The Riemann Hypothesis — Current State of Knowledge

⚡ Key Takeaways

One-line statement: All nontrivial zeros of the Riemann zeta function $\zeta(s)$ lie on the critical line $\mathrm{Re}(s) = \tfrac{1}{2}$ .

Key result in words: If true, the Riemann Hypothesis (RH) pins down the distribution of prime numbers to the smallest conceivable error, equivalent to $\pi(x) = \mathrm{Li}(x) + O(\sqrt{x}\,\log x)$ .

Where it applies: Analytic number theory, the distribution of primes, $L$ -functions, random matrix theory, quantum chaos, arithmetic geometry.

Who/when: Conjectured by Bernhard Riemann in his 1859 memoir Über die Anzahl der Primzahlen unter einer gegebenen Größe. Still open as of 2024; it is one of the seven Clay Millennium Prize Problems ($1,000,000).

💡 The Core Insight

Primes look random, but their randomness is structured. The structure is encoded in a single analytic object — $\zeta(s)$ — whose zeros act as a hidden frequency spectrum. The "explicit formula" expresses the deviation of $\pi(x)$ from its smooth approximation as a sum of oscillating terms $x^{\rho}$ , one per zero $\rho$ . A zero at $\rho = \beta + i\gamma$ contributes a wave of amplitude proportional to $x^{\beta}$ . The Riemann Hypothesis says all these amplitudes share the same exponent $\beta = \tfrac{1}{2}$ — the primes are as evenly distributed as randomness permits, with no anomalously large clustering. Everything else in the subject is a corollary or a reformulation of this single equidistribution-of-error statement.

DefinitionThe Riemann Zeta Function

For $\mathrm{Re}(s) > 1$ , define

1

\quad \zeta(s) = \sum_{n=1}^{\infty} \frac{1}{n^s} = \prod_{p \text{ prime}} \left(1 - p^{-s}\right)^{-1}.

Derivation hint: The Euler product follows from unique factorization — expand each geometric factor $(1-p^{-s})^{-1} = \sum_{k\ge 0} p^{-ks}$ and multiply; each $n$ appears exactly once via its prime factorization. Here $s = \sigma + it$ with $\sigma = \mathrm{Re}(s)$ , $t = \mathrm{Im}(s)$ , and $p$ ranges over primes. The series and product converge absolutely for $\sigma > 1$ .

DefinitionAnalytic Continuation and the Functional Equation

The completed zeta function

2

\quad \xi(s) = \tfrac{1}{2}\, s(s-1)\, \pi^{-s/2}\, \Gamma\!\left(\tfrac{s}{2}\right) \zeta(s)

is entire and satisfies the symmetric functional equation

3

\quad \xi(s) = \xi(1-s).

Derivation hint: Use the theta function $\theta(t) = \sum_{n\in\mathbb{Z}} e^{-\pi n^2 t}$ , apply the Poisson summation–derived modular relation $\theta(1/t) = \sqrt{t}\,\theta(t)$ , and split the Mellin integral $\int_0^\infty (\theta(t)-1)\,t^{s/2}\,\frac{dt}{t}$ at $t=1$ ; the symmetry of the relation produces $s \leftrightarrow 1-s$ . The factors $\Gamma(s/2)$ and the pole structure of $\zeta$ combine so that the zeros of $\Gamma(s/2)$ at $s = 0, -2, -4, \dots$ force the trivial zeros of $\zeta$ at $s = -2, -4, -6, \dots$ (the $s=0$ pole is cancelled by the factor $s$ ). The completed $\xi(s)$ has zeros exactly at the nontrivial zeros of $\zeta$ .

TheoremThe Riemann Hypothesis

Every zero $\rho$ of $\xi(s)$ — equivalently every nontrivial zero of $\zeta(s)$ — satisfies

4

\quad \mathrm{Re}(\rho) = \tfrac{1}{2}.

Derivation hint: This is the conjecture itself — there is no derivation. By the functional equation [3], zeros come in pairs $\rho, 1-\rho$ ; with the reflection $\overline{\zeta(\bar s)} = \zeta(s)$ they form quadruples $\rho, \bar\rho, 1-\rho, 1-\bar\rho$ symmetric about the line $\sigma = \tfrac{1}{2}$ . RH asserts each quadruple degenerates onto that line.

