OctoInformBeta

OpenAI preprint, September 2026 · §9. Changing the predual

Theorem 9.4: The changed predual has no infinite-dimensional norm-separable Banach quotient

The continuum hypothesis gives a real counterexample.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

The separable quotient problem concerns bounded linear surjections from an infinite-dimensional Banach space onto an infinite-dimensional norm-separable Banach space. Changing the predual modifies a closed subspace of a dual space by perturbing the quotient map that defines it.

The continuum hypothesis, abbreviated CH, is the equality |ℝ|=ℵ1|\mathbb R|=\aleph_1.✦ Let ℬℝ\mathcal B_{\mathbb R} be the fixed ambient class of infinite-dimensional real Banach spaces. A real normed space ZZ has a separable quotient when there are an infinite-dimensional norm-separable real Banach space YY and a bounded linear surjection T:Z→YT:Z\to Y, with the same ambient size convention for ZZ and YY.✦ Write 𝒮(Z)\mathcal S(Z) for the collection of such pairs (Y,T)(Y,T). The assertion SQℝ\mathrm{SQ}_{\mathbb R} says that 𝒮(Z)\mathcal S(Z) is nonempty for every Z∈ℬℝZ\in\mathcal B_{\mathbb R}.✦

The Norming Families

The coordinate set is Γ=ω1\Gamma=\omega_1, the set of countable ordinals. Fix a bijection π:Γ→ℕ×Γ\pi:\Gamma\to\mathbb N\times\Gamma.✦ The color of a∈Γa\in\Gamma is c(a)=π(a)1c(a)=\pi(a)_1, and Ck={a:c(a)=k}C_k=\{a:c(a)=k\} is its kkth color class.✦ Each CkC_k is unbounded and has cardinality ℵ1\aleph_1. The color embedding ιk:Γ→Ck\iota_k:\Gamma\to C_k is given by ιk(a)=π−1(k,a)\iota_k(a)=\pi^{-1}(k,a).✦

A rational array uu on Γ\Gamma is a rational-valued function with finite support.✦ The operation families are the pure families pkp_k, one for each k∈ℕk\in\mathbb N, and the mixed family μ\mu.✦ Their parameter sfs_f, family index ifi_f, and source color cfc_f are specified by (1). fsfifcfμ201pkk+2k+1k+1\begin{array}{c|ccc} f&s_f&i_f&c_f\\ \hline \mu&2&0&1\\ p_k&k+2&k+1&k+1 \end{array}(1) Two arrays have successive supports for a family ff if every coordinate in the first support precedes every coordinate in the second. For f=μf=\mu, every color in the first support must also be smaller than every color in the second.

For a natural parameter ss, the depth dnd_n, exponent scale ts,nt_{s,n}, and accumulated length bs,nb_{s,n} are the sequences in (2). The associated block lengths and normalization factors are Ls,nL_{s,n} and ms,nm_{s,n} in the same display. dn=n+11,(ts,0,bs,0)=(d02+2d0+2,0),bs,n+1=bs,n+2Ls,n,Ls,n=2s(s+1)ts,n,ts,n+1=ts,n+(bs,n+1+1)(dn+1+1)+1+dn+1+dn+12,ms,n=2(s−1)(s+1)ts,n+dn.\begin{aligned} d_n&=n+11,& (t_{s,0},b_{s,0})&=(d_0^2+2d_0+2,0),\\ b_{s,n+1}&=b_{s,n}+2L_{s,n},& L_{s,n}&=2^{s(s+1)t_{s,n}},\\ t_{s,n+1}&=t_{s,n}+(b_{s,n+1}+1)(d_{n+1}+1) +1+d_{n+1}+d_{n+1}^2,& m_{s,n}&=2^{(s-1)(s+1)t_{s,n}+d_n}. \end{aligned}(2) All sequence indices start at zero; the corresponding positive index is n+1n+1. Subtraction in natural-number parameters stops at zero. A family uses Lf(j)=Lsf,j−1L_f(j)=L_{s_f,j-1} and mf(j)=msf,j−1m_f(j)=m_{s_f,j-1} at weight j≥1j\geq1. The primal exponent is sfs_f, the path exponent is qf=(sf+1)/sfq_f=(s_f+1)/s_f, and the source exponent is qf′=sf+1q_f'=s_f+1, the primal exponent of pcfp_{c_f}. The underlying function qs=(s+1)/sq_s=(s+1)/s for positive ss has value zero at s=0s=0.✦

