The \(1/N\) expansion of tensor models beyond perturbation theory (Q2510672): Difference between revisions

Matrix models play a foundational role in modern mathematical physics and tensor models are a more recent generalization by studying, instead, probability distributions and moments of objects with more than two indices. 't Hooft's \(1/N\) expansion of matrix models was a ground breaking work which lay at the heart of gauge theories and quantum gravity. The present author had some seminal papers which only in the last few years found the analogue of the \(1/N\) for tensor models as well as a powerful universality property -- that in the large \(N\) limit invariant tensor models respecting a uniform bound on the cumulants become Gaussian. However, there remain many mathematical challenges. As with matrix models, the \(1/N\) expansion is inherently perturbative, can one go beyond this and show, for instance, Borel resummability? The present nice work answers this in the affirmative by rigorously analysing the most general quartic invariant probability measure for the random tensor model. Using a version of the so-called Loop Vetex Expansion (LVE) in the combination \(\lambda / N^{D-1}\) where \(\lambda\) is a coupling constant and \(D\) is the rank of the tensor, the author shows that the cumulants can be written as an explicit series in \(1/N\) plus bounded remainders. The subleading terms can be recast into a counting problem of trees decorated by a finite number of loop edges. The perturbative expansion is shown to be Borel resummable and the cummulants obey an uniform scaling bound. This is a good step toward analytic continuation and understanding such behaviour as phase transitions.