DefinitionCounting Zeros and the Riemann–von Mangoldt Formula

Let $N(T)$ be the number of nontrivial zeros with $0 < \gamma \le T$ (counted with multiplicity, $\rho = \beta + i\gamma$ ). Then

5

\quad N(T) = \frac{T}{2\pi}\log\frac{T}{2\pi} - \frac{T}{2\pi} + \frac{7}{8} + S(T) + O\!\left(\tfrac{1}{T}\right),

where $S(T) = \tfrac{1}{\pi}\arg\zeta(\tfrac12 + iT)$ is the (small, $O(\log T)$ ) fluctuation term. Derivation hint: Apply the argument principle to $\xi(s)$ over a rectangle $[-1,2]\times[0,T]$ ; the main term is $\tfrac{1}{2\pi}\Delta\arg\xi$ , computed from Stirling's approximation of $\Gamma(s/2)$ , and $S(T)$ collects the contribution of $\arg\zeta$ along the right edge. This shows there are infinitely many nontrivial zeros, with density $\sim \tfrac{1}{2\pi}\log\tfrac{T}{2\pi}$ near height $T$ .

ProofConnecting Zeros to Primes: the Explicit Formula

We sketch the von Mangoldt explicit formula, the engine making RH equivalent to optimal prime distribution.

Step 1 — Logarithmic derivative. Take the logarithm of the Euler product [1] and differentiate:

6

\quad -\frac{\zeta'(s)}{\zeta(s)} = \sum_{n=1}^{\infty} \frac{\Lambda(n)}{n^s}, \qquad \mathrm{Re}(s) > 1,

where $\Lambda(n) = \log p$ if $n = p^k$ and $0$ otherwise (the von Mangoldt function). Derivation hint: $\log\zeta(s) = -\sum_p \log(1-p^{-s}) = \sum_p \sum_k \frac{p^{-ks}}{k}$ ; differentiate term by term, and $\frac{d}{ds}p^{-ks} = -k(\log p)p^{-ks}$ reproduces the weights $\Lambda(p^k)=\log p$ .

Step 2 — Perron / contour integration. Define the Chebyshev function $\psi(x) = \sum_{n \le x} \Lambda(n)$ . By Perron's formula,

7

\quad \psi(x) = \frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty} \left(-\frac{\zeta'(s)}{\zeta(s)}\right)\frac{x^s}{s}\, ds, \qquad c > 1.

Derivation hint: This is the inverse Mellin transform of [6]; the kernel $x^s/s$ acts as a smoothed step function picking out $n\le x$ via the residue at $s=0$ of $x^s/s$ .

Step 3 — Move the contour and collect residues. Shifting the contour to the left, we sweep past poles of $-\zeta'/\zeta$ : the pole of $\zeta$ at $s=1$ (residue $x$ ), zeros of $\zeta$ (each a simple pole of $-\zeta'/\zeta$ with residue $-x^\rho/\rho$ ), and the trivial zeros. The result:

8

\quad \psi(x) = x - \sum_{\rho} \frac{x^{\rho}}{\rho} - \frac{\zeta'(0)}{\zeta(0)} - \tfrac{1}{2}\log\!\left(1 - x^{-2}\right).

Derivation hint: Each nontrivial zero $\rho$ contributes $-\mathrm{Res}_{s=\rho}\frac{\zeta'}{\zeta}\cdot\frac{x^s}{s} = -x^\rho/\rho$ ; the trivial zeros sum to the $\log(1-x^{-2})$ term; the pole at $s=1$ gives the dominant $+x$ .

Step 4 — Read off RH's meaning. Each term $x^\rho/\rho$ has magnitude $x^{\beta}/|\rho|$ where $\beta = \mathrm{Re}(\rho)$ . If RH holds, $\beta = \tfrac12$ for all $\rho$ , so the oscillating sum is $O(x^{1/2}(\log x)^2)$ , giving

9

\quad \psi(x) = x + O\!\left(\sqrt{x}\,(\log x)^2\right) \iff \pi(x) = \mathrm{Li}(x) + O\!\left(\sqrt{x}\,\log x\right).

