Alphabeta Math
TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-5.6-terra)audited 2026-09-14
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Dunford--Pettis for real L1 on a finite measure space

Statement

Assume the Axiom of Choice. Let (S,A,μ) be a finite measure space, and let K be a subset of the real Banach space L1(μ) of almost-everywhere equivalence classes. Then K is relatively weakly compact if and only if it is uniformly integrable. Here uniform integrability is equivalently the conjunction of L1 boundedness and uniform absolute continuity:

supfKf1<,ε>0 δ>0 fK EAμ(E)<δEfdμ<ε.

Facts & Assumptions

[A1]

The Axiom of Choice holds (The Axiom of Choice).

[L3]

On a finite measure space, uniform integrability is exactly L1 boundedness plus the displayed uniform absolute continuity (On a finite measure space, uniform integrability is equivalent to L^1-boundedness plus uniform absolute continuity).

[L4]

Under the principles supplied by [L1], Eberlein--Smulian identifies relative weak compactness with relative weak sequential compactness, and every weakly convergent sequence is norm bounded (Relative weak compactness and three sequential notions, Eberlein–Šmulian theorem, Weakly convergent sequences are norm bounded, A strictly increasing index map satisfies nkk).

[L5]

An individual integrable function has absolutely continuous integral, and under DC a nonempty complete metric space is a Baire space (Absolute continuity of the integral, Under Dependent Choice, a nonempty complete metric space is not a countable union of closed sets with empty interior).

[L6]

Under Countable Choice, real L2(μ) is reflexive. Under the ultrafilter lemma and HB, its closed ball is weakly compact (Reflexivity of Lp for one less p less infinity, Reflexive iff unit ball weakly compact).

[L7]

Holder's inequality gives the bounded inclusion L2L1 on a finite measure space. Since a finite measure is sigma-finite, every member of (L1) is integration against a member of L (Holder's inequality for integrals, including the endpoint cases, On a sigma-finite measure space, every bounded linear functional on Lp is integration against a unique Lq function).

[L8]

Under HB the canonical map JE:EE is linear and isometric. The weak and weak-star topologies are their evaluation initial topologies, and weak-star addition and scalar multiplication are continuous; weak-star space is Hausdorff (Relative Hahn–Banach makes the canonical bidual map an isometry, Weak topology on a normed space, The weak-star topology from finite evaluations, Basic weak star neighborhoods).

Proof

technique · direct

Given: AC, a finite measure space, E=L1(μ;R), and KE.

1.1

Expose every choice principle used below. By [L1], AC supplies Countable Choice for [L2] and [L6], DC for Baire and Eberlein--Smulian, the ultrafilter lemma for Alaoglu and the compactness forms in [L4], [L6], and [L9], and HB for [L4], [L6], and [L8]. No additional choice principle will be left implicit.

givenA1L1
1.2

Fix the Banach and weak-topology conventions. By [L2], E is the real Banach space of classes, not the raw class of integrable representatives. Its weak topology is σ(E,E). By [L8], JE is an isometry and a homeomorphism from weak E to its image in E with the relative weak-star topology: the identity (JEf)(Λ)=Λ(f) makes the two evaluation families identical.

givenL2L8
1.3

Dispose of the empty and null cases. If K=, it is relatively weakly compact and uniformly integrable vacuously. If μ(S)=0, then E={0} and every subset of E is finite, weakly compact, and uniformly integrable. Hence below we may assume K and μ(S)>0.

givenL2L3L4
2.1

A relatively weakly compact family is norm bounded. Suppose K is relatively weakly compact but not norm bounded. Using AC choose fnK with fn1>n. By Eberlein--Smulian in [L4], a subsequence fnj converges weakly in E. The weak-sequence boundedness theorem in [L4] makes its norms bounded, whereas strict increase of the indices gives njj and hence fnj1>njj, a contradiction.

A1L4step 1.1step 1.2choose
2.2

Set up the Baire argument for a weakly null sequence. Let gn0 in E, and let X={[1A]:AA}E with the L1 metric. This set is closed: if a sequence of indicator classes converges in L1, [L2] supplies a subsequence of representatives converging almost everywhere to a representative u of the limit; outside the countable union of the null sets on which those representatives differ from their indicators, u is a pointwise limit of zeros and ones and therefore equals an indicator. Thus X is complete and nonempty.

L2L5L7step 1.1step 1.2

For η>0 define

Xm={[1A]X:Agndμη for every nm}.

Each Xm is closed. Indeed, L1 convergence of indicators means μ(AjA)0, and [L5] applied to the fixed gn gives AjAgn0 for every n. Also mXm=X, because hAh is a bounded functional by the endpoint Holder inequality in [L7], and hence weak nullity gives Agn0 for each fixed A.

2.3

Uniform integrability gives weakly compact truncation approximants. For the reverse implication assume K is uniformly integrable and fix ε>0. By [L3] choose M>0 so that {f>M}f<ε for all fK. The truncation TMf=max(M,min(f,M)) is well defined on classes, measurable, and

L3L6L7step 1.1step 1.2step 1.3construct

fTMf1{f>M}fdμ<ε.

