The Goldbach Conjecture: 282 Years of Adding Primes

In June 1742, the German mathematician Christian Goldbach wrote a letter to his more famous friend, Leonhard Euler. Among other observations, Goldbach proposed two statements about how primes combine with addition:

Every integer greater than 2 can be written as a sum of three primes.
Every integer greater than 2 can be written as a sum of one or more primes.

Euler replied with a related, simpler statement: every even integer greater than 2 is the sum of two primes. This single sentence has become one of the most famous unsolved problems in mathematics. 282 years later, it remains open.

The conjecture is striking because of how elementary it is. Anyone who can add small numbers can verify it for the first few cases: 4 = 2+2, 6 = 3+3, 8 = 3+5, 10 = 3+7 = 5+5, 12 = 5+7, 14 = 3+11 = 7+7, 16 = 3+13 = 5+11. Computers have checked it up to about 4×10¹⁸ without finding a counterexample. Mathematicians have made significant partial progress. No proof.

This article is about what Goldbach’s conjecture says, why it’s so hard, and what we know about it.

The statements

Strong (binary) Goldbach conjecture: Every even integer $n > 2$ can be written as $n = p + q$ where $p$ and $q$ are primes.

Weak (ternary) Goldbach conjecture: Every odd integer $n > 5$ can be written as $n = p + q + r$ where $p, q, r$ are primes.

The strong version implies the weak: if every even number is a sum of two primes, then every odd number $> 5$ can be written as 3 plus an even number $\geq 4$ , which is a sum of two primes, giving 3+prime+prime — three primes total.

The weak version was proven by Harald Helfgott in 2013 (building on Ivan Vinogradov’s 1937 result for sufficiently large odd integers, plus extensive computer verification for the remaining small cases). The weak conjecture is now a theorem.

The strong version is still a conjecture.

How many representations?

For any specific even number, you can ask: how many ways can it be written as a sum of two primes?

n	Representations as p + q	Count
4	2+2	1
6	3+3	1
8	3+5	1
10	3+7, 5+5	2
12	5+7	1
14	3+11, 7+7	2
16	3+13, 5+11	2
18	5+13, 7+11	2
20	3+17, 7+13	2
22	3+19, 5+17, 11+11	3
24	5+19, 7+17, 11+13	3
30	7+23, 11+19, 13+17	3
100	3+97, 11+89, 17+83, 29+71, 41+59, 47+53	6
1000	…	28
1,000,000	…	5,402

The count grows roughly like $n / (\log n)^2$ . This makes intuitive sense: there are roughly $n / \log n$ primes up to $n$ (by the Prime Number Theorem), and you’re picking pairs. The probability heuristic says counterexamples should be vanishingly rare as $n$ grows.

For Goldbach to fail, you’d need a “lucky” even number where zero representations exist — meaning no pair of primes adds up. As $n$ grows, this becomes overwhelmingly improbable. But “overwhelmingly improbable” isn’t a proof.

Goldbach’s comet

Plot the number of representations of $n$ as a function of $n$ , and you get the Goldbach comet — a striking visualization:

Each dot is a single even number $n$ at horizontal position $n$ and vertical position equal to the count of prime-pair representations. The plot shows the comet shape: the cloud of points grows upward as $n$ grows, with thick lower density and sparser high counts.

Crucial observation: the lower envelope never touches zero. Every even number plotted has at least one representation. If Goldbach were to fail, a single dot would appear on the $n$ -axis itself. Up to roughly $4 \times 10^{18}$ , no such dot exists.

Why this is so hard

If Goldbach’s conjecture is almost certainly true (numerically), why can’t we prove it?

The fundamental difficulty: primes are not a structured set in any algebraic sense. They’re defined by what they’re not (not products of smaller numbers) rather than what they are. There’s no formula generating primes. They appear essentially randomly within the constraints of the multiplication table.

This makes statements about primes very hard to prove. Most modern proofs about primes use analytic methods — properties of complex functions like the Riemann zeta function — to extract information about primes from their generating series. These methods work for some questions (Prime Number Theorem) but seem inadequate for additive questions about primes.

Goldbach is fundamentally an additive question: when do you get an even number by adding two primes? Multiplicative questions (about products of primes) connect naturally to analytic tools; additive questions don’t, in general.

The mathematician Ivan Vinogradov, in 1937, found a way to attack the ternary Goldbach conjecture for sufficiently large numbers using the circle method — an analytic technique that breaks Fourier integrals into “major” and “minor” arcs. The major arcs contribute the main term; minor arcs are bounded as error terms.

Vinogradov showed:

For large enough odd $n$ , the number of representations of $n$ as a sum of three primes grows as expected.
“Large enough” was effective but enormous — beyond direct computation.

Helfgott’s 2013 work extended this to handle all odd $n > 5$ , partly by improving the analytical bounds and partly by extending direct computation up to the threshold. The result: weak Goldbach is now a theorem.

For binary Goldbach, the circle method gives weaker results. The “main term” arguments work, but the “error term” estimates are too loose to rule out exceptions. New techniques would be needed — and so far, no one has them.

Partial results

Significant progress has been made on related questions, even without a full proof.

Chen’s theorem (1973): Every sufficiently large even integer is the sum of a prime and a number that’s either prime or a product of two primes. So Goldbach is “almost true” — instead of two primes, you get one prime and a “near-prime.”

Schnirelmann’s theorem (1930): Every integer $n \geq 2$ is a sum of at most $C$ primes, for some constant $C$ . Schnirelmann showed $C \leq 800{,}000$ . Vinogradov improved this. Helfgott’s work showed $C \leq 4$ .

Estermann’s theorem (1938): “Almost all” even integers satisfy Goldbach’s conjecture — the density of exceptions, if any exist, has measure zero.

