Alphabeta Math
TheoremStatement: AI-adaptedProof: AI-generatedSession-authored (Fable 5 assisted)precheck passaudited 2026-07-27
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.

Every metric space has a completion, constructed as the equivalence classes of its Cauchy sequences

Statement

Let (X,d)(X,d) be a metric space (Metric space: d(x,y)=0d(x,y) = 0 iff x=yx = y, symmetry, and the triangle inequality; pseudometric and ultrametric) and let C\mathcal{C} be the set of all Cauchy sequences in XX (Cauchy sequence in a metric space). Then:

  1. For all x=(xn)x = (x_n) and y=(yn)y = (y_n) in C\mathcal{C} the real sequence (d(xn,yn))n\big(d(x_n,y_n)\big)_n converges, so ρ(x,y)  :=  limnd(xn,yn)\rho(x,y) \;:=\; \lim_{n} d(x_n,y_n) is a single well-determined real (The Cauchy criterion from the least-upper-bound property: in a complete ordered field every Cauchy sequence converges, A sequence has at most one limit).
  2. The relation xy:ρ(x,y)=0x \sim y :\Longleftrightarrow \rho(x,y) = 0 is an equivalence relation on C\mathcal{C}. Write X^:=C/ ⁣\widehat{X} := \mathcal{C}/\!\sim for the set of its classes and [x][x] for the class of xx.
  3. d^([x],[y]):=ρ(x,y)\widehat{d}\big([x],[y]\big) := \rho(x,y) does not depend on the chosen representatives, and d^\widehat{d} is a metric on X^\widehat{X}.
  4. The map ι:XX^\iota : X \to \widehat{X} sending pp to the class of the constant sequence at pp is an isometric embedding with dense image (Isometry, isometric embedding, and the subspace metric on a subset, Interior, closure, boundary, limit point, isolated point and dense subset of a metric space).
  5. (X^,d^)(\widehat{X}, \widehat{d}) is complete.

Consequently ((X^,d^),ι)\big((\widehat{X},\widehat{d}), \iota\big) is a completion of (X,d)(X,d) (A completion of a metric space: a complete metric space together with an isometric embedding onto a dense subspace), and every metric space has a completion.

The notation is kept honest. A Cauchy sequence in XX need not converge in XX, so no symbol limnxn\lim_n x_n appears anywhere below; the only limits taken are limits of real sequences, and each is written only after its existence has been proved. The equivalence relation is defined and verified here rather than cited, as was done for The integers as equivalence classes of pairs of naturals, so that the construction is self-contained and its transitivity argument is visible at the point of use.

Facts & Assumptions

Given: A metric space (X,d)(X,d); the set C\mathcal{C} of Cauchy sequences in XX; elements x=(xn)x = (x_n), y=(yn)y = (y_n), z=(zn)z = (z_n) of C\mathcal{C}; a real ε>0\varepsilon > 0.

[A1]

Cauchyness: for every real ε>0\varepsilon > 0 there is KK with d(xn,xm)<εd(x_n,x_m) < \varepsilon for all n,mKn,m \ge K (Cauchy sequence in a metric space, The rationals embed densely in the reals).

[L1]

Reverse triangle inequality: d(u,w)d(v,w)d(u,v)|d(u,w) - d(v,w)| \le d(u,v) (The reverse triangle inequality d(x,z)d(y,z)d(x,y)|d(x,z) - d(y,z)| \le d(x,y) in any metric space); with the triangle inequality for the absolute value (Basic properties of the absolute value) this gives the quadrilateral estimate d(u,v)d(u,v)d(u,u)+d(v,v)|d(u,v) - d(u',v')| \le d(u,u') + d(v,v').

[L2]

Every Cauchy sequence of reals converges, and the limit of a real sequence is unique, which licenses the notation limnan\lim_n a_n for a sequence already known to converge (The Cauchy criterion from the least-upper-bound property: in a complete ordered field every Cauchy sequence converges, A sequence has at most one limit, Limits and Cauchy sequences of reals).

[L3]

Limits of reals preserve non-strict inequalities holding eventually, and behave additively (Limits preserve non-strict inequalities, Algebra of limits: sums, scalar multiples, products and quotients); a constant sequence converges to that constant.

[L6]

Countable choice: a family (Ak)kN(A_k)_{k \in \mathbb{N}} of nonempty sets admits kakk \mapsto a_k with akAka_k \in A_k (The Axiom of Countable Choice (ACω\mathrm{AC}_\omega)).

[L7]

For every real η>0\eta > 0 there is a natural KK with 1/(K+1)<η1/(K+1) < \eta (For every ε>0\varepsilon > 0 in a complete ordered field there is a natural n1n \ge 1 with 1/n<ε1/n < \varepsilon).

Proof

technique · constructive
1.1

The quadrilateral estimate of [L1] gives d(xn,yn)d(xm,ym)d(xn,xm)+d(yn,ym)|d(x_n,y_n) - d(x_m,y_m)| \le d(x_n,x_m) + d(y_n,y_m) for all n,mn,m.

