamine and ferrocyanide with rGO/Au microelectrodes

(A) CVs at an rGO/Au μECoG in DA-PBS at varying [DA]. Scan rate, 4 V s−1.

(B) The peak oxidation current of the CVs in (A) vs. [DA]. The corresponding equation of best fit and R is included for the linear regime of 1 μM ≤ [DA] < 50 μM.

(C) CVs at an rGO/Au μECoG in 1X PBS and 10 mM Fe(CN)6. Scan rate, 50 mV s−1.

Table S4 provides an overview of some comparable sensors, demonstrating that the rGO/Au electrodes exhibit comparable or better sensitivity than modified glassy carbon electrodes (GCE) (Wang et al., 2014; Thamilselvan et al., 2019), many carbon fiber electrodes (CFEs) (Davis et al., 2020; Hejazi et al., 2020a, 2020b; Patel et al., 2020), and even modified CNTs (Palomäki et al., 2018). In contrast to sensitivity, LODs differ greatly among the various electrode types, but importantly, the LOD of the rGO/Au electrodes here is sufficiently low to support their use for the detection of DA at biologically relevant levels (1 nM–20 μM) (Labib et al., 2016; Palomäki et al., 2018; Matt and Gaskill, 2020; Patel et al., 2020).

By applying the Randles-Sevcik equation (Equation 3), it is also possible to estimate the effective electrochemical surface area of the rGO/Au electrodes, A (Kang et al., 2012; Wang et al., 2014). For this analysis to hold true, it is critical that the reaction considered be a case of reversible outer sphere electron transfer (Ngamchuea et al., 2014; Muhammad et al., 2020), and so the Fe(CN)6 redox event was considered (Figure 4C). Accordingly, we use the value of i determined from the CV of Fe(CN)6 at a concentration of 10 mM, as well as the diffusion coefficient D = 7.63 × 10−6 cm2 s−1, which is standard from literature. Under these considerations, the calculated A of the rGO/Au electrodes is 0.0143 mm2, which corresponds to a ∼6x increase in surface area compared to the geometric size of the contact (GSA = 0.0025 mm2). This finding supports the hypothesis that the enhancement of the electrochemistry at the rGO/Au electrode interface compared to uncoated Au arises from the larger effective surface area of the VC-rGO coating. This high area enhancement factor may also indicate that there are more sites for molecules to adsorb to the rGO/Au surface, and since many biomolecules beyond DA—including serotonin (5-HT), glucose, and hydrogen peroxide (H2O2)—must adsorb to the electrode surface before they can be detected (Labib et al., 2016), this implies that the rGO/Au electrodes could be useful platforms for biochemical sensing studies involving a wide variety of physiologically relevant species. It will also be important to conduct interference studies in the future, so as to define the selectivity of the rGO/Au microelectrodes toward different species with similar redox potentials (Wang et al., 2014; Yang et al., 2014; Zhang et al., 2017). Looking toward long-term applications, as well, susceptibility to biofouling during DA or other molecule detection will have to be explored (Hanssen et al., 2016),(Debiemme-Chouvy and Cachet, 2018).

In vivo recording experiments with the rGO/Au μECoG arrays

Finally, to demonstrate the utility of the rGO/Au μECoG arrays for in vivo electrophysiology, we conducted a somatosensory evoked potential (SSEP) study in an adult rat (Figure 5). A single whisker was stimulated mechanically with a piezoelectric capillary tube while recording neural activity with the μECoG array across a roughly 750 × 750 μm region of the cerebral cortex sensitive to input from this whisker. Evoked potentials were averaged over 100 separate stimulation presentations, to provide a truly representative mean response for each targeted whisker. Upon whisker stimulation, robust SSEPs (Zhu and Connors, 1999; Luhmann and Khazipov, 2018; Vilarchao et al., 2018) appear across the rGO/Au μECoG array (Figure 5A, left panel). Control recordings subsequently repeated over the same area of cortex with the whisker removed from the capillary tube confirm that the evoked waveforms are not artifacts caused by the whisker stimulation system (Figure 5A, middle panel). Following subtraction of the noise from the SSEPs, we observe clean stimulus-evoked neural activity for each channel of the rGO/Au μECoG array (Figure 5A, right panel), peaking at ∼10–15 ms after whisker deflection. This response latency is in excellent agreement with previous reports (Benison et al., 2007; Kheradpezhouh et al., 2017). Furthermore, high-density mapping with the rGO/Au μECoG array evidence a clear spatial gradient in the SSEPs across the sampled area (Figure 5B), with a maximum at the site of the cortical representation for the stimulated whisker (Figure 5C). The distribution of evoked neural activity also clearly shifted in a manner consistent with the cortical representations map when a different whisker was deflected (see Figure S8).

