Alphabeta Math
LemmaStatement: AI-adaptedProof: AI-adaptedPipeline-generatedprecheck passaudited 2026-10-08
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.

Length in a metric target: lower semicontinuity and arc-length reparametrization

Statement

Let (X,d) be a metric space (Metric space: d(x,y)=0 iff x=y, symmetry, and the triangle inequality; pseudometric and ultrametric) and let a<b be real numbers. A path in X is a map γ ⁣:[a,b]→X. A partition of [a,b] is a finite sequence a=t0<t1<⋯<tm=b; the polygonal sum of γ over that partition is ∑i=1md(γ(ti−1),γ(ti)), and the length of γ is the supremum L(γ):=sup⁡{ ∑i=1md(γ(ti−1),γ(ti)):a=t0<⋯<tm=b }∈[0,∞] For a singleton interval [u,u], its only partition is the one-term sequence u, its polygonal sum is the empty sum 0, and its path length is 0. For nondegenerate intervals the supremum is taken in the extended reals (The extended real line R‾=R∪{−∞,+∞}, its order, and the arithmetic that is left undefined, Every subset of R‾ has a least upper bound and a greatest lower bound in R‾, agreeing with the real supremum and infimum on nonempty sets bounded in R, Upper bound, least upper bound, and strict upper bound); γ is rectifiable if L(γ)<∞. For [u,v]⊆[a,b] write γ∣[u,v] for the restriction of γ to [u,v], a path on [u,v], and L(γ∣[u,v]) for its length.

Chord bound and additivity at the initial point. For a≤u≤v≤b one has d(γ(u),γ(v))≤L(γ∣[u,v]), and for a≤v≤w≤b one has L(γ∣[a,w])=L(γ∣[a,v])+L(γ∣[v,w]); in particular s(t):=L(γ∣[a,t]) is nondecreasing on [a,b].

(i) Lower semicontinuity. If γk,γ ⁣:[a,b]→X are paths with sup⁡t∈[a,b]d(γk(t),γ(t))→0 (uniform convergence, Uniform convergence, and the topology of uniform convergence: the metric topology of the uniform metric on YX and on C(X,Y)), then L(γ)≤lim inf⁡k→∞L(γk), the limit inferior being taken in [0,∞] (Limit superior and limit inferior of a nonnegative extended-real sequence).

(ii) Arc-length parametrization. If γ is continuous (Continuity of a map between metric spaces, at a point and globally, in the ε-δ form) and rectifiable, with L:=L(γ)<∞, then s(t):=L(γ∣[a,t]) defines a continuous nondecreasing surjection s ⁣:[a,b]→[0,L] with s(a)=0 and s(b)=L; there is a unique map γˉ ⁣:[0,L]→X with γ=γˉ∘s; and γˉ is 1-Lipschitz with L(γˉ∣[r,q])=q−r for all 0≤r≤q≤L. If L=0 then γ is constant, [0,L]={0}, and γˉ is that constant.

(iii) Equicontinuity of bounded arc-length families. Call a path γ ⁣:[a,b]→X arc-length parametrized when L(γ)<∞ and L(γ∣[s,t])=t−sb−a L(γ)for all a≤s≤t≤b. If M<∞ and the paths γk ⁣:[0,1]→X are arc-length parametrized with L(γk)≤M for every k, then each γk is M-Lipschitz and the family (γk) is equicontinuous and uniformly equicontinuous (Equicontinuity at a point, uniform equicontinuity, and pointwise boundedness of a family of maps between metric spaces).

(iv) Application to polyhedral gluings. If X carries the chain metric d of an isometric polyhedral gluing with hypotheses (H1)-(H3) (Abstract isometric polyhedral gluings and the chain metric), then d is a metric on X (The chain metric is a metric, its topology is the weak topology, and the space is proper and complete (1)), and clauses (i)-(iii) apply verbatim to paths in (X,d).

No compactness, completeness, convexity or local structure of X is used in (i)-(iii), and no Euclidean-target theorem on arc length is invoked: the statements are proved in the stated metric generality.

Facts & Assumptions

Given: A metric space (X,d), real numbers a<b, and paths γ,γk ⁣:[a,b]→X as in the Statement, together with the partition sums, the length L and the restrictions γ∣[u,v] defined there.

