Choosing an Electrode for Your Experiment

The right electrode is the one whose geometry and placement fit the target, the measurement, the duration of the experiment, the subject and the recording system it will be read through. Six questions, answered in order, select it from the catalog of 216 silicon probe and 53 thin-film grid designs — penetrating or surface, which structure, spikes or field potentials, hours or months, mouse or macaque, and which headstage — and the answers are a probe name.

The brain is not one size fits all, and neither is the electrode. Every design in the catalog was drawn against a target, and the wrong choice is not a worse recording but a different experiment: a probe whose sites straddle two structures answers a question nobody asked, and a grid placed where a shank was needed records rhythms when the study needed units. This guide is the selection in the order the decisions bind. It assumes the background in How an Electrode Records Neural Activity; the package, which is chosen after the design, is in Choosing a Probe Package.

Penetrating or surface — does the target lie inside the tissue or on it?#

This is the first axis and it eliminates half the catalog. A silicon probe penetrates — brain, spinal cord, peripheral nerve, ganglion or heart — and records at known depths along a trajectory; it is the choice whenever the structure of interest is beneath a surface, whenever single neurons are the measurement, or whenever a depth profile is the point. A thin-film grid rests on a surface — the cortex under the dura or on the pia, the skull, the epicardium, the outside of a nerve — and records the summed field potential of the population beneath each site; it is the choice when the question is about a region's activity, rhythms or maps rather than its cells, when the tissue must not be entered, or when the preparation must last months with the least intervention.

The two are not competing answers to one question. Individual spikes are out of reach from a surface, so a grid cannot substitute for a probe where units are needed; and a probe samples a cylinder about 200 µm across, so it cannot substitute for a grid where a centimetre of cortex must be mapped at once. Studies that need both use both, often on the same headstage.

Which structure — how deep, how thick, how wide?#

The target's anatomy sets three numbers, and each maps to a catalog field. Its depth below the surface sets the shank length: 3 mm reaches rodent cortex, 5 mm the rodent hippocampus and striatum, 10 mm deep rodent structures and primate cortex, and beyond about 10 mm the mechanics change and a Vector design — a short array on a 70 or 110 mm support body — replaces a bare shank. Its thickness sets the site span: about 1 mm covers the mouse cortex, 2 mm the rat's, 2 to 3 mm the primate's, and a hippocampal profile that must cross CA1 and the dentate gyrus wants 3 mm or more. Its width sets whether one shank is enough or several are needed, and at what separation: 200 µm places shanks in adjacent columns, 400 to 500 µm in different areas.

Read the numbers in the species you are using. Two shanks 200 µm apart sit within one cortical column in a rat and across two in a mouse; a 32-site span at 50 µm pitch (1.55 mm) profiles a full mouse cortex and two thirds of a rat's. Small structures do not have proportionally thinner layers, so the densest available pitch is often not the right one for a mouse.

Catalog examples: rodent cortical laminar profile, A1x32-5mm-50-177 (32 sites, 50 µm pitch, 1.55 mm span, 5 mm shank); rat hippocampus through CA1 and dentate, A1x32-10mm-100-177 (3.1 mm span on a 10 mm shank); four adjacent columns at once, A4x8-5mm-100-200-177 (four shanks at 200 µm, eight sites each at 100 µm pitch). Every catalog design's geometry can be compared in the design space.

Spikes or field potentials — what is the measurement?#

The measurement sets site area and site pitch, because they decide how local a site's view is. For single units, choose small sites — 177 µm² is the catalog's working standard — and a pitch inside the detection radius so that one neuron appears on several sites at once: a tetrode cluster, a polytrode's 20 to 25 µm inter-site spacing, or a Buzsaki layout's staggered tip. That shared view is what a spike sorter separates units on, and it is why a 32-site polytrode in 0.8 mm of tissue yields more isolated units than 32 sites spread over 3 mm. For a depth profile of the field potential, or for spikes and LFP together across layers, choose a linear layout at 50 or 100 µm so that each site samples its own tissue; larger sites — 703 µm² — give a cleaner field potential and are the choice where stimulation or population signals matter more than unit detail. On a grid the same logic gives the site diameter: 25 to 100 µm for micro-ECoG that resolves columns, 500 µm to 1 mm for regional signals and skull-surface EEG.

