Exact-scale Euclidean Ramsey configurations

Source and verification

This article explains family 172 of the October 6, 2026 OpenAI collection. Research conclusions are attributed to its manuscripts. The review covers statements, introductions, and selected proof mechanisms; it is not an independent verification of the complete proofs.

🏷️ Congruent monochromatic configurations

A finite nonempty set $A$ in Euclidean space is Euclidean Ramsey if, for every number of colors $r\ge 2$, there is a dimension $D$ such that every coloring of ${\mathbb{R}}^D$ contains a monochromatic congruent copy of $A$. Congruence preserves all distances at their original scale. Similarity, which permits rescaling, would define a different problem.

The coloring is arbitrary. Consequently, an approximate copy cannot be repaired using continuity of the colors. Exact distances are required at the final step.

The manuscript claims a complete criterion in terms of exact relations among the coordinates of $A$. This post emphasizes its geometric consequences and the exact-scale mechanism; it does not develop a separate course in the algebra used by the criterion. (OpenAI, 2026)

🏷️ Sphericity and sufficient conditions

A set is spherical if all its points lie on a sphere in its affine span. The manuscript’s criterion implies that every Euclidean Ramsey set is spherical, but also produces spherical sets that are not Ramsey.

Geometric consequences claimed in the manuscript

Every nonempty subset of a finite transitive Euclidean configuration is Ramsey. Every nonempty set of at most five points on a circle is Ramsey. More generally, a spherical set is Ramsey if evaluation at its points gives linearly independent functionals on quadratic polynomials. There are nine-point and twelve-point spherical configurations that are not Ramsey. (OpenAI, 2026)

A finite configuration is transitive when, for any two of its points, an isometry preserving the entire configuration sends the first to the second. This definition needs no further group-theoretic machinery for the stated consequence.

The quadratic criterion has an accessible circle example. Restrict quadratic polynomials in two coordinates to the unit circle, writing $(x,y)=(\cos \theta ,\sin \theta )$. Their restrictions span

$$\begin{array}{c}1,\cos \theta ,\sin \theta ,\\ \cos 2\theta ,\sin 2\theta .\end{array}$$

A nonzero linear combination has at most four distinct zeros on the circle: substitute $z=e^{i\theta }$ and multiply by $z^2$ to obtain a nonzero polynomial of degree at most four. Evaluation at five distinct points is therefore invertible on this five-dimensional space. Fewer points inherit independence by extending them to five points. The claimed quadratic criterion then gives the circle conclusion.

This deduction uses the criterion as an input. The elementary zero count alone does not prove the Ramsey property.

🏷️ The exact relation criterion

Choose coordinates $A=\{a_1,\dots ,a_s\}\subset {\mathbb{R}}^d$ spanning their affine space, and write $p_i=(1,a_i{)}^{\mathsf{T}}$. Let $F$ be the field generated over $\mathbb{Q}$ by these coordinates. To state the criterion, use paired expressions $x\otimes y$ that are bilinear over $\mathbb{Q}$, while keeping the left and right copies of $F$ distinct. Their space is $B=F{\otimes }_{\mathbb{Q}}F$.

There is a multiplication map $m:B\to F$ sending $x\otimes y$ to $xy$. An identity among paired expressions is stronger than the identity obtained after applying $m$; the map can lose information.

The manuscript asserts that $A$ is Ramsey exactly when there is a $(d+1)\times (d+1)$ matrix $P$ with entries in $B$ satisfying

$$\begin{array}{c}(p_i\otimes 1{)}^{\mathsf{T}}P(1\otimes p_i)=0\\ (1\le i\le s),\\ m(P_{\alpha \beta })={\delta }_{\alpha \beta }(1\le \alpha ,\beta \le d).\end{array}$$

The constant row and column, indexed by zero, have no additional multiplication conditions. Applying $m$ to the first identities gives

$$\Vert a_i{\Vert }^2+\ell \cdot a_i+c=0$$

for suitable $\ell ,c$. Completing the square recovers a common sphere. Requiring the identities before multiplication is the extra condition that separates Ramsey configurations from merely spherical ones.