Figure 5

In vivo recording with an rGO/Au μECoG array

(A) Left: total evoked potential, middle: noise, right: resulting evoked neural response recorded across an rGO/Au μECoG array. Time on the x axis is relative to onset of mechanical stimulation of a contralateral whisker. Each trace is the mean of 100 evoked responses on each channel, after excluding the 10% largest and 10% smallest responses.

(B) Map of the evoked neural responses on the rGO/Au μECoG array. The colormap represents peak amplitudes of the evoked responses, and overlaid traces show the mean stimulus-evoked response recorded at each channel.

(C) Top: approximate location of the stimulated whisker, highlighted in blue. Bottom: estimated location of the rGO/Au μECoG array relative to the whisker representations in barrel cortex, with the representation corresponding to the stimulated whisker highlighted in blue.

While the in vitro impedance of the rGO/Au arrays at 1 kHz is in the range of ∼20–80 kΩ, the in vivo impedance of the array over vS1 increases to ∼50–500 kΩ (Figures S9A and S9B). The lack of correlation between the SSEP amplitudes and the impedance distribution on the array, though, confirms that the spatial organization of the SSEPs reflects the vibrissal cortical whisker representations, and is not dependent on the individual channel impedance (p = 0.876 for Stim Trial 1, p = 0.613 for Stim Trial 2, Figure S9C). A voltage divider effect due to the 50–500 kΩ electrode impedances in series with the input impedance of the recording system (nominally 16 MΩ at 1 kHz for the Intan RHS system) is expected to produce only about a ∼2% difference in signal amplitudes across the array (Nelson et al., 2008). This finding, coupled with the anatomically accurate mapping of SSEP distributions, highlights the advantage of a high-density, microscale array for mapping cortical activity with high spatial resolution.

Conclusion

In this work, we have presented a novel method to enhance the recording, stimulation, and biochemical detection properties of neural microelectrodes using reduced graphene oxide coatings. Specifically, we have demonstrated a novel approach for realizing rGO films that leverages the advantageous kinetics of vitamin C to complete the reduction process in a safer, fully biocompatible, non-destructive, and highly scalable fashion. The VC reduction process is specifically optimized for compatibility with the soft, conformable, polymeric substrates that are commonly used in implantable medical devices. We have also designed the rGO processing and film deposition method for easy integration into microfabrication process flows for producing neural microelectrodes. The VC-rGO coatings demonstrated in this work are sufficiently conductive, highly stable, and have significantly improved electrochemical properties over bare metallic electrodes, with values exceeding even PEDOT:PSS and CNT coatings.

The stability of the rGO-coated electrode impedance under continuous charge injection, coupled with the stability of the VC-rGO DC conductivity over time in both air and electrolytic environments, suggests that the VC-rGO films are highly versatile and potentially stable for long-term applications. These findings, along with the excellent charge storage and charge-transfer characteristics, suggest that rGO-coated electrodes are excellent candidates for chronic recording and stimulation studies in vivo, although further investigation is required. We have already demonstrated high-density cortical recordings of stimulus-evoked potentials in a rat model, which evidence the ability of this technology to map spatial gradients of neural activity on the microscale. Future work may seek to complete simultaneous recording and stimulation experiments in vivo under a similar experimental setup, wherein the rGO/Au arrays may be used not only for recording electrophysiology but also for direct electrical stimulation of neural tissue. We also demonstrated the in vitro detection of dopamine using the rGO/Au μECoG arrays with sensitivity of 29 nA μM−1 and a limit of detection of 63 nM. These values are comparable to standard electrode technologies used for in vivo biochemical studies, such as carbon fiber electrodes, and thus may enable future multimodal experiments involving analytical sensing of biologically relevant chemical species alongside recording and stimulation.

