Highly conformable chip-in-foil implants for neural applications

The study addresses the need for neural implants that combine high functionality and spatial resolution with long-term biocompatibility. While silicon (Si)-based integrated circuits enable high integration density, their mechanical stiffness can provoke foreign body responses and tissue encapsulation, degrading chronic performance. Conformable implants that match tissue mechanics improve electrode–tissue contact and reduce invasiveness. Polyimide (PI) substrates are highlighted for chemical stability, biocompatibility, low water uptake, and excellent dielectric and mechanical properties, enabling sub-micrometer thin, flexible, and durable neural interfaces compatible with microfabrication. The authors propose hybrid chip-in-foil bioelectronic devices that embed small Si dice into flexible PI foils to extend coverage over larger brain areas while preserving mechanical compliance. Given anatomical variability (e.g., curvature differences across species and brain regions) and application-specific constraints, modular arrangements of circuit components are sought. The research questions focus on: (1) establishing design rules that ensure mechanical compliance and integrity at the die–substrate interface during conformation to curved brain surfaces and (2) developing a scalable transfer and microfabrication process that enables modular chip placement and reliable interconnection for chronic implant use.

The paper situates its contribution within ongoing efforts in neurotechnology including brain pacemakers, retinal implants, and brain–computer interfaces, emphasizing cross-disciplinary advances in neuroscience, materials, and engineering. Prior work shows that stiff Si induces foreign body responses, whereas mechanically compliant, conformable implants reduce this effect and enhance efficacy through intimate tissue contact. Polyimide has been widely used for flexible neural implants due to its chemical stability, biocompatibility, low water uptake, and robust electrical insulation; long-term stability of PI-based systems has been demonstrated. Intrinsically flexible organic electronics have improved but currently lack the integration density and complexity of mature Si technologies. Geometric and mechanical design factors (implant thickness, substrate properties, curvature of target tissue, surface tension of interstitial fluids) govern conformability. Fenestrations in flexible substrates can further reduce bending stiffness and improve conformation and tissue response compared to solid substrates. These prior insights motivate hybrid designs that combine Si IC functionality with PI substrate compliance.

Design assessment by finite element modeling (FEM): The team developed an FEM to simulate mechanical behavior of chip-in-foil implants conforming to a curved brain surface. The target anatomy was the rat somatosensory cortex, approximated as a cylinder with a 3 mm radius of curvature. Simulations considered a PI substrate thickness of 10 µm and a Si die thickness of 100 µm with a quadratic base shape and length l_de = 400 µm. The die was rotated by α_z = 10° relative to the curvature axes to account for process and implantation tolerances. Edge fillet radii at die corners were varied between 1 and 50 µm. The model quantified delamination height at the die–PI interface during conformation and mapped mechanical stress distribution in the PI around die edges and corners. The largest radial distance from the die edge to any point showing delamination was subtracted from die edge length to define an effective die length/area usable for pad placement. Simplifications included modeling the PI as a solid foil (neglecting thin-film metal interconnects with much higher Young’s modulus and substrate fenestrations) and neglecting metal areas on the die except at pads due to their small relative area. Notes on stiffness: the enhanced thickness ratio of metal to PI was ~0.03, with bending stiffness dominated by the PI (thickness dependence to the third power). Anticipated fenestration opening-to-substrate area ratios between 0.1 and 0.2 were noted to further reduce effective bending stiffness.

Transfer and microfabrication process development: A batch transfer process was designed to place, align, interconnect, and thin multiple Si dice within PI substrates. The general flow comprised: (1) Dice placement into a transfer carrier substrate containing etched cavities sized for the dice; cavity sidewalls were coated with a copper sacrificial layer. (2) Fabrication of the flexible PI-based substrate and metallization: photoresist with a sharp undercut was patterned on die surfaces; sputtered metal interconnects and contact pads were defined by lift-off, enabling continuous interconnects bridging the gap between die and carrier. Vias were formed by dry etching the PI to contact die pads, creating PI/metal/PI stacks, as confirmed by SEM cross-sections. (3) Backside thinning: the carrier and embedded dice were mounted on grinding tape, and the die backside was ground to the desired thickness (down to tens of micrometers; e.g., prototype dice thinned to ~24 µm). (4) Device release to yield flexible chip-in-foil implants. The process supports arbitrary die shapes and sizes placed at independent target positions on the flexible substrate determined by their positions on the fabrication wafer.

