Advanced Dipole Simulation
The Cortical Dipole
Four terms get used almost interchangeably when people describe "brain activity" — action potential, neurotransmitter release, electric field, current source density. They aren't four parallel channels. They're four stages of one event, and a QEEG amplifier is only ever looking at the last two. This page walks the chain, then lets you build the field yourself.
The chain
one event, four vantage pointsFour stages
what each one actually isAction potential
The all-or-none spike. Past threshold, voltage-gated Na⁺ channels snap open and K⁺ channels restore the resting potential — a stereotyped ~1 ms pulse that regenerates itself down the axon at full amplitude. It's the fastest signal in the brain, and the one most people picture when they think "neurons talk." It is not what your EEG amplifier sees: individual spikes are too brief, too scattered across differently oriented axons, and too asynchronous to sum into a field a scalp electrode can resolve.
Neurotransmitter release
The handoff. An arriving spike opens voltage-gated Ca²⁺ channels at the terminal; the Ca²⁺ influx triggers vesicles to fuse and dump transmitter into the cleft. Binding postsynaptic receptors opens or closes ion channels, producing a graded postsynaptic potential — depolarizing (EPSP) if excitatory, hyperpolarizing (IPSP) if inhibitory. Unlike the spike, these decay over tens of milliseconds and, crucially, summate — across nearby synapses and across time. That summation is what builds a signal large enough to leave the cell.
Electric field
Current entering a dendrite has to exit somewhere else along the cell to complete its circuit — a sink in one place, a source in another, separated by the cell's length. That separation is a dipole. Pyramidal cells are unusually good generators: long, and packed in parallel columns all facing the same way. When a patch of them depolarizes together, their dipoles sum, and the current spreading through the tissue around them is the field an electrode records. One neuron's field is undetectable; tens of thousands, synchronized, is what a QEEG is looking at.
Current source density
The un-blurring. A single electrode records a volume-conducted sum of every nearby source and sink — it can't say where in the tissue the current actually started. CSD estimates that, by taking the spatial second derivative of the potential across closely spaced sites (classically a depth electrode spanning cortical layers). Nothing new is happening physically; it's the same field, read with enough spatial resolution to say which layer, or which patch of cortex, is doing the work.
Where the rhythm comes from
worth asking before stage 03Neurotransmitter release is the messenger, not the clock. Each cycle in the chambers below, when the modeled synapse fires, that's a real event — an arriving spike triggering vesicle release, exactly as Stage 02 above describes. But swapping which neurotransmitter is involved wouldn't change the rhythm's frequency, and release itself doesn't set the tempo. The pacing comes from further up the chain: recurrent excitatory–inhibitory loops (pyramidal cells drive fast GABAergic interneurons, which inhibit them back — the round-trip delay, set mostly by GABA-A decay kinetics, is a major source of gamma-band rhythms), thalamocortical loops (thalamic relay neurons carry intrinsic pacemaker channels — T-type Ca²⁺, Iₕ — that let them burst rhythmically on their own, a big contributor to alpha and to sleep spindles), and intrinsic membrane conductances that can make even a single cell's potential want to oscillate with no rhythmic input at all. The chambers below don't model any of that circuitry: the oscillation is imposed directly, as a sine wave, so the frequency slider can freely scrub delta through gamma and show you the resulting field — not derived from a pacemaker. That's a deliberate simplification for teaching the field and CSD, not a claim about how the brain actually generates a given band.
And yes — the isopotential field oscillates right along with it. The contour geometry (nested loops, always perpendicular to the field lines) is fixed by the source arrangement; what changes frame to frame is each contour's strength and, every half cycle, each pole's identity, because every pole's charge is the same fixed ratio times one shared, signed drive value. As the drive swings toward its peak the contours expand outward; as it crosses zero they collapse and briefly vanish — the same real zero-crossing behavior as the field lines, not a rendering gap; on the far side of zero, amber and blue swap, because whichever pole was a source a moment ago is now a sink. The dashed 0V line is the one thing that stays put through all of this, since its position depends only on the poles' fixed geometry, not their shared sign. Track a single fixed point under that shifting contour map — exactly what one scalp electrode does — and its value over time is the raw EEG waveform in the strip chart below.
