Thin-Film Grids for Surface Recording

A thin-film grid is a flexible 12 µm polyimide electrode array that records from the surface of the cortex, the skull, the heart or a peripheral nerve rather than penetrating it. Because its sites sit outside the tissue, it records the field potential of the population beneath each site and never individual spikes, so two decisions select a design — how large the sites are and where the array is placed — across four catalog families and 53 designs.

A grid is the surface counterpart of the silicon probe. The same thin-film process lays platinum sites and traces on a polymer film instead of a silicon shank, the finished array bends to whatever it rests on, and it mounts on a package and mates to the same headstages. What changes is what the surface implies, and this guide is about those changes: what the signal is, what the two decisions are, and which family answers which surface. It assumes the background in How an Electrode Records Neural Activity.

What does a surface electrode record?#

The local field potential of the population under each site, and nothing smaller. A spike is visible only within about a hundred micrometres of a site, and on a surface the sites are separated from the neurons by the pia, the dura or the skull, so single units are out of reach; what remains is the summed synaptic and population current of the tissue beneath, which does not fade with distance the way a spike does and is present on every site at once. That signal is the point of a grid, not a limitation of it. It reports rhythms, state, evoked responses and their spatial pattern across a region — the map a shank cannot make, because a shank samples a cylinder about 200 µm across.

Three consequences follow for the electrode. Sites can be large, because they are averaging a population, and a large site has lower impedance and a cleaner field potential — which is why grid sites run from 25 µm to 1 mm rather than the 15 µm of a probe site. The reference matters more, because slow potentials travel, and a distant reference adds signal from everywhere between it and the site: the reference decision is worked through for grids in Thin Film Grids: EEG Referencing. And contact is the design problem: nothing anchors a grid but its fit, so conformity to a curved or moving surface is what every family is engineered for.

How large should the sites be?#

Site diameter is the first selection axis, and across the ECoG catalog it spans fortyfold. A 25 µm site reads a local patch at high impedance and, at a 300 µm pitch, resolves cortical columns — micro-ECoG. A 1 mm site averages a wide patch at low impedance and favours the low-frequency field potential over anything resolved in space. Between them, 100 to 300 µm sites at 500 µm to 1 mm pitch map evoked activity across an area. Pick the diameter from the spatial scale of the claim the study will make, then pick the site count that covers the area: catalog ECoG grids run 4 to 256 sites at diameters of 25, 30, 40, 50, 60, 80, 100, 200, 300 and 1,000 µm.

The other families fix the diameter for their surface. EEG arrays use 500 µm platinum discs on every design, because the skull has already blurred the signal to that scale. Cardiac grids use 575 µm sites, sized to the epicardial field potential. Nerve cuffs place their sites around the circumference, so the number of sites and the cuff diameter, not the site size, are the choice.

Where will the array be placed?#

Placement is the second decision, and it is made before the array is ordered because the families differ by surface.

On the cortex, an ECoG grid goes epidurally on intact dura or subdurally on the pia. Epidural placement leaves the dura intact — lower cortical-injury risk, less bleeding and a better-behaved chronic preparation — at the cost of a signal the dura attenuates and spatially blurs. Subdural placement gives higher amplitude and finer resolution, including high-gamma activity, and costs a dural opening and a preparation that scars more. The film follows the curvature and the movement of the cortex, and that is what keeps the sites in contact; the surgical decision and procedure are in the ECoG surgical guide. Catalog examples: the 16-site E16-300-CL5-25 micro-ECoG grid (300 µm pitch, 25 µm sites, 5 mm cable); the 128-site E128-200-CL8-40; the 16-site E16-1200-1500-CL10-300 for regional mapping at 300 µm sites.

On the skull, an EEG array records with no bone removed, no dura opened and no cortex exposed, which is what makes a preparation that lasts months. Skull and scalp spread the signal, so it measures rhythms and state — sleep and wake staging, seizure detection, cortical oscillations, long-duration monitoring — rather than local circuit activity. The difficulty moves to alignment and fixation: layouts are dimensioned from bregma for a specific skull, so a rat design on a mouse places sites off the target areas entirely, and the bone must be dry at placement because the coupling through it is part of the measurement. Nine catalog designs, five for mouse and three for rat, with 6 to 32 sites and cables of 7, 8, 10 or 50 mm; the 30-site mouse designs exist at all three short cable lengths so the connector sits where the headcap allows. Catalog examples: EEG-M-30-A-CL10 for mouse; EEG-R-16-CL10-func for rat, with a bregma fiducial. The procedure is in the EEG surgical guide.

On the heart, a cardiac surface grid sits on the epicardium and deforms with it. On a beating surface the limit is rarely amplitude but motion artifact — an array that loses and regains contact through the cycle produces signal that cannot be separated from the mechanics — so the designs address deformation directly: accordion layouts fold along one axis, and segmented 8×8 layouts divide the array so regions flex independently. Five catalog designs with 16, 64 or 128 sites of 575 µm on a 150 mm cable; CS8x8-seg-4000-CL150-575 is the segmented array, CS2x2x4-accord-4000-CL150-575 the accordion.

Around a nerve, a nerve cuff is pre-curved during fabrication to wrap a peripheral nerve from the outside. Nothing is inserted through the epineurium, so the axons are not damaged on placement; the trade is selectivity, because the signal is summed across fascicles. Curvature cannot be changed afterwards, so the diameter is chosen rather than adjusted — a cuff sized for a larger nerve sits loose and moves, one sized too small compresses — and the target nerve is measured before selecting. Catalog cuffs are pre-curved to 350 µm, 900 µm and 3 mm, with 3 to 24 sites: C16-3000-CL50 is a 16-site cuff at 3 mm, C3-CL50 a three-site cuff. A sieve variant places holes through the array for a transected nerve to regenerate through, which is a different experiment rather than a different mounting.

Surface and depth together#

A grid pairs with a penetrating probe in the same animal — a surface map and a laminar profile from one preparation, on the same headstage and the same clock. The surface array goes on first, and a probe passes through a slot or beside the grid. The Matrix 3D platform can be combined with an ECoG array by design. This is the arrangement for a study that needs both the region's pattern and the column's units, and it is the reason the two families share a package vocabulary and a recording chain.

What to order#

Decide the surface, then the site size, then the site count that covers the area, then the cable length against the headcap — on a mouse a real constraint, decided with the package. ECoG cables run 5 to 100 mm in 14 standard lengths; packages, connector options and per-design site pitch are catalog data and are read from the product record rather than from this guide. Where the catalog has no layout for the surface, the same process draws one: footprint, site diameter, site count, cable length and curvature are all customisable. Hold a grid to the same standard as a probe — the determinants in What Makes a Thin-Film Electrode Good apply to a polymer array as they do to silicon — and choose it against the target with Choosing an Electrode for Your Experiment. NeuroNexus products are for research use only.