Let ω {\displaystyle \omega } be a free ultrafilter on the natural numbers and let R be the hyperfinite type II1 factor with trace τ {\displaystyle \tau } . One can construct the ultrapower R ω {\displaystyle R^{\omega }} as follows: let l ∞ ( R ) = { ( x n ) n ⊆ R : sup n | | x n | | < ∞ } {\displaystyle l^{\infty }(R)=\{(x_{n})_{n}\subseteq R:\sup _{n}||x_{n}||<\infty \}} be the von Neumann algebra of norm-bounded sequences and let I ω = { ( x n ) ∈ l ∞ ( R ) : lim n → ω τ ( x n ∗ x n ) 1 2 = 0 } {\displaystyle I_{\omega }=\{(x_{n})\in l^{\infty }(R):\lim _{n\rightarrow \omega }\tau (x_{n}^{*}x_{n})^{\frac {1}{2}}=0\}} . The quotient R ω = l ∞ ( R ) / I ω {\displaystyle R^{\omega }=l^{\infty }(R)/I_{\omega }} turns out to be a II1 factor with trace τ R ω ( x ) = lim n → ω τ ( x n + I ω ) {\displaystyle \tau _{R^{\omega }}(x)=\lim _{n\rightarrow \omega }\tau (x_{n}+I_{\omega })} , where ( x n ) n {\displaystyle (x_{n})_{n}} is any representative sequence of x {\displaystyle x} .
Connes' embedding problem asks whether every type II1 factor on a separable Hilbert space can be embedded into some R ω {\displaystyle R^{\omega }} .
A positive solution to the problem would imply that invariant subspaces exist for a large class of operators in type II1 factors (Uffe Haagerup); all countable discrete groups are hyperlinear. A positive solution to the problem would be implied by equality between free entropy χ ∗ {\displaystyle \chi ^{*}} and free entropy defined by microstates (Dan Voiculescu). In January 2020, a group of researchers8 claimed to have resolved the problem in the negative, i.e., there exist type II1 von Neumann factors that do not embed in an ultrapower R ω {\displaystyle R^{\omega }} of the hyperfinite II1 factor.
The isomorphism class of R ω {\displaystyle R^{\omega }} is independent of the ultrafilter if and only if the continuum hypothesis is true (Ge-Hadwin and Farah-Hart-Sherman), but such an embedding property does not depend on the ultrafilter because von Neumann algebras acting on separable Hilbert spaces are, roughly speaking, very small.
The problem admits a number of equivalent formulations.9
Hadwin, Don (2001). "A Noncommutative Moment Problem". Proceedings of the American Mathematical Society. 129 (6): 1785–1791. doi:10.1090/S0002-9939-01-05772-0. JSTOR 2669132. https://doi.org/10.1090%2FS0002-9939-01-05772-0 ↩
Ji, Zhengfeng; Natarajan, Anand; Vidick, Thomas; Wright, John; Yuen, Henry (2020). "MIP*=RE". arXiv:2001.04383 [quant-ph]. /wiki/ArXiv_(identifier) ↩
Castelvecchi, Davide (2020). "How 'spooky' is quantum physics? The answer could be incalculable". Nature. 577 (7791): 461–462. Bibcode:2020Natur.577..461C. doi:10.1038/d41586-020-00120-6. PMID 31965099. https://doi.org/10.1038%2Fd41586-020-00120-6 ↩
Hartnett, Kevin (4 March 2020). "Landmark Computer Science Proof Cascades Through Physics and Math". Quanta Magazine. Retrieved 2020-03-09. https://www.quantamagazine.org/landmark-computer-science-proof-cascades-through-physics-and-math-20200304/ ↩
Ji, Zhengfeng; Natarajan, Anand; Vidick, Thomas; Wright, John; Yuen, Henry (27 September 2020). "Quantum soundness of the classical low individual degree test". arXiv:2009.12982 [quant-ph]. /wiki/ArXiv_(identifier) ↩
Ji, Zhengfeng; Natarajan, Anand; Vidick, Thomas; Wright, John; Yuen, Henry (November 2021). "MIP* = RE". Communications of the ACM. 64 (11): 131–138. doi:10.1145/3485628. S2CID 210165045. https://doi.org/10.1145%2F3485628 ↩
Isaac Goldbring (October 2022), "The Connes Embedding Problem: A Guided Tour" (PDF), Bulletin of the American Mathematical Society, 58 (4): 503–560, doi:10.1090/bull/1768, S2CID 237940159 https://www.ams.org/journals/bull/2022-59-04/S0273-0979-2022-01768-5/S0273-0979-2022-01768-5.pdf ↩