Equilateral Dimension¶

Measure how large an exactly pairwise-equidistant subset a metric space can support, with the distance scale and attainment convention stated explicitly.

Version: v2 · 2026-08-30 · History
Domain-specific #: 1778
Origin domain: mathematics
Subdomain: metric geometry
Aliases: Equilateral number

Core Idea¶

The Equilateral Dimension of a metric space measures the largest size of an exactly equidistant configuration that the space can support. A subset S of (X,d) is lambda-equilateral, for lambda>0, when.

d(x,y)=lambda for every distinct x,y in S`.

For a fixed distance lambda, the careful convention writes e_lambda(X) for the largest cardinality when a largest set exists and writes e_lambda(X)=infinity when no finite largest size exists. In a general metric space, the distance scale matters. A scale-free version may be stated as.

e(X)=sup_{lambda>0} e_lambda(X),

using infinity as the unbounded sentinel.^[1] This formulation prevents “maximum” from silently claiming an optimizer that may not exist.

In a normed vector space, scaling an equilateral set changes lambda without changing its size. The distance parameter can therefore be suppressed and one writes e(X). In finite-dimensional normed spaces the value is finite and a largest equilateral set exists. Standard anchors are e(l_2^n)=n+1, realized by a regular simplex, and e(l_infinity^n)=2^n, realized by the vertices of a cube; every n-dimensional normed space has e(X)<=2^n.^[1]

The word “dimension” is conventional but potentially misleading. The value does not count independent coordinates or degrees of freedom. It records an extremal cardinality under one exact pairwise-distance constraint, and it depends on the full ambient metric: two norms on the same vector space can have different equilateral dimensions.

Structural Signature¶

Sig role-phrases:

the ambient metric space (X,d) — the carrier and its declared distance function
the positive distance scale lambda — the common distance required of every distinct pair, essential in general metric spaces
the candidate subset S — a set of points tested for the equilateral condition
the all-pairs equality constraint — d(x,y)=lambda for every distinct x,y in S, with no approximate tolerance
the cardinality objective |S| — the quantity maximized or bounded
the extremal operator — maximum when attained, otherwise a supremal or unbounded convention
the attainment status — records whether a largest set actually exists rather than conflating unbounded finite examples with an infinite witness
the ambient-metric dependence — changing the metric can change feasibility and the resulting value even when the point set is unchanged
the scale-invariance option — normed spaces permit rescaling and hence suppression of lambda; arbitrary metric spaces generally do not

Recognition test. State (X,d), state whether lambda is fixed or ranges over positive values, require exact equality for every distinct pair, and specify whether the reported value is an attained maximum, a finite supremum, or the infinity convention. If the object counts coordinates, allows unequal distances within a tolerance, tests only local maximality, omits the ambient metric, or infers an infinite equilateral set merely from arbitrarily large finite ones, the signature fails.

What It Is Not¶

Not linear or topological dimension. It does not count basis vectors, independent parameters, or local coordinates.
Not metric dimension in graph localization. Resolving-set metric dimension asks how many landmarks distinguish vertices by distance vectors.
Not the size of a maximal equilateral set. A set can be inclusion-maximal yet smaller than the greatest possible equilateral set.
Not a packing number. Packing requires pair distances at least a threshold; equilateral dimension requires every pair distance to equal one value.
Not a kissing number. Tangency to a central ball and pairwise nonoverlap impose a different geometry.
Not the number of equidistant points from one center. Equal radii to a center do not make the points pairwise equidistant.
Not approximate equidistance. Allowing a tolerance defines a different stability or coding problem.
Not automatically an attained cardinality. In noncompact or infinite settings, the convention infinity can mean unbounded finite sizes without any infinite equilateral subset.

Scope of Application¶

Equilateral dimension organizes extremal questions across several parts of metric and discrete geometry.

