TheoremBase

Without any choice, assigns to each bounded real sequence a canonical subsequence converging to its limit superior; recursion then gives nested index sequences, the k-th making the k-th column converge, and their diagonal makes every column converge.

Proof

Each result cited below is universally quantified over the data in its own statement, and is applied to the data named where it is cited. The proof uses no form of choice: at each stage a specific subsequence is singled out by a formula, and the sequences built in stages are given by the recursion of Recursion on the Natural Numbers Starting at One §recursion. By The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness §reals, R\mathbb{R} is an ordered field, with the rules of Rules of Arithmetic and Order in an Ordered Field carried by the setting. Natural numbers are read in R\mathbb{R} as in The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness §identification, this reading being that of The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness §embedding, and the order agrees whether formed in N\mathbb{N} or in R\mathbb{R}, by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness §agreement. In particular 0<n0<n in R\mathbb{R} for every n∈Nn\in\mathbb{N}, by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares §naturals. N\mathbb{N} carries its order, with the facts of Arithmetic and Order of the Natural Numbers; strictly increasing sequences in N\mathbb{N} are as in Monotone Sequences §monotone, and Σ\Sigma denotes the set of strictly increasing σ∈NN\sigma\in\mathbb{N}^{\mathbb{N}}, with NN\mathbb{N}^{\mathbb{N}} the set of maps of Sets and Maps: Ordinary Notation §maps; Σ\Sigma is a set, being a subset of NN\mathbb{N}^{\mathbb{N}} given by a property, by Sets and Maps: Ordinary Notation §set-builder. By Basic Properties of Functions: Equality, Composition, Identity, Inverse and Restriction §identity, idN\mathrm{id}_{\mathbb{N}} is a map N→N\mathbb{N}\to\mathbb{N} with idN(u)=u\mathrm{id}_{\mathbb{N}}(u)=u for every u∈Nu\in\mathbb{N}, so idN∈Σ\mathrm{id}_{\mathbb{N}}\in\Sigma by Arithmetic and Order of the Natural Numbers §successor, and Σ\Sigma is closed under composition by Monotone Sequences and Subsequences: Comparison of All Terms, Growth of the Indices, and Subsequences of Subsequences §composition.

Step 1 (A canonical convergent subsequence). Let bb be a bounded sequence in R\mathbb{R}. By Limits of Sequences of Real Numbers: Uniqueness, Boundedness, Constants, Tails, Arithmetic, Quotients, Absolute Values, Finite Sums, Order, Squeezing, Domination and Subsequences §two-sided it is bounded above and bounded below, so by Bounded Sequences of Real Numbers §bounded its set of terms has an upper bound M+M^{+} and a lower bound M−M^{-}. For n∈Nn\in\mathbb{N} let Tn={z∈R:z=bm for some m∈N with n≤m}T_{n}=\{z\in\mathbb{R}: z=b_{m}\text{ for some }m\in\mathbb{N}\text{ with }n\le m\}, a subset of R\mathbb{R} by Sets and Maps: Ordinary Notation §set-builder. It contains bnb_{n}, as n≤nn\le n by Arithmetic and Order of the Natural Numbers §partial-order, and is bounded above by M+M^{+}, so it has a supremum by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness §supremum, which is unique by Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order §supremum; let UnU_{n} be this supremum. If n≤n′n\le n', then Tn′⊆TnT_{n'}\subseteq T_{n}, so UnU_{n} is an upper bound of Tn′T_{n'} and Un′≤UnU_{n'}\le U_{n}; and M−≤bn≤UnM^{-}\le b_{n}\le U_{n}. Since UnU_{n} is defined for every n∈Nn\in\mathbb{N}, the set {z∈R:z=Un for some n∈N}\{z\in\mathbb{R}: z=U_{n}\text{ for some }n\in\mathbb{N}\}, a subset of R\mathbb{R} by Sets and Maps: Ordinary Notation §set-builder, is nonempty and bounded below by M−M^{-}, so it has an infimum by The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness §infimum, which is unique by Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order §supremum; let LbL_{b} be this infimum.