The criterion is independent of the congruent representative chosen in the affine span. It depends on exact coordinate relations, not on approximate floating-point measurements of an arbitrary real configuration.

🏷️ Exact cancellation and scale synchronization

The proof translates the criterion into a finite rational combination of pairs of affine functions. The function values cancel at every point of $A$, while their gradient data recover the Euclidean metric.

A Gaussian construction and a discretization within a fixed space of relations then produce two lists of affine functions. Evaluating the lists at $a_i$ gives vectors $f_i,g_i$ such that

$$\begin{array}{rl}\Vert f_i-f_j{\Vert }^2& =\Vert a_i-a_j{\Vert }^2,\\ \Vert g_i-g_j{\Vert }^2& =q\Vert a_i-a_j{\Vert }^2,\end{array}$$

where $q>0$ can be arbitrarily small, and each $g_i$ is a coordinate permutation of $f_i$. The permutation may depend on $i$, so there is no contradiction between equal coordinate multisets and different interpoint distances. All these identities are exact.

The next stage records copies as paths, with weight equal to the sum of their squared scales. A finite word-coloring theorem produces a monochromatic combinatorial line, but the number $t$ of variable positions is not known in advance. If each position carried a copy at squared scale one, the resulting distances would acquire a factor $t$.

The source prepares synchronized paths at weights $1,1/2,\dots ,1/n$ before applying the word theorem of length $n$. Once $t$ is known, choosing weight $1/t$ gives total squared scale $t(1/t)=1$. An additional correction argument ensures that the paths have compatible endpoints while making the correction weight negligible relative to the larger path.

That endpoint construction is a substantial part of the source’s algebraic proof. Here its role is explicit: it makes the finite word argument return an exactly congruent configuration at the required scale. Approximation alone would not suffice for arbitrary colorings.

🏷️ Key steps and proof outlines

Preserve exact cancellation while improving a quadratic inequality

The criterion supplies a finite signed stencil with zero sums on every class of affine functions taking the same value at each point $a_i$. Its second-moment matrix is $C$, and a matrix $U$ converting stencil coordinates to Euclidean gradients satisfies $UC{U}^{\mathsf{T}}=I_d$. The “Two-scale configurations” section turns this exact data into two coordinate lists with very different metric scales. (OpenAI, 2026)

The analytic step constructs a nonnegative smooth function $b$ with

$$\int (D_{C}b{)}_{-}(x)\Vert Ux{\Vert }^{2}dx<\eta \int b(x)dx,D_C=\frac{1}{2}\sum _{j,l}{C}_{jl}{\partial }_j{\partial }_l.$$

Here $r_{-}=\max (-r,0)$. Although $C$ need not be positive semidefinite, choose a positive covariance $S$ with $S+C>0$ and $\mathrm{tr}(US{U}^{\mathsf{T}})$ very small. To do this, make $S$ small on $\mathrm{im}{U}^{\mathsf{T}}$ and large on $\ker U$; a Schur-complement estimate ensures positivity of $S+C$.

For centered Gaussian density ${\varphi }_Q$ with covariance $Q$, differentiation gives $\frac{d}{dt}{\varphi }_{S+tC}=D_C{\varphi }_{S+tC}$. Therefore

$$b_{\ast }:={\int }_0^1{\varphi }_{S+tC}dt⟹D_C{b}_{\ast }={\varphi }_{S+C}-{\varphi }_S.$$

Its negative part is bounded by ${\varphi }_S$, whose weighted second moment is the small trace above. Smooth compact truncation preserves the strict inequality because Gaussian derivatives decay rapidly.