Overall, our novel method of producing rGO coatings relies on the safe and efficient chemical reduction properties of VC. In addition to addressing the safety and polymer compatibility concerns posed by other chemical and thermal reduction processes, the VC reduction method is a biocompatible, readily scalable, cost-effective procedure to realize rGO films for applications in bioelectronics. Furthermore, the resulting rGO films may serve as an improvement over other complex, time-consuming surface enhancement techniques, resulting in electrode coatings that significantly improve the recording, stimulating, and biochemical sensing properties of electrode interfaces for the nervous system.

Limitations of the study

In this work, we show a scalable and safe process to fabricate rGO-coated neural microelectrodes and demonstrate the feasibility of high-resolution neural recordings in vivo, alongside very promising charge delivery and dopamine (DA) sensing performance in vitro. Further experiments are needed to characterize stimulation performance and sensitivity for neurochemical sensing in vivo. The recording studies in this report were also conducted acutely, so future work ought to explore the chronic stability of these arrays in vivo, especially in light of the excellent in vitro stability results we discuss. In terms of biosensing, susceptibility of VC-rGO coatings to biofouling, as well as specificity in physiological conditions with multiple interfering chemical species, remain important aspects to be further explored. Additional work will also elucidate the ability of VC-rGO to detect other biologically relevant neurotransmitters and analytes beyond DA, such as serotonin, acetylcholine, and uric acid.

STAR★Methods

Key resources table

Resource availability

Lead contact

For further information or requests for resources, reagents, and other matters, please connect with the lead contact and corresponding author, Dr. Flavia Vitale (vitalef@pennmedicine.upenn.edu).

Materials availability

This study did not generate any new or unique reagents. All materials used in this study (including reagents and animals) are commercially available, and details for purchasing them may be found in the Key resources table (KRT).

Experimental model and subject details

For in vivo animal studies, one male rat (Crl:SD, 225 g) was used for an acute neural recording session with the rGO/Au μECoG array. Anesthesia was induced with 5% isoflurane in oxygen. The anesthetized rat was placed in a stereotaxic frame with his body resting on a heating pad. Buprenorphine SR was administered for systemic analgesia. A midline scalp incision site was shaved, cleaned, and injected with line block of 0.25% bupivacaine. Respiratory rate and pedal and palpebral reflexes were checked every 10 min to verify depth of anesthesia, which was maintained with 3–3.5% isoflurane. The scalp was incised to expose the dorsal skull and a craniotomy of the right parietal bone was performed to expose the vibrissal representations of primary somatosensory cortex (i.e., barrel cortex). Two 00–90 skull screws with attached gold pins were implanted: one in the left frontal bone (recording ground) and one in the left occipital bone (recording reference). The rGO/Au μECoG array was glued to the bottom of a custom designed 3D-printed rod that was mounted in a stereotaxic microdrive. The microdrive was angled 30° from vertical, and the array was lowered to press gently into the pial surface of the barrel cortex. Individual contralateral whiskers were deflected using a custom stimulation system. The whiskers were cut to a length of approximately 1 cm and inserted at their natural angle into a glass capillary tube (0.23 mm inner diameter) that was epoxied to a piezoelectric bender (Thorlabs, Inc.). Each deflection consisted of a pulse with 2 ms rise time, 2 ms plateau, and 2 ms fall time that was smoothed to prevent mechanical ringing. These whisker stimuli were delivered at 0.5 Hz while wideband neural signals were recorded at 20 kHz per channel using an Intan RHS system (Intan Technologies). To distinguish evoked neural activity from any noise introduced by the stimulation system, whisker stimulation recordings were interleaved with control recordings, in which the same stimuli were delivered but with the whisker removed from the capillary tube. All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania (IACUC Protocol No. 806241).

Method details

Preparation of GO + VC solutions

Highly concentrated GO (>5 mg mL−1) was purchased from Graphene Supermarket® and was used for all experiments. VC in powder form was purchased from Sigma Aldrich, and VC solutions were made at the appropriate concentrations by dissolving powder in deionized water over 30 min under continuous magnetic stirring at room temperature. All GO + VC mixtures were made by combining GO and VC solutions in a 1:1 volume ratio and vortex mixing for 5 min.