Prototyping and demonstration: Prototype PI-based devices with chains of three Si dice (e.g., 390 × 390 µm², 24 µm thick) interconnected by sputtered thin-film metal lines were fabricated. Larger dice (1620 × 1620 µm²) embedded in PI were shown conforming to a spherical surface (wetted glove) with bending radius ~6 mm. SEM analyses verified void-free PI–die interfaces and the imprint topography from 200 nm metal structures on the die after mechanical shearing tests. Placement accuracy was investigated, though specific numerical tolerances are not provided in the excerpt.

• FEM-derived design rules: Adding edge fillets to die corners improves die–substrate integrity and increases the effective area available for pad placement. Mechanical stress concentrates at die corners across fillet radii from 1–50 µm; routing interconnects near corners should be avoided. Contact pads should be placed with clearance from the die rim to mitigate delamination risks during conformation.
• Quantitative FEM results (PI thickness 10 µm, die 400 µm length, 100 µm thick, α_z = 10°): maximum interfacial delamination occurred along die edges. For r_edge = 20 µm, maximum delamination height reached 24 nm up to 12.8 µm from the die edge; for r_edge = 50 µm, maximum delamination was 14.1 nm up to 8.2 µm from the edge. Effective usable area (after accounting for delamination) was 87.8% of total die area for r_edge = 20 µm and 92.3% for r_edge = 50 µm. For the die with a 50 µm fillet radius, despite a smaller geometric area (0.158 mm²), the effective area (0.146 mm²) exceeded that of the die with a 20 µm fillet due to reduced delamination impact under misalignment.
• Process achievements: A microfabrication transfer process successfully placed, aligned, interconnected, thinned, and released multiple dice embedded in PI, enabling arbitrary die geometries and independent positioning. Continuous interconnects over die–carrier gaps were realized via lift-off; PI/metal/PI via structures to die pads were confirmed by SEM, and void-free interfaces were observed. Fabricated devices demonstrated conformability to curvilinear surfaces with bending radii on the order of millimeters (e.g., ~6 mm).

The results directly address the study goals by establishing practical design rules that enhance mechanical compliance and structural integrity of chip-in-foil implants. FEM analyses reveal that edge filleting of dice reduces interfacial delamination and increases the effective area suitable for reliable pad placement, thereby improving the likelihood of long-term adhesion and electrical reliability under curvature. Identifying corner regions as stress hotspots informs routing strategies that minimize interconnect failure by keeping lines away from high-stress zones. The quantified relationships between fillet radius, delamination extent, and effective die area help balance die geometry against usable interconnect/pad area, accounting for realistic misalignment during fabrication or implantation. On the manufacturing side, the demonstrated transfer process enables modular distribution of multiple, miniaturized dice across PI foils, reducing local bending stiffness compared to monolithic dice and improving conformability to neural tissues. Together, the design rules and process advances form a foundation for scalable, reliable, high-density neural interfaces that better match the biomechanics of brain surfaces, supporting chronic implantation objectives.

This work formulates and validates design rules for hybrid chip-in-foil neural implants and demonstrates a compatible transfer and microfabrication process. FEM studies show that edge fillets at die corners reduce delamination and expand the effective area for pad placement, while interconnect routing should avoid corner regions due to stress concentrations. Contact pads should be set back from die rims to prevent delamination during conformal bending. The developed process enables arbitrary die shapes and modular placement with reliable interconnects and thinning, yielding flexible PI-based prototypes that conform to curvilinear surfaces. These contributions advance the development of scalable, high-channel-count, mechanically compliant neural interfaces suitable for chronic applications. Future work could extend the mechanical model to include thin-film metal and fenestrations, quantify placement accuracy and long-term adhesion under physiological conditions, and assess chronic in vivo performance and electrical reliability.

• Modeling simplifications: The FEM treated the PI as a solid foil, neglecting thin-film metal interconnect layers and fenestrations, which could affect local stiffness and stress distributions. Metal features on die surfaces were also neglected except at pads.
• Geometric approximations: The brain surface was approximated as a cylinder with a 3 mm radius; more complex anatomies and interfacial fluid effects (surface tension, CSF layer) were not explicitly modeled.
• Fixed parameters: Simulations used a die rotation misalignment of 10°; broader variation could further refine design rules. Specific material property variations, viscoelasticity, and time-dependent effects were not detailed.
• Experimental scope: While prototypes demonstrated conformability and process feasibility, quantitative data on three-dimensional placement accuracy, long-term adhesion, interconnect durability under cyclic bending, and chronic in vivo performance are not provided in the excerpt.
• Electrical characterization: Detailed electrical performance metrics (e.g., contact resistance stability, noise) are not reported here.