Rational approximation occurs inside the space of exact identities

Convolve the original finite stencil $\nu$ with a sampled version of $b$, rather than independently perturbing the desired moments. If $\nu$ has zero total mass and first moment, and second moment $C$, then

$$\sum _{\lambda }(\nu \ast d)(\lambda )\lambda {\lambda }^{\mathsf{T}}=\left(\sum _{\beta }d(\beta )\right)C.$$

All cross terms vanish exactly. The cancellation sums on evaluation classes also remain zero under convolution. Taylor expansion and Riemann sums transfer the strict negative-part inequality to sufficiently fine finite samples. Only then approximate the finitely many sampled input weights $d(\beta )$, then define $\mu =\nu \ast d$ by rationals. Strict inequalities persist by continuity, while the convolution identities remain exact for every choice of those coefficients.

Clear denominators and expand positive and negative weights into repeated affine rows. At each point $a_i$, the two row lists have exactly the same multiset of evaluated values. Thus their evaluation vectors are coordinate permutations. Their gradient Gram matrices differ by the identity, and the negative list has small Gram matrix. Adding the same positive semidefinite Gram correction to both lists and normalizing gives squared metric scales $1$ and $q=\eta /(1+\eta )$. The permutation can depend on $i$, so it does not preserve distances between different labels.

Synchronize all possible word lengths in advance

The source’s endpoint construction turns this scale separation into two finite paths with the same endpoint and weights $\lambda >0$ and $\mu \ge 0$, with $\rho =\mu /\lambda <1/n$. A path’s weight is the sum of its copy factors’ squared scales. Common factors contribute zero. The algebraic endpoint construction is used here as an input; the scale synchronization itself is elementary.

For $1\le t\le n$, set

$${\alpha }_t=\frac{1/t-\rho }{1-\rho }.$$

These numbers partition $[0,1]$. On an interval of length $l$, scale both available paths by $\sqrt{l/\lambda }$; their weights become $l$ and $\rho l$, while their endpoints still agree. Concatenate intervals in one fixed order. For target $1/t$, use the larger path on intervals before ${\alpha }_t$ and the smaller path afterward. Every choice has the same endpoint, and its weight is exactly

$${\alpha }_t+\rho (1-{\alpha }_t)=1/t.$$

No interchange of factors is required.

The final monochromatic copy has exact scale

A finite word-coloring theorem now supplies a monochromatic combinatorial line with some nonzero number $t\le n$ of variable positions. Expand each variable position using the already prepared weight-$1/t$ path. The simultaneous realization lemma chooses every common string identically across all configuration labels, while copy factors use prescribed one-letter strings. All resulting strings have aligned equal lengths.

Embed their positions in mutually orthogonal Euclidean coordinate blocks. Squared distances add across blocks; common strings contribute zero, and the $t$ variable positions contribute

$$t\cdot \frac{1}{t}\Vert a_i-a_j{\Vert }^2=\Vert a_i-a_j{\Vert }^2.$$

This proves exact congruence despite the unpredictable value of $t$. Finally compactness reduces finite-support infinite-dimensional space to some finite dimension: any finite set of forbidden-copy conditions involves only finitely many coordinates. The proof thus uses approximation to achieve a strict auxiliary inequality, while preserving every identity needed for the arbitrary-coloring conclusion.

See Also

• Plane coloring fixes dimension two and forbids a single monochromatic distance; the present problem lets dimension grow while fixing an entire configuration.
• Borsuk’s covering problem also translates distance restrictions into a finite covering obstruction.

🏷️ Manuscripts and verification scope

• A classification of finite Euclidean Ramsey configurations — pinned manuscript and companion source.

The family scope page describes selected formalization coverage. The page lists the tensor criterion, spherical necessity, transitive and quadratic sufficient conditions, circle configurations of at most five points, and specific nine- and twelve-point obstructions. No Lean build, axiom audit, or manuscript-to-formal-statement equivalence check was performed for these notes.

OpenAI collection provides prerequisites and the reading sequence; coverage register records completed and pending families.

🐻  OpenAI 2026. A classification of finite Euclidean Ramsey configurations.