The electrode is a cell, not a wire
You already know the decoders. But every decoder — Kalman, Riemannian, or a deep network — sits downstream of a physical transducer: a metal or polymer contact bathed in ionic fluid, separated from the neuron by living tissue. The single most useful reframing in this track is that a recording site on a microelectrode array is not a wire, it is an electrochemical half-cell. Its behaviour is governed by chemistry, not just geometry.
For recording, the site converts ionic currents in tissue into electronic currents in the wire across the electrical double layer; the raw signal is the extracellular action potential, tens to a few hundred microvolts. For stimulation it runs the same physics in reverse. Both directions are limited by one operating point: the interfacial impedance and how charge crosses it.
Impedance sets the noise floor
A spike of tens of microvolts competes with noise, and a large share of that noise is thermal (Johnson–Nyquist noise) generated by the resistive part of the electrode impedance itself. The mean-square voltage noise over a bandwidth \Delta f is set by the real part of the site impedance:
Thermal noise power of a recording site scales with the real part of its impedance — halve the resistive impedance and you cut noise power in half.
This creates the field's founding tension. A smaller site gives better spatial resolution but has higher impedance and therefore more noise — a hard trade between resolution and SNR. The entire materials programme is a way to break that trade: lower the impedance without changing the geometric size of the contact.
Charge injection sets the stimulation ceiling
For writing — microstimulation — the limit is different: how much charge you can push per pulse without triggering irreversible chemistry that damages tissue or corrodes the electrode. Charge crosses the interface two ways, captured by faradaic vs capacitive charge injection: capacitive (charging/discharging the double layer, fully reversible) and faradaic (electron transfer through a redox reaction, safe only if the reaction reverses on the opposite phase).
The safe envelope is bounded below by the electrode's charge injection capacity and above by two hard limits: the water window (potentials that avoid electrolysis) and the empirical Shannon–McCreery tissue-damage line. That line couples charge density D per phase and charge Q per phase:
The Shannon–McCreery safety relation (D in µC/cm² per phase, Q in µC per phase). Higher charge per pulse is only safe at lower charge density — which is exactly what more electrode area buys you.
Bare platinum has only a modest charge storage capacity. The entire field of stimulation materials — iridium oxide, PEDOT:PSS — exists to push injectable charge up by roughly an order of magnitude while keeping every electron reversible.
The real ceiling: the tissue writes back
Even a chemically perfect electrode provokes a biological reaction. Insertion and chronic presence trigger the foreign-body response: reactive gliosis walls the site off in a glial scar, and peri-electrode neuronal loss removes the very cells you wanted to hear. Over weeks impedance rises and amplitude falls — chronic signal degradation — driven partly by ongoing tissue micromotion against a stiff shank.