Finite-dimensional normed spaces. The unit-ball geometry determines which large exact equilateral configurations exist. Bounds for l_p^n, general Minkowski spaces, and Banach–Mazur neighborhoods are central examples.^[1]
Euclidean and convex geometry. Regular simplices realize the Euclidean value n+1; cube vertices realize the l_infinity value 2^n. Upper-bound proofs connect equilateral sets to convex hulls, packing, linear algebra, and topology.^[1]
General metric spaces. Tori, Hamming spaces, manifolds, projective spaces, and compacta support distinct equilateral-number problems because their metrics impose different exact-distance relations.^[2]^[3]
Infinite-dimensional spaces. Arbitrarily large finite equilateral sets need not assemble into one infinite equilateral set. The invariant therefore carries an existence distinction rather than only a growth statement.^[3]
Coding and configuration problems. An equilateral set is a constant-distance code in the declared metric. Results can transfer literally when the code metric and exact equality condition are preserved.

The abstraction should not be applied to similarity scores that fail the metric axioms, to noisy empirical distances without an exactness policy, or to a point cloud without specifying whether the ambient space or only the observed finite set is being measured.

Clarity¶

Three questions must be separated.

Fixed scale or any scale? e_lambda(X) asks about one distance. In a normed space, homothety makes all positive scales equivalent. In an arbitrary metric space they can differ, so e(X) must declare whether it takes a supremum across lambda.^[1]^[3]

Maximum or supremum? If a largest feasible set S exists, its cardinality is a maximum. If sets of every finite size exist but no infinite equilateral set exists, there is no maximizing set even though the standard extended value is infinity. “Maximum” is then only conventional shorthand and should not be read as attainment.

Maximum or maximal? A maximum-size set dominates all feasible sizes. A maximal set merely cannot be enlarged while retaining its existing points. Different maximal equilateral sets can have different sizes, so extension results do not by themselves determine e(X).

Compactness resolves part of the issue: for fixed lambda>0, a compact metric space cannot contain arbitrarily large lambda-equilateral sets, because such a set is lambda-separated.^[1] It does not by itself make a single bound uniform over all lambda approaching zero.

Manages Complexity¶

The invariant compresses a space's entire exact-distance feasibility landscape into a cardinal bound. It lets researchers compare metrics, state extremal theorems, separate construction problems from upper-bound problems, and quantify how a norm's unit-ball geometry supports symmetric configurations.

The signature also decomposes proofs. A lower bound exhibits an equilateral set and checks all pair distances. An upper bound shows that any set satisfying the equality constraints has limited cardinality. An attainment argument then establishes that the bound is realized. Keeping these obligations separate prevents a construction from being mistaken for an optimum.

Metric dependence provides a controlled comparison interface. The same vector space R^n can be equipped with l_2, l_1, or l_infinity; different values reveal geometry of the norm rather than a change in algebraic dimension. Conversely, metric embeddings or distortions need not preserve exact equilateral dimension unless they preserve the relevant distances exactly.

Abstract Reasoning¶

Construct for lower bounds. Propose coordinates or combinatorial objects, calculate every pair distance, and count the set. A regular simplex, signed basis set, or cube vertex set supplies a certificate only for its own metric.

Prove upper bounds globally. Translate equal distances into linear independence, rank, convexity, volume, or combinatorial restrictions. A proof must cover every feasible equilateral set, not only a symmetric candidate.

Audit scale. In a normed space, rescale to lambda=1. In a general metric space, retain lambda and avoid merging results at different distances without an explicit supremum.

Separate unboundedness from witnesses. A sequence of finite sets with sizes tending to infinity proves an unbounded extended value. It does not produce an infinite equilateral set unless a compactness, limit, or selection argument supplies one.

Compare metrics on the same carrier. Hold the point-vector set fixed and change only the norm. Any change in the extremal configuration then diagnoses metric geometry rather than algebraic dimension.

Test exactness under transfer. Approximate isometries can turn equality into a range of distances. If exact equality is lost, the transferred question is about almost-equilateral sets, not this invariant.

Knowledge Transfer¶

The abstraction transfers literally among metric spaces when four items survive: a genuine metric, an exact common distance, a cardinality objective, and a declared attainment convention. This supports movement between normed spaces, discrete cubes, tori, manifolds, and graph metrics without changing the logical role.

What does not transfer is a numerical value or its geometric explanation. The cube construction is equilateral for l_infinity but not for Euclidean distance. A constant-distance error-correcting code and a regular simplex instantiate the same signature, yet their upper-bound tools and interpretations differ.