[F1]

As a metric space, (X,d) satisfies d(x,x)=0, d(x,y)=d(y,x) and d(x,z)≤d(x,y)+d(y,z) for all x,y,z∈X, and d(x,y)≥0. Metric space: d(x,y)=0 iff x=y, symmetry, and the triangle inequality; pseudometric and ultrametric, Nonnegativity of a metric is a consequence of the other axioms, not an axiom

[F2]

Reverse triangle inequality: ∣d(x,z)−d(y,z)∣≤d(x,y) for all x,y,z∈X. The reverse triangle inequality ∣d(x,z)−d(y,z)∣≤d(x,y) in any metric space

[F3]

The extended real line R‾ is totally ordered, its order restricts to that of R, and every subset of R‾ has a least upper bound and a greatest lower bound in R‾; a least upper bound of a set is an upper bound of it lying below every upper bound, and a greatest lower bound is a lower bound lying above every lower bound. The extended real line R‾=R∪{−∞,+∞}, its order, and the arithmetic that is left undefined, Every subset of R‾ has a least upper bound and a greatest lower bound in R‾, agreeing with the real supremum and infimum on nonempty sets bounded in R, Upper bound, least upper bound, and strict upper bound, Greatest lower bound (infimum)

[F4]

A sequence of maps into a metric space converges uniformly when one index serves every point of the domain: for every real ε>0 there is K with d(fk(x),f(x))<ε for every x and every k≥K. Uniform convergence, and the topology of uniform convergence: the metric topology of the uniform metric on YX and on C(X,Y)

[F5]

For a sequence (ak) in [0,+∞] the limit inferior is the supremum of the tail infima, lim inf⁡kak=sup⁡N∈Ninf⁡k≥Nak, all suprema and infima taken in R‾. Limit superior and limit inferior of a nonnegative extended-real sequence

[F6]

Real sequences: xk→x means that for every real ε>0 there is K with ∣xk−x∣<ε for all k≥K; a finite sum of convergent real sequences converges to the sum of the limits; and if xk≤yk from some index on, then lim⁡kxk≤lim⁡kyk whenever both limits exist. Limits and Cauchy sequences of reals, Algebra of limits: sums, scalar multiples, products and quotients, Limits preserve non-strict inequalities

[F7]

Archimedean property and translation: for every real ε>0 there is a natural number n≥1 with 1/n<ε, and x<y implies x+z<y+z for reals x,y,z. For every ε>0 in a complete ordered field there is a natural n≥1 with 1/n<ε, Order is preserved by adding a constant and by adding inequalities

[F8]

The real line is the complete ordered field: every nonempty set of reals that is bounded above has a real least upper bound. Every interval is connected, and a continuous real-valued map on a connected space assumes every value between any two of its values. Complete ordered field (least-upper-bound property), The connected subspaces of R with its usual topology are exactly the order-convex subsets, the published characterisation transported by the identification of the two descriptions of "open in R", A real-valued continuous map on a connected space has order-convex image, so it takes every value between any two of its values

[F9]

Metric continuity: γ is continuous at t when for every real ε>0 there is a real δ>0 with d(γ(t′),γ(t))<ε for every t′ with ∣t′−t∣<δ. Continuity of a map between metric spaces, at a point and globally, in the ε-δ form

[F10]

Uniform equicontinuity of a family of maps between metric spaces asks for one δ serving every member of the family and every pair of points within δ; uniform equicontinuity implies equicontinuity. Equicontinuity at a point, uniform equicontinuity, and pointwise boundedness of a family of maps between metric spaces

[F11]

For an isometric polyhedral gluing with hypotheses (H1)-(H3), the chain metric candidate is a metric on the gluing and its metric topology is the weak topology. Abstract isometric polyhedral gluings and the chain metric, The chain metric is a metric, its topology is the weak topology, and the space is proper and complete

Proof

Given: A metric space (X,d), real numbers a<b, a path γ ⁣:[a,b]→X and, where the clause says so, paths γk as in the Statement.