Every TMf belongs to L2 and has L2 norm at most R=Mμ(S). By [L6], RBL2 is weakly compact. The inclusion I:L2E satisfies Ih1μ(S)h2 by [L7] and is weak-to-weak continuous: if ΛE, [L7] writes Λ(h)=hg for some gL; finiteness of μ puts gL2, so this is an L2-continuous functional. Therefore Cε=I(RBL2) is weakly compact and KCε+εBE.

3.1

Apply Baire to obtain one uniform tail neighborhood. The Baire theorem applied to the complete nonempty space X and its closed cover (Xm) gives N, an indicator [1A0], and ρ>0 such that every indicator whose L1 distance from [1A0] is below ρ belongs to XN.

L5step 1.1step 2.2
3.2

Put the uniformly integrable family into a compact bidual closure. Uniform integrability gives a bound C for f1, fK. Let G=JE(K)w in E. Every zG satisfies zC: for ΛE, every weak-star neighborhood of z meets JE(K), so z(Λ)CΛ by letting the neighborhood radius tend to zero. Hence GCBE. Banach--Alaoglu and [L9] make that ball weak-star compact; G, being closed in it, is weak-star compact.

L3L8L9step 1.1step 1.2step 2.3
4.1

Derive uniform absolute continuity for every weakly null sequence. Fix a desired ε>0 and run steps 2.2--3.1 with η=ε/8. If μ(A)<ρ, put B1=A0A and B2=B1A. Both indicators are within μ(A)<ρ of 1A0, so for nN,

L5step 3.1

Agn=B1gnB2gn2η.

Apply this to A{gn0} and A{gn<0}. Their measures are below ρ, and the two signed integrals have absolute value at most 2η, whence Agn4η=ε/2 for nN. For the finitely many n<N, [L5] supplies a common positive δρ for which every corresponding integral is below ε. Thus μ(A)<δ implies Agn<ε for every n: every weakly null sequence has uniformly absolutely continuous integrals.

4.2

Trap the bidual closure in compact neighborhoods of the canonical image. For each ε>0, the set JE(Cε) is weak-star compact, because Cε is weakly compact and JE is weak-to-weak-star continuous. The product JE(Cε)×εBE is compact by [L9], and weak-star addition is continuous by [L8]. Hence

L8L9step 1.2step 2.3step 3.2

Sε:=JE(Cε)+εBE

is weak-star compact and therefore weak-star closed in the Hausdorff weak-star space. Step 2.3 gives JE(K)Sε, so its weak-star closure satisfies GSε for every ε>0.

5.1

Every weakly convergent sequence is uniformly integrable. If fnf, then gn=fnf is weakly null. Step 4.1 gives uniform absolute continuity of (gn), and [L4] gives norm boundedness. The individual function f has absolutely continuous integral by [L5], so AfnAgn+Af makes (fn) uniformly absolutely continuous as well. It is norm bounded by the triangle inequality. Thus [L3] makes the entire sequence (fn) uniformly integrable, including its finite initial segment.

L3L4L5step 1.1step 4.1
5.2

Show that the compact bidual closure actually lies in JE(E). First JE(E) is norm closed. Indeed, if y is in its norm closure, AC chooses xnE with yJExn<1/(n+1). Isometry makes (xn) Cauchy; completeness of E gives xnx, and then JExnJEx=y.

A1L2L8step 1.1step 4.2choose

Now let zG. From GS1/(n+1), AC chooses cnJE(C1/(n+1))JE(E) with zcn1/(n+1). Hence z belongs to the norm closure of JE(E), which is JE(E). Therefore GJE(E).

6.1

Complete the relatively-weakly-compact-to-UI implication. Assume K is relatively weakly compact. If its integrals were not uniformly absolutely continuous, AC would supply ε0>0, fnK, and AnA with μ(An)<1/(n+1) but Anfnε0. By [L4], a subsequence fnj converges weakly. Step 5.1 makes that subsequence uniformly integrable and hence uniformly absolutely continuous by [L3]. But njj makes μ(Anj)0, contradicting the displayed lower bound. Thus K is uniformly absolutely continuous; step 2.1 supplies norm boundedness, so [L3] makes K uniformly integrable.

A1L3L4step 1.1step 2.1step 5.1choose
6.2

Complete the UI-to-relatively-weakly-compact implication. Assume K is uniformly integrable. Step 3.2 makes G weak-star compact and step 5.2 puts it inside JE(E). Since JE is the weak-to-relative-weak-star homeomorphism of step 1.2, JE1(G) is weakly compact. Because ambient weak-star closure agrees with relative closure once GJE(E), this inverse is exactly Kw. Thus K is relatively weakly compact in the sense of [L4].

L4L8L9step 1.2step 3.2step 5.2
7.1

Combine both directions and account for all boundaries. [A1, L3, step 1.3, step 6.1, step 6.2] Steps 6.1 and 6.2 prove the two implications; step 1.3 covers empty K and null measure spaces. Zero truncation levels are unnecessary because uniform integrability permits positive M, and arbitrary positive ε is retained in the bidual intersection argument. The theorem is specifically for real L1; no complex-duality conclusion is silently used. AC is spent only as itemized in step 1.1 and for the explicit countable selections in the norm-boundedness argument, step 5.2, and step 6.1.

A1step 1.1step 1.3step 6.1step 6.2

Depends on

Used by

Dependency tree · two levels

131 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources