in vitro Reactive-Accelerated-Aging (RAA) Assessment of Tissue-Engineered Electronic Nerve Interfaces (TEENI)

Abstract

A reactive-accelerated-aging (RAA) soak-test has been employed to challenge microfabricated neural interface devices against an aggressive environment that mimics worst-case chronic physiological inflammation. The RAA tests were able to determine the ability of different materials to increase the adhesive strength of the polyimide and platinum-gold-platinum metallization thin-film interface. It was found that a 3-day RAA soak-test at 87 °C in phosphate buffered saline with 10 to 20 mM hydrogen peroxide resulted in adhesive failure of the metal-polyimide interface when titanium was used as the primary adhesion promotor. The addition of hydrogenated amorphous silicon carbide was able to eliminate the onset of adhesive failure of the metal-polyimide interface during 7-day RAA soak tests. However, sporadic cracking of the silicon carbide layer resulted in a minority of broken metal interconnects that resulted in failed electrodes. These tests have demonstrated the ability of RAA soak tests to provide rapid in vitro assessment of microfabricated neural interfaces and thereby reduce the time needed to develop synthetic methods to fabricate chronically reliable devices.

I. Introduction

Over 20 million Americans suffer from some form of peripheral-nerve injury that can result in loss of motor or sensory function [1]. Implantable neural electronic interfaces aim to stimulate and/or record neural activity to advance fundamental understanding and develop therapies for neurological disease or injury. The development of these devices is currently driven by the need for high-channel count, high-channel density, and robust devices that can operate throughout the lifetime of the patient to enable various functions, for example, motor and sensory control of artificial limbs for the amputee population. Existing electrode arrays typically employ conductive materials (i.e., metals and polymers) that are encapsulated by electrical insulators (i.e., polymers and ceramics) with exposed electrode sites to interface with neural tissue. However, these neural interfaces can suffer from a host of abiotic failure mechanisms (e.g., material degradation, corrosion, and delamination) that can result in failure of the implant [2]. Better designs and material selections can impede common failure mechanisms and increase device durability during chronic implantation.

To rapidly develop neural electronic interfaces with improved reliability over chronic time periods, researchers use in vitro high-temperature accelerated soak tests in simulated body fluids (e.g., phosphate buffered saline: PBS) to simulate chemical degradation and assess long-term robustness of the interface prior to in vivo validation. These experiments allow for iterative testing of material combinations, fabrication methods, and device designs. Due to the exponential Arrhenius relationship, increasing the temperature from typical physiological (~37 °C) to a much higher temperature (e.g., 87 °C) results in an acceleration factor of ~32x [3]. However, this acceleration factor is only valid if the failure mechanisms are faithfully captured by the accelerated soak test being implemented [3]. For implantable neural interfaces, a key driver of polymer-metal-interface failure is the fact that activated immune cells create an aggressive chemical environment around implants. By releasing digestive enzymes and reactive oxygen species (ROS), cells attempt to breakdown foreign objects [4–5]. However, when PBS is heated to 87 °C, the concentration of oxygen drops precipitously resulting in a nearly oxygen-free solution that will not come close to mimicking the aggressive chemical environment surrounding implants [6–7]. Fortunately, a compelling alternative has recently been developed by Takmakov who led a team that created a reactive-accelerated-aging (RAA) soak test that uses hydrogen peroxide as a source of ROS to mimic the aggressive chemical environment created by immune cells in response to chronic implants. With the RAA soak test, they were able to achieve implant interface degradation equivalent to 6 months in vivo in only 7 days at 87 °C [8].

Herein, we describe the fabrication of polyimide (PI)-based microelectrode devices used for our hybrid tissue-engineered electronic neural interfaces (TEENI), which consist of multi-electrode polyimide-based “threads” embedded into a biodegradable hydrogel composite scaffold that is sutured to the ends of a transected nerve and facilitates axonal regeneration [9–12]. RAA soak testing was used to assess the ability of different TEENI microelectrode device designs to impede abiotic failure by mimicking a worst-case physiological environment. The effect of different adhesion layers including titanium and amorphous hydrogenated silicon carbide (a-SiC:H) on the device’s electrical and physical performance were evaluated. The biodegradable hydrogel composite scaffold was not employed for these studies.

