Can every Banach space have a separable infinite-dimensional quotient?

← All digests

Family 323 · Functional analysis
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-25

1. The question

Every infinite-dimensional Banach space contains a separable infinite-dimensional closed subspace: take the closed span of countably many linearly independent vectors. Can one always obtain such a space as a quotient instead?

A quotient collapses a closed subspace to zero. It describes the information retained by a bounded linear surjection, rather than information found by looking inside the original space. The separable quotient problem asks whether every infinite-dimensional Banach space admits a separable infinite-dimensional quotient.

Family 323 claims that the unrestricted statement is independent of ZFC, relative to the consistency of a measurable cardinal. Under the continuum hypothesis it constructs counterexamples; under a different set-theoretic measure hypothesis it proves the positive statement. For spaces of density exactly $\aleph_1$, it claims independence assuming only consistency of ZFC.

2. Background and essential definitions

A Banach space is a complete normed vector space. It is separable if it has a countable norm-dense subset. Its density character is the smallest cardinality of a norm-dense set.

If $M\subseteq X$ is closed, the quotient $X/M$ has norm

$$\|x+M\|=\inf_{m\in M}\|x+m\|.$$

It is again a Banach space. A bounded linear surjection $Q:X\to Y$ identifies $Y$, up to equivalent norm, with $X/\ker Q$ by the open mapping theorem.

The dual $X^*$ is the Banach space of bounded linear functionals on $X$. A predual of a space $E$ is a Banach space $X$ with $X^*$ isomorphic to $E$. Different preduals can give different evaluation behavior even when their duals are isomorphic.

ZFC denotes the usual set-theory axioms with choice. The continuum hypothesis, CH, says $\mathfrak c=2^{\aleph_0}=\aleph_1$. A measurable cardinal is a large-cardinal assumption stronger than ZFC alone in the consistency hierarchy.

M1 uses the condition that $\mathfrak c$ is real-valued measurable: there is a probability measure on every subset of a set of size $\mathfrak c$, vanishing on singletons and additive over disjoint families of size less than $\mathfrak c$. This is substantially stronger than ordinary Lebesgue measure on a sigma-algebra of measurable sets.

Relative independence means that appropriate models of the axioms give opposite answers. It does not mean a particular specified Banach space simultaneously has and lacks a quotient inside one fixed model.

3. Development history and previous partial results

The problem is traditionally attributed to Banach and Mazur, although historical accounts note that it is not explicitly stated in Banach’s 1932 book. Rosenthal’s 1969 work is an early written formulation.

Many positive cases are known. Johnson–Rosenthal developed weak-star basic-sequence methods and showed that every separable infinite-dimensional Banach space has an infinite-dimensional quotient with a Schauder basis. A Schauder basis permits unique norm-convergent coordinate expansions.

Argyros, Dodos and Kanellopoulos proved the positive answer for every dual Banach space. Another useful criterion, based on Hagler–Johnson, says that an unconditional basic sequence in $X^*$ yields a separable infinite-dimensional quotient of $X$. Unconditionality means convergence is stable under sign changes and rearrangements in the appropriate basis sense.

Set theory already influenced positive results. Let $\mathfrak b$ be the smallest size of a family of functions $\mathbb N\to\mathbb N$ unbounded under eventual domination. Saxon–Sánchez Ruiz proved the quotient statement for infinite-dimensional spaces with density below $\mathfrak b$. Thus $\mathfrak b>\aleph_1$ gives the positive assertion at density $\aleph_1$.

Earlier large-cardinal consistency results covered certain large density ranges. M1 claims a positive implication covering all infinite-dimensional spaces from real-valued measurability of the continuum, together with a new CH construction giving failure already at the smallest nonseparable density.

4. What the new manuscripts claim

For either scalar field $\mathbb F=\mathbb R$ or $\mathbb C$, let $\mathrm{SQ}_{\mathbb F}$ denote the assertion that every infinite-dimensional Banach space over $\mathbb F$ has a separable infinite-dimensional quotient.

M1 states that ZFC proves both implications

$$\mathfrak c\text{ real-valued measurable} \quad\Longrightarrow\quad \mathrm{SQ}_{\mathbb F},$$

and

$$\mathrm{CH}\quad\Longrightarrow\quad\neg\mathrm{SQ}_{\mathbb F}.$$

The CH counterexamples have norm density $\aleph_1$. A standard model construction from a measurable cardinal supplies the positive measure hypothesis. Consequently, if ZFC with a measurable cardinal is consistent, the unrestricted quotient assertion is independent of ZFC. Consistency of its failure needs only consistency of ZFC, because CH is available in the constructible model.

For the restricted statement that every space of density exactly $\aleph_1$ has such a quotient, M1 claims independence relative to consistency of ZFC alone. The positive side uses the earlier density-below-$\mathfrak b$ theorem together with a model of Martin’s axiom where $\mathfrak b>\aleph_1$. The negative side uses M1’s CH example.

