Chain Reactions, Criticality, and Control

Each fission releases two or three neutrons. If, on average, exactly one of those neutrons goes on to cause another fission, the reaction is self-sustaining at constant power: each generation of fissions is the same size as the last. Physicists call this state critical, and despite how the word sounds in headlines, a critical reactor is simply a reactor that is on. If slightly more than one neutron per fission causes a new fission, power rises (supercritical — also normal, briefly, whenever operators raise power); slightly fewer, and power falls (subcritical — the state of a shutdown reactor).

The whole craft of reactor design is making "exactly one" easy to hold. The bookkeeping is brutal at first glance: of the ~2.4 neutrons from a typical U-235 fission, some escape the core entirely, some are captured by U-238, by structural metal, by the coolant, or by fission fragments, and only the survivors cause new fissions. The fuel's enrichment, the core's geometry, and the materials around it are all chosen so the balance lands near 1.0 — and then operators trim it precisely with control rods, made of neutron-absorbing materials like boron carbide or silver-indium-cadmium. Push the rods in, and they soak up neutrons, tipping the balance below one; withdraw them, and the balance rises. Pressurized-water reactors also dissolve boron directly in the coolant and adjust its concentration over the fuel cycle — a slow chemical throttle alongside the mechanical one.

Two physical gifts make this controllable at all. The first is delayed neutrons. A small fraction of fission neutrons — well under one percent — are not released instantly but seconds to minutes later, as certain fission fragments decay. Reactors are designed so that the chain reaction cannot be sustained on prompt neutrons alone; it needs the delayed ones to reach criticality. That means the reactor's response time is set not by the microsecond life of a prompt neutron but by the leisurely seconds of the delayed group — slow enough for control systems, and human operators, to keep up.

The second gift is negative feedback. In a well-designed reactor, getting hotter makes the chain reaction weaker, automatically. Hot fuel absorbs more neutrons uselessly in U-238 (the Doppler effect); hot water is less dense and moderates neutrons less effectively. Both effects push power back down without anyone doing anything. US licensing requires this self-stabilizing behavior at power. Its absence — a design that, under certain conditions, got more reactive as coolant boiled away — is a core part of what made Chernobyl's RBMK design dangerous in a way Western light-water reactors are not (the safety page takes this up properly).

Shutting down is the fast direction. Every power reactor has a scram system: on a trip signal, control rods drive or drop into the core in seconds, and the chain reaction stops. What cannot be scrammed away is decay heat — the 6–7%-and-falling afterglow from fission fragments described on the previous page. A scrammed reactor is out of the chain-reaction business but still very much in the cooling business, which is why so much of a plant's safety equipment — emergency diesel generators, multiple independent cooling injection systems — exists for the hours and days after shutdown.

It's worth saying plainly, because it is one of the most common public misconceptions: a power reactor physically cannot explode like a nuclear weapon. Weapon physics requires near-pure fissile material slammed together in microseconds; power-reactor fuel is a few percent U-235, arranged to be barely critical with the help of a moderator and delayed neutrons. The worst failure modes of a power reactor are loss-of-cooling accidents — overheating, fuel damage, and in extreme cases chemical (hydrogen) explosions, as at Fukushima — serious, but a categorically different physical phenomenon.