Foldable three dimensional neural electrode arrays for simultaneous brain interfacing of cortical surface and intracortical multilayers

The study addresses the need for neural interfaces that can simultaneously and continuously record electrophysiological activity from both the brain surface (cortex) and intracortical regions to elucidate complex 3D neural network dynamics. Existing methods—ECoG arrays (surface), Utah arrays (multi-shank intracortical), and Michigan probes (multichannel depth recording)—provide complementary but spatially and structurally isolated data, often limited to planar, 2D alignments that cannot capture volumetric signal propagation. Rigid silicon-based devices also induce inflammatory responses due to mechanical mismatch with soft brain tissue, compromising chronic recording. The authors hypothesize that a flexible, integrated, foldable 3D electrode array combining surface and penetrating shanks can overcome these limitations to enable simultaneous, high-fidelity mapping of 3D neural transmission from intracortical layers to the cortical surface with reduced tissue response.

The introduction reviews three principal electrophysiological modalities: (1) ECoG for high-resolution, large-area surface recordings; (2) Utah arrays for broadened intracortical measurements; and (3) Michigan probes for layer-specific depth recordings. While recent efforts have targeted 3D recordings in deep brain regions (e.g., multi-shank interfaces, injectable mesh electrodes), these approaches lack concurrent surface activity measurement, limiting insights into intracortical-to-surface propagation. Attempts to combine separate ECoG arrays with penetrating probes suffer from temporal misalignment and increased invasiveness, while rigid devices exacerbate tissue responses due to mechanical mismatch. Flexible electronics have shown improved biocompatibility and conformal contact, reducing immune responses and tethering forces, motivating integrated, flexible 3D architectures to record correlated surface and intracortical activity simultaneously.

Device design and fabrication: A foldable, flexible 3D electrode array integrates 9 ECoG surface electrodes (500 µm spacing) and four penetrating shanks (P1–P4), each with 6 electrodes (20 × 20 µm2 sites; 200 µm spacing; three per side, 400 µm inter-side spacing). The probe thickness is <20 µm (bilayer folded), shank length ~1.5 mm (1.2 mm recording depth). Fabrication begins with spin-coated polyimide (PI) layers on glass, Au/Cr interconnect patterning via photolithography, device outline/hinge definition by RIE, and SU-8 encapsulation (exposing recording sites). A pop-up assembly is achieved by transferring the patterned 2D device to PVA tape using a PDMS stamp, selectively sputtering 50 nm SiO2 on target bonding regions, and bonding onto a pre-stretched elastomer via UV/O plasma activation to form Si–O–Si bonds. Releasing the elastomer induces a controlled 3D pop-up geometry with tailored hinges (etched PI thickness ratio t2/t1 ~0.42) to minimize tip diameter and emulate conventional shank geometry. Mechanical modeling: 3D finite element analysis (FEA) of PI/SU-8 bilayers evaluated bending deformation and effective bending stiffness as a function of hinge PI thickness, incorporating elastic-plastic PI and linear elastic SU-8 properties. Simulations included crack notches common to fabrication and predicted tip curvature and stress distributions, corroborated by experimental SEM and optical imaging. Temporary stiffening for insertion: Penetrating shanks were coated with biocompatible/bioresorbable PEG 4000 (~40 µm thickness) to provide sufficient stiffness for insertion; dissolution in biofluids restores flexibility and electrical performance. Electrochemical characterization: Electrochemical impedance spectroscopy (EIS) was performed in ACSF and PBS at 36.5 °C using a three-electrode setup. Pt black electroplating (chronoamperometry, 0.2 V for 30 s) on Au sites reduced impedance (~1.3 MΩ to ~20 kΩ at 1 kHz). Randles equivalent circuit fitting extracted parameters (e.g., Cdl ~10 µF/cm2 for Au, ~34 µF/cm2 for Pt black). PEG dissolution over time restored pre-coating impedance values. Mechanical reliability: Bending (wrapped on 5 mm radius rod) and stretching (0–40% strain) tests for 1000 cycles monitored EIS stability. Biocompatibility and cytotoxicity: PC12 neuron-like cells and NIH 3T3 fibroblasts were used for live/dead assays and extract cytotoxicity (ISO 10993-5). Primary cortical neurons (E18 rat) were cultured on glass, PI with Pt black, and SU-8 with Pt black after PDL coating; viability quantified via calcein-AM and PI staining. In vivo histology (GFAP, Iba1) at 2 days post-implant assessed glial response. In vivo recording: Adult C57BL/6 mice (8 weeks) underwent craniotomy over S1 barrel cortex (AP −1.5 mm, ML −3 mm). PEG-coated shanks were inserted at 0.2 mm/s with robotic stereotaxic control. After cement fixation and 2-day recovery, simultaneous intracortical and surface recordings were acquired using an Intan RHD 2132 system. Unidirectional total whisker stimulation (right whiskers, 1 s every 5 s) evoked responses. Signal processing: Raw data sampled at 20 kS/s per channel with 0.1–6 kHz bandpass and 60 Hz notch filters. LFPs were low-pass filtered at 300 Hz. Custom MATLAB spike sorting used polarity and amplitude thresholds; ISIs <2 ms were excluded. Firing rates computed over stimulation epochs. 3D heatmap generation: Device STL geometry imported to MATLAB; LFP amplitudes, firing rates, and neuron counts mapped onto electrode coordinates to visualize spatiotemporal propagation across front, back, and top views for defined frames (100 ms windows).