Derivation hint: Bound $\sum_\rho x^\rho/\rho$ using the zero-density [5]; pass from $\psi(x)$ to $\pi(x)$ by partial summation against $\mathrm{Li}$ . Conversely, an error of size $x^{1/2+\varepsilon}$ forces $\beta \le \tfrac12 + \varepsilon$ , so the error bound and RH are equivalent. $\blacksquare$

∎

BR

Bernhard Riemann

1826–1866

Born in Breselenz, Hanover, Riemann was a student of Gauss and Dirichlet at Göttingen. His entire output on number theory consists of a single eight-page memoir (1859), written upon election to the Berlin Academy, in which he introduced the complex-analytic study of $\zeta(s)$ , proved the functional equation, gave the explicit formula relating primes to zeros, and stated — almost in passing, as "very likely" (sehr wahrscheinlich) — that all nontrivial zeros lie on the critical line. He admitted he could not prove it and "set it aside." Riemann's Nachlass (unpublished notes), partly decoded by C. L. Siegel in 1932, revealed astonishing computational depth, including the Riemann–Siegel formula now used to verify zeros numerically. He died of tuberculosis at 39. His paper reshaped number theory into an analytic discipline and posed the problem that has organized the field for 165 years.

🧠 Intuition

Think of $\log\zeta(s)$ as a "music box" for the primes. The prime power $p^k$ contributes a pure tone of frequency $\log p^k$ . The function $\psi(x)$ is the running total of how loud the primes have been up to $x$ ; on average it grows like $x$ (the prime number theorem). The deviation $\psi(x) - x$ is a sound wave, and the explicit formula [8] decomposes it into harmonics indexed by the zeros: a zero at height $\gamma$ on the line $\sigma=\beta$ produces a wave $\propto x^{\beta}\cos(\gamma \log x)/|\rho|$ . If all zeros sit on $\sigma = \tfrac12$ , every harmonic shares the same envelope $\sqrt{x}$ — none dominates, and the error stays as small as a random walk of these frequencies allows. A single off-line zero at $\sigma = 0.7$ would produce a harmonic growing like $x^{0.7}$ , eventually drowning out the others and creating large, systematic biases in prime gaps. RH is the statement: the prime spectrum is perfectly balanced. The location $\sigma = \tfrac12$ is precisely the axis of symmetry imposed by the functional equation [3] — RH says nature chose the most symmetric possible configuration.

Figure

The critical strip: trivial zeros, the critical line, and a quadruple of nontrivial zeros symmetric about

\sigma=\tfrac12

ExampleMagnitude of the leading oscillation from the first zero

Given: The first nontrivial zero is $\rho_1 = \tfrac12 + i\gamma_1$ with $\gamma_1 \approx 14.134725$ . Estimate the contribution of the pair $\rho_1, \bar\rho_1$ to $\psi(x) - x$ at $x = 10^6$ , assuming RH.

Steps: 1. The paired term is $\dfrac{x^{\rho_1}}{\rho_1} + \dfrac{x^{\bar\rho_1}}{\bar\rho_1} = 2\,\mathrm{Re}\!\left(\dfrac{x^{\rho_1}}{\rho_1}\right)$ . 2. $|\rho_1| = \sqrt{0.25 + 14.1347^2} \approx 14.143$ . 3. $x^{\rho_1} = x^{1/2} e^{i\gamma_1 \log x}$ . With $x = 10^6$ : $x^{1/2} = 10^3 = 1000$ , and $\log x = 6\ln 10 \approx 13.8155$ , so $\gamma_1 \log x \approx 14.1347 \times 13.8155 \approx 195.28$ rad. 4. Reduce modulo $2\pi$ : $195.28 / (2\pi) \approx 31.08$ , fractional $0.08 \times 2\pi \approx 0.503$ rad, $\cos(0.503)\approx 0.876$ (the phase of $x^{\rho_1}$ ; the $1/\rho_1$ factor adds a small phase $-\arctan(2\gamma_1)\approx -\pi/2$ , but for a magnitude estimate we take the envelope). 5. Amplitude envelope $= \dfrac{2 x^{1/2}}{|\rho_1|} = \dfrac{2\times 1000}{14.143} \approx 141.4$ .

Answer: $\boxed{|{\text{first-zero term}}| \lesssim 141 \text{ (in units of } \psi)}$ . The full deviation $\psi(10^6) - 10^6$ is a superposition of such terms; each is bounded by $\sim 2\sqrt{x}/|\rho|$ , confirming the $O(\sqrt{x}\,\log^2 x)$ scale of [9].