Outside mathematics, “equally separated options” is usually analogical unless the domain supplies a validated metric and exact equality is meaningful. Survey scales, semantic similarities, and noisy physical measurements commonly lack that precision. The disciplined transfer question is not whether the language sounds geometric, but whether all pair distances can be audited against one declared lambda.

Examples¶

Canonical: the same plane under two norms¶

In Euclidean l_2^2, the points

(0,0), (1,0), (1/2,sqrt(3)/2)

have all pairwise distances 1, forming an equilateral set of size three. No four points in the Euclidean plane are pairwise equidistant, so e(l_2^2)=3. In l_infinity^2, the four square vertices (1,1), (1,-1), (-1,1), and (-1,-1) have pairwise distance 2; the general bound is attained, so e(l_infinity^2)=4.^[1]

Mapped back: R^2 with each chosen norm is the ambient metric space; 1 or 2 is the distance scale; the triangle or square vertices form the candidate subset; direct calculation verifies the all-pairs equality constraint; sizes three and four are the cardinality objective; and the different answers demonstrate ambient-metric dependence despite unchanged algebraic dimension.

Applied / In Practice: unbounded finite sets without attainment¶

Let X be the disjoint union of finite blocks B_n, one block of size n for every positive integer. Give distinct points in the same block distance 1. Give points in different blocks B_i and B_j distance 2+|i-j|. This is a metric: the cross-block part obeys the triangle inequality inherited from the integer line with a positive offset, and within-block distance 1 is smaller than any cross-block detour.

Each B_n is a 1-equilateral set of size n, so sizes are unbounded. Yet every distance-one pair lies within one finite block, so no infinite 1-equilateral set exists and no largest finite one exists. Under the standard convention e_1(X)=infinity, but the value is not attained.

Mapped back: the block union with the declared d is the ambient metric space; lambda=1 fixes the distance scale; each B_n is a candidate subset satisfying the all-pairs equality constraint; |B_n|=n makes the cardinality objective unbounded; and the absent largest set is the attainment status that a bare “maximum” would conceal.

Structural Tensions¶

T1: Maximum versus supremum. Familiar finite spaces have optimizers, but arbitrary spaces need not. Diagnostic: Is a largest equilateral set exhibited or proved to exist, or is only an upper bound or unbounded sequence known?

T2: Fixed scale versus any scale. Normed spaces permit rescaling, while general metrics may support different sizes at different distances. Diagnostic: Is lambda fixed, suppressed by a justified similarity, or ranged over explicitly?

T3: Exact equality versus approximation. Exact equilateral sets are brittle under noise and distortion. Diagnostic: Are all pair distances mathematically equal, or has a tolerance silently changed the problem?

T4: Maximum versus maximal. An unextendable configuration need not have greatest size. Diagnostic: Does the proof compare against every equilateral set, or only show that one set cannot absorb another point?

T5: Algebraic dimension versus metric geometry. The same vector space can have different equilateral dimensions under different norms. Diagnostic: Which metric, rather than merely which coordinate dimension, supports the claimed value?

T6: Cardinal compression versus structural explanation. One number compares spaces but hides the shapes and distances of extremizers. Diagnostic: Are a realizing construction, upper-bound mechanism, and attainment status retained alongside the value?

T7: Domain autonomy versus prime reduction. Cardinality and Metric explain size and distance, yet they do not impose one exact all-pairs distance or an extremal-attainment protocol. Diagnostic: After naming those parents, does the equilateral feasibility test still provide independent construction, bounding, and failure reasoning? If yes, the domain node remains autonomous.

Structural–Framed Character¶

Criterion 1 — Vocabulary travels (0.25). Distance, cardinality, and extremal sets travel broadly, but equilateral dimension remains an exact metric-geometric invariant rather than a generic phrase.

Criterion 2 — Evaluative weight (0.0). A larger value is not intrinsically better or more legitimate; it is a formal property of the metric space.

Criterion 3 — Institutional origin (0.0). Mathematical convention chooses notation and the infinity sentinel, but no institution makes a set equilateral.

Criterion 4 — Human-practice boundedness (0.0). The abstraction applies to mathematical objects without roles, norms, or institutional practices.

Criterion 5 — Import versus recognition (0.25). Literal recognition is possible wherever a metric is supplied; empirical domains must import and validate that metric before using the invariant.