L1
2.1

Given a real ε>0\varepsilon > 0, [A1] supplies KxK_x and KyK_y with d(xn,xm)<ε/2d(x_n,x_m) < \varepsilon/2 and d(yn,ym)<ε/2d(y_n,y_m) < \varepsilon/2 for indices beyond them; with K:=max{Kx,Ky}K := \max\{K_x,K_y\} the sequence (d(xn,yn))n\big(d(x_n,y_n)\big)_n satisfies d(xn,yn)d(xm,ym)<ε|d(x_n,y_n) - d(x_m,y_m)| < \varepsilon for n,mKn,m \ge K, so it is a Cauchy sequence of reals.

step 1.1A1L8
3.1

By [L2] that sequence converges and its limit is unique, so ρ(x,y):=limnd(xn,yn)\rho(x,y) := \lim_n d(x_n,y_n) is a single well-determined real: claim 1.

step 2.1L2construct
4.1

ρ\rho is nonnegative and symmetric, and satisfies ρ(x,z)ρ(x,y)+ρ(y,z)\rho(x,z) \le \rho(x,y) + \rho(y,z): the terms d(xn,yn)d(x_n,y_n) are nonnegative, d(xn,yn)=d(yn,xn)d(x_n,y_n) = d(y_n,x_n), and d(xn,zn)d(xn,yn)+d(yn,zn)d(x_n,z_n) \le d(x_n,y_n) + d(y_n,z_n) for every nn, and all three pass to the limit.

step 3.1L3L4
5.1

Claim 2: \sim is reflexive since d(xn,xn)=0d(x_n,x_n) = 0 for every nn; symmetric since ρ\rho is; and transitive, since ρ(x,y)=ρ(y,z)=0\rho(x,y) = \rho(y,z) = 0 gives 0ρ(x,z)0+0=00 \le \rho(x,z) \le 0 + 0 = 0. So \sim is an equivalence relation and X^=C/ ⁣\widehat{X} = \mathcal{C}/\!\sim is defined.

step 4.1L4construct
6.1

Claim 3, well-definedness: if xxx \sim x' and yyy \sim y' then, by [L1] applied termwise, d(xn,yn)d(xn,yn)d(xn,xn)+d(yn,yn)|d(x_n,y_n) - d(x'_n,y'_n)| \le d(x_n,x'_n) + d(y_n,y'_n); passing to the limit gives ρ(x,y)ρ(x,y)ρ(x,x)+ρ(y,y)=0|\rho(x,y) - \rho(x',y')| \le \rho(x,x') + \rho(y,y') = 0, so ρ(x,y)=ρ(x,y)\rho(x,y) = \rho(x',y'). Hence d^([x],[y]):=ρ(x,y)\widehat{d}([x],[y]) := \rho(x,y) is a well-defined function on X^×X^\widehat{X} \times \widehat{X}.

step 3.1step 5.1L1L3
7.1

d^\widehat{d} is a metric: symmetry and the triangle inequality are step 4.1 read on classes, and d^([x],[y])=0\widehat{d}([x],[y]) = 0 says ρ(x,y)=0\rho(x,y) = 0, which says xyx \sim y, which says [x]=[y][x] = [y]. This completes claim 3.

step 4.1step 5.1step 6.1L4
8.1

Claim 4: for pXp \in X the constant sequence at pp is Cauchy, so ι(p):=[(p)n]\iota(p) := [(p)_n] is defined, and d^(ι(p),ι(q))=limnd(p,q)=d(p,q)\widehat{d}(\iota(p),\iota(q)) = \lim_n d(p,q) = d(p,q), a constant sequence; so ι\iota is an isometric embedding.

step 7.1L3construct
9.1

Density: let [x]X^[x] \in \widehat{X} and let ε>0\varepsilon > 0 be real. By [A1] there is NN with d(xn,xm)<ε/2d(x_n,x_m) < \varepsilon/2 for all n,mNn,m \ge N; in particular d(xn,xN)ε/2d(x_n,x_N) \le \varepsilon/2 for all nNn \ge N, so d^([x],ι(xN))=limnd(xn,xN)ε/2<ε\widehat{d}([x], \iota(x_N)) = \lim_n d(x_n,x_N) \le \varepsilon/2 < \varepsilon. Hence every ball around [x][x] meets ι[X]\iota[X], that is ι[X]\iota[X] is dense, completing claim 4.

step 8.1A1L3L5
10.1

Claim 5: let (ξ(k))(\xi^{(k)}) be a Cauchy sequence in (X^,d^)(\widehat{X},\widehat{d}). For each kk the set Ak:={pX:d^(ξ(k),ι(p))<1/(k+1)}A_k := \{\, p \in X : \widehat{d}(\xi^{(k)}, \iota(p)) < 1/(k+1) \,\} is nonempty by step 9.1, so [L6] supplies akAka_k \in A_k for every kk, that is a sequence (ak)(a_k) in XX with d^(ξ(k),ι(ak))<1/(k+1)\widehat{d}(\xi^{(k)}, \iota(a_k)) < 1/(k+1).