A type I expression ee for ff is a weight j≥1j\geq1 and a list of nonzero rational children x0,…,xℓ−1x_0,\ldots,x_{\ell-1} with 1≤ℓ≤Lf(j)1\leq\ell\leq L_f(j) and pairwise successive supports for ff.✦ Its normalized value is the array in (3). val⁡(e)=1mf(j)∑h<ℓxh.\operatorname{val}(e)=\frac1{m_f(j)}\sum_{h<\ell}x_h.(3)✦

The coherent finite closure system ℱ\mathcal F assigns to each β<ω1\beta<\omega_1 an increasing sequence ℱn(β)\mathcal F_n(\beta) of finite subsets of [0,β][0,\beta], each containing β\beta, whose union is [0,β][0,\beta]. Its coherence condition is (4). a∈ℱn(β)⇒ℱn(a)=ℱn(β)∩[0,a].a\in\mathcal F_n(\beta) \quad\Longrightarrow\quad \mathcal F_n(a)=\mathcal F_n(\beta)\cap[0,a].(4) A predecessor closure family GG below β\beta is a family assigning a finite set Gn(a)G_n(a) to each a<βa<\beta and n∈ℕn\in\mathbb N. A predecessor closure family GG below β\beta is coherent when the endpoint, monotonicity, exhaustion, and restriction conditions above hold below β\beta.✦ A coherent endpoint extension (Hn)(H_n) of a predecessor closure family GG below β\beta is an increasing sequence of finite sets Hn⊆[0,β]H_n\subseteq[0,\beta] containing β\beta and exhausting [0,β][0,\beta], such that Gn(a)=Hn∩[0,a]G_n(a)=H_n\cap[0,a] whenever a<βa<\beta belongs to HnH_n.✦ The chosen endpoint construction assigns a sequence of finite sets to a predecessor family. The system ℱ\mathcal F is defined by recursion along ω1\omega_1, applying that construction to the families already assigned to the predecessors.✦

The positive-indexed finite closure Fp(β)F_p(\beta) below β\beta is the finite set given by Fp(β)=ℱp−1(β)F_p(\beta)=\mathcal F_{p-1}(\beta) for p≥1p\geq1; F0(β)=ℱ0(β)F_0(\beta)=\mathcal F_0(\beta) as well.✦ The closure entry rank ρ(a,b)\rho(a,b) is the first positive pp for which a∈Fp(b)a\in F_p(b) when a≤ba\leq b, and is zero when a>ba>b.✦

A finite path code cc is a finite list of triples consisting of a natural weight, natural metadata, and a finite set of pairs in ℕ×ℚ\mathbb N\times\mathbb Q.✦ For a finite coordinate set AA, the order position of aa is pos⁡A(a)=|{b∈A:b<a}|\operatorname{pos}_A(a)=|\{b\in A:b<a\}|.✦ The positional code ac⁡A(u)\operatorname{ac}_A(u) of an array uu relative to a finite coordinate set AA is the set given by ac⁡A(u)={(pos⁡A(a),u(a)):a∈A}\operatorname{ac}_A(u)=\{(\operatorname{pos}_A(a),u(a)):a\in A\}, retaining zero coefficients.✦ Let M(c)M(c) be the largest metadata entry of a code cc, with M(⌀)=0M(\varnothing)=0.✦ The coding map σ\sigma sends cc to the natural-number pairing of its canonical natural-number code and M(c)+1M(c)+1.✦ It is injective, and σ(c)\sigma(c) exceeds every metadata entry of cc.

A raw infinite path is a sequence P=(ui,ji,pi)i≥0P=(u_i,j_i,p_i)_{i\geq0} of nonzero rational arrays, declared weights, and metadata.✦ Its prefix support SnS_n, endpoint ana_n, closure AnA_n, and code cnc_n are defined by (5). Sn=⋃i≤nsupp⁡ui,an=max⁡Sn,An=Fpn(an),cn=[(ji,pi,ac⁡An(ui))]i=0n.\begin{aligned} S_n&=\bigcup_{i\leq n}\operatorname{supp}u_i,& a_n&=\max S_n,& A_n&=F_{p_n}(a_n),\\ c_n&=\bigl[(j_i,p_i,\operatorname{ac}_{A_n}(u_i))\bigr]_{i=0}^{n}. \end{aligned}(5) The raw path is admissible when its supports are successive in the ordinal order, its metadata are strictly increasing and positive, and the conditions in (6) hold. 1≤jn≤pn,ρ(a,b)≤pn(a,b∈Sn,a≤b),jn+1=σ(cn)(n≥0).\begin{gathered} 1\leq j_n\leq p_n,\qquad \rho(a,b)\leq p_n\quad(a,b\in S_n,\ a\leq b),\\ j_{n+1}=\sigma(c_n)\qquad(n\geq0). \end{gathered}(6)✦ Prefix validity of length ℓ\ell means these conditions restricted to the entries i<ℓi<\ell, with the coding rule required only when n+1<ℓn+1<\ell; entries beyond this prefix are unrestricted.✦

