Large Set (Combinatorics)¶

A large set of positive integers has a divergent sum of reciprocals, a criterion finer than merely being infinite or having positive density.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13372
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomain: Additive Combinatorics → Mathematics
Aliases: Large Set of Integers, Reciprocal Large Set

Core Idea¶

In this specific combinatorial-number-theory usage, a set A of positive integers is large when its reciprocal sum Σ_{a∈A}1/a diverges. It is small when that sum converges. This is a criterion of weighted abundance, not cardinality: both the primes and the squares are infinite, but reciprocal primes diverge while reciprocal squares converge. Nor does “large” mean positive natural density; the primes have natural density zero yet remain large.^[1]^[2]

The criterion gains meaning as a threshold in other problems. For a set of positive integer exponents, the classical Müntz–Szász form says the span of 1 and x^a (a∈A) is dense in continuous functions on [0,1] exactly when the reciprocal sum diverges. In additive combinatorics, a large set is known to contain a nontrivial three-term arithmetic progression, while Erdős's proposed arbitrary-length conclusion remains open in general. The primes' arbitrary-length progressions are a separately proved special case, not proof of the full conjecture.^[2]^[3]^[4]^[5]

Structural Signature¶

Sig role-phrases:

Positive-integer set: a distinct-index subset A⊆{1,2,...} is being classified.
Reciprocal weight: each selected integer a contributes positive mass 1/a.
Divergence test: unbounded partial sums define large; finite total defines small.
Downstream hypothesis: approximation or progression results may take large status as a premise, with their own further assumptions and proof status.
Density comparator: ordinary frequency may help intuition but does not replace the reciprocal-sum test.^[2]^[1]^[3]

Condensed: integer subset + reciprocal mass → divergent/large or convergent/small → setting-specific consequences.

What It Is Not¶

It is not “infinite”: squares, powers of two and many other infinite sets have convergent reciprocal sums. It is not “positive density”: primes have density zero but their reciprocals diverge. It is not a direct measure of the largest element, the number of elements alone, or Lebesgue measure. “Every large set has arithmetic progressions of every length” is an open conjecture, not a definition or theorem. In Müntz approximation, the claim here includes exponent 0 through the constant function 1 and restricts to positive integer exponents; dropping those qualifications changes the theorem's form.^[1]^[2]^[3]

Scope of Application¶

Elementary comparisons calibrate the boundary. For even numbers, Σ_{n≥1}1/(2n)=(1/2)Σ_{n≥1}1/n, so the set is large. For squares, Σ_{n≥1}1/n² converges, so the set is small. A subset of a small set remains small; a superset of a large set remains large; adding or removing finitely many terms cannot change either outcome. These follow from positivity and ordinary convergence comparison rather than from a new named theorem.^[2]

Prime numbers are the instructive sparse case. Tao's authored number-theory notes derive Euler's divergence of Σ_p1/p; their natural density is zero. Bloom and Sisask's original result establishes the three-term progression implication for any large set. Green and Tao separately prove arbitrarily long progressions within the primes. Those statements are compatible with Erdős's broader conjecture still being unsettled for longer progressions in arbitrary large sets.^[1]^[3]^[4]^[5]

In approximation theory, Borwein and Erdélyi's research paper states the classical Müntz theorem in its introduction and supplies a proof in its broader treatment. For a sequence of distinct positive integer exponents, bounded away from zero automatically, and the constant included, the same reciprocal threshold is exact for density of the associated powers. “Dense” here means uniform approximation of every continuous function on [0,1] by finite linear combinations of the allowed powers; it does not mean the exponent set itself is dense among integers.^[2]

Clarity¶

One must specify what is being summed. If A={2,4,6,...}, summing 1/a over members yields half the harmonic series and diverges. If A={1,4,9,...}, it yields a p-series with exponent 2 and converges. Both sets are endless and have increasing entries; only one is large under this definition. The primes show that even asymptotic density cannot be substituted: frequency tends to zero while reciprocal mass accumulates without bound.^[1]

Manages Complexity¶

The large/small distinction classifies diverse integer subsets with one exact numerical test. It permits comparison and monotonicity arguments without carrying a detailed counting function in every proof. The test is coarse in another way: it says nothing directly about exact progression length, distribution in short intervals, or a basis of functions until a separate theorem links those matters. Its value is as a reusable hypothesis, not as a universal structure detector.

Abstract Reasoning¶

Because all summands are positive, enlarging a large set preserves divergence and thinning a small set preserves convergence. Finite changes add a finite number to the total and are irrelevant. But thinning a large set can cross the boundary: retain only squares among positive integers and the reciprocal sum converges. Thus largeness is a property of the infinite tail, not of the first million terms or mere unboundedness.^[2]

For approximation, take allowed exponents A={2,4,6,...}. The reciprocal sum diverges, so the stated integer Müntz–Szász theorem predicts density of span{1,x²,x⁴,...} in C[0,1]. This is also intuitively compatible with x↦x² being a continuous one-to-one transformation on [0,1]; the theorem gives the exact criterion for more general integer exponent subsets. Replace evens by square exponents {1,4,9,...} and the sum converges, so the theorem denies density of that restricted power span.^[2]

Knowledge Transfer¶

The reciprocal test transfers unchanged from exponent sets to additive-combinatoric sets: A is large or small by the same series. The conclusion does not transfer automatically. Müntz–Szász is an established if-and-only-if approximation theorem under its stated hypotheses; three-term progression existence is established by Bloom–Sisask; arbitrary-length progressions for all large sets remain a conjecture. Green–Tao settles one important large set, the primes, by a separate argument.^[2]^[3]^[4]