ExampleCounting zeros up to height

T

Given: Use the Riemann–von Mangoldt formula [5] to estimate the number of nontrivial zeros with $0 < \gamma \le T$ for $T = 1000$ , ignoring the bounded $S(T)$ term.

Steps: 1. Main term: $N(T) \approx \dfrac{T}{2\pi}\log\dfrac{T}{2\pi} - \dfrac{T}{2\pi} + \dfrac{7}{8}$ . 2. Compute $\dfrac{T}{2\pi} = \dfrac{1000}{6.2832} \approx 159.155$ . 3. $\dfrac{T}{2\pi} = 159.155$ ; $\log(159.155) \approx 5.0699$ . 4. First term: $159.155 \times 5.0699 \approx 806.9$ . 5. Subtract $\dfrac{T}{2\pi} = 159.155$ : $806.9 - 159.16 = 647.7$ . 6. Add $7/8 = 0.875$ : $648.6$ .

Answer: $\boxed{N(1000) \approx 649 \text{ zeros}}$ . The exact verified count is $649$ — all of them confirmed (by computation) to lie on the critical line, consistent with RH up to far greater heights.

⚠️ What RH does and does NOT say

RH is not "proved for the first $N$ zeros, so probably true." Numerical verification (over $10^{13}$ zeros checked, all on the line) is evidence, not proof. The functions involved have logarithmically slow features ( $S(T)$ is provably unbounded), so finite checks cannot decide the conjecture. The Skewes-number phenomenon (where $\pi(x) - \mathrm{Li}(x)$ first turns positive only near $x \approx 1.4\times 10^{316}$ ) is a cautionary tale: number-theoretic crossovers can hide astronomically far out.
RH does not assert there are no zeros off the line — it asserts none exist anywhere. A single counterexample anywhere in the strip refutes it.
The "trivial zeros" $-2,-4,\dots$ are irrelevant to RH. RH concerns only zeros in $0<\sigma<1$ .
RH $\ne$ the Prime Number Theorem (PNT). PNT is the unconditional statement $\zeta(1+it)\ne 0$ (no zeros on the edge $\sigma=1$ ), proved 1896. RH is the vastly stronger statement that they all sit on $\sigma=\tfrac12$ .
The Generalized RH (GRH) for Dirichlet $L$ -functions, and the Grand RH for automorphic $L$ -functions, are separate (stronger) conjectures; proving RH for $\zeta$ alone does not prove them.

CounterexampleThe Davenport–Heilbronn function: zeros off the line are possible for similar-looking functions

RH is special to $\zeta$ and genuine $L$ -functions; it is not a generic feature of Dirichlet-series-with-functional-equations. The Davenport–Heilbronn function is a linear combination of two Dirichlet $L$ -functions:

f(s) = \tfrac{1}{2}\left(1 - i\,\kappa\right) L(s,\chi) + \tfrac12\left(1 + i\,\kappa\right) L(s,\bar\chi),

where $\chi$ is a character mod 5 and $\kappa$ a specific real constant. It satisfies a functional equation of Riemann type ( $f(s)\leftrightarrow f(1-s)$ ) yet has infinitely many zeros off the critical line (and indeed in $\sigma>1$ ). The reason: $f$ lacks an Euler product. This shows that the functional equation alone is insufficient — the Euler product (multiplicativity of arithmetic) is essential to RH. Any proof strategy that ignores the Euler product is doomed by this counterexample.