A finite path for ff is a raw infinite path with a positive length ℓ\ell and a valid prefix of that length, together with type I expressions for the first ℓ\ell original pieces. Their values and weights agree with uiu_i and jij_i, and the original pieces are successive for ff. Its value is the finite sum ∑i<ℓui\sum_{i<\ell}u_i of the original uncropped rational pieces.✦

For a base set DD and a family ff, the norming construction starts with K0(D,f)=DK_0(D,f)=D. Each next stage retains the previous one and adds values of type I expressions whose children belong to the previous stage, and values of path operations whose children of every original uncropped piece on every path belong to that stage. The signed monochromatic base, pure norming sets, and full norming set are those in (7); δa\delta_a denotes the rational unit array at aa. Dk={0}∪{δa,−δa:a∈Ck},Kk=⋃n≥0Kn(Dk,pk),Dpk=Dk,Dμ=⋃k≥0Kk,Kf=⋃n≥0Kn(Df,f),K=Kμ.\begin{aligned} D_k&=\{0\}\cup\{\delta_a,-\delta_a:a\in C_k\},\\ K_k&=\bigcup_{n\geq0}K_n(D_k,p_k),\\ D_{p_k}&=D_k,& D_\mu&=\bigcup_{k\geq0}K_k,\\ K_f&=\bigcup_{n\geq0}K_n(D_f,f),& K&=K_\mu. \end{aligned}(7) An infinite path for ff is an admissible raw infinite path with a type I expression at each position, having the same value and weight as its raw piece, with the raw pieces successive for ff and every child in KfK_f.✦ Its prefix through index nn is the finite path of length n+1n+1 with the same raw data and the same original expressions at those positions.✦ The path index set IfI_f consists of the raw infinite paths arising from infinite paths for ff.✦ For each a∈Ifa\in I_f, choose a representative PaP_a with exactly those raw data.✦

The Norming Completion

A norming set on Γ\Gamma is a set 𝒦\mathcal K of rational arrays containing every δa\delta_a and satisfying |u(a)|≤1|u(a)|\leq1 for u∈𝒦u\in\mathcal K and a∈Γa\in\Gamma.✦ Its bounded function space is ℓ∞(𝒦)\ell_\infty(\mathcal K) with the supremum norm.✦ The coordinate evaluation vector eae_a is the function ea(u)=u(a)e_a(u)=u(a) in this space.✦ The finite-support evaluation map sends x∈c00(Γ,ℝ)x\in c_{00}(\Gamma,\mathbb R) to ∑ax(a)ea\sum_a x(a)e_a.✦ The norming completion E𝒦E_{\mathcal K} is the closed linear span of these vectors, with the inherited norm; the finite-support inclusion is the same evaluation map with range in E𝒦E_{\mathcal K}.

For u∈𝒦u\in\mathcal K, the norming evaluation functional ϕu∈E𝒦*\phi_u\in E_{\mathcal K}^* is point evaluation at uu.✦ The coordinate functional ea*e_a^* on E𝒦E_{\mathcal K} is the norming evaluation functional given by ea*=ϕδae_a^*=\phi_{\delta_a}.✦ For the full norming set KK, set E=EKE=E_K. The coordinate predual, dual quotient, and quotient map are defined in (8), with norm closure taken in the usual real continuous dual E*E^*. X0=span⁡ℝ{ea*:a∈Γ}¯,W=E*/X0,Q:E*→W.X_0=\overline{\operatorname{span}_{\mathbb R}\{e_a^*:a\in\Gamma\}}, \qquad W=E^*/X_0,\qquad Q:E^*\longrightarrow W.(8)

The Perturbation

For an infinite path PP with raw pieces uiu_i, its path partial sum is the functional HP,n=ϕ∑i≤nuiH_{P,n}=\phi_{\sum_{i\leq n}u_i} on EE.✦ The path functional ΦP∈E*\Phi_P\in E^* of an infinite path PP is chosen as the pointwise limit on EE of its path partial sums HP,nH_{P,n}.✦ The path coset of a∈Ifa\in I_f is wa=QΦPa∈Ww_a=Q\Phi_{P_a}\in W.✦

