How fission reactors work

How a fission reactor sustains a chain reaction, what the moderator, coolant and control rods do, why delayed neutrons make control possible, where decay heat comes from, and what passive safety does and does not mean.

Guide · Updated 2026-09-23

Every nuclear power station operating today runs on fission: the splitting of a heavy atomic nucleus into two lighter ones. This guide explains what happens inside the reactor, what each major part does, and which physical facts decide whether a safety claim holds up.

Fission

A uranium-235 nucleus that absorbs a neutron becomes unstable and usually splits into two fragments of unequal size. The fragments fly apart at high speed and stop within a fraction of a millimetre of the fuel, turning their energy into heat. One fission of uranium-235 releases about 193 MeV of energy and, on average, about 2.4 new neutrons. Plutonium-239 releases slightly more energy per fission and about 2.9 neutrons.

Almost all of that energy ends up as heat in the fuel. A reactor is, in engineering terms, a way of making heat in a controlled place and carrying it away to a turbine.

The chain reaction

The new neutrons from each fission can go on to cause more fissions. Reactor physicists track this with the effective multiplication factor, written k: the number of fissions in one generation divided by the number in the generation before.

If k is below 1, the reaction dies away. The reactor is subcritical.
If k equals 1, each generation is the same size and power holds steady. The reactor is critical.
If k is above 1, the population grows. The reactor is supercritical, which is how an operator raises power.

“Critical” is the normal operating state of every power reactor. News coverage often uses “went critical” to mean an accident. It does not. A reactor reaching criticality for the first time is a commissioning milestone, and a reactor at steady power is critical all the time.

The moderator

Neutrons are born fast, with energies of a few million electronvolts. Uranium-235 is far more likely to fission when struck by a slow neutron than a fast one. Most reactors therefore slow their neutrons down on purpose, using a moderator: a material with light nuclei that the neutrons bounce off, losing energy with each collision.

Ordinary water, whose hydrogen nuclei are almost the same mass as a neutron, slows neutrons in the fewest collisions but also absorbs some of them.
Heavy water, in which the hydrogen is deuterium, absorbs far fewer neutrons. That is why a heavy-water reactor can run on natural, unenriched uranium.
Graphite, which is carbon, absorbs few neutrons but needs more collisions, so graphite cores are large.

A reactor whose neutrons are slowed this way is a thermal reactor. A fast reactor has no moderator and uses the neutrons at close to the energy they were born with. Fast reactors need fuel with a higher share of fissile material, but they can convert more uranium-238 into plutonium and can fission heavier elements that thermal reactors cannot. Spectrum decides a great deal downstream, from fuel enrichment to the damage done to structural materials, so a claim made about one kind of reactor should not be transferred to the other.

The coolant

The coolant carries heat out of the core. In the most common design, the pressurised water reactor, ordinary water does both jobs at once: it is the moderator and the coolant. Other designs separate the two, or drop the moderator altogether. Coolants in use or in development include ordinary water, heavy water, helium, carbon dioxide, liquid sodium, liquid lead and molten salt. The reactor types guide goes through each.

Control

Operators change k by adding or removing materials that absorb neutrons.

Control rods contain strong neutron absorbers such as boron or hafnium. Inserting them lowers k; withdrawing them raises it. A full insertion to shut the reactor down is called a scram or trip.
Soluble boron is dissolved in the coolant of pressurised water reactors and adjusted slowly as the fuel is used up.
Control drums, rotating cylinders with an absorber on one face, are used in several small reactor designs in place of rods.

Delayed neutrons

If every neutron appeared the instant its fission happened, a reactor would be almost impossible to control. In a thermal reactor the time from one neutron generation to the next is of the order of a ten-thousandth of a second. A small excess in k would multiply power many times over in under a second.

What makes control possible is that a small fraction of neutrons is not released at the moment of fission. Some fission fragments decay radioactively a little later and emit a neutron as they do. These delayed neutrons arrive between a fraction of a second and about a minute after the fission; the six standard groups have half-lives from 0.23 to 55.7 seconds. For uranium-235 the delayed fraction, written β, is about 0.0064, or 0.64% of all fission neutrons.