(A) For every positive ε∈R\varepsilon\in\mathbb{R} and every u∈Nu\in\mathbb{N} there is m∈Nm\in\mathbb{N} with u<mu<m and ∣bm−Lb∣<ε|b_{m}-L_{b}|<\varepsilon. By Arbitrary Positive Slack, and Approximation of Suprema and Infima, in the Real Numbers §epsilon-below choose n0∈Nn_{0}\in\mathbb{N} with Un0<Lb+εU_{n_{0}}<L_{b}+\varepsilon, and put n=n0+un=n_{0}+u, so n0<nn_{0}<n and u<nu<n by Arithmetic and Order of the Natural Numbers §difference and Arithmetic and Order of the Natural Numbers §commutative. Then Lb≤Un≤Un0<Lb+εL_{b}\le U_{n}\le U_{n_{0}}<L_{b}+\varepsilon. By Arbitrary Positive Slack, and Approximation of Suprema and Infima, in the Real Numbers §epsilon-above there is m∈Nm\in\mathbb{N} with n≤mn\le m and Un−ε<bmU_{n}-\varepsilon<b_{m}; thus Lb−ε≤Un−ε<bm≤Un<Lb+εL_{b}-\varepsilon\le U_{n}-\varepsilon<b_{m}\le U_{n}<L_{b}+\varepsilon, so Lb−ε<bm<Lb+εL_{b}-\varepsilon<b_{m}<L_{b}+\varepsilon by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares §mixed. Subtracting LbL_{b} by Rules of Arithmetic and Order in an Ordered Field §order-sum gives −ε<bm−Lb<ε-\varepsilon<b_{m}-L_{b}<\varepsilon, so ∣bm−Lb∣<ε|b_{m}-L_{b}|<\varepsilon by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares §absolute-strict. Finally u<mu<m: by Arithmetic and Order of the Natural Numbers §partial-order, n≤mn\le m means n<mn<m or n=mn=m, and in the first case u<mu<m follows from u<nu<n by the strict transitivity in the same clause.

For j,u∈Nj,u\in\mathbb{N} the set Eb(j,u)={m∈N:u<m, ∣bm−Lb∣<1/j}E_{b}(j,u)=\{m\in\mathbb{N}:u<m,\ |b_{m}-L_{b}|<1/j\} is nonempty by (A), 1/j1/j being positive: 0<j0<j (shown at the start), so j≠0j\neq0 by Uniqueness of Least and Greatest Elements, Properties of the Strict Order, and Trichotomy for Total Orders §strict-characterization, 1/j=1⋅j−1=j−11/j=1\cdot j^{-1}=j^{-1} by Negatives, Differences, Reciprocals and Quotients §reciprocal and the commutativity and unit law of Commutative Rings §ring, and 0<j−10<j^{-1} by Rules of Arithmetic and Order in an Ordered Field §positive-reciprocal, so it has a least element by Arithmetic and Order of the Natural Numbers §well-order, which is unique by Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order §least; let μb(j,u)\mu_{b}(j,u) be this least element. With g(n,u)=μb(n+1,u)g(n,u)=\mu_{b}(n+1,u), a map g:N×N→Ng:\mathbb{N}\times\mathbb{N}\to\mathbb{N} by Maps and Relations Given by Formulas §binary, Recursion on the Natural Numbers Starting at One §recursion gives exactly one map σb:N→N\sigma_{b}:\mathbb{N}\to\mathbb{N} with σb(1)=μb(1,1)\sigma_{b}(1)=\mu_{b}(1,1) and σb(j+1)=μb(j+1,σb(j))\sigma_{b}(j+1)=\mu_{b}(j+1,\sigma_{b}(j)) for every j∈Nj\in\mathbb{N}. Then σb(j)<σb(j+1)\sigma_{b}(j)<\sigma_{b}(j+1), so σb∈Σ\sigma_{b}\in\Sigma; and, as every j∈Nj\in\mathbb{N} is 11 or a successor by Arithmetic and Order of the Natural Numbers §predecessor, ∣bσb(j)−Lb∣<1/j|b_{\sigma_{b}(j)}-L_{b}|<1/j for every jj. Since 1/j→01/j\to0 by Completeness of the Real Numbers for Sequences: Monotone Convergence, the Bolzano-Weierstrass Theorem and Cauchy Sequences §reciprocals, the subsequence b∘σbb\circ\sigma_{b} converges to LbL_{b} by Limits of Sequences of Real Numbers: Uniqueness, Boundedness, Constants, Tails, Arithmetic, Quotients, Absolute Values, Finite Sums, Order, Squeezing, Domination and Subsequences §domination. Since LbL_{b} is the unique infimum (Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order §supremum) and σb\sigma_{b} the unique map above, σb\sigma_{b} is determined by bb alone.