- Integrated 3D array: 9 ECoG surface electrodes plus 4 penetrating shanks × 6 electrodes = 33 channels enabling simultaneous surface and intracortical recordings with a foldable pop-up geometry; probe thickness <20 µm; shank length ~1.5 mm; recording depth ~1.2 mm. - Mechanical optimization: Hinge PI thickness reduction (t2/t1 ~0.42) minimized tip head diameter and increased pop-up height; FEA matched experimental deformation; localized cracking at the extreme tip did not affect recording sites. - Temporary stiffening and insertion: PEG 4000 coating (~40 µm) provided sufficient rigidity for vertical insertion; full insertion into 0.7% agarose without failure; PEG dissolution restored flexibility and electrical properties. - Electrochemical performance: Pt black reduced |Z| from ~1.3 MΩ to ~20 kΩ at 1 kHz; Randles circuit fits with Cdl ≈ 10 µF/cm2 (Au) and ≈ 34 µF/cm2 (Pt black). During PEG dissolution in ACSF, impedance returned to ~20 kΩ at 1 kHz. - Mechanical/electrical stability: Stable EIS over 1000 bending/stretching cycles (details in Supplementary). - Biocompatibility: High PC12 and primary cortical neuron viability on Pt black, PI, and SU-8 surfaces vs. glass; NIH 3T3 extract tests showed negligible cytotoxicity even at 100% extracts; in vivo histology at 2 days showed reduced glial response compared with rigid implants. - In vivo recording: Simultaneous LFPs and single-unit spikes recorded from intracortical and surface electrodes during unidirectional total whisker stimulation under anesthesia. LFP changes synchronized with spike activity; increased firing rate and spike counts during stimulation vs. spontaneous activity. - Spatiotemporal mapping: 3D heatmaps showed propagation patterns from intracortical sites (notably P3) to adjacent shanks and surface regions. Two stimulation windows identified: Stim1 (peak intervals ~10.8 ms; amplitudes 250–800 µV) and Stim2 (intervals ~6.5 ms; amplitudes 500–800 µV). Surface electrode E5 exhibited largest LFP amplitude and spike counts. Heatmap color ranges: LFP −50 to 800 µV; firing rate 0–70 Hz; neuron count 0–4.

By integrating flexible surface electrodes with multiple penetrating shanks in a single foldable 3D platform, the device overcomes the planar and modality-segregated limitations of conventional ECoG and penetrating probes. The mechanically compliant architecture, temporary PEG stiffening for insertion, and optimized hinge design enable reliable intracortical access with reduced tissue damage and immune response. Simultaneous recordings demonstrate correlated dynamics between intracortical single-unit activity and cortical surface LFPs, directly addressing the research goal of mapping signal propagation across brain layers in three dimensions. The spatiotemporal 3D heatmaps reveal distinct propagation patterns associated with different whisker stimulations, validating the platform’s capability to resolve neural transmission pathways. These findings suggest utility for studying circuit mechanisms and potentially aiding diagnosis or therapy monitoring in neurological disorders where multi-layer interactions are critical.

The work introduces a foldable, flexible 3D neural electrode array that unifies ECoG and intracortical recordings in a single, integrated system. The platform features mechanically optimized pop-up shanks, PEG-enabled insertion, low-impedance Pt black electrodes, and demonstrated biocompatibility. In vivo, it simultaneously captures LFPs and single-unit activity across cortical surface and intracortical layers during whisker stimulation, enabling 3D spatiotemporal mapping of neural signal propagation. Future directions include integrating active multiplexing (e.g., flexible transistor arrays) to increase channel counts and reduce wiring complexity, and scaling the number of shanks to broaden coverage while maintaining minimal invasiveness and insertion stability.

- Limited channel count due to passive electrode array architecture and spatial constraints limits recording density and coverage. - Insertion stability and reliability, while improved with PEG stiffening, require further substantiation for chronic applications. - Acute recordings with short-term histology were demonstrated; long-term chronic stability, gliosis progression, and signal longevity were not fully evaluated within this study. - Heatmap visualization ranges and neuron count estimates depend on spike sorting thresholds and may be sensitive to algorithmic parameters.