Rigidity and defect actions in Landau-Ginzburg models (Q664323): Difference between revisions

This paper shows that the correspondence between monoidal categories of defect lines in \(\mathcal{N}=2\) conformal field theory (CFT) and in Landau-Ginzburg models (LG) \(\underline{\text{cannot}}\) straightforwardly be extended to the level of rigid and pivotal monoidal categories. Physically, LG-models arise as infrared fixed points of CFT under renormalization group flow. Defect lines in CFT give rise to a monoidal category that is rigid (has a `good' notion of dual objects) and has pivotal structure (an isomorphism between an object and its double dual). When the potential \(W\) is a complex polynomial with an isolated singularity at the origin, defect lines in LG-models also give rise to a monoidal category \(\mathrm{MF}_{\text{bi}}(W)\) [\textit{N. Carqueville} and \textit{I. Runkel}, J. Phys. A, Math. Theor. 43, No. 27, Article ID 275401, 33 p. (2010; Zbl 1201.81098)]. Reversing the orientation of a defect line induces a duality operation on \(\mathrm{MF}_{\text{bi}}(W)\), providing it also with a rigid and pivotal structure which Section 2 computes explicitly for \(W=x^d\). Once rigid and pivotal structures have been established, they can be used to compute the actions of defects on bulk fields. One might expect the action on the LG side to agree with the one on the CFT side- but it doesn't. It differs from it by a phase. Remark 3.9 justifies this discrepancy in light of the observation that the assignment of defect operators to defect conditions in LG-models should factor through the relevant notion of a Grothendieck ring, but that such a factorization behaves quite differently in the CFT case and in LG-models. For quantities independent of pivotal structures, however, the actions of defects on bulk fields on the LG side and on the CFT side are shown to agree. Section 4 suggests that, in light of the aforementioned results, perhaps what we should really be studying is the \(\underline{\text{bicategory}}\) of LG-models, whose objects are pairs \((R,W)\) of polynomial rings \(R\) and potentials \(W\in R\), whose 1-morphisms between \((R,W)\) and \((R^\prime,W^\prime)\) are matrix factorizations \(X\) of \(W\otimes_{\mathbb{C}}1+1\otimes_{\mathbb{C}}W^\prime\), and whose 2-morphisms between \(X\) and \(Y\) are elements of \(\mathrm{Hom}_{\text{MF}(W\otimes_{\mathbb{C}}1+1\otimes_{\mathbb{C}}W^\prime)}(X,Y)\).