Examples¶

Even exponents in Müntz approximation¶

Let A={2,4,6,...}. Its reciprocal sum is (1/2)Σ1/n, hence divergent. Include exponent zero by adding the constant 1. Under the positive-integer Müntz–Szász theorem, finite linear combinations of 1,x²,x⁴,... are dense in C[0,1]. If exponents were only perfect squares, the reciprocal sum would converge and that theorem's density condition would fail. The comparison is a specific use of the threshold, not a declaration that every infinite power family is complete.^[2]

Mapped back: the even integers are the positive-integer set; 1/(2n) is each reciprocal weight; harmonic divergence makes the set large; uniform approximation is the downstream conclusion; density 1/2 is a comparator, not the theorem's formal hypothesis.

Primes in the progression question¶

The prime set has divergent reciprocal sum by Euler's theorem, despite zero natural density. Bloom–Sisask therefore supplies the general three-term large-set conclusion; an explicit short instance is 3,5,7. Green–Tao proves arbitrary prime progression lengths by a separate theorem. None of this settles Erdős's claim for every large set at length four and beyond.^[1]^[3]^[4]^[5]

Mapped back: primes are the integer set; 1/p terms carry reciprocal weight; their sum diverges; arithmetic progressions are the downstream question with a proved three-term and separately proved prime-only arbitrary-length result; zero density shows why ordinary frequency is not the criterion.

Structural Tensions¶

No intrinsic opposed-cost tension is established by this reciprocal-summability class. Zero natural density and a divergent reciprocal sum can coexist mathematically, as for primes; there is no choice to balance between them. Diagnostic: is a claim measuring |A∩[1,N]|/N or the partial reciprocal sum, and does it silently equate those different quantities? The status of a downstream theorem is likewise a proof boundary, not a competing pressure within a large set: the positive-integer Müntz conclusion and the three-term progression result are established, while arbitrary lengths for every large set remain conjectural. Diagnostic: which exponent assumptions, progression length and theorem are actually proved for this use?^[1]^[2]^[3]^[5]

Structural–Framed Character¶

The entry is strongly structural: a positive series either converges or diverges, independent of taste. Evaluative weight enters when researchers choose reciprocal divergence as the useful notion of “large” for an approximation or progression problem; the adjective itself is disciplinary framing. Human proof practice defines and applies the criterion, but no institution makes primes large by decree. Its vocabulary travels literally between combinatorial number theory and Müntz approximation when the underlying positive-integer reciprocal sum is the same. Calling an organization or data set “large” by volume is ordinary language, not this identity. Its character: an exact reciprocal-summability class whose broad name conceals a precise and sometimes counterintuitive sparseness threshold.

Structural Core vs. Domain Accent¶

The skeletal relation is an infinite collection classified by whether individually decreasing weights retain unbounded total mass. No verified live prime for such weighted largeness has been established; an eventual prime would need independent non-number-theoretic cases and a stable boundary. The domain-bound mechanism is reciprocal weights of positive integers, convergence comparison, and theorems using that exact series. Large Set fails the prime bar because replacing 1/a with another weighting changes membership and downstream results; the adjective alone is not a portable abstraction.

This entry is a kind of Set and Membership.

The live Set and Membership prime is the strict parent: membership in a positive-integer subset precedes the reciprocal-divergence differentia. “Small set” is the convergence contrast and the second frozen requested-title route, not a separately admitted node in this bundle. No unverified catalog edge to asymptotic density or arithmetic progression is created.

Relationships to Other Abstractions¶

Current abstraction Large Set (Combinatorics) Domain-specific

Parents (1) — more general patterns this builds on

Large Set (Combinatorics) is a kind of Set and Membership Prime

A reciprocal-large set is a set of positive integers with divergent reciprocal sum.

Hierarchy path (1) — routes to 1 parentless root

Large Set (Combinatorics) → Set and Membership

Neighborhood in Abstraction Space¶

Large Set (Combinatorics) sits in a sparse region of the domain-specific corpus (81^st percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Number-Theoretic Properties & Tests (20 abstractions)

Nearest neighbors

Yule–Simon Distribution — 0.83
Average Order of an Arithmetic Function — 0.82
Blackwell–Girshick Equation — 0.82
Kolmogorov's Three-Series Theorem — 0.82
Least Common Multiple — 0.82

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Infinite sets can be small. Positive-density sets are large but large sets can have zero density. Müntz's criterion requires the constant function and the stated exponent conditions; a broad version for arbitrary real exponents uses a differently qualified series. Erdős's arbitrary-length large-set progression statement is a conjecture, whereas the three-term result and prime-only arbitrary-length result are theorems.

References¶

[1] Terence Tao, 254A notes on elementary multiplicative number theory, Euler's reciprocal-prime divergence. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[2] Borwein and Erdélyi, “The Full Müntz Theorem in C[0, 1] and L1[0, 1]”, introduction, pp. 1–2 (classical bounded-away-from-zero reciprocal criterion); author-hosted research paper. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k

[3] Bloom and Sisask, “Breaking the logarithmic barrier in Roth's theorem”, original three-term case. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[4] Green and Tao, “The primes contain arbitrarily long arithmetic progressions”, original theorem. registry ↩a ↩b ↩c ↩d

[5] Paul Erdős, original problem discussion, reciprocal-divergence progression conjecture. registry ↩a ↩b ↩c ↩d