Assume CH. For each family ff, choose an injection λf0:If↪Γ\lambda_f^0:I_f\hookrightarrow \Gamma. The colored path labelling λf\lambda_f associated with the injection λf0\lambda_f^0 is the map on IfI_f given by λf(a)=ιcf(ιif(λf0(a)))\lambda_f(a)=\iota_{c_f}(\iota_{i_f}(\lambda_f^0(a))).✦ Finite color synthesis and finite path synthesis are the linear maps in (9). Uf0(c)=∑a∈Ifc(a)eλf(a)∈E,Vf0(c)=∑a∈Ifc(a)wa∈W,c∈c00(If,ℝ).U_f^0(c)=\sum_{a\in I_f}c(a)e_{\lambda_f(a)}\in E, \qquad V_f^0(c)=\sum_{a\in I_f}c(a)w_a\in W, \qquad c\in c_{00}(I_f,\mathbb R).(9) For an index set II and an exponent pp, the finitely supported sequence embedding c00(I,ℝ)→ℓp(I)c_{00}(I,\mathbb R)\to\ell_p(I) sends each coefficient family to the same coordinate function.✦ Color synthesis Uf:ℓqf′(If)→EU_f:\ell_{q_f'}(I_f)\to E and path synthesis Vf:ℓqf(If)→WV_f:\ell_{q_f}(I_f)\to W are the continuous linear extensions of Uf0U_f^0 and Vf0V_f^0 along these embeddings.

For conjugate exponents pp and qq, the dual coefficient sequence of g∈ℓp(I)*g\in\ell_p(I)^* has coordinate g(𝒆a)g(\mathbf e_a) at a∈Ia\in I, where 𝒆a\mathbf e_a is the unit sequence.✦ It belongs to ℓq(I)\ell_q(I). For conjugate exponents pp and qq and a bounded linear map T:ℓp(I)→ET:\ell_p(I)\to E, the dual analysis operator of TT is the bounded linear map from E*E^* to ℓq(I)\ell_q(I) that sends g∈E*g\in E^* to the coefficient sequence of g∘Tg\circ T.✦ The color analysis operator AfA_f is the dual analysis operator of UfU_f.✦ The family perturbation operator BfB_f, perturbation weights τn\tau_n, and total perturbation BB are given in (10). (Afg)(a)=g(eλf(a)),Bf=VfAf,f0=μ,fk+1=pk,τn=132(12)n,B=∑n=0∞τnBfn:E*→W.\begin{aligned} (A_fg)(a)&=g(e_{\lambda_f(a)}),& B_f&=V_fA_f,\\ f_0&=\mu,&f_{k+1}&=p_k,\\ \tau_n&=\frac1{32}\left(\frac12\right)^n,& B&=\sum_{n=0}^{\infty}\tau_n B_{f_n}:E^*\longrightarrow W. \end{aligned}(10) The sum defining BB converges in operator norm. The changed predual XX is the real linear subspace of E*E^* defined in (11), with the inherited norm. X=ker⁡(Q−B)={g∈E*:Qg=Bg}.X=\ker(Q-B)=\{g\in E^*:Qg=Bg\}.(11)✦

The Claim

Theorem 1. Assume the continuum hypothesis. The changed predual XX in (11) is complete and has the properties (12) and (13). dim⁡ℝX=∞.\dim_{\mathbb R}X=\infty.(12)✦ 𝒮(X)=⌀.\mathcal S(X)=\varnothing.(13)✦ For the fixed ambient class, (14) holds. ∃Z∈ℬℝ𝒮(Z)=⌀.\exists Z\in\mathcal B_{\mathbb R}\quad \mathcal S(Z)=\varnothing.(14)✦

What Is Checked

The assertions concern the explicit kernel of the quotient map minus the path perturbation under CH, and the universal real assertion over the fixed ambient class of spaces and quotient targets. The dual E*E^* uses its usual operator norm, and WW is complete.✦

For the full norming set, every δa\delta_a belongs to KK✦, and every coefficient of every u∈Ku\in K has absolute value at most one.✦ Every raw prefix support SnS_n is nonempty.✦ For each infinite path, its finite prefix sum belongs to KK✦, and ΦP\Phi_P has norm at most one. The source and path exponents qf′q_f' and qfq_f are Hölder conjugates.✦ For a general dual analysis operator associated with TT, its operator norm is at most ∥T∥\|T\|.