← All problems

Is every Banach space with Markov type 2 superreflexive?

For a metric space (X,d)(X,d) and p≥1p\geq 1, say that XX has Markov type pp if there is a finite constant KK such that, for every finite stationary reversible Markov chain (Zt)t≥0(Z_t)_{t\geq 0} on {1,…,n}\{1,\ldots,n\}, every map f:{1,…,n}→Xf:\{1,\ldots,n\}\to X, and every integer t≥1t\geq 1, one has

E d(f(Zt),f(Z0))p≤Kpt E d(f(Z1),f(Z0))p.\mathbb{E}\,d(f(Z_t),f(Z_0))^p\leq K^p t\,\mathbb{E}\,d(f(Z_1),f(Z_0))^p.

Here stationarity means the distribution of ZtZ_t is independent of tt, and reversibility means πiaij=πjaji\pi_i a_{ij}=\pi_j a_{ji} for the stationary probabilities πi\pi_i and transition probabilities aija_{ij}. The least admissible KK is the Markov type pp constant. For a Banach space, take d(x,y)=∥x−y∥d(x,y)=\|x-y\|. Is every Banach space with Markov type 22 superreflexive? More generally, is every Banach space with nontrivial Markov type superreflexive? A Banach space is superreflexive if it admits an equivalent uniformly convex norm. AimPL: Metric embeddings Markov chains in smooth Banach spaces and Gromov hyperbolic metric spaces Keith Ball, The Ribe Programme (2011)

Background

Keith Ball introduced Markov type in connection with Lipschitz extension: it gives a metric inequality on the distance a stationary reversible chain travels after many steps, and can serve as a domain condition for extending Lipschitz maps into suitably convex Banach spaces. In particular, Markov type 2 expresses diffusive-scale control, with the expected squared distance after tt steps bounded by a constant times tt times the one-step expected squared distance. The workshop report traces Ball’s superreflexivity question to his 1992 paper on Lipschitz extensions. Markov chains in smooth Banach spaces and Gromov hyperbolic metric spaces Metric embeddings workshop summary (2025) Keith Ball, The Ribe Programme (2011)

The question connects a chain-based metric inequality to the geometry of Banach spaces. Superreflexivity is equivalent to the existence of an equivalent uniformly convex norm; Bourgain’s metric characterization, as recounted by Ball, describes it through the absence of uniformly well-behaved copies of arbitrarily large binary trees. This makes tests by finite graphs and other finite metric spaces natural avenues for studying whether Markov type forces superreflexivity. Keith Ball, The Ribe Programme (2011) AimPL: Metric embeddings

References

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 1

Claimed by OpenAI. Claims that every real Banach space with Markov type exponent greater than one is superreflexive and admits an equivalent uniformly smooth norm.See full solutionHide full solution

Scope relative to this problem: The source claims every real Banach space with a Markov type exponent greater than 1 is superreflexive and admits an equivalent uniformly smooth norm. This includes Markov type 2 and addresses the nontrivial-type question in the source real setting. It is not a result about exponent 1, arbitrary metric spaces or equality with maximal Markov type.

GitHub repository: https://github.com/openai/math

Manuscript: https://github.com/openai/math/blob/adc7f1241b42e322a6451854ab7e4b4c146bf78a/preprints/Nontrivial-Markov-Type-Forces-Superreflexivity-September-23-2026/paper.pdf

  • OpenAI-327-01-Nontrivial-Markov-Type-Forces-Superreflexivity.pdf356,997 bytesOpen