Linnik’s theorem (1953): Every sufficiently large odd integer is the sum of a prime and a power of 2.

Helfgott’s theorem (2013): Every odd integer $> 5$ is the sum of three primes (weak Goldbach proven).

Each of these is a major result. None proves the strong Goldbach conjecture. They establish “almost-Goldbach” results in different precise senses.

Computational verification

How far have computers checked the conjecture?

The first explicit verifications, by hand, went up to about $10^4$ in the 19th century. By 1938, Pipping had verified up to $10^5$ . By 1965, Stein and Stein reached $2 \times 10^6$ . The numbers grew with computing power.

Modern computational records:

1998: Deshouillers, te Riele, Saouter — up to $4 \times 10^{14}$
2003: Richstein — up to $2 \times 10^{16}$
2013: Oliveira e Silva, Herzog, Pardi — up to $4 \times 10^{18}$

Beyond $4 \times 10^{18}$ , the computational cost becomes prohibitive even with modern hardware. Each new factor of 10 roughly doubles the required compute time. We have no realistic hope of brute-force verifying Goldbach for arbitrarily large $n$ .

But even checking up to $4 \times 10^{18}$ — about 4 billion billion — has not produced a single counterexample. If Goldbach is false, the counterexample is astronomically far out, in regions where statistical heuristics predict counterexamples should be essentially impossible.

What makes Goldbach famous

Goldbach’s conjecture is famous for several reasons:

Simplicity: anyone who knows primes can understand the statement. Most outstanding mathematical problems require a graduate degree to even state.

Age: 282 years is unusually long for an open problem at its level of fame. Compare: Fermat’s Last Theorem (proven after 358 years), Riemann Hypothesis (still open after 165 years).

Computational suggestion of truth: unlike the Riemann hypothesis (where numerical evidence is suggestive but not overwhelming), Goldbach is verified to $4 \times 10^{18}$ — the conjecture is essentially “known true” empirically.

Resistance to proof: despite enormous effort, the conjecture has resisted all approaches. This itself is intriguing — it suggests the problem requires fundamentally new methods.

Pop culture: Apostolos Doxiadis’s 1992 novel Uncle Petros and Goldbach’s Conjecture brought public attention to the problem. The novel describes a fictional mathematician obsessed with the conjecture; it inspired Faber and Faber’s $1 million prize offer.

What proving it might require

What kind of new mathematics might prove Goldbach?

One possibility: a deeper understanding of additive structures of primes. Are there hidden patterns in how primes distribute additively? The current “primes are essentially random” model gives heuristic predictions that match observations, but doesn’t enable proof.

Another possibility: a breakthrough in the Riemann hypothesis and related conjectures about the Riemann zeta function. Many additive questions about primes (Goldbach, twin primes, prime gaps) are connected via analytic number theory. A proof of the Riemann hypothesis might unlock new techniques.

A third possibility: an entirely new framework. Some mathematical advances come from genuinely novel ideas that don’t fit into existing approaches. Wiles’s proof of Fermat used modular forms and Galois representations — tools developed for entirely different purposes that turned out to be the right ones for Fermat. Goldbach might require something analogous, currently not yet developed.

For now, the conjecture remains open.

What Goldbach’s conjecture teaches

The deepest lesson of the Goldbach conjecture is that mathematical truth and mathematical proof can be very different. The conjecture is almost certainly true in any reasonable sense — verified for 4×10^18 cases without exception, supported by powerful heuristic arguments. But mathematics requires proof, not evidence, and proof remains elusive.

This is the same lesson as the Collatz conjecture, the twin prime conjecture, and many other unsolved problems: numerical patterns that hold for trillions of cases can still be open mathematical questions. The discipline demands certainty in a strong sense.

For students of mathematics, Goldbach is one of the cleanest demonstrations of the difference between conjecture (mathematically believed) and theorem (mathematically proved). The two are not the same.

For working mathematicians, Goldbach is a benchmark of difficulty. It looks elementary; it remains unsolved. Anyone who claims to have a simple proof of Goldbach is almost certainly mistaken — but the conjecture has attracted countless amateur attempts over the years, and many published “proofs” have turned out to contain subtle errors.

For everyone else, Goldbach is a reminder that simple-sounding questions can be deeply hard. The simplest question in number theory remains open after 282 years. Some of the most important unfinished business of human knowledge is, in form, just arithmetic.

282 years. 4 billion billion verified cases. Zero counterexamples. No proof. That’s the state of the Goldbach conjecture in 2025. Maybe the proof will come in 2026, or 2050, or 2200. Maybe it won’t come at all. We don’t know. That uncertainty is itself one of the most striking mathematical facts of our time.

Frequently asked

Is the Goldbach conjecture actually unproven?

Yes. Despite being 282 years old, despite computer verification for every even number up to about 4×10¹⁸, despite enormous effort by mathematicians, no proof exists. It is widely believed to be true, but mathematical 'widely believed' is not the same as proven. The problem appears to require fundamentally new methods that haven't been developed.

What's the difference between strong and weak Goldbach?

The strong (or 'binary') Goldbach conjecture says every even integer > 2 is the sum of two primes. The weak (or 'ternary') Goldbach conjecture says every odd integer > 5 is the sum of three primes. The weak version was proven by Harald Helfgott in 2013, building on prior work by Vinogradov. The strong version remains open.

How much money is offered for solving it?

In 2000, Faber and Faber publishers offered $1 million for a proof, valid for two years after the publication of their novel Uncle Petros and Goldbach's Conjecture. No one collected. The Clay Mathematics Institute did not include Goldbach in their Millennium Problems — it's considered hard but not at the level of the Riemann Hypothesis or P vs NP.