Why the factorial wants to grow up
The factorial is the most innocent function in mathematics: 4! = 4 * 3 * 2 * 1 = 24, and that is the whole story. But it has a frustrating limitation — it only speaks to whole numbers. There is no obvious meaning to (1/2)! or (3.7)!, let alone (i)!. You might shrug and say these are nonsense questions, but the entire toolkit you have built on the earlier rungs — power series, contour integrals, analytic continuation — exists precisely to chase a discrete pattern out into the continuous plane and discover it had a soul all along. The factorial is begging to be such a function.
The trouble is that infinitely many smooth curves pass through the dots (1, 1), (2, 2), (3, 6), (4, 24), and so on. You could bend a curve through them in countless ways, so 'the' factorial in between is not unique on cleverness alone. The miracle is that one of those interpolations is forced on us by a single extra demand of regularity, and once you find it, every classical formula about factorials reappears, deepened. That distinguished interpolation is the gamma function, written Gamma(z), and the convention it obeys is Gamma(n) = (n - 1)! — shifted by one, for historical reasons we will simply live with.
Euler's integral: a factory that runs on real numbers
Here is Euler's stroke of genius. Instead of trying to draw a curve through the dots, he found an integral whose value already equals (n - 1)! at the integers but makes perfect sense for any z in between. This is the Euler integral for gamma: Gamma(z) = integral from 0 to infinity of t^(z-1) e^(-t) dt. The integrand is a tug-of-war — the power t^(z-1) pushes the value up as t grows, while the decay e^(-t) eventually crushes everything to zero — and the integral measures who wins over the whole positive axis.
Gamma(z) = integral from 0 to infinity of t^(z-1) e^(-t) dt ( valid for Re z > 0 )
check the integers:
Gamma(1) = integral of e^(-t) dt = 1 = 0!
Gamma(2) = integral of t e^(-t) dt = 1 = 1!
Gamma(3) = integral of t^2 e^(-t) dt = 2 = 2!
one integration by parts on integral of t^z e^(-t) dt gives
Gamma(z + 1) = z * Gamma(z)But there is honest fine print, and it matters. This integral only converges when the real part of z is positive — written Re z > 0. Near the lower limit t = 0 the integrand behaves like t^(z-1), and that has a finite integral only when Re z > 0; push z to the left of the imaginary axis and the integral diverges at zero. So Euler's formula is not yet a function on the whole plane: it is a perfectly good holomorphic function on the right half-plane, and nothing more. Extending it past that wall is the job of the next section — and it is exactly the kind of job analytic continuation was invented for.
The recurrence: a factorial in disguise
The single most important property of gamma falls out of one integration by parts, and it is worth doing in your head once. Integrate Gamma(z + 1) = integral of t^z e^(-t) dt by parts, differentiating t^z and integrating e^(-t). The boundary term vanishes at both ends (t^z is killed at 0, and e^(-t) is killed at infinity), and what survives is z times the original integral. The result is the gamma functional equation: Gamma(z + 1) = z * Gamma(z). This is the engine of the whole subject.
Stare at that recurrence and the factorial leaps out. It says: to go up by one in the argument, multiply by the current number. That is exactly what the factorial does — (n+1)! = (n+1) * n!. Starting from the known value Gamma(1) = 1 and applying the rule again and again gives Gamma(2) = 1 * Gamma(1) = 1, then Gamma(3) = 2 * Gamma(2) = 2, then Gamma(4) = 3 * Gamma(3) = 6, marching out the factorials one step at a time. Gamma is not merely a curve through the factorial dots; it carries the factorial's defining rule in its bones, now for every z, not just the integers.
The recurrence also lets you compute genuinely new values, the ones the plain factorial could never reach. The crown jewel is Gamma(1/2) = sqrt(pi), a fact you can prove by turning the Euler integral into a Gaussian integral. From it the recurrence spins off a whole ladder: Gamma(3/2) = (1/2) Gamma(1/2) = sqrt(pi)/2, then Gamma(5/2) = (3/2) Gamma(3/2) = 3 sqrt(pi)/4, and so on. In everyday language, (1/2)! = Gamma(3/2) = sqrt(pi)/2 — the factorial of one-half is half the square root of pi. That a question about half-integers should answer itself with pi is the first sign that gamma is doing something far deeper than interpolation.
Off the wall: continuation onto the whole plane
The Euler integral stops at the wall Re z > 0, but the recurrence does not care about walls — and that is the key that unlocks the rest of the plane. Rearrange Gamma(z + 1) = z * Gamma(z) into Gamma(z) = Gamma(z + 1) / z. The right-hand side makes sense for Re z > -1 (everywhere except z = 0), because the numerator only needs z + 1 to sit in the original half-plane. So this rearranged formula defines gamma one full unit further to the left, and it must agree with Euler's integral wherever both are defined, since two holomorphic functions that match on an overlap are forced equal by the uniqueness of analytic continuation.
- Start with Euler's integral, which defines Gamma(z) on the right half-plane Re z > 0.
- Use Gamma(z) = Gamma(z + 1) / z to extend one strip to the left, reaching Re z > -1 except the new gap at z = 0.
- Apply the same step again to reach Re z > -2 except z = 0, -1, then again for z = 0, -1, -2, and so on indefinitely.
- The result is a single function on the whole plane minus the points z = 0, -1, -2, -3, ... where divisions by zero have piled up.
Now look at what the leftover points cost. At z = 0 the formula divides Gamma(1) = 1 by z, so gamma blows up like 1/z — a simple pole. Step left again and z = -1 inherits a pole from the division by (z + 1), and so on: gamma has a simple pole at every non-positive integer 0, -1, -2, -3, ... and nowhere else. A function whose only singularities are poles is a meromorphic function, so the finished gamma is meromorphic on the whole plane. Crucially it has no zeros at all — a fact that becomes a star player when this story reaches the Riemann zeta function later in the rung.
Which interpolation, and what comes next
Return to the worry that infinitely many curves pass through the factorial dots. What pins down gamma as 'the' one? The honest answer is that you must impose an extra regularity condition, and the cleanest is the Bohr-Mollerup theorem: among all functions that satisfy the recurrence f(z + 1) = z f(z), start at f(1) = 1, and are logarithmically convex on the positive axis, gamma is the unique one. Drop the convexity and other interpolations sneak back in. So gamma is not magically forced by the dots alone — it is forced by the dots plus one reasonable smoothness demand, and that is the honest statement.
There is a second face of gamma you should know exists, because the rest of the rung leans on it. Instead of an integral, you can write gamma as an infinite product over all its poles at once — Weierstrass's product, built from the elementary factors you met when constructing functions with prescribed zeros. That product form makes the absence of zeros obvious (a product of never-zero factors is never zero), exposes the famous Euler-Mascheroni constant, and is the natural launchpad for the reflection formula Gamma(z) Gamma(1 - z) = pi / sin(pi z) and for Stirling's asymptotics — the two headline results of the very next guide.
Step back and see what you have. A childhood operation that only knew whole numbers has become a meromorphic function on the entire plane, carrying the factorial's recurrence everywhere, with simple poles precisely at the non-positive integers and no zeros anywhere. Every move that built it — an integral that converges on a half-plane, a functional equation that extends it, the rigid uniqueness of continuation — was a tool you already owned. The gamma function is the first great reward for climbing this far, and the doorway to the zeta function and the primes that the rest of this rung will open.