The two consistency strengths should not be conflated. M1 does not establish that the unrestricted positive statement has exactly the consistency strength of a measurable cardinal; that cardinal is a sufficient assumption for the stated relative-independence conclusion.

5. Why the advance matters

Importance: 5/5. If correct, this would resolve a central existence question in Banach-space theory by locating it at a set-theoretic boundary. The simple analogy between subspaces and quotients would fail in some models even though it succeeds for broad familiar classes.

The density-$\aleph_1$ result is especially sharp in location: it places counterexamples at the first possible nonseparable density, while established axioms yielding a larger bounding cardinal give a positive answer at that same density.

The real and complex conclusions also match. The phenomenon would not be an artifact of choosing one scalar field. The manuscript includes explicit transfers between the two settings rather than assuming that realification automatically preserves every quotient issue.

6. The main idea

Test quotients through dual evaluation

For a closed subspace $F\subseteq X^*$, define

$$R_F:X\to F^*,\qquad (R_Fx)(f)=f(x).$$

M1 establishes a common criterion: a separable infinite-dimensional quotient exists exactly when some infinite-dimensional closed $F$ has norm-separable evaluation image $R_F(X)$. In the necessary direction $F$ can itself be chosen separable.

Separability of $F$ does not imply separability of $F^*$ or of the evaluation image. Also, an image with a useful topology is not automatically an onto map to a Banach quotient; the source supplies the additional surjectivity step. Both constructions use this criterion in opposite directions.

Use a measure to force an unconditional sequence

For the positive implication, assume a counterexample $X$ and choose a separable infinite-dimensional subspace of its dual. The measure hypothesis first forces its evaluation image to have full continuum density; a smaller dense set would permit a basic-sequence construction yielding a quotient.

The large image then supplies a family of dual functionals with controlled norms that vanishes almost everywhere at each fixed vector of $X$. Sequential bidual approximations and Borel coding allow an integration argument to select an unconditional basic sequence.

The source uses ordinary Borel Fubini after coding; it does not assume an unrestricted interchange theorem for arbitrary power-set measures. The resulting unconditional sequence contradicts the established dual criterion for a separable quotient.

Construct a dual with many separating paths

For the negative implication, M1 builds a dual space $E=X_0^*$ with coordinates indexed by $\omega_1$, the first uncountable ordinal. Successive finitely supported norming functionals form paths with bounded partial sums, hence define functionals on $E$.

Integer weights and coding make different paths share only finitely many relevant weights. A block argument produces uncountably many norm-separated restrictions of these path functionals on every infinite-dimensional subspace of $E$. The goal is to prevent the evaluation image required by the quotient criterion from ever becoming separable.

Change the predual while retaining the dual

CH permits assigning all the needed paths to the available coordinates. A small perturbation of a quotient map involving $E^*/X_0$ produces a new predual $X$, with $X^*$ isomorphic to $E$, whose evaluations approximate the path information.

For every separable infinite-dimensional subspace of $X^*$, those evaluations are nonseparable. The common criterion therefore excludes every separable infinite-dimensional quotient of $X$.

This does not contradict the positive theorem for dual Banach spaces: the counterexample is the new predual $X$, not the dual space $E$ used to organize its construction. The distinction between a space and its dual is part of the mechanism.

7. Scope, limitations, and open questions

The unrestricted independence claim is relative to consistency of ZFC with a measurable cardinal. It is not an unconditional proof that ZFC is consistent, nor a claim that the positive statement requires a measurable cardinal in every possible model.

The sharper ZFC-consistency-only conclusion concerns spaces of density exactly $\aleph_1$. It cannot be promoted to the unrestricted statement merely because the negative examples already occur there.

All separability and density conditions on evaluation images use the norm topology, unless the source explicitly says otherwise. Weak or weak-star separability would not replace the criterion. The quotient must be infinite-dimensional and arise through a bounded linear surjection, not just a nonlinear map or a dense nonclosed image.

The manuscript’s consistency and construction claims are taken as stated. Determining the optimal consistency strength of the unrestricted positive assertion, or describing additional geometric classes with a ZFC positive answer, remains beyond its stated conclusion.

8. Sources and further reading

• Saxon and Sánchez Ruiz, Barrelled Countable Enlargements and the Bounding Cardinal, for the earlier density criterion.
• Argyros, Dodos and Kanellopoulos, Unconditional families in Banach spaces, for the positive dual-space result.
• M1’s final section states the Solovay measure-model input and separates the unrestricted and density-$\aleph_1$ consistency bounds.

Source manuscripts

• M1. OpenAI, Relative independence of the separable quotient problem (September 23, 2026).

Sources consulted on 25 October 2026 at repository revision 1581d4c. Download the manuscript citations.