Get AI summaries of any video or article — Sign up free
But what is the Riemann zeta function? Visualizing analytic continuation thumbnail

But what is the Riemann zeta function? Visualizing analytic continuation

3Blue1Brown·
5 min read

Based on 3Blue1Brown's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.

TL;DR

The zeta function is initially defined by ζ(s)=∑_{n=1}^∞ 1/n^s only for Re(s)>1, where the series converges.

Briefing

The Riemann zeta function becomes understandable once its “analytic continuation” is treated as a rigid, geometry-driven extension: start with a complex-valued function defined by an infinite series only where it converges (real part of s greater than 1), then extend it into the rest of the complex plane in exactly one way that preserves angles. That uniqueness is what makes the famous values—like ζ(−1) = −1/12—and the location of its zeros meaningful, including the million-dollar Riemann hypothesis about where the non-trivial zeros lie.

For real inputs, the zeta function is initially defined by the series ζ(s) = Σ_{n=1}^∞ 1/n^s, which converges only when Re(s) > 1. When s = 2, the sum becomes 1 + 1/4 + 1/9 + … and approaches π^2/6, illustrating why the function is tied to deep constants. But plugging in negative integers breaks the series: ζ(−1) would formally correspond to 1 + 2 + 3 + …, which diverges, and ζ(−2) would correspond to 1 + 4 + 9 + …, which also diverges—so the “trivial zeros” at negative even numbers cannot be read off from the original series.

The bridge to the full function is complex analysis. When s is complex (say s = 2 + i), each term n^{−s} splits into n^{−Re(s)} times a rotation coming from n^{−i·Im(s)}. That rotation has magnitude 1, so it doesn’t blow up the term sizes; it only turns them around in the complex plane. As a result, the series still converges for Re(s) > 1, but the partial sums trace a spiral-like path toward a complex value. Visually, varying s across the right half-plane produces a transformed “grid” of inputs mapped to outputs elsewhere in the complex plane.

The key geometric constraint is analyticity. Complex functions that have derivatives everywhere preserve the angles between intersecting curves—an angle-preserving property that can be pictured by how a grid’s lines meet after transformation. Because this constraint is so strong, there is essentially only one analytic way to extend a function beyond where its original series definition works. That “jigsaw puzzle” logic is analytic continuation: extend the zeta function into the left half-plane by demanding that the extension remain analytic, which forces a unique answer.

Once extended, the zeta function has zeros at negative even integers (the trivial zeros) and additional “non-trivial zeros” inside the critical strip. Riemann’s hypothesis claims every non-trivial zero lies on the critical line where Re(s) = 1/2. The Clay Math Institute’s million-dollar prize is tied to proving this placement. The same analytic continuation also yields ζ(−1) = −1/12, giving the striking formal identity 1 + 2 + 3 + … = −1/12—despite the fact that the original series diverges—because the extension is uniquely determined by the convergent region. The upshot: the zeta function’s extended values aren’t arbitrary “regularizations”; they are the only angle-preserving continuation consistent with the convergent series.

Cornell Notes

The Riemann zeta function starts as a convergent series ζ(s)=∑_{n=1}^∞ 1/n^s only when Re(s)>1. For complex s, the imaginary part doesn’t change term magnitudes; it rotates each term in the complex plane, so the series still converges and the sums spiral to a complex value. To define ζ(s) where Re(s)≤1, mathematicians use analytic continuation: extend the function in the only way that stays analytic (angle-preserving for intersecting curves). This uniqueness explains how ζ(−1)=−1/12 and why ζ has trivial zeros at negative even integers, even though the original series diverges there. The same continuation produces the non-trivial zeros in the critical strip, where the Riemann hypothesis predicts they all lie on Re(s)=1/2.

Why does the series definition of ζ(s) fail for negative integers, and what replaces it?

