The cell membrane is a lipid bilayer. Its core is a layer of water-repelling tails facing each other, so a charged ion cannot cross on its own. The way through is a protein that spans the membrane, and this figure animates a cross-section holding six of them, together with the calculation of the membrane potential.
At rest the only channel open is the K⁺ leak channel. The membrane is tens of times more permeable to K⁺ than to Na⁺, so the potential settles near -70 mV, close to the equilibrium potential of K⁺. A resting potential is not the absence of activity. It is the value set by whichever channels happen to be open.
When a stimulus brings the potential up to about -55 mV, voltage-gated Na⁺ channels open. Na⁺ flows in, the potential rises further, and that opens still more channels, so from there it runs away like a ball down a slope. Four domains have to move together, which makes the open probability proportional to m³.
A millisecond or so after the rise, the same Na⁺ channel closes itself. An IFM motif swings up from the cytoplasmic side and plugs the pore, and it stays there until the membrane repolarizes. This inactivation is what the refractory period really is, and it is also why an action potential does not travel back the way it came.
Voltage-gated K⁺ channels then open, late. All four subunits must open before current flows, so the open probability goes as n⁴ and the current lags the Na⁺ current by a few milliseconds. K⁺ leaving the cell pulls the potential back down, and because the channels are slow to close, it overshoots the resting level into an afterhyperpolarization. The lower box of the recording shows this order directly, in the offset between gNa and gK.
Every one of these channels assumes that a concentration difference already exists across the membrane. What builds it is the Na⁺/K⁺-ATPase, which pumps out three Na⁺ and takes in two K⁺ for each ATP. It turns only a few hundred times a second, but there are many of them, and together they account for about half of the ATP a neuron spends. Block it with ouabain and the gradient decays slowly, the resting potential drifts up, and in the end the cell can no longer fire.
The remaining two do not respond to voltage at all. The nicotinic ACh receptor opens only when the molecule binds, and it passes Na⁺ and K⁺ without telling them apart. At the resting potential the inward Na⁺ flux dominates, so the cell depolarizes, and once the threshold is crossed the voltage-gated channels take over from there. Aquaporin passes water and not protons. It is a pore with nothing to do with the ion pathways.
The equilibrium potentials in the readout panel come from the Nernst equation E = (61.5/z) log₁₀([out]/[in]), for a mammalian neuron at 37 °C. The membrane potential in the figure integrates the Hodgkin–Huxley model (squid giant axon, 6.3 °C, E_Na = +50 mV, E_K = -77 mV) in steps of 0.025 ms, so its equilibrium potentials differ from the table. The gates are drawn straight from m³h and n⁴.
Click a protein in the cross-section to switch the description to it, or pick one from the Proteins panel at the top left. The Controls panel at the bottom left fires stimuli, changes playback and speed, and applies blockers. Panels can be dragged by their header and resized from any edge or corner. Double-click a title to pin the panel to the top right.