The aggregate 0.1 yields structural under the documented rubric. The exact metric and extremal package still fails the substrate-free prime bar.

Structural Core vs. Domain Accent¶

What is skeletal. A relation constrains every pair in a candidate collection, cardinality measures the collection, and an extremal operator asks how far the constraint can be sustained. Metric, Cardinality, Constraint, and Optimization express pieces of this pattern.

What is domain-bound. The relation must be one declared metric, every distinct pair must have exactly the same positive distance, the scale must be handled correctly, and the output must distinguish maximum, supremum, and unbounded nonattainment. These are specialist conditions of metric geometry.

Why this does not clear the prime bar. Unrelated substrates do not preserve the identity merely by having “balanced” or “equally different” items. They must first supply a genuine metric and the exact all-pairs equality problem. The broader reasoning is already represented by its primes; Equilateral Dimension remains a reusable domain-specific invariant.

Equilateral Dimension instantiates prime:cardinality by strict subsumption. Its output is the greatest or supremal size of a subset satisfying a declared condition. Cardinality alone does not state that condition or the attainment convention.

It also presupposes prime:metric through strict composition. The ambient metric determines the equal-distance relation and can change the result on the same carrier. Metric alone does not pose an extremal subset problem.

prime:dimension is a name neighbor, not a proposed parent. The live prime counts degrees of freedom under an independence criterion, whereas equilateral dimension counts the size of an exact-distance configuration; neither entails the other. prime:optimization and prime:constraint describe the problem form, but their roles are already internal to the cardinality-plus-metric specialization and are declined for minimality. prime:measure is broader size vocabulary but its additive set-function identity does not fit this invariant.

Relationships to Other Abstractions¶

Current abstraction Equilateral Dimension Domain-specific

Parents (2) — more general patterns this builds on

Equilateral Dimension is a kind of Cardinality Prime

Equilateral Dimension instantiates prime:cardinality by strict subsumption.
Equilateral Dimension presupposes Metric Prime

Equilateral Dimension instantiates prime:cardinality by strict subsumption.

Hierarchy paths (6) — routes to 3 parentless roots

Equilateral Dimension → Cardinality → Bijectivity → Function (Mapping)

Show alternative paths (5)

Neighborhood in Abstraction Space¶

Equilateral Dimension sits in a sparse region of the domain-specific corpus (63^rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Metric Geometry & Approximation (13 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Linear dimension. Counts basis elements or degrees of freedom. Tell: Is the output an independent-coordinate count or an equilateral-set size?
Graph metric dimension. Counts resolving landmarks. Tell: Must landmarks distinguish distance vectors, or must all selected pairs have equal distance?
Maximal equilateral set. Cannot be extended while retaining its members. Tell: Is global greatest size proved?
Packing number. Requires pairwise separation of at least r. Tell: Are unequal distances above the threshold permitted?
Kissing number. Counts nonoverlapping neighbors tangent to a central ball. Tell: Is there a privileged center and tangency constraint?
Equiangular lines. Equalizes angles between lines rather than distances between points. Tell: Are the objects projective lines or metric points?
Regular simplex. A construction realizing some values, not the invariant itself. Tell: Is one configuration named or the ambient maximum sought?
Clique number. Can encode a fixed-distance problem only after constructing a distance graph for a chosen lambda. Tell: Is the underlying object the metric space or an auxiliary graph?

References¶

[1] Konrad J. Swanepoel, “Equilateral Sets in Finite-Dimensional Normed Spaces”, in Seminar of Mathematical Analysis, 195–237, 2004; full text. Defines lambda-equilateral sets and e_lambda, explains scale suppression in normed spaces, and surveys constructions and upper bounds. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[2] Noga Alon, “Equilateral Sets in the Torus”, author manuscript, 2024. Defines the equilateral number with the unbounded-size infinity convention and studies exact values and bounds for several torus metrics. registry ↩

[3] Tomasz Kobos and Konrad Swanepoel, “Equilateral Dimension of the Planar Banach–Mazur Compactum”, Proceedings of the American Mathematical Society 153(10), 4423–4436, 2025; accepted manuscript. States the general metric invariant, distinguishes scale behavior outside normed spaces, and discusses arbitrarily large finite sets without infinite attainment. registry ↩a ↩b ↩c