connection · RH within the web of mathematics

Prime Number Theorem: PNT is the assertion that $\zeta(s)\ne0$ on $\sigma=1$ ; RH sharpens the error term to optimal size via [9]. RH $\Rightarrow$ the strongest unconditional prime-gap bounds.
Dirichlet L-functions and Generalized Riemann Hypothesis: GRH extends RH to all $L(s,\chi)$ , controlling primes in arithmetic progressions; it implies the validity of fast primality and discrete-log algorithms.
<a class="wiki-link" href="/wiki/random-matrix-theory">Random Matrix Theory</a> (GUE): The Montgomery–Odlyzko law states the pair correlation of zeros matches the Gaussian Unitary Ensemble, suggesting the zeros are eigenvalues of some self-adjoint operator (the Hilbert–Pólya conjecture).
Weil Conjectures and the function-field analogue: Over finite fields, the analogue of RH (for zeta functions of curves/varieties) was proved by Weil (1948) and Deligne (1974). The function-field zeros are eigenvalues of Frobenius acting on cohomology — a true spectral interpretation that the number-field case still lacks.
Selberg Trace Formula: Provides a genuine spectral analogue (eigenvalues of the Laplacian on a hyperbolic surface) where the "Selberg zeta function" satisfies an exact RH, modeling the hoped-for arithmetic situation.
Mertens Function and $\mu(n)$ : RH $\iff M(x) = \sum_{n\le x}\mu(n) = O(x^{1/2+\varepsilon})$ , recasting it as a square-root-cancellation statement for the Möbius function.

🔬 Concrete consequences and uses

⏳ History

- 1737 — Euler. Discovers the Euler product

\zeta(s)=\prod_p(1-p^{-s})^{-1}

and evaluates

\zeta(2)=\pi^2/6

, first linking the analytic function to primes.

- 1859 — Riemann. Eight-page memoir: analytic continuation, functional equation [3], explicit formula, and the hypothesis that all nontrivial zeros have

\mathrm{Re}=\tfrac12

.

- 1896 — Hadamard & de la Vallée Poussin (independently). Prove

\zeta(1+it)\ne0

, yielding the Prime Number Theorem

\pi(x)\sim\mathrm{Li}(x)

. A "zero-free region" near

\sigma=1

is established, the unconditional backbone.

- 1900 — Hilbert. Lists RH as part of Problem 8 in his famous Paris address, cementing it as a central open problem.

- 1914 — Hardy. Proves infinitely many zeros lie exactly on the critical line (no statement they are all there).

- 1932 — Siegel. Decodes Riemann's Nachlass, recovering the Riemann–Siegel formula, the basis of all subsequent numerical verification.

- 1942 — Selberg. Shows a positive proportion of zeros lie on the line; Levinson (1974) raises it to

>1/3

, Conrey (1989) to

>2/5

.

- 1972 — Montgomery & Dyson. Pair-correlation conjecture links zero spacings to the GUE of random matrix theory, born from a chance Princeton tea conversation.

- 1974 — Deligne. Proves the Weil conjectures, establishing the function-field analogue of RH, the strongest evidence that an arithmetic-spectral proof should exist.

- 2000 — Clay Mathematics Institute. Names RH one of the seven Millennium Prize Problems (

1,000,000). Verification by Gourdon–Demichel (2004) confirms the first

10¹³$ zeros on the line.

open-problem · The state of the assault on RH (2024)

No proof, no disproof. Despite 165 years and verification of $>10^{13}$ zeros, RH remains open.
The Hilbert–Pólya program seeks a self-adjoint operator $H$ whose eigenvalues are the imaginary parts $\gamma$ (forcing them real, hence zeros on the line). The Berry–Keating $H=xp$ heuristic and Connes' noncommutative-geometry trace formula are the leading incarnations; none yet yields the operator.
The de Branges approach via Hilbert spaces of entire functions has gaps; the necessary positivity conditions remain unverified and a counterexample to the precise formulation was identified.
Function-field success vs. number-field failure. Deligne's proof uses étale cohomology and Frobenius eigenvalues — tools that do not exist over $\mathbb{Q}$ . Constructing the "field with one element" $\mathbb{F}_1$ or an arithmetic site (Connes–Consani) to import these ideas is a major active program.
Equivalent reformulations abound (Robin's inequality on $\sigma(n)$ , the Nyman–Beurling closure criterion in $L^2(0,1)$ , the Báez-Duarte sequence, the Lagarias inequality $\sigma(n)\le H_n + e^{H_n}\log H_n$ ), but none has cracked the core difficulty: encoding the Euler product's positivity into a spectral or variational statement.
The barrier of $S(T)$ . Since $S(T)=\Omega_\pm(\log T/\log\log T)$ (Montgomery, Selberg), the fluctuations grow without bound; any proof must control these unbounded oscillations, which finite computation cannot.

The Riemann Hypothesis — Current State of Knowledge

Formal Statement

Why It (Would) Hold — The Explicit Formula

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also