Heating a channel open
The best-supported magnetic-write method is magnetothermal stimulation. Superparamagnetic nanoparticles are delivered near neurons that express a heat-sensitive channel (classically TRPV1). An alternating magnetic field at hundreds of kilohertz drives the particles through hysteresis-like loss, dissipating heat; the local temperature rise opens the channel and depolarises the cell. This has been demonstrated in vivo in rodents, driving behaviour from deep targets without an implanted light source.
Linear-response (Rosensweig) volumetric heating power of superparamagnetic particles in a field of amplitude H_0 and frequency f. Loss peaks when the drive matches the magnetic relaxation time τ — nanoparticle size and field frequency must be tuned together.
The honest caveat is thermal geometry. A single 10-nanometre particle dissipates only femtowatts, and the steady temperature it raises falls off as $1/r$ — negligible at the membrane. Useful gating needs dense particle clusters or a slow bulk temperature step, which raises off-target heating and safety questions bounded by the same tissue-heating budget you know from implants.
Steady-state temperature rise at distance r from a spherical heat source of total power Q in a medium of thermal conductivity k. Because ΔT scales with Q and falls as 1/r, a lone tiny particle barely warms its neighbourhood — clustering, not single particles, does the work.
Force, mechanics and electricity
Two cousins avoid heat. Magnetomechanical stimulation uses larger particles whose torque or pull in a field mechanically gates a stretch-sensitive channel. Magnetoelectric nanoparticles convert a magnetic field into a local electric field — in principle a wireless micro-electrode — but coupling coefficients are small and in-vivo demonstrations remain early and contested.
The magnetogenetics controversy
Now the field's cautionary tale. A set of high-profile papers proposed ferritin-based magnetogenetics: a fusion protein tethering the iron-storage protein ferritin to a mechano- or thermo-sensitive channel, so that a modest magnetic field alone would gate it. Behaviourally, the constructs seemed to work. Physically, they should not.
The thermodynamic objection (after Meister): the magnetic energy m·B of a ferritin-scale moment in laboratory fields is orders of magnitude below the thermal energy k_B T, and the gradient force ∇(m·B) is far below thermal (Langevin) forces — so a field cannot deterministically pull, twist or heat the channel open.
This is the heart of the ferritin-magnetogenetics controversy: a large body of theory says the claimed forces and energies are far too small, replication has been inconsistent, and no agreed mechanism explains the behavioural effects. The lesson is not that magnetism is useless — magnetothermal works — but that an intriguing behavioural readout is not a mechanism. Hold the whole family to the k_B T ruler.
Reading with magnetism
The read side is quieter but cleaner. Neuronal action currents produce a picotesla-scale magnetic field; a nitrogen-vacancy diamond magnetometer can sense it optically and label-free, and has resolved the magnetic signature of a single firing axon in a dish. The prize is a recording that needs no penetrating electrode and no genetic modification.
The catch is depth and sensitivity: the field falls steeply with distance and competes with ambient noise, so in-vivo, whole-brain magnetic reading remains a hard open problem. It belongs, for now, alongside the other emerging sensing modalities — a beautiful physics demonstration awaiting a path to the intact brain.