II. Materials and Methods

A. Electrode Device Microfabrication

TEENI neural electrodes were microfabricated in a class 100/1000 cleanroom using photolithography, thin-film deposition, and dry etching processes adapted from [9]. Briefly, 5 μm of BPDA-PDA polyimide (U-Varnish S, UBE Ind.) was spin coated on a 100-mm-diameter HMDS coated Si wafer and cured at 350 °C under N₂. The PI surface was activated in a reactive-ion-etching (RIE) O₂ plasma and then coated by a 250-nm-thick film of stoichiometric a-SiC:H formed by plasma enhanced chemical vapor deposition (PECVD) at 300 °C. Electrode sites, interconnect traces, test structures, and connector pads were formed by a sputtered 400-nm-thick Ti+Pt/Au/Pt+Ti metal stack that was patterned by a lift-off process using image-reversal photoresist (nLOF 2035, MicroChemicals GmbH). The metal structures were sealed by a 2^nd 250-nm-thick film of a-SiC:H that was then coated with an aminopropyl triethoxysilane silane agent (APTES) (VM-652, HD Microsystems) followed by 5 μm of polyimide, deposited and cured as previously discussed. Electrode sites, connector pads, and thread-set geometry were patterned by RIE dry etching using O₂ (to remove PI) and SF₆ (to remove Ti and a-SiC:H) plasmas.

B. Electrochemistry and in vitro RAA

Devices were characterized by electrical impedance spectroscopy (EIS) using a PGSTAT302N (Metrohm Autolab) and a probe station. EIS spectra were obtained using a 3-electrode setup with a TEENI working electrode, Pt wire counter electrode, and a 3 M KCl Ag/AgCl reference electrode in a 10 mM PBS electrolyte from 100 kHz to 10 mHz. No dc offset with respect to the open circuit potential was applied. RAA soak tests were performed in 10 to 20 mM H₂O₂ PBS at 87 °C to assess in vitro electrode durability according to [8]. H₂O₂ concentration ([H₂O₂]) was maintained by replenishing 8.75 mM of H₂O₂ per hour via a syringe pump. [H₂O₂] was verified by measuring the absorbance at 407 nm using a titanium (IV) oxysulfate (Milipore Sigma) colorimetric assay with a dilution factor of 15 in conjunction with Beer’s law. Three microelectrode arrays (per device design) with a total of 44 electrodes ranging in size from 200 to 16,000 μm² were tested.

III. Results and Discussion

A. Microelectrode Design and Fabrication

The TEENI microelectrode device is composed of two layers of PI (10 μm total) used to isolate buried metal interconnects that electrically connect exposed electrode sites and test structures to 200-μm-diameter disk contact pads that can interface with a percutaneous connector. The implant region that transects the nerve contains 3 or 4 80-μm-wide PI “threads” spaced 160 μm apart interspersed with exposed electrodes and test-structures (Fig. 1). The primary abiotic failure mechanisms that the device can suffer from include: (1) PI-PI delamination/ degradation leading to shorting between adjacent metal structures, (2) PI-metal delamination resulting in changes to electrode area or failure, and (3) metallic fracture resulting in open circuits. The adhesive strength between polymers and metals are notoriously poor due to the lack of significant chemical bond formation. However, research by Cogan [13] and Stieglitz [14] has shown that a-SiC:H and amorphous carbon can improve the polymer-metal interfacial strength. Three different TEENI device designs were fabricated to probe these failure mechanisms (Fig. 1). Design-A did not use a-SiC:H and resulted in only PI-PI or PI-metal interfaces. Design-B used a conformal a-SiC:H adhesion layer producing both PI-SiC-PI and PI-SiC-metal interfaces. Design-C used SF₆ patterned a-SiC:H that resulted in PI-PI, PI-SiC-PI, and PI-SiC-metal interfaces. All designs maintained a 4 μm overlap of PI around exposed metal sites, and design-C used a 6 μm overlap of a-SiC:H around all metal.

Figure 1:

Implant region of a three thread-set TEENI microelectrode device with 14 functional electrodes and 2 test structures. (A) Design-A, (B) Design-B, and (C) Design-C fabrication cross-sections and images of 1,600 μm² electrodes showing the different adhesion layers and device designs used to probe the various abiotic failures of material interfaces.

During fabrication, qualitative assessments could be made about the differing adhesive strengths between material interfaces and fabrication methodologies. Metal traces on design-A would sporadically delaminate during lift-off due to residual film stress and poor interfacial adhesion between PI and Ti-Pt/Au/Pt-Ti. Delamination was also evidenced during initial testing for designs-B and C, when a-SiC:H was used with Pt/Au/Pt stacks without Ti. However, delamination during lift-off was completely eliminated with the addition of a Ti layer due to its strong adhesion to the native SiO₂ layer on a-SiC:H. Furthermore, PI-SiC delamination occurred at the top PI layer if the APTES agent was not used. The authors attempted to use PECVD deposited amorphous carbon as outlined by Ordonez et al. [14] instead of APTES to improve this interface without success.