1.1F1F3F7

(Chord bound, additivity at the initial point, monotonicity.) For a≤u≤v≤b if u=v both the chord and the length are 0 by the singleton convention and [F1]; if u<v the sequence u,v is a partition of [u,v] with polygonal sum d(γ(u),γ(v)), so d(γ(u),γ(v))≤L(γ∣[u,v]) because L(γ∣[u,v]) is the least upper bound of all polygonal sums [F3]. For a≤v≤w≤b let first P be a partition of [a,w]; inserting v if it is absent gives a partition P′ of [a,w] whose polygonal sum is at least that of P by the triangle inequality [F1], and which is the union of a partition of [a,v] and a partition of [v,w]. Hence every polygonal sum of γ∣[a,w] is at most L(γ∣[a,v])+L(γ∣[v,w]), so L(γ∣[a,w])≤L(γ∣[a,v])+L(γ∣[v,w]) [F3]. Conversely, if both summands are finite, then for every real ε>0 there are partitions of [a,v] and of [v,w] with sums exceeding L(γ∣[a,v])−ε/2 and L(γ∣[v,w])−ε/2, and their union is a partition of [a,w], so L(γ∣[a,w])≥L(γ∣[a,v])+L(γ∣[v,w])−ε; as the real ε>0 is arbitrary, [F7] gives L(γ∣[a,w])≥L(γ∣[a,v])+L(γ∣[v,w]). If L(γ∣[a,v])=+∞, then every partition of [a,v] extends by w to a partition of [a,w] with sum no smaller, since the added chord is nonnegative, so L(γ∣[a,w])≥L(γ∣[a,v])=+∞ and the two sides agree; and if L(γ∣[a,v])<∞ while L(γ∣[v,w])=+∞, then for every real M there is a partition of [v,w] with sum exceeding M, and adjoining any fixed partition of [a,v] yields a partition of [a,w] with sum exceeding M, so L(γ∣[a,w])=+∞. The degenerate cases v=a or v=w follow directly from the singleton convention. This proves additivity, and monotonicity of s follows from s(w)=s(v)+L(γ∣[v,w])≥s(v) for v≤w because lengths are suprema of sums of nonnegative terms [F1, F3].

1.2F1F2F3F4F5F6F7

(Lower semicontinuity.) Suppose u∈R‾ is an upper bound of the tail infima ℓN:=inf⁡k≥NL(γk) of [F5]; I show u≥L(γ). Assume u<L(γ). Then u is a real number, because u=+∞ contradicts u<L(γ)≤+∞ and u=−∞ is impossible as all L(γk)≥0 forces every ℓN≥0 [F1, F3]. Since L(γ) is the least upper bound of the polygonal sums of γ [F3] and u<L(γ), some partition P of [a,b] has polygonal sum S(P)>u. By [F7] choose a natural number n≥1 with 1/n<S(P)−u. For each partition point ti, the convergence sup⁡td(γk(t),γ(t))→0 gives d(γk(ti),γ(ti))→0 [F4]; consequently Sk(P):=∑id(γk(ti−1),γk(ti)) converges to S(P) by [F2] and [F6]. So there is K with Sk(P)>S(P)−1/n>u for all k≥K. Since Sk(P)≤L(γk) by [F3], this gives ℓK=inf⁡k≥KL(γk)≥S(P)−1/n>u, contradicting that u is an upper bound of the ℓN. Hence every upper bound u of the tail infima satisfies u≥L(γ), and since the least such upper bound is lim inf⁡kL(γk) [F5], L(γ)≤lim inf⁡kL(γk).

2.1step 1.1

(The arclength function.) From now on assume that γ is continuous and that L:=L(γ)<∞; write s(t):=L(γ∣[a,t]). Then s(a)=0, s(b)=L, and s is nondecreasing by [step 1.1]; moreover s(w)−s(v)=L(γ∣[v,w]) for all a≤v≤w≤b, by additivity at the initial point [step 1.1].

2.2F1F10step 1.1

(Equicontinuity of bounded arc-length families.) Let γk ⁣:[0,1]→X be arc-length parametrized with L(γk)≤M<∞. For 0≤s≤t≤1 the chord bound [step 1.1] and the definition of arc-length parametrized give d(γk(s),γk(t))≤L(γk∣[s,t])=(t−s)L(γk)≤M(t−s), so every γk is M-Lipschitz. If M=0 then every γk is constant by separation [F1] and the family is uniformly equicontinuous with any δ>0. If M>0 and ε>0 is real, take δ:=ε/M>0; then d(γk(s),γk(t))≤M∣s−t∣<ε for all k and all s,t with ∣s−t∣<δ, so the family is uniformly equicontinuous, hence equicontinuous [F10].