A reactor is run so that it needs those delayed neutrons to stay critical. That stretches the effective time between generations from about a ten-thousandth of a second to around a tenth of a second: 0.64% of neutrons, delayed by an average of about 13 seconds, add about 0.08 seconds to the average. As a result power changes on a timescale of seconds to minutes that rods and operators can follow. The margin matters: a reactor that became critical on its prompt neutrons alone, called prompt critical, would lose that slowing effect. Reactivity is often quoted in “dollars” for this reason, where one dollar equals β.

Plutonium-239 releases a markedly smaller delayed fraction than uranium-235, so a core fuelled heavily with plutonium has a smaller margin to prompt criticality. This is one reason core designs differ between fuels.

Feedback: temperature and void coefficients

A well-designed reactor pushes back when it heats up. The reactivity coefficients say how k changes when the fuel gets hotter, when the moderator gets hotter, or when bubbles or voids form in the coolant.

In uranium fuel, hotter fuel absorbs more neutrons in uranium-238 without fissioning, which lowers k. This is the Doppler effect, and it acts within the fuel almost instantly.
In light-water reactors, water boiling away removes moderator, so k falls. The void coefficient is negative: if the water boils or is lost, power drops.
The RBMK reactors at Chernobyl had a large positive void coefficient. Boiling raised reactivity, which raised power, which caused more boiling. That feedback contributed to the 1986 accident.
A CANDU heavy-water reactor has a positive void coefficient that is small enough for its control systems to handle. Large sodium-cooled fast reactors can also have a positive sodium void coefficient, which designers offset with core shape and other feedbacks.

A negative coefficient is not automatic. It depends on the design, and on where the core is in its fuel cycle. A claim that a reactor is “inherently safe” rests on the actual numbers, and a reader is entitled to ask for them.

Decay heat

Shutting a reactor down stops the chain reaction within seconds. It does not stop the heat. The fission fragments are radioactive, and their decay keeps producing heat for days, months and years.

At the moment of shutdown, decay heat is about 6.5% of the power the reactor was running at. It falls to roughly 1 to 1.5% after an hour, depending on how long the reactor had been running, 0.4% after a day and 0.2% after a week. For a reactor producing 3,000 MW of heat, a size typical of a large power station, that is:

Time after shutdown	Share of previous power	Heat to remove
At shutdown	about 6.5%	about 195 MW
1 hour	about 1 to 1.5%	about 30 to 45 MW
1 day	about 0.4%	about 12 MW
1 week	about 0.2%	about 6 MW

The third column is 3,000 MW multiplied by the second. The exact figures depend on how long the reactor had been running and what fuel it used. The removal of this heat, not the chain reaction, was the problem at Three Mile Island in 1979 and at Fukushima Daiichi in 2011: in both cases the reactors had shut down, and it was the loss of cooling afterwards that damaged the fuel.

Decay heat scales with reactor power. A 15 MW thermal microreactor has about one two-hundredth of the decay heat of a 3,000 MW core, which is why small reactors can more plausibly lose that heat to their surroundings without pumps.

Passive safety

A passive safety feature performs its job using physics that needs no outside power, no pumps and no operator action: gravity, natural circulation of hot coolant rising and cool coolant sinking, conduction of heat through solid material, or thermal expansion that pushes k down as the core warms.

An example of passive shutdown was demonstrated in April 1986 at EBR-II, a sodium-cooled fast reactor in Idaho. Operators switched off the primary coolant pumps with the reactor at full power and did not insert the control rods. Power fell to near zero within about 300 seconds, driven by the expansion of the hot fuel and core structure, and the fuel was not damaged.

Passive safety has limits that the phrase can hide.

It covers a stated set of events. A design is analysed against particular accidents, such as loss of power or loss of coolant flow. A claim of passive safety is a claim about those events, not about every possible one.
It usually covers a stated time. Many designs claim a period, such as three days, during which no action is needed. After that, someone must refill a water pool or restore power.
It depends on the numbers. Passive shutdown depends on negative reactivity coefficients. Passive cooling depends on the decay heat being small enough, and the heat path good enough, to keep the fuel below its damage temperature.
It does not remove the need for regulation, maintenance or security. A passively safe reactor still holds radioactive material that must be contained.

When a company says its reactor is “walk-away safe”, the questions to ask are: for which events, for how long, and what are the reactivity coefficients and the decay heat removal path that support it. The dataset records these claims as the company’s own words and marks them as claims.

Sources

Fission · technology
Water-cooled · technology
Pressurised water reactor · technology
Low-enriched uranium · glossary