Step 2 (Nested index sequences, by recursion). For k∈Nk\in\mathbb{N} and τ∈Σ\tau\in\Sigma let ck,τc^{k,\tau} be the sequence i↦aτ(i),ki\mapsto a_{\tau(i),k}, a map N→R\mathbb{N}\to\mathbb{R} by Sets and Maps: Ordinary Notation §maps, as aτ(i),k∈Ra_{\tau(i),k}\in\mathbb{R} for every i∈Ni\in\mathbb{N}. It is bounded: if ∣am,k∣≤M|a_{m,k}|\le M for every m∈Nm\in\mathbb{N}, then ∣aτ(i),k∣≤M|a_{\tau(i),k}|\le M for every ii. By Step 1 and Maps and Relations Given by Formulas §binary, (k,τ)↦τ∘σck+1,τ(k,\tau)\mapsto\tau\circ\sigma_{c^{k+1,\tau}} is a map N×Σ→Σ\mathbb{N}\times\Sigma\to\Sigma. As Σ\Sigma is a set (shown at the start), Recursion on the Natural Numbers Starting at One §recursion applies: there is a map N→Σ\mathbb{N}\to\Sigma, k↦τkk\mapsto\tau_{k}, with

τ1=σc1,id,τk+1=τk∘ρk,where ρk=σck+1,τk∈Σ.\tau_{1}=\sigma_{c^{1,\mathrm{id}}},\qquad\tau_{k+1}=\tau_{k}\circ\rho_{k},\quad\text{where }\rho_{k}=\sigma_{c^{k+1,\tau_{k}}}\in\Sigma .

For every k∈Nk\in\mathbb{N} the sequence (aτk(i),k)i∈N(a_{\tau_{k}(i),k})_{i\in\mathbb{N}} converges: it is c1,id∘σc1,idc^{1,\mathrm{id}}\circ\sigma_{c^{1,\mathrm{id}}} if k=1k=1, and cl+1,τl∘σcl+1,τlc^{l+1,\tau_{l}}\circ\sigma_{c^{l+1,\tau_{l}}} if k=l+1k=l+1, which converge by Step 1; every kk is of one of these forms by Arithmetic and Order of the Natural Numbers §predecessor.

Step 3 (Nesting). Let PP be the set of j∈Nj\in\mathbb{N} such that for every k∈Nk\in\mathbb{N} with k≤jk\le j there is π∈Σ\pi\in\Sigma with τj=τk∘π\tau_{j}=\tau_{k}\circ\pi. Then 1∈P1\in P: k≤1k\le1 forces k=1k=1 by Arithmetic and Order of the Natural Numbers §least and Arithmetic and Order of the Natural Numbers §partial-order, and π=idN\pi=\mathrm{id}_{\mathbb{N}} works, as τ1=τ1∘idN\tau_{1}=\tau_{1}\circ\mathrm{id}_{\mathbb{N}} by Basic Properties of Functions: Equality, Composition, Identity, Inverse and Restriction §identity-neutral. Let j∈Pj\in P and k≤j+1k\le j+1. If k=j+1k=j+1, take π=idN\pi=\mathrm{id}_{\mathbb{N}}, again by Basic Properties of Functions: Equality, Composition, Identity, Inverse and Restriction §identity-neutral. Otherwise k<j+1k<j+1, hence k≤jk\le j by Arithmetic and Order of the Natural Numbers §successor and Arithmetic and Order of the Natural Numbers §trichotomy; with π\pi from j∈Pj\in P, τj+1=(τk∘π)∘ρj=τk∘(π∘ρj)\tau_{j+1}=(\tau_{k}\circ\pi)\circ\rho_{j}=\tau_{k}\circ(\pi\circ\rho_{j}) and π∘ρj∈Σ\pi\circ\rho_{j}\in\Sigma, by Monotone Sequences and Subsequences: Comparison of All Terms, Growth of the Indices, and Subsequences of Subsequences §composition. So j+1∈Pj+1\in P, and P=NP=\mathbb{N} by Arithmetic and Order of the Natural Numbers §induction.