The dipole chamber
stage 03, built by handField of a single pyramidal neuron
Schematic, not a calibrated biophysical simulation — field lines are traced from a simple two- or multi-point-charge model chosen to show the shape of a dipole field, not measured amplitudes or realistic tissue conductivity. Layer boundaries and geometry are illustrative proportions, not a specific atlas. The Amplitude/Neurons readouts turn the recruited population size and frequency band into an order-of-magnitude scalp microvolt estimate — a literature-ballpark per-neuron moment, linear summation when synchronized vs. square-root summation when not, and a rough band-amplitude trend — so the numbers land in the right physical neighborhood, not a calibrated forward model. Real scalp amplitude depends heavily on head geometry, electrode reference, and individual state.
Controls
Reading stage 04
the depth profileWhat the picture on the left actually is. A real laminar depth electrode — a probe inserted straight down through the cortical column, with contacts spanning layers I through VI. That's the only way to directly measure current source density, and it's mostly an animal or invasive-human (e.g. epilepsy-surgery) technique, not something a clinical EEG can do. It's shown here because it's the clearest way to see why a sink/source pattern exists physiologically, before asking what a scalp electrode can and can't recover of it.
What a real, human, scalp qEEG gives you instead. No depth access at all. A standard clinical montage uses 19 scalp electrodes (the 10-20 system), each sitting outside the skull and picking up a heavily blurred, volume-conducted sum of activity from a wide patch of underlying cortex — near and far, superficial and deep, all mixed into one number per channel at each instant. There's no physical way to un-mix that back into "this came from layer IV" the way the probe on the left can.
The inverse problem, and sLORETA. Working backward from those 19 scalp voltages to "where in the three-dimensional brain did this activity actually originate" is called the inverse solution — and it's fundamentally underdetermined: infinitely many different arrangements of sources inside the head could produce the exact same 19 scalp readings, so there's no way to simply undo the blur. You have to add an assumption to pick one plausible answer out of infinitely many. sLORETA (standardized Low-Resolution Electromagnetic Tomography) is the assumption most qEEG software reaches for: treat neighboring cortex as likely co-active, and solve for the smoothest three-dimensional current-density map — across a template cortical grid of thousands of voxels — that could have produced the recorded scalp pattern. It's a real, validated estimate, not a guess, but it's inferred from 19 blurred numbers, nowhere near the anatomical precision of the six clean layers on the left. sLORETA gets its own tool later in this series; this chamber is about the physiology underneath it.
Even short of a full sLORETA solve, the same second-derivative idea sharpens things directly at the scalp: a surface Laplacian, computed across the 19-channel cap, reduces the reference-electrode and volume-conduction smearing that otherwise inflates apparent connectivity between nearby channels. That matters directly for coupling and coherence metrics — iCOH, phase-amplitude coupling, anything built on comparing two channels — because volume conduction can manufacture correlation between electrodes that are really picking up the same underlying source.
The CSD chamber
stage 04, built by handTripolar field — input at layer IV
Live: the same tripolar source, sampled at all 19 positions by true distance from Cz. Color = source/sink; size and brightness = strength relative to the strongest channel this instant. This single-column model has no left-right or front-back asymmetry, so equidistant channels always match exactly — real cortical sources rarely do, but the falloff-with-distance it demonstrates does generalize.
Same picture language as the dipole chamber above, generated by a different geometry: one pyramidal cell, synapse at the chosen layer (sink), return current at two flanking points (sources) — a tripolar pattern, not the simple two-pole dipole up top, because the input lands partway down the dendrite rather than at its tip. The strip below is the actual lesson, plotted by depth (pia on the left, white matter on the right) rather than by time: raw field is that geometry spatially blurred to stand in for volume conduction — a broad hump spread across several layers; CSD is the true finite-difference second derivative computed from that blurred trace, not read back from the neuron directly, and comes out sharply localized right at the input layer. Same instant, same underlying current — the width difference between the two curves is what "un-blurring" buys you. Layer boundaries, blur kernel, and geometry are illustrative, not measured — as is the nA/mm³ scale on the "CSD at input layer" readout, which is sized for one modeled cell, not the thousands of synchronized cells behind a published population CSD figure.
