Okay, some of you who are experts might have noticed that I simplified things quite a bit here. We need to talk about something called "reactivity," which can be expressed in a lot of different ways; we're going to discuss just one of them.
There's a very general factor in reactor physics called the effective neutron multiplication factor, or k. This is the number of neutrons from one fission that directly cause another fission. When k=1, the reaction is self-sustaining, or "critical," and power output is steady and stable. If we want to increase power, we simply increase k very, very slightly beyond critical, and power will increase until we bring k back to 1. Likewise, we can reduce k slightly below 1 to begin reducing power, bringing k back to 1 when we're at the power level we want. Expressing reactivity in terms of k gives us a good reference point at criticality.
Of course, it can't be that simple, and it's not. As it turns out, there are two ways for a neutron to be released in a nuclear fission chain reaction, and we give neutrons different names based on their origins.
Prompt Neutrons are neutrons which result directly from an atom of U-235 splitting into fragments. These neutrons are released immediately upon the collision that caused the fission. Because of this quick action, if a reactor were to go critical from only the fissions caused by prompt neutrons, it would be impossible for humans or even computers to control. The tiniest change of reactivity would cause an incredibly rapid change in power level.
Delayed Neutrons are released when an atom of U-235 splits into pieces which are themselves unstable. At some point soon after the fission takes place, these unstable fission products spontaneously decay, releasing neutrons. Because there is a short period of time between the splitting of the original atom and the release of the delayed neutrons, changes in reactivity cause a more gradual change in power. In fact, the reactor responds four or five orders of magnitude more slowly to reactivity changes on delayed neutrons.
The ratio of prompt neutrons to delayed neutrons is expressed as β, and the relationship between k, β, and criticality is divided into ranges.
When k < 1, the reactor is subcritical.
When k = 1, the reactor is delayed critical.
When 1 < k < 1/(1-β), the reactor is delayed supercritical.
When k > 1/(1-β), the reactor is prompt supercritical.
Power reactors operate in the delayed critical to delayed supercritical range of reactivities. Nuclear weapons operate in the prompt supercritical range. Fuel in a nuclear power reactor is not sufficiently enriched to form a critical mass and cannot explode in the way that a nuclear weapon does, but it can run away and produce so much heat that it melts its own fuel; this is called a meltdown.
In the odd vernacular of nuclear reactor engineers, the range of reactivity between k=1 and k=1/(1-β) is divided into "dollars" and "cents." One dollar is the amount of reactivity required to go from critical to prompt critical. Go over a dollar and things get ugly.