Past the electrode
Every channel you met in Volumes I and II ultimately touches tissue with metal or accepts the optical price of getting light through it. Electrodes scar, drift, and cannot be placed everywhere at once; optical writing is exquisite but shallow. This track asks a different question: what if the interface were carried not by a wire but by a field, a molecule, or a piece of engineered DNA already living inside the neuron?
The appeal is minimal invasiveness. A magnetic field passes through the skull as if it were not there; a molecule can be delivered by a virus and expressed by the cell itself; a molecular recorder writes to the genome and needs no readout cable at all. These are, in principle, the channels that could reach deep structures across large volumes without a craniotomy — the holy grail hiding behind the invasiveness–signal-quality tradeoff.
Three families
Magnetic. Magnetic nanoparticles convert an external field into local heat or force to gate ion channels (magnetothermal stimulation, magnetogenetics); on the read side, nitrogen-vacancy diamond magnetometry senses the tiny magnetic field of neuronal currents. Molecular. Engineered receptors respond to a designer drug (chemogenetics / DREADDs); genetically-encoded sensors light up when calcium, voltage or a neurotransmitter changes. Genetic. A DNA molecular ticker tape or CRISPR recorder writes activity into the genome to be sequenced later.
Notice a structural fact that will haunt the whole track: almost every one of these channels is read-only or write-only, and almost every one is slow. This is a very different regime from the millisecond, bidirectional electrode. It is why these methods are, so far, tools for neuroscience and candidates for therapy — not yet a general-purpose interface.
Why depth favours fields
The single strongest argument for the magnetic and molecular route is attenuation. Light entering brain tissue is scattered and absorbed, so its usable fluence falls off roughly exponentially with depth; even in the near-infrared window the effective attenuation length is on the order of a millimetre.
Diffusive light fall-off in tissue: fluence drops with depth z on the scale of the effective attenuation length set by absorption μ_a and reduced scattering μ_s'. This is the barrier optogenetics fights against.
A low-frequency magnetic field, by contrast, passes through non-magnetic tissue essentially unattenuated — brain is magnetically transparent. That is the whole temptation of magnetogenetics: reach anywhere, no fibre, no craniotomy. The figure shows the flagship magnetic-write concept, magnetothermal stimulation, which we dissect next.
The read/write asymmetry, restated
You met the read/write asymmetry in Volume II: reading intent is easier than writing a well-formed percept. Molecular channels sharpen it into a slogan. Reading at the molecular level can be extraordinarily specific — a sensor for one neurotransmitter in one cell type — but is usually slow and often destructive. Writing at the molecular level is coarse in time and hard to localise, and it almost always requires first installing machinery by gene delivery. Specificity and control rarely come together in the same molecule.