Spray-coating techniques

Prior to spray-coating, the surface of the target substrate was activated with an air plasma (Harrick PDC-32G Basic Plasma Cleaner, p = 305 mTorr, plasma power = 18 W, exposure time = 2 min). The substrates used in this report included precleaned microscope slides (Fisherbrand 12-550-A3, 25x75x1 mm), cut into 25 mm × 25 mm squares, and 3″ diameter Si wafers coated with parylene-C and prepatterned with Au contacts. After plasma activation, samples were placed on a hot plate at 150°C and preheated for 5 min. (Figure 1A). Subsequently, 3 mL of the GO + VC solutions were spray-cast onto the sample using a commercial airbrush (Gocheer) connected to a TC-20 air compressor unit (Figure 1B). The average spraying pressure was ∼0.25 MPa (36 psi), and the average distance between the airbrush and substrate was 12–15 cm. Following spray-coating of the GO + VC solutions, samples were left on the hot plate at 150°C in order to complete the reduction step and produce VC-rGO films (Figure 1C). To optimize the reduction process, VC concentration and film heating time at 150°C were varied between 1 and 50 mM and 2.5–30 min, respectively.

Film characterization

Sheet resistance: R, of VC-rGO films was measured with a Loresta-AX MCP-T370 handheld low resistivity four-point probe (Nittoseiko Analytech Co., Ltd.). Profilometry was performed with a Tencor KLA P-7 Stylus Profiler (KLA Corp.; scan speed = 50 μm s−1; sampling frequency, f = 200 Hz; stylus applied force, f = 2 mg) to determine VC-rGO film thickness, τ. After measuring the values of sheet resistance and average thickness, the DC conductivity of the films, σ, was calculated using the relationship σ = (Rs·τ)−1.

Raman: spectra were collected using an NT-MDT Ntegra Raman-NSOM system, with a 532 nm excitation laser. The effective wavelength range was limited to 180–2,580 cm−1. Raman spectra were averaged across N = 6 separate scans, and then fitted with a Lorentzian function in MATLAB to determine peak positions and intensities for the peaks specific to graphitic carbon (i.e., the D and G bands).

Imaging: a JSM-7500F scanning electron microscope (SEM; JEOL, Ltd.) with a 3 keV accelerating voltage was used for imaging VC-rGO thin films on glass substrates. A Bruker Icon atomic force microscope (AFM; Bruker Corp.) was used to characterize the flake morphology of GO + VC and VC-rGO thin films on glass substrates, and ImageJ was used to help identify individual flakes in the collected micrographs.

Monitoring the DC conductivity over 1 month

VC-rGO films were made with the optimized VC reduction method (50 mM VC, 15 min at 150°C), followed by aging in air at room temperature, or submerged in 1X PBS at room temperature. DC conductivity of the films was measured at the initial (t) and final (30 days) timepoints using a 4-point probe (Figure S3). To quantify the total conductivity retained after aging, %Δσ, we defined the following parameter:where σ0 is the DC conductivity at t, and σ is the DC conductivity at the final timepoint.

Fabrication of μECoG devices

rGO/Au μECoG arrays were fabricated following previously published methods (Driscoll et al., 2020; Murphy et al., 2020). Briefly, a Si wafer was coated with 4 μm of parylene-C and patterned with electrode traces consisting of 10nm/100nm Ti/Au using photolithography, electron-beam deposition, and lift-off. 1% Micro-90 cleaning solution was spin-cast onto the wafer as an anti-adhesion layer, and a 3 μm-thick sacrificial parylene-C layer was deposited. Openings over the Au traces were made in the sacrificial parylene layer via photolithography and reactive ion etching (RIE). Next, 3 mL of the GO + VC mixture was spray-cast onto the wafer. The sacrificial parylene layer was then manually removed, leaving behind GO + VC films only in the areas defined by the previous patterning steps. The GO + VC films were subsequently reduced by heating the entire wafer on a hot plate at 150°C for 15 min. After the reduction step, a top encapsulation layer of 4 μm of parylene-C was deposited, and device outlines were patterned via photolithography and lift-off techniques. A final RIE step removed residual parylene, and the rGO/Au microelectrodes were finally peeled off from the Si wafer for characterization and use. Plain Au electrodes were made using a similar fabrication flow, only excluding the sacrificial parylene patterning, GO + VC spray-coating, and reduction steps.