Step 4 (The diagonal). Put nj=τj(j)n_{j}=\tau_{j}(j) for j∈Nj\in\mathbb{N}. We have j<j+1j<j+1 by Arithmetic and Order of the Natural Numbers §successor and j+1≤ρj(j+1)j+1\le\rho_{j}(j+1) by Monotone Sequences and Subsequences: Comparison of All Terms, Growth of the Indices, and Subsequences of Subsequences §index; by Arithmetic and Order of the Natural Numbers §partial-order, the latter means j+1<ρj(j+1)j+1<\rho_{j}(j+1) or j+1=ρj(j+1)j+1=\rho_{j}(j+1), so j<ρj(j+1)j<\rho_{j}(j+1), using the strict transitivity in that clause in the first case. Hence Monotone Sequences and Subsequences: Comparison of All Terms, Growth of the Indices, and Subsequences of Subsequences §index gives nj=τj(j)<τj(ρj(j+1))=τj+1(j+1)=nj+1n_{j}=\tau_{j}(j)<\tau_{j}(\rho_{j}(j+1))=\tau_{j+1}(j+1)=n_{j+1}; so (nj)(n_{j}) is strictly increasing, and (anj,k)j(a_{n_{j},k})_{j} is a subsequence of (am,k)m(a_{m,k})_{m} for every kk. Fix k∈Nk\in\mathbb{N}; by Step 2, (aτk(i),k)i(a_{\tau_{k}(i),k})_{i} converges to some L∈RL\in\mathbb{R}. Let ε∈R\varepsilon\in\mathbb{R} be positive. First choose, by Convergent Sequences of Real Numbers §converges, I∈NI\in\mathbb{N} with ∣aτk(i),k−L∣<ε|a_{\tau_{k}(i),k}-L|<\varepsilon for every i≥Ii\ge I; then put J=k+IJ=k+I, so that k<Jk<J and I<JI<J by Arithmetic and Order of the Natural Numbers §difference and Arithmetic and Order of the Natural Numbers §commutative. Let j≥Jj\ge J. By Arithmetic and Order of the Natural Numbers §partial-order, J≤jJ\le j means J<jJ<j or J=jJ=j, so k<jk<j from k<Jk<J, using the strict transitivity in that clause in the first case, and hence k≤jk\le j by the same clause. So by Step 3 there is π∈Σ\pi\in\Sigma with τj=τk∘π\tau_{j}=\tau_{k}\circ\pi, whence nj=τk(π(j))n_{j}=\tau_{k}(\pi(j)). Moreover j≤π(j)j\le\pi(j) by Monotone Sequences and Subsequences: Comparison of All Terms, Growth of the Indices, and Subsequences of Subsequences §index, so J≤π(j)J\le\pi(j) by the transitivity in Arithmetic and Order of the Natural Numbers §partial-order; as I<JI<J, this gives I<π(j)I<\pi(j) in the same way as above, and so I≤π(j)I\le\pi(j). Hence ∣anj,k−L∣<ε|a_{n_{j},k}-L|<\varepsilon. By Convergent Sequences of Real Numbers §converges, (anj,k)j(a_{n_{j},k})_{j} converges to LL. As kk was arbitrary, (nj)(n_{j}) is as required.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…