Definition
An invariant is any property that does not change when a system is subjected to a specified class of transformations. The key word is "specified": an invariant is always defined relative to a particular set of operations. What remains fixed under one transformation may change under another. This relativity is not a weakness — it is the concept's power. By naming exactly which properties survive which operations, invariants reveal the essential structure that underlies apparently diverse situations.
In geometry, the area of a figure is invariant under rigid motions (translation, rotation, reflection) but not under scaling. The length of a vector is invariant under rotation but not under stretching. The cross-ratio of four points on a line is invariant under projective transformations — and this fact turns out to characterize the whole of projective geometry. Felix Klein's Erlangen programme (1872) used exactly this move: define a geometry by its transformation group, then study the invariants of that group. Every classical geometry is an invariant theory in disguise.
In physics, conservation laws are invariant quantities. Energy is conserved because the laws of physics are invariant under time translation (Noether's theorem). Momentum is conserved because they are invariant under spatial translation. The deep connection between symmetry and conservation is one of the most productive discoveries in the history of science: to find what is conserved, look for what the laws are invariant under.
Why it matters
How it works
Invariants as problem-solving tools
Many mathematical competition problems are solved by finding an invariant. A sequence of moves is applied repeatedly; the question is whether a certain state can be reached from a given starting state. If a quantity can be shown to remain the same under each individual move, then any reachable state must share that quantity with the starting state. If the target state has a different value of the quantity, it is unreachable — without examining every possible sequence of moves. The invariant does the work of ruling out a vast space of possibilities at once.
This heuristic generalizes. Whenever a complex process unfolds through many steps, asking 'what does each step preserve?' can collapse the analysis. If you find something genuinely conserved, it constrains every intermediate state and the final state simultaneously.
Loop invariants and program correctness
In computer science, a loop invariant is a logical assertion that is true before the loop begins, remains true after each execution of the loop body, and — crucially — is useful enough that when the loop terminates it implies the desired output condition. Proving a loop invariant is one of the principal techniques for establishing that an algorithm is correct, not merely that it runs without crashing. The technique transforms a question about the behavior of an iterating process into a question about a single static condition.