Experiment
Keep a chain reaction steady
Split one atom's core and it splits the next. Build it wrong first: pile up uranium and watch the chain fizzle. Then make it grow, and hold it steady with control rods.
From the Short "Why Does a Lump of Uranium Just Sit There?". Coming soon on YouTube
Split one
Hit a uranium atom's core, its nucleus, with a neutron, a tiny particle. The nucleus splits in two and throws out two or three more neutrons, to split others.
Fire a neutron and follow the chain. How many neutrons come out of each split?
Pile up uranium
Real uranium is a mix. Over 99 in 100 of its nuclei are a kind that rarely splits (grey). Each round of splits is called a generation. New neutrons come out fast, at around 20,000 km a second.
Fire a neutron into one lump. Then try a big pile. Can you keep a chain going?
How much uranium
Slow them down
Now the uranium is in fuel rods. Add more of the kind that splits. Then, oddly, slow the neutrons down: bouncing off water, they lose speed, and a slow one splits that kind hundreds of times more easily.
Get the chain to grow. Try each change on its own too.
The kind that splits
Around the rods
Hold it steady
Now the chain grows. Control rods soak up neutrons. Slide them in until each split causes just one more, on average. That's called critical.
Hold the chain steady for 5 generations in a row.
The real trick: heat is speed
Back to one split, slowed right down. Both pieces carry positive charge, so they push each other apart hard. One split gives about 50 million times the heat of burning one carbon atom.
Split it and follow the two pieces. Where does their speed go?
What's going on
Strictly, it's the nucleus, the tiny core of the atom, that splits. A uranium-235 nucleus that takes in a neutron (a particle with no electric charge) splits into two smaller pieces of unequal size and throws out about 2.4 new neutrons on average, so two or three. If, on average, more than one of those goes on to cause another split, the chain grows. Fewer than one, and it dies away.
Natural uranium is about 99.3% uranium-238 and 0.7% uranium-235. Uranium-238 rarely splits, and it readily swallows neutrons of middling speed. New neutrons come out fast, at around 20,000 km a second. So a lump of natural uranium on its own can't keep a chain going at any size: too many neutrons are swallowed or escape. It just sits there, only very slightly warm from ordinary radioactive decay.
Most power reactors use uranium enriched to about 3 to 5% uranium-235, and slow the neutrons down. Slow neutrons are hundreds of times more likely to split uranium-235 than fast ones. In ordinary water they bounce off hydrogen nuclei, which weigh almost the same as a neutron, so each bounce takes a big share of the speed: about 18 to 20 bounces bring one down to about 2 km a second. A material that slows neutrons is called a moderator. Some designs run on natural uranium by slowing the neutrons with heavy water or graphite instead of ordinary water.
Control rods hold materials that swallow neutrons, such as boron, cadmium or hafnium; many reactors also dissolve boron in the water. "Critical" means each split leads, on average, to exactly one more: the steady state a reactor runs in all the time. Growing is called supercritical, and dying away subcritical.
The heat is mostly speed. Of the energy from each split, around 80 to 85% is the speed of the two pieces, which fly apart because both carry positive charge and repel each other. They start at around 10,000 to 14,000 km a second and stop within roughly 5 to 10 thousandths of a millimetre, so their energy turns into heat right where the split happened. The rest comes from the fast neutrons, from gamma rays and from the later radioactive decay of the pieces; a little is carried off by neutrinos and lost. Atom for atom, one split gives about 50 million times the heat of burning one carbon atom. In most reactors the same water that slows the neutrons carries the heat away to make steam, which spins a turbine.
Simplified: time is slowed down hugely, and nuclei are drawn huge and few. The drawing shows far more of the kind that splits than the real 0.7% or 4%. Each neutron's chances (split a nucleus, get swallowed, escape, or be soaked up by a control rod) are a rough guide set so the chain behaves as the Short describes, not reactor figures, and every swallowed neutron is drawn going into the kind that rarely splits. The water slows neutrons in fewer bounces than the real 18 or so. When the chain is big, only some neutrons are drawn and the count is worked out from the average; the count stops at a million to fit the chart. Split nuclei are put back so the fuel never runs out. Left out: a small share of neutrons come out seconds late, which is what makes a real reactor's chain slow enough to control, and in water-cooled reactors hotter water slows fewer neutrons, so the chain eases off by itself. In step 5 the speeds come from each piece's share of the energy and its mass, and the slowing down is a smooth rough curve.