The series ζ(s)=∑_{n=1}^∞ 1/n^s converges only when Re(s)>1. At negative integers, terms grow instead of shrinking (e.g., for s=−1 the terms become 1/n^{−1}=n, giving 1+2+3+… which diverges). The full zeta function on the left half-plane is defined instead by analytic continuation: extend the function from the convergent region to Re(s)≤1 in a way that remains analytic (has derivatives everywhere).

How does a complex exponent turn into a “rotation” rather than a blow-up?

For s=a+bi, n^{−s}=n^{−a}·n^{−bi}. The factor n^{−a} controls size, while n^{−bi} lies on the unit circle because it has absolute value 1. As b changes, n^{−bi} rotates around the unit circle, so the series can converge for Re(s)=a>1 even though the terms acquire complex phases.

What does “analytic” mean geometrically, and why does it force a unique continuation?

Analytic functions preserve angles between intersecting curves. If two lines intersect in the input plane, their images under an analytic function intersect with the same angle (with a caveat that at points where the derivative is zero, angles can be multiplied by an integer). Because this angle-preserving constraint is extremely restrictive, extending an analytic function beyond its original domain becomes a rigid “jigsaw puzzle” with only one compatible solution.

How do trivial zeros and ζ(−1)=−1/12 arise despite divergence of the original series?

Trivial zeros occur at negative even integers (e.g., ζ(−2)=0), and ζ(−1)=−1/12 holds for the analytically continued function. The original series diverges at these inputs, so the equalities cannot come from term-by-term summation. Instead, they follow from the unique analytic continuation from the region Re(s)>1, which determines the function’s values everywhere it can be extended analytically.

What is the critical strip and what does the Riemann hypothesis claim about zeros?

The non-trivial zeros lie in the critical strip, a vertical region where the extended zeta function equals zero beyond the trivial zeros at negative even integers. Riemann hypothesized that all non-trivial zeros lie on the critical line Re(s)=1/2. Proving this would confirm a deep pattern tied to prime numbers and unlock many results dependent on the hypothesis.

Review Questions

  1. What role does the condition Re(s)>1 play in the convergence of ζ(s), and what changes when s is complex?
  2. Explain analytic continuation using the angle-preserving property of analytic functions. Why does this make the extension essentially unique?
  3. How can ζ(−1)=−1/12 be consistent with the divergence of 1+2+3+…? What mechanism produces the value?

Key Points

  1. 1

    The zeta function is initially defined by ζ(s)=∑_{n=1}^∞ 1/n^s only for Re(s)>1, where the series converges.

  2. 2

    For complex s=a+bi, the imaginary part contributes only rotations (unit-modulus factors), so convergence depends mainly on a=Re(s).

  3. 3

    Analytic functions preserve angles between intersecting curves, providing a geometric intuition for having derivatives everywhere.

  4. 4

    Requiring the extension to remain analytic turns analytic continuation into a rigid, essentially unique “jigsaw puzzle” beyond the series’ convergence region.

  5. 5

    The analytically continued zeta function has trivial zeros at negative even integers and satisfies ζ(−1)=−1/12 even though the original series diverges there.

  6. 6

    Non-trivial zeros lie in the critical strip, and the Riemann hypothesis predicts they all sit on the critical line Re(s)=1/2.

  7. 7

    The million-dollar prize is tied to proving the non-trivial zeros’ location on the critical line, a claim with far-reaching consequences for mathematics.

Highlights

Complex exponents n^{−(a+bi)} split into a size factor n^{−a} and a pure rotation n^{−bi} of magnitude 1, letting the series converge for Re(s)>1.
Analytic continuation isn’t arbitrary: demanding analyticity forces an angle-preserving extension, effectively locking in one possible continuation.
ζ(−1)=−1/12 comes from the unique analytic continuation of a function originally defined only where its series converges.
The Riemann hypothesis concerns the non-trivial zeros in the critical strip, predicting they all lie on Re(s)=1/2.
The “trivial zeros” at negative even integers are consistent with the continuation even though the original series diverges at those inputs.

Topics

Mentioned

  • Bernard Riemann