A Soft, High-Density Neuroelectronic Array

Abstract

Techniques to study brain activities have evolved dramatically, yet tremendous challenges remain in acquiring high-throughput electrophysiological recordings minimally invasively. Here, we develop an integrated neuroelectronic array that is filamentary, high-density and flexible. Specifically, with a design of single-transistor multiplexing and current sensing, the total 256 neuroelectrodes achieve only a 2.3 × 0.3 mm2 area, unprecedentedly on a flexible substrate. A novel single-transistor multiplexing acquisition circuit further reduces noise from the electrodes, decreased the footprint of each pixel, and potentially increased the device lifetime. The filamentary neuroelectronic array also integrates with a rollable contact pad design, allowing the device to be injected through a syringe, enabling potential minimally invasive array delivery. Successful acute auditory experiments in rats validate the ability of the array to record neural signals with high tone decoding accuracy. Together, these results establish soft, high-density neuroelectronic arrays as promising devices for neuroscience research and clinical applications.

Fig. 1. Overview of the soft, high-density neuroelectronic array.

a Optical image of a soft, high-density neuroelectronic array with a human hair floating on the water surface. Inset: high-magnification optical images of electrode area. b SEM image of a device showing checkboard design and rectangular electrode sites. c Schematic of a device structure with an explosive view of a device, showing four different PI layers, three metal layers (Au), and a doped Si NM layer. d Circuit diagram of 2 × 3 array showing the sensing scheme of the device.

Fig. 2. Fabrication and electrical performance of soft, high-density neuroelectronic arrays.

a Cross-section schematics (top) and optical images (bottom) of key fabrication steps, including doping, transfer printing, isolation of n-doped Si NMs, gate dielectric deposition and metal 1 patterning for input selects, metal 2 & 3 patterning for output signals, and lift-off as the final step. b Representative I-V characteristics of a Si NM transistor used in the device. Inset: Optical image of a representative test transistor. c Mobility histogram of 256 transistors with an average value of 583.7 ± 48.3 cm²/Vs. d Comparison of this work, other flexible bioelectronics, and Neuropixels in electrode size and density.

Fig. 3. Bench testing of soft, high-density neuroelectronic arrays.

a A circuit diagram for a pixel in the active array, having a Si transistor as a multiplexer. b Corresponding optical image of the circuit diagram in A from a microscope. c V_t histogram from 256 transistors with an average value of 1.56 ± 0.19 V. d Bench recording of a 10 Hz, 4 mV_rms sine-wave input voltage (black) and corresponding output current (red). e Power spectra density (PSD) of the recorded sine-wave output in (d). f SNR histogram of a 256-ch array with an average value of 30.1 ± 10.4 dB. g SNR heatmap of a 256-ch array.

Fig. 4. In vivo results of a soft, high-density neuroelectronic array.

a Schematic of the device implanted on the surface of rodent primary auditory cortex (left), averaged responses from three independent channels, and annotations of the electrode positions in the array (right). b Single-trial neural responses to three successive tone stimuli from the tone task. Data were filtered from 1–100 Hz with a 6th order Butterworth filter. c The soft neuroelectronic array achieved a decoding accuracy of 51% compared to 35% accuracy from the passive array in a tone discrimination task.