3.1F1F2F3F9step 2.1

(s is continuous.) Fix t∈[a,b) and a real δ>0; put ε:=δ/2. Since L is the least upper bound of the polygonal sums, choose a partition P of [a,b] with polygonal sum S(P)>L−ε [F3], and insert t into P if absent (the sum only increases [F1]); write A and B for the sums of the parts of P on [a,t] and on [t,b], so that A+B=S(P), and let t+ be the successor of t in P when t<b. By continuity of γ at t [F9] choose η∈(0,t+−t] with d(γ(t),γ(t′))<ε for ∣t′−t∣<η; I claim s(t+h)−s(t)≤δ for every h∈(0,η). Let R be a partition of [t,t+h]; then the union of the parts of P on [a,t], of R, and of the part of P on [t+,b] is a partition of [a,b] whose polygonal sum is A+Sum(R)+d(γ(t+h),γ(t+))+(B−d(γ(t),γ(t+)))≤L, so Sum(R)≤L−S(P)+d(γ(t),γ(t+))−d(γ(t+h),γ(t+))≤L−S(P)+d(γ(t),γ(t+h))<ε+ε=2ε by the reverse triangle inequality [F2]. Taking the supremum over R gives s(t+h)−s(t)=L(γ∣[t,t+h])≤2ε=δ [step 2.1]. The same argument applied to partitions of [t−h,t] gives left-continuity at every t∈(a,b]; hence s is continuous.

4.1F8step 2.1step 3.1

(s is surjective.) The interval [a,b] is connected [F8] and s ⁣:[a,b]→R is continuous [step 3.1] with s(a)=0 and s(b)=L [step 2.1]; by the intermediate value theorem [F8], for every real r with 0≤r≤L there is t∈[a,b] with s(t)=r.

5.1F1step 1.1step 2.1step 4.1

(Factorisation through s.) If a≤u≤v≤b satisfy s(u)=s(v), then L(γ∣[u,v])=s(v)−s(u)=0 [step 2.1], so d(γ(u),γ(v))≤0 by the chord bound [step 1.1] and hence γ(u)=γ(v) by separation [F1]. Since s is surjective [step 4.1], there is therefore a well-defined and unique map γˉ ⁣:[0,L]→X with γ=γˉ∘s, namely γˉ(r):=γ(t) for any t with s(t)=r. If L=0 then s vanishes identically and the same argument with u=a, v=b shows that γ is constant; then [0,L]={0} and γˉ is that constant.

6.1F8step 1.1step 2.1step 3.1step 4.1step 5.1

(γˉ is 1-Lipschitz and has unit speed.) Given 0≤r≤q≤L, choose by [step 4.1] points u,v∈[a,b] with s(u)=r and s(v)=q, and relabel so that u≤v; then d(γˉ(r),γˉ(q))=d(γ(u),γ(v))≤L(γ∣[u,v])=s(v)−s(u)=q−r by the chord bound [step 1.1] and [step 2.1], so γˉ is 1-Lipschitz. If r=q, the singleton-interval convention gives L(γˉ∣[r,r])=0=q−r; hence assume r<q for the length identity. Define, for each real ρ with 0≤ρ≤L, the number uρ:=sup⁡{t∈[a,b]:s(t)≤ρ}, which exists by the least-upper-bound property [F8] and satisfies s(uρ)=ρ: indeed s(uρ)≤ρ because points of the set approach uρ from below and s is continuous [step 3.1], while if uρ<b then every t>uρ has s(t)>ρ and continuity gives s(uρ)≥ρ, and if uρ=b then s(uρ)=L≤ρ≤L. Also uρ≤uρ′ for ρ≤ρ′, and γˉ(ρ)=γ(uρ) [step 5.1]. Now let r=ρ0<⋯<ρN=q be a partition of [r,q]; its polygonal sum for γˉ is ∑id(γ(uρi−1),γ(uρi))≤∑iL(γ∣[uρi−1,uρi])=∑i(ρi−ρi−1)=q−r by the chord bound [step 1.1] and [step 2.1], so L(γˉ∣[r,q])≤q−r. Conversely let v0<⋯<vN be a partition of [ur,uq]; its polygonal sum for γ is ∑id(γˉ(s(vi−1)),γˉ(s(vi))), the points s(vi) form a nondecreasing sequence from r=s(v0) to q=s(vN), and after deleting repetitions this is a partition of [r,q] whose polygonal sum for γˉ is the same number; hence every polygonal sum of γ∣[ur,uq] is at most L(γˉ∣[r,q]), and q−r=L(γ∣[ur,uq])≤L(γˉ∣[r,q]) [step 2.1]. Therefore L(γˉ∣[r,q])=q−r.

7.1F11step 1.2step 2.2step 6.1∎

(Application.) Let X carry the chain metric d of an isometric polyhedral gluing with (H1)-(H3); by [F11] this d is a metric on X, so clauses (i)-(iii), whose statements and proofs mention only the metric space (X,d), hold verbatim for paths in (X,d); in particular the constants a<b are arbitrary reals and no hypothesis beyond the metric axioms was used.

Depends on

Used by

Dependency tree · two levels

83 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