A function that outgrows its own definition
In the previous guide you met the Riemann zeta function through its two faces for Re s > 1: the sum zeta(s) = sum over n of 1/n^s and the Euler product over the primes. But that whole story lives in a single half-plane. The sum diverges the instant Re s drops to 1 (there it is the harmonic series), and the product needs Re s > 1 to converge too. So as a formula, zeta is trapped to the right of the line Re s = 1. The astonishing claim of this rung is that zeta is really defined almost everywhere — on the whole plane except the single point s = 1 — and the two formulas were only ever a keyhole view of it.
How can a function have meaning where its only formula breaks down? This is the central magic of complex analysis, and you have already seen the engine in the analytic continuation guide back in the Continuation rung. The idea is jigsaw puzzle pieces. A holomorphic function known on one patch — even a tiny disk — can sometimes be extended to a neighbouring patch so that it stays holomorphic and agrees on the overlap. Fit enough pieces and the original keyhole view grows into a portrait covering a vast region. Zeta is the most celebrated example of a function that lives this way.
Warm-up: how gamma walks left across the plane
Before tackling zeta, recall how the gamma function from the first two guides in this rung escaped its own half-plane — it is the cleanest model of the whole trick. Euler's integral Gamma(z) = integral from 0 to infinity of t^(z-1) e^(-t) dt converges only for Re z > 0. Yet we know Gamma lives on the whole plane as a meromorphic function, with simple poles at 0, -1, -2, ... and nowhere else. What carried it leftward was not a cleverer integral but a single algebraic identity: the functional equation Gamma(z+1) = z Gamma(z).
Read that equation backwards — Gamma(z) = Gamma(z+1)/z — and it becomes a continuation machine. The right-hand side makes sense whenever z+1 sits in the known region, that is for Re z > -1 (with z = 0 excluded). So one rearrangement extends Gamma one full strip to the left, and the apparent blow-up at z = 0 is revealed as exactly a simple pole. Apply it again, Gamma(z) = Gamma(z+2)/(z(z+1)), and you reach Re z > -2, exposing the pole at z = -1. Each step both extends the domain and explains a pole. This is continuation by an identity, the gentlest version of what zeta will need.
Continuing zeta past the wall at Re s = 1
Zeta needs the same spirit but a sharper tool, because its sum does not satisfy a one-line shift identity like gamma's. The first step out of the half-plane is a small miracle. Form the alternating cousin eta(s) = sum over n of (-1)^(n-1)/n^s, called the Dirichlet eta function. Its terms partly cancel, so this sum converges in the much larger half-plane Re s > 0. A short manipulation — pull out the even terms — gives the bridge eta(s) = (1 - 2^(1 - s)) zeta(s). Solve for zeta and you have a formula valid for Re s > 0 (except where 1 - 2^(1-s) = 0).
The manipulation is a one-line subtraction worth seeing. The even-indexed terms of zeta are sum over even n of 1/n^s = 2^(-s) sum over m of 1/m^s = 2^(-s) zeta(s). Subtracting twice this even part from the full sum flips the sign on the evens, which is exactly eta: zeta(s) - 2 * 2^(-s) zeta(s) = eta(s), that is eta(s) = (1 - 2^(1-s)) zeta(s). Rearranged, zeta(s) = eta(s)/(1 - 2^(1-s)) — and the right-hand side now makes sense for Re s > 0, because eta converges there.
This single bridge already does enormous work. It continues zeta to the whole half-plane Re s > 0 minus the trouble spots, and it isolates the only pole zeta will ever have. The factor 1 - 2^(1-s) vanishes at s = 1, where eta(s) is finite (eta(1) = log 2), so zeta blows up there — a simple pole at s = 1 with residue 1. The other zeros of 1 - 2^(1-s) at s = 1 + 2 pi i k / log 2 are not poles of zeta: there eta also vanishes, and the apparent singularities cancel, leaving zeta perfectly finite. Always check whether a factor's zero is a genuine pole or a cancelling one; here all but s = 1 cancel.
Riemann's functional equation: folding the plane in half
The eta bridge reaches Re s > 0, but the left half of the plane still seems unreachable. The instrument that conquers it all at once is the functional equation, and it is far more than a continuation device — it is a deep symmetry. It relates the value of zeta at s to its value at the mirror point 1 - s, reflected across the central line Re s = 1/2. The cleanest form uses the completed zeta xi(s) = pi^(-s/2) Gamma(s/2) zeta(s), built by gluing a gamma factor onto zeta. Then the whole statement is simply xi(s) = xi(1 - s): the completed function is unchanged by the reflection s -> 1 - s.
xi(s) = pi^(-s/2) * Gamma(s/2) * zeta(s)
xi(s) = xi(1 - s) (symmetric form)
zeta(s) = 2^s pi^(s-1) sin(pi s/2) Gamma(1-s) zeta(1-s) (unsymmetric form)Notice that the gamma function is not optional decoration. Riemann's proof runs through a theta function theta(t) = sum over n of e^(-pi n^2 t), which obeys its own beautiful transformation theta(1/t) = sqrt(t) theta(t). Feeding theta into a gamma-type (Mellin) integral manufactures exactly xi(s), and the theta symmetry t -> 1/t becomes the reflection s -> 1 - s. So zeta's s-to-(1-s) symmetry is the analytic shadow of a theta function's t-to-(1/t) symmetry — number theory tied to the geometry of lattices through the gamma function you studied first.
Now the continuation is complete in a single sweep. Given any s with Re s < 0, its mirror 1 - s has Re(1-s) > 1, where the ordinary sum defines zeta(1-s) perfectly. The unsymmetric form zeta(s) = 2^s pi^(s-1) sin(pi s/2) Gamma(1-s) zeta(1-s) then hands you zeta(s) directly. The whole left half-plane is filled in by reflecting known values from the right. Two formulas — the eta bridge for the middle band and the functional equation for the far left — together cover the entire plane minus the lone pole at s = 1.
Reading the consequences off the equation
The functional equation does not just continue zeta — it tells you exactly what the continued function looks like in the left half-plane. Look at the unsymmetric form and watch the factor sin(pi s/2). It vanishes at the even integers, and at s = -2, -4, -6, ... none of the other factors blows up to rescue the value, so zeta itself vanishes there. These are the trivial zeros of zeta: an infinite ladder of zeros marching out along the negative real axis, handed to you for free by the sine in the equation.
- Compute a left-plane value from a right-plane one. To find zeta(-1), set s = -1 so 1 - s = 2, and read the right side: 2^(-1) pi^(-2) sin(-pi/2) Gamma(2) zeta(2).
- Plug in the known pieces: sin(-pi/2) = -1, Gamma(2) = 1! = 1, and the famous zeta(2) = pi^2/6.
- Multiply through: (1/2) * pi^(-2) * (-1) * 1 * (pi^2/6) = -1/12. So zeta(-1) = -1/12.
- Read the meaning honestly: -1/12 is the VALUE of the continued zeta at s = -1. The divergent sum 1 + 2 + 3 + ... does not 'equal' -1/12; that slogan is shorthand for zeta(-1), the analytic continuation evaluated off the region where the sum converges.
With the trivial zeros and the lone pole accounted for, everything genuinely mysterious is squeezed into the vertical band 0 < Re s < 1, the critical strip. The equation's s <-> 1 - s symmetry means any other zeros come in pairs reflected across the central line Re s = 1/2 — the critical line. The next and final guide in this rung follows those zeros, reconnects them through the Euler product to the primes, and arrives at the conjecture that every one of them sits exactly on that line. That conjecture, the Riemann hypothesis, remains OPEN to this day.