Catalog examples: unit isolation in one structure, A1x32-Poly2-5mm-50s-177 or the Buzsaki Buz32-sp-5mm; a laminar spike-and-LFP profile, A1x32-5mm-50-177; two dense columns, A2x32-5mm-25-200-177 (25 µm pitch on two shanks at 200 µm); a micro-ECoG map, the 16-site E16-300-CL5-25 grid (300 µm pitch, 25 µm sites).

Hours or months — how long must the electrode work?#

Duration decides the preparation, and the preparation constrains the design. An acute session — hours, the probe removed at the end and reused — can use any design that reaches the target, on a 15 µm substrate that displaces the least tissue, and it is the faster, cheaper way to learn what to build a chronic study around. A chronic implant — days to months, the animal awake and moving — needs the same neurons on more than one day or behaviour the animal can only produce untethered, and it changes three things: the substrate thickness may go to 50 µm where a long shank must resist buckling on insertion; the site metal and dielectric must hold their impedance over weeks, which is what iridium sites and silicon-carbide insulation are for; and the package must be light enough to be worn and robust enough to survive being worn. The design is chosen first, the package second, and the package is committed at implant — it cannot be revisited afterwards.

A surface grid changes the calculus for the longest studies. A skull-surface EEG array involves no craniotomy and no cortical exposure, and is the choice for a preparation held over months where rhythms and state are the measurement; an epidural ECoG grid on intact dura survives chronically better than a subdural one, at the cost of attenuation. The surgical decisions that follow from duration are in the acute and chronic surgical guides and their ECoG and EEG counterparts.

Mouse or macaque — what can the subject carry, and how far is the target?#

The subject sets two constraints the target does not. The first is depth and trajectory: in a large animal a deep target is tens of millimetres from the surface and a slender silicon shank cannot get there straight, which is what the Vector design's rigid support body is for, and a chronic chamber and microdrive replace the stereotaxic holder. The second is load: a 30 g mouse is not a small rat, and an implant the animal cannot carry produces no behaviour worth recording. Channel count, connector size and cable stiffness all add to the head weight, so the site count is budgeted against the subject at selection, alongside the geometry, and not after the package is chosen. The catalog is concentrated where the subjects are: 198 of 216 silicon designs have shanks of 10 mm or less, which covers the rodent brain; the rest reach further.

Grids scale with the subject too: EEG layouts are dimensioned from bregma for mouse or rat and a layout drawn for one put on the other samples different structures, and ECoG grids run from a few millimetres for rodent cortex to tens of millimetres for a primate or large-animal surface.

Catalog examples: deep target in a primate, V1x32-Edge-10mm-100-177 on a 110 mm support body; rat skull-surface EEG, the 16-site EEG-R-16-CL10-func array with bregma fiducial.

Which headstage — what will read it?#

The last question closes the loop to the recording chain, and it is asked last because the design should not be bent to the electronics. Site count fixes the connector and the headstage channel count: a 16- or 32-site design mates to a 16- or 32-channel headstage, a 64-site design to a 64-channel one, and a 256-site or larger design needs a high-density package and a headstage to match. Site impedance should be at least ten times below the headstage's input impedance, which every NeuroNexus headstage is built to for catalog site areas. Where the count runs past what a passive probe and headstage can serve, the SiNAPS active-pixel designs put the amplifier under each site and read 256 to 1,024+ channels through one cable. Stimulation, optical delivery and drug delivery are package options on the same designs, and each brings its own instrument requirement.

The system also decides what the study can know during the session rather than after it: a recording chain that reports impedance, noise and unit yield while the animal is on the rig catches a wrong choice in the first minutes rather than in analysis weeks later, which is what Radiens is for.

Putting the answers together#

The six answers are a probe name. A laminar spike-and-LFP profile of rat somatosensory cortex, chronic, head-fixed, on a 32-channel headstage is A1x32-5mm-50-177 on a chronic package; unit isolation in mouse CA1 across a few sessions is a polytrode or Buzsaki design at 25 µm on a light chronic package; a map of evoked activity across a primate motor cortex is a 128-site ECoG grid on a high-density package; a regional EEG study held for three months in mice is a skull-surface array with no craniotomy at all. Where the catalog has no design at the target, the same process draws one: geometry, materials and package are all customisable, and a custom design is a conversation about the target rather than a compromise on it.

Judge the candidate before ordering by the determinants in What Makes a Thin-Film Electrode Good, choose its package with Choosing a Probe Package, and see the full range of designs in The Electrode Design Space. Ask Oaks with the target, the measurement and the subject, or request a quote with the probe name. NeuroNexus products are for research use only.