Characterization of μECoG devices in vitro

Completed μECoG arrays were imaged using a Keyence VHX-6000 Digital Light Microscope (Keyence Corp.). The devices were then loaded into zero-insertion force (ZIF) connectors soldered to custom Omnetics printed circuit boards (PCBs; DigiKey Electronics). A Gamry Reference 600 potentiostat/galvanostat/ZRA (Gamry Instruments, Inc.) was used for all electrochemical characterization in vitro. In particular, we conducted electrochemical impedance spectroscopy (EIS), cyclic voltammetry (CV), chronopotentiometry and continuous charge-injection tests to measure the electrochemical impedance, cathodal charge storage capacity (CSCC), safe charge injection limit (Q), total charge injection capacity (CIC), and long-term pulsing stability of the μECoG devices. All in vitro testing was conducted in 1X phosphate buffered saline (PBS; pH 7.4, Quality Biological) at room temperature, using a standard three-electrode cell featuring a graphite rod counter electrode (1-cm diameter; Bio-Rad Laboratories, Inc.), and an aqueous Ag/AgCl reference (3M KCl; Sigma Aldrich). Specifically, EIS was measured with a 10 mVrms input voltage, sweeping the frequency from 1–105 Hz. Impedance spectra were subsequently fitted with equivalent circuit models in Gamry’s EChem Analyst software (see Fitting procedure details below), to obtain values of charge-transfer resistance (R) and double-layer capacitance (C). CVs were completed first in Au’s safe operating window (Ganji et al., 2017a; Ganji et al., 2017b), then in the rGO operating window, at a scan rate of 200 mV s−1. The first scan was always completed in the cathodic direction (i.e., towards the negative potential limit first). Values of Q were determined from both CVs and impedance fitting results, using the equation (Cogan, 2008; Apollo et al., 2015),where V is the size of the electrode’s full operating voltage window (also known as the electrochemical window or the water window), and GSA is the electrode geometric surface area. During chronopotentiometry experiments, voltage transients were recorded in response to biphasic current pulses of 7.5 μA amplitude and 650 μs duration, with the cathodic pulse delivered first and a 20-μs interpulse interval (Ganji et al., 2017b). In the continuous charge-injection tests, charge was injected through the electrodes at 500 μC cm−2 per phase (current amplitude, 1.56 μA phase−1) at 130 Hz over 24 h (Dalrymple et al., 2019), and the electrode impedance was measured every 260,000 pulses. The starting open circuit potential (OCP) was 0 V for all measurements described above.

Fitting procedures for μECoG devices in 1X PBS

A modified Randles cell was used to fit the impedance spectra of the bare Au electrodes (Franks et al., 2005; Boehler et al., 2020). The coating model used to describe the rGO/Au electrodes describes coatings on metallic or Si electrodes (Pan et al., 1996). For both models, we used a constant phase element (CPE) in place of the traditional double-layer capacitance, C, in order to account for the non-ideal behavior of the interfacial double-layer (Kochowski and Nitsch, 2002).

C values reported in the main manuscript and in Table S2 were calculated from CPE parameters using the equation (Hirschorn et al., 2010; Bera et al., 2017).where R is the solution/spreading resistance with units of Ω, Y is the CPE admittance with units of S , and α is the unitless CPE constant (α = 1 represents a perfect capacitor, α = 0 is a perfect resistor).

In addition to C, it is also possible to calculate the theoretical value of the solution resistance, R, using the equation (Franks et al., 2005)where ρ is the solution resistivity (72 Ω cm for 1X PBS), and l is the electrode side-length (here, 50 μm). For the Au and rGO/Au electrodes in this work, R = 6.354 kΩ, which is in excellent agreement with experimentally fitted values of R, supporting the validity of our proposed models (Table S2).

Evaluating the total charge injection capacity

The total charge injection capacity (CIC) of a stimulating technology defines the maximum charge that an electrode can inject into a solution (or into tissue) such that all reactions that occur are entirely reversible. In terms of electrochemical parameters, the CIC can be defined as the total charge (current × time [pulse width]) at which the maximum excursion/polarization potential, E, exceeds the water window limits of the electrode material, either in the anodic or cathodic phase (Ganji et al., 2017b; Cisnal et al., 2018; Driscoll et al., 2021). To determine the CIC, then, the polarization potential, E, must first be defined. E is best summarized by the equation (Cisnal et al., 2018).where E is the initial offset voltage (typically 0 V), and ΔE is the polarization potential of the electrode across the electrode-electrolyte interface. ΔE can be determined from voltage transient experiments, as it is equivalent to the voltage transient response, ΔV, after subtracting the access voltage, V, which is the instantaneous potential change at the beginning of, or immediately after, current pulse delivery (Figure S6).