B. EIS and RAA

EIS characterization of as-fabricated devices revealed expected dependencies between electrical impedance and exposed electrode area and typical phase-angle behavior. Devices were subjected to 3-day and 7-day RAA experiments at 87 °C, which corresponds to ~3 and ~7 months in vivo respectively, and observed changes to EIS behavior were correlated to abiotic failure mechanisms. After 3-day RAA, design-A devices with no a-SiC:H showed widespread PBS penetration between the PI-metal-PI interface (Fig. 2 A1) and partial delamination of the exposed electrode metal as evidenced by warping/bubbling of the metal layer. The EIS spectra for 44 electrodes decreased by −32 ± 9% (arithmetic average ± stdv) at 1 KHz caused by partial delamination of the PI-metal interface resulting in an increase in the active electrode area that can interface with the electrolyte (Fig. 3). Furthermore, 3 of 44 electrodes had significant cracks along the exposed metal electrode sites resulting in open circuits. After 7-day RAA, design-A devices showed extensive delamination and/or complete loss of exposed metal from the underlying PI (Fig. 2 A2) rendering further EIS characterization using a probe station difficult. Ti-PI interfacial strength is primarily due to mechanical interlocking instead of strong covalent bonding, and RAA experimentation was effective at highlighting the weakness of Ti as an effective adhesion layer to PI for neural interfacial implants. No PI-PI delamination was observed at either 3-day or 7-day RAA.

Figure 2:

Optical images of (top) 3-day RAA and (bottom) 7-day RAA showing worst-case failures experienced. Design-A (A1) PBS penetration between metal-PI interface and (A2) metal delamination from the underlying PI. Design-B (B1) no optically detectable damage and (B2) a-SiC:H cracking leading to PBS penetration. Design-C (C1) no optically detectable damage and (C2) a-SiC:H cracking and PBS penetration between PI-SiC interface.

Figure 3.

EIS Bode impedance spectra after 3-day RAA. (A) Design-A demonstrated decreased impedance and significant overlap between differently sized electrodes indicating PBS penetration and damage to the device. (B) Design-B impedance was consistent after 3-day RAA suggesting little to no damage to the device. Insets show impedance at 1 KHz as a function of electrode area. Design-A electrodes showed a significant shift in impedance from the pre-soak values while both Design B and C showed minimal changes in impedance.

Both design-B and C that used a-SiC:H adhesion layers between PI and metal exhibited an increased ability to resist abiotic failure. After 3-day RAA, there was no optically detected damage to any electrode sites, metal traces, or material interfaces (Fig. 2 B1 & C1), and EIS data showed minimal change after soaking with −0.9 ± 1.9% and −7.6 ± 9.5% average change at 1 KHz for design-B and C, respectively (Fig. 3). After 7-day RAA, there was still no optically detected metal-SiC delamination confirming a-SiC:H’s ability to effectively chemically bond with Ti through Ti-C and/or Ti-Si. However, the PECVD deposited a-SiC:H formed a highly compressive (~ −350 MPa) film that led to sporadic fracturing of the a-SiC:H layer at high radius of curvature areas (e.g. around electrode sites and along curved metal traces) during RAA (Fig. 2 B2 and C2). PBS was able to penetrate along these cracks and partially delaminate the weaker SiC-PI interface. A PECVD deposited amorphous carbon layer has been reported to increase the interfacial strength between a-SiC:H and PI [14]. Unfortunately, the authors were unable to successfully fabricate robust devices using this PECVD carbon layer; however, additional testing is required to verify these results.

EIS analysis of 7-day RAA soaked devices confirmed that a-SiC:H cracking led to breaks in some metal traces resulting in open-circuits on 5/44 and 4/44 electrodes for design-B and C, respectively. Another 2/44 and 6/44 electrodes experienced a significant decrease in impedance with an unusual bimodal -phase plot indicating capacitive behavior at high frequencies and resistive at low with a local maximum between 10–100 Hz. There was sporadic optical discoloration at the PI-SiC-metal interface around some electrode sites. However, this discoloration was not completely correlated with the 8 electrodes that experienced the unusual EIS behavior therefore the exact source of the discoloration and EIS performance remains elusive. Even after 7-day RAA, no metal delamination was evidenced suggesting that the SiC-metal interface was not compromised, and the EIS spectra of the majority remaining 37/44 and 34/44 “good” electrodes changed by only −7.2 ± 7.4% and −11.7 ± 6.8% at 1 KHz from the pre-soak values for designs-B and C, respectively.

IV. Conclusions and Future Work

We have described the fabrication of polyimide-based microelectrode devices that are used to produce tissue-engineered electronic neural interfaces (TEENI). Devices with different adhesion layers/ designs were challenged against an aggressive reactive accelerated aging in vitro test. It was shown that a-SiC:H in conjunction with titanium can provide a dramatic increase in device durability against oxidative degradation; however, cracking of the a-SiC:H resulted in sporadic failure of a minority of active electrode sites. Future work will focus on addressing the cracking issue and performing longitudinal EIS analysis during RAA to determine the kinetics of different failure mechanisms and to provide for more rapid device design iteration in preparation for in vivo implants.