step 9.1L6L7choose
11.1

(ak)(a_k) is Cauchy in XX: since ι\iota is isometric, d(ak,al)=d^(ι(ak),ι(al))1/(k+1)+d^(ξ(k),ξ(l))+1/(l+1)d(a_k,a_l) = \widehat{d}(\iota(a_k),\iota(a_l)) \le 1/(k+1) + \widehat{d}(\xi^{(k)},\xi^{(l)}) + 1/(l+1); given a real ε>0\varepsilon > 0, choose KK so large that 1/(K+1)<ε/31/(K+1) < \varepsilon/3 and d^(ξ(k),ξ(l))<ε/3\widehat{d}(\xi^{(k)},\xi^{(l)}) < \varepsilon/3 for all k,lKk,l \ge K, and then d(ak,al)<εd(a_k,a_l) < \varepsilon for all k,lKk,l \ge K. So a:=(ak)Ca := (a_k) \in \mathcal{C} and [a]X^[a] \in \widehat{X}.

step 8.1step 10.1L4L7L8
12.1

ξ(k)[a]\xi^{(k)} \to [a]: given a real ε>0\varepsilon > 0 take KK as in step 11.1 for ε/3\varepsilon/3, so that d(ak,an)<ε/3d(a_k,a_n) < \varepsilon/3 for all k,nKk,n \ge K, enlarged if necessary so that also 1/(K+1)<ε/31/(K+1) < \varepsilon/3. For kKk \ge K we have d(ak,an)ε/3d(a_k,a_n) \le \varepsilon/3 for all nKn \ge K, hence d^(ι(ak),[a])=limnd(ak,an)ε/3\widehat{d}(\iota(a_k),[a]) = \lim_n d(a_k,a_n) \le \varepsilon/3, and therefore d^(ξ(k),[a])d^(ξ(k),ι(ak))+d^(ι(ak),[a])<ε/3+ε/3<ε\widehat{d}(\xi^{(k)},[a]) \le \widehat{d}(\xi^{(k)},\iota(a_k)) + \widehat{d}(\iota(a_k),[a]) < \varepsilon/3 + \varepsilon/3 < \varepsilon.

step 10.1step 11.1L3L4L5
13.1

So every Cauchy sequence in (X^,d^)(\widehat{X},\widehat{d}) converges in it, which is claim 5; with claims 1 to 4 this exhibits ((X^,d^),ι)\big((\widehat{X},\widehat{d}),\iota\big) as a completion of the arbitrary metric space (X,d)(X,d).

step 3.1step 5.1step 7.1step 9.1step 12.1discharge-construct

Remarks

  • What the new points are. A point of X^\widehat{X} is a class of Cauchy sequences of XX, two sequences being identified exactly when the distance between their nn-th terms tends to 00. The old points reappear as the classes of constant sequences, and the new ones are the classes of Cauchy sequences that had nowhere to go. This is the same move that builds R\mathbb{R} out of Q\mathbb{Q}, and the resulting completion of Q\mathbb{Q} really is R\mathbb{R} (The completion of Q\mathbb{Q} under the usual metric is R\mathbb{R} ).
  • Where choice is spent, and where it is not. Only at step 10.1, which selects one point of XX per natural number: that is exactly ACω\mathrm{AC}_\omega (The Axiom of Countable Choice (ACω\mathrm{AC}_\omega)). Claims 1 to 4 are choice free. The selection could not be avoided by a uniqueness argument, because a point of XX within 1/(k+1)1/(k+1) of ξ(k)\xi^{(k)} is in general far from unique.
  • The quadrilateral estimate does most of the work. Step 1.1 is used twice: at step 2.1, to show (d(xn,yn))n\big(d(x_n,y_n)\big)_n is Cauchy, and at step 6.1, to show ρ\rho is independent of representatives. It is the reverse triangle inequality (The reverse triangle inequality d(x,z)d(y,z)d(x,y)|d(x,z) - d(y,z)| \le d(x,y) in any metric space) applied twice and added. The triangle inequality for ρ\rho, and hence for d^\widehat{d}, does not use it: step 4.1 gets that from the triangle inequality of dd applied termwise and passed to the limit.
  • Nothing here needs XX to be nonempty. If X=X = \emptyset then C=\mathcal{C} = \emptyset, X^=\widehat{X} = \emptyset, and the empty metric space is complete, vacuously. The construction degenerates correctly rather than requiring a hypothesis.

Depends on

Used by

Dependency tree · next 3 levels

Direct dependencies and their dependencies through the next three levels: 115 results over 32 levels. An arrow runs from a result to what uses it, and this result sits at the bottom with a heavier outline. Click the chart to enlarge it.

Sources