Dopamine and ferrocyanide detection in vitro

Dopamine and potassium ferrocyanide were purchased in powder form from Sigma Aldrich and test solutions were made by dissolving the powders in 1X PBS by vigorous hand shaking, followed by vortex mixing for 5 min. For DA sensing, CVs were completed from −0.4 V to +1.0 V at a scan rate of 4 V s−1 in a three electrode configuration, with a carbon rod counter electrode and aqueous Ag/AgCl reference. For detecting Fe(CN)6, the same three-electrode configuration was used, but CVs were completed from 0.0 V to +0.6 V at a scan rate of 50 mV s−1. In both instances, the first scan was always completed in the cathodic direction, the initial OCP was 0 V, and no nitrogen or argon bubbling was used before, during, or after the CVs were measured. Different rGO/Au μECoG arrays were used for DA detection and Fe(CN)6 detection, and these arrays were also separate from those used for EIS measurements, so as to avoid any potential complications that might have been caused by DA adsorption.

Peak oxidation currents were determined using standard techniques after background subtraction (Elgrishi et al., 2018; Espinoza et al., 2019), and a calibration curve was constructed for DA in the 1–50 μM concentration range (Venton and Cao, 2020). The sensitivity of the rGO/Au electrodes for DA detection was calculated from the slope, m, of the fitting curve (in units of nA μM−1), and the limit of detection (LOD) was calculated using the equation (Palomäki et al., 2018; Manbohi and Ahmadi, 2019),where s is the standard deviation of the background current in amperes.

To estimate the effective electrochemical surface area of the rGO/Au μECoG electrodes, A, we applied the simplified Randles-Sevcik equation at room temperature to the Fe(CN)6 data (Kang et al., 2012; Wang et al., 2014),where i is the background-subtracted peak current in amperes, A is the electrode area in cm2, C is the concentration of the solution in mol cm−3, n is the number of electrons transferred in the Fe(CN)6 redox event, D is the diffusion coefficient of Fe(CN)6 in cm2 s−1, and is the scan rate in V s−1.

Quantification and statistical analysis

The average values cited in this work were defined as the sum of the relevant measurement terms, Σ(terms), divided by the total number of relevant measurement terms, Nterms:

The standard deviations in this work were calculated as the square root of the variance, where the variance was defined as the sum of the squared differences between each relevant measurement term and the average value of that same measurement term, Σ(terms-μ)2, then divided by the total number of relevant measurement terms, Nterms:

The average values and standard deviations for each relevant measurement term are cited in the main manuscript, and are also provided in the Figures and Tables of the main manuscript and Supplemental information.

The only statistical tests applied in this study were Student’s t-tests used to compare values of R and C between Au and rGO/Au μECoG fitting results. These analyses were run across n = 4 separate electrodes, with significance defined as a p-value p < 0.05. Statistical results are cited in the main manuscript, and can also be found in Figure S4.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Highly Concentrated Graphene Oxide (HCGO, GO)	Graphene Supermarket	SKU: SKU-HCGO-W-60ML
L-ascorbic acid (Vitamin C, VC)	Millipore Sigma; Sigma Aldrich	CAS: 50-81-7 Product No.: A4403
1X Phosphate Buffered Saline Solution, pH 7.4 (1X PBS, PBS)	Quality Biological Inc.	Item No.: 114-058-101 UNSPSC: 12161700
Dopamine hydrochloride (Dopamine, DA)	Millipore Sigma; Sigma Aldrich	CAS: 62-31-7 Product No.: H8502
Potassium ferricyanide(iii) (Ferricyanide, Fe(CN6, ferrocyanide, Red prussiate)	Millipore Sigma; Sigma Aldrich	CAS: 13746-66-2 Product No.: 702587
Experimental models: organisms/strains
Rat: Adult Male Sprague Dawley, Crl:CD(SD), 225g	Charles River Laboratories	RGD ID: 734476 Ontology ID: RS:0000064 University of Pennsylvania IACUC Protocol No. 806241
Software and algorithms
MATLAB	MATLAB	Custom scripts were created to handle plotting and presentation of the data collected.