residue theorem
The residue theorem is the grand unifying payoff of contour integration: to integrate a function around a closed loop, you do not need to trace the curve at all — you just find the singularities trapped inside, compute one number (the residue) at each, add them up, and multiply by 2pi i. The entire global integral collapses into a tiny local bookkeeping exercise. It turns calculus that would be hopeless on the real line into routine arithmetic.
Precisely: let f be holomorphic on an open set except for isolated singularities, and let gamma be a positively-oriented simple closed contour enclosing finitely many of them and passing through none. Then the contour integral of f over gamma equals 2pi i times the sum of the residues of f at the singularities inside gamma. (More generally, each residue is weighted by the winding number of gamma about that singularity.) Cauchy's integral theorem is the special case with no enclosed singularities, giving 0.
Its reach is enormous. It evaluates real definite integrals like the integral of 1/(1 + x^2) over the whole line, sums series such as sum 1/n^2, computes Fourier and Laplace transforms, and underlies the argument principle that counts zeros minus poles. The essential discipline is correct accounting: include EVERY singularity inside the contour and EXCLUDE every one outside, watch the orientation, and remember that a singularity sitting exactly on the contour makes the theorem inapplicable as stated.
To integrate 1/(z^2 + 1) counterclockwise around the circle |z| = 2: the integrand has simple poles at z = i and z = -i, both inside. Residue at i is 1/(2i), residue at -i is -1/(2i); they sum to 0, so the integral is 2pi i times 0 = 0. The same machinery, using only the upper pole and a semicircle, yields the real integral of 1/(x^2+1) over the whole line equals pi.
Summing residues to integrate, and to crack a real integral.