Multi-Electrode Array (MEA)

Tens of thousands of microscopic ears pressed against a living neural network, each listening to a single whisper of electricity at the same moment.

An MEA replaces the single recording electrode — the workhorse of classical electrophysiology — with hundreds to tens of thousands of electrodes packed into a sensor area a few millimetres across. Neurons or brain slices are placed directly on the chip surface; action potentials crossing a nearby electrode induce a millivolt-scale extracellular voltage deflection that the chip amplifies and digitises in real time. The same electrodes can inject current pulses to stimulate cells. Because the array records simultaneously from many sites, it captures not just single-cell firing but network-level phenomena: synchrony, propagating waves, burst dynamics, and the spatial structure of spontaneous activity. Modern CMOS-based MEAs take this further by integrating on-chip amplifiers and a reconfigurable switch matrix so that any subset of tens of thousands of electrodes can be routed to readout channels at will — enabling sub-cellular resolution tracking of action potential propagation along individual axons.

Classical passive MEAs etch metal electrodes (titanium nitride, platinum, or indium-tin oxide) onto a glass substrate. Electrode count tops out around 256 because every electrode needs its own wire off-chip. CMOS-MEA designs break that barrier by fabricating the electrode array on a silicon integrated circuit that also contains on-chip amplifiers, analogue-to-digital converters, and a programmable switch matrix. The MaxWell Biosystems MaxOne, derived from ETH Zurich CMOS research, packs 26,400 platinum electrodes into a 3.85 × 2.10 mm² sensing area with a 17.5 µm pitch — sub-cellular resolution that allows the same neuron's axon to be tracked across multiple consecutive electrodes. At any moment, up to 1,024 of those 26,400 electrodes are routed to readout channels; the selection can be updated between experiments without moving the chip. Neuropixels-derived architectures extend the principle to silicon probes and have inspired in-vitro chip variants with 960 simultaneous recording channels along a 10 mm shank. The fundamental tradeoff is spatial coverage vs. simultaneous channel count: recording from the entire sensor area at full density would require bandwidth and power budgets that no current chip meets, so researchers must choose which region or cellular population to interrogate in each session.

In plain terms

Imagine a chessboard where every square has a tiny microphone buried in it. You scatter living nerve cells across the board, and overnight they wire themselves together and start talking in electrical pulses. The microphones pick up every conversation happening at every square at the same time, and a computer logs it all. That's a multi-electrode array: a listening grid for networks of real neurons.

Also called: microelectrode array, HD-MEA, CMOS-MEA.

Why it matters

Where it lives

Key numbers

Surprising facts

Open questions

A short history

The first multi-electrode recording from neural tissue was demonstrated by Thomas et al. in 1972 using a 30-electrode array on a silicon substrate. Commercial passive MEA dishes (64–256 electrodes) became laboratory workhorses in the 1990s through manufacturers such as Multi Channel Systems and Axion Biosystems. The leap to CMOS integration was made systematically at ETH Zurich, culminating in the Ballini et al. 2014 chip with 26,400 electrodes published in IEEE Journal of Solid-State Circuits — the architecture that became the MaxWell Biosystems MaxOne product. Litke and colleagues at UC Santa Cruz independently pushed to 65,536 simultaneously recording electrodes for retinal preparations (Nature Neuroscience Methods, 2017). Neuropixels probes (imec, 2017) brought high-density silicon recording to in vivo neuroscience and cross-pollinated in-vitro MEA design with probe-based multichannel architectures.

Sources

Related