Finite-contact field component
Experimental finite disk stimulation of a reconstructed human neuron.
Explore the field gallery → Compare disks, squares, rectangles and rings at three sizes, with shared potential and electric-field scales.
The verified disk potential drives the laboratory’s human cable solver in this separate benchmark. Full membrane replay uses the same mechanisms and step-averaged waveform as Cortex Laboratory. Disk radius: 25 µm. Conductivity: 0.3 S/m. Return: distant grounded boundary.
One cathodic-first biphasic pulse at 5 ms: 200 µs per phase, 100 µs gap. Each reconstruction is centered without scaling or trimming. Its entire membrane must fit within r ≤ 250 µm and z ≤ 250 µm and remain at least 25 µm above the insulating plane. This conservative plane guard also preserves the contact-distance guardrail.
Loading benchmark…
Scientific evidence for this configuration
Loading scoped evidence…
Validation results and limitationsDrag to orbit; scroll to zoom; click a branch to inspect it. Neuron: blue −100 mV → orange +40 mV. Gold: whole neuron excluded. Tissue slices: blue negative / orange positive applied potential, saturating at ±10 mV. Cylinder outlines the validated region; the plane is insulating and the small disk conducts.
Run the cell to inspect membrane voltage and ionic currents.
Recording traces and electric-field magnitude are unavailable: matching recording physics and field gradients have not been verified. Threshold crossings, waveform evidence, and propagation are distinct; the finite field does not resolve existing neural convergence failures.
Download complete package (8.8 MB ZIP) · SHA-256 checksum
The archive contains 78 files: the numerical solver and mesh generator, browser/Worker sampler, generated potential field, tests, examples, reports, plots, reproducible commands and INTEGRATION_HANDOFF.md.
Source and reproducibility
Integration verification and remaining limits · Mesh and timestep results. Browser visual verification for this integration is pending; the Chromium results below describe the supplied standalone package.
Start with the Cortex integration handoff. The unpacked source and artifacts are also available in the repository.
Original integration handoff (historical)
Read docs/cortex/FINITE_CONTACT_HANDOFF.md and the package in workbench/cortex-finite-contact-field/. Inspect the current target branch. Integrate the disk as an explicit experimental half-space/distant-return field mode. Preserve the separate point-source baseline, 25 µm guardrail and unresolved neural convergence limitations. Implement worker loading, SI coordinate transforms, cached compartment potentials and explicit invalid-region handling. Resolve or disable incompatible recording and electric-field outputs. Run integration and browser tests. Do not deploy further changes without authorization.
Exactly what is verified
An equipotential conducting disk of radius 25 µm lies in an insulating plane and injects prescribed total current into homogeneous isotropic tissue of conductivity 0.3 S/m. A grounded cylinder of radius and height 50 mm approximates a distant return. Current density is not imposed uniformly. This is a benchmark contact, not a validated Utah, Neuralink or Neuropixels implant.
| Measurement | Recorded result |
|---|---|
| Maximum independent-reference potential error | 0.7352% across 2,002 tested tissue points |
| Final mesh / domain change | 0.3071% / 0.6600% |
| Weak current imbalance | 1.31 × 10⁻¹⁴ |
| Raw-gradient return-current error | 4.0115% — separately reported; no electric-field sampler |
| Final acceptance gates | 19 pass; prior domain and coarse-flux failures retained |
| Runtime verification | Node.js, Chromium main thread and Chromium Web Worker |
| 100,000-point Worker sampling | 10.7 ms on the measured AMD EPYC host |
| Final offline solve | 3.21 s; 671 MiB peak process RSS |
The binary field is 956,564 bytes. Accepted tissue coverage is r ≤ 250 µm and 0 < z ≤ 250 µm in the disk frame, excluding a 0.5 µm neighborhood of the sharp rim. This is empirical numerical coverage, not a universal safe distance or a uniform error proof. The ideal rim has singular current density despite finite potential.
See the full validation report and figures for tolerances, absolute errors, convergence, interpolation, neighboring-compartment differences and hardware details.
Integration boundaries
Cortex currently uses full-space bipolar point sources. The disk package uses an insulating half-space and a distant return. It must be introduced as a distinct physical configuration; independently placed disk fields do not constitute a solved multi-contact implant. Recording physics and displayed fields must remain consistent with the selected conductor.
Arbitrary contacts, insulating shanks, interface impedance/capacitance and multi-contact generation remain incomplete. Whole-neuron collision checks, neural spatial/timestep convergence, stimulation thresholds and biological selectivity require separate verification. Field accuracy does not resolve those problems.
Archive integrity
SHA-256:
74f9905905c7281c19e978ba5a0cea76784e327edc4b4f8b2daae79ff41407e0After extracting, run sha256sum -c SHA256SUMS inside the package directory. Read README.md for exact reproduction commands.