Action potential in 3D: a membrane patch with Na⁺ and K⁺ channels

Cut-away bilayer with gated channels, ions and the whole neuron, all driven by the 2D page's Hodgkin–Huxley model.

Switch to 2D

Presets

Playback

Stimulus and membrane

Camera

Layers

Quality

Model check

The science

What you are looking at

A cut-away of a patch of nerve membrane: a lipid bilayer (cream heads, amber tails) with ten ion channels drawn in it – five voltage-gated Na⁺ channels (orange protein) and five K⁺ channels (teal protein) – between the extracellular fluid (above, Na⁺ rich) and the cytoplasm (below, K⁺ rich). A glass micro-electrode injects the stimulus current. In the distance the whole neuron is drawn for context and glows with the membrane potential.

The model: Hodgkin–Huxley (1952)

C dV/dt = Iext − gNa m³h (V−ENa) − gK n⁴ (V−EK) − gL (V−EL) with gating dx/dt = φ[αx(V)(1−x) − βx(V)x] for x = m, h, n. The code is the 2D page's module, copied verbatim (squid giant axon, C = 1 µF/cm², gNa = 120, gK = 36, gL = 0.3 mS/cm², ENa = +50, EK = −77, EL = −54.4 mV; temperature scales all rates by 3(T−6.3)/10).

How the channels are tied to the equations

Each Na⁺ channel has three activation flaps (gold, m) on the outside and one inactivation particle on a tether (magenta, h) on the inside; each K⁺ channel has four activation flaps (green, n). Every gate subunit carries a fixed random number u. A subunit is open when its gating variable is larger than u, so for many channels the fraction of conducting Na⁺ channels is exactly m³h and of K⁺ channels n⁴. The page tests this with 6000 virtual channels. The ten drawn channels are therefore one random sample of the ensemble that the equations describe; their flickering is the visual counterpart of the smooth averages. Ions leave the pool and cross a channel at a rate proportional to the driving force |V−E| while that channel conducts, in the direction of the current.

The “available Na⁺ channels” slider scales gNa; blocked channels stay shut in the picture.

Honest simplifications

  • The model is a single uniform patch (space clamp): there is no spike travelling along the axon, so the whole cell is shown with one V(t). The neuron is a drawing for orientation, not a simulation of propagation.
  • Ion counts, sizes and the ten channels are illustrative and not to scale; the model keeps concentrations and reversal potentials fixed, so the ion pools do not run down.
  • Squid axon constants; mammalian neurons have other channel kinetics. Channels are shown as rigid barrels with flaps – real gating involves voltage-sensor movement and is not mechanically like this.

Checks run in the page

  • The three checks of the 2D page: rest potential by Newton iteration, RK4 step halving, zero net current at the steady-state gates.
  • The channel rule used by the scene gives m³h and n⁴ open fractions within 4.5 binomial standard deviations at five times.
  • The rotation of the m and n gate nodes in the Blender model, read back from the node quaternions, is linear in the gate probability.

Credits

Blender 5.2 neuron, channels with named gate nodes, lipid and micro-electrode with Cycles-baked textures; Three.js r160 animation driven by the Hodgkin–Huxley integration.

Loading model...
Loading Blender model...
Membrane potential V(t)
Gating variables m, h, n
Ionic currents I_Na and I_K
All-or-none: peak V vs stimulus current
Made with 🟠 Blender · Three.js