The opsin toolkit
Writing starts with the actuator. The opsin toolkit has grown from the original blue-light cation channel (channelrhodopsin-2, which excites) to include inhibitory pumps and anion channels (which silence). Two engineered classes matter most at the frontier: red-shifted opsins, which absorb longer, less-scattered, deeper-penetrating light and free up the blue-green band for imaging; and soma-targeted opsins, engineered to stay in the cell body so that stray light on nearby axons and dendrites does not cause off-target firing.
Opsin photocurrent saturates with photon flux. To drive a spike you need enough open channels to reach threshold; opsin kinetics and the tissue power budget set how many cells you can push, and how fast.
Why single cells need two photons
One-photon (conventional) illumination cannot address single cells in depth. Blue light both scatters and is absorbed along the whole path, so a single fibre or LED activates an entire cone of tissue — a column, not a cell. Two-photon excitation fixes this because the excitation rate depends on the square of intensity: only at the tight focus, where photons are dense enough for a molecule to absorb two at once, does meaningful excitation occur. That quadratic dependence is what confines a write in all three dimensions.
Two-photon excitation rate scales as intensity squared — the physical origin of optical sectioning and the single-cell axial confinement that one-photon light cannot achieve.
Holographic parallel writing
Scanning a single focus from cell to cell is serial and slow — too slow to paint a pattern within one neural timescale. Two-photon holographic stimulation solves this with computer-generated holography: a spatial light modulator reshapes the wavefront to place many diffraction-limited spots at once, anywhere in a 3D volume. Combined with temporal focusing for axial tightness, this drives tens to hundreds of chosen cells with millisecond, near-single-spike timing.
'Playing in' activity — and its honest ceiling
With holographic control you can write a defined ensemble — activate a specific set of cells in a specific order — and, in mouse cortex, bias or even evoke a percept and drive behaviour from that imposed pattern. That is a landmark: causal, cell-resolved writing into a working brain. But note the honest ceiling. Choosing cells and driving them to spike is solved; knowing which pattern reproduces a natural computation is the unsolved biomimetic-write problem. We can write fluently in an alphabet whose grammar we do not yet know.