BPS states and the \(P = W\) conjecture (Q2875829)

The paper under review gives a new cohomological definition of the refined Gopakumar-Vafa invariants of local curves. Using this new definition and the \(P=W\) conjecture, the authors derive via geometric engineering the HRV formula.NEWLINENEWLINELet \(X\) be a smooth projective curve and let \(r\geq 1\) and \(e\in\mathbb{Z}\) be coprime integers. On one hand, there is the \textit{character variety} \(\mathcal{C}_r^e(X)\), which is a moduli space of representations of \(\pi_1(X)\). It is a smooth quasi-projective variety and its rational cohomology comes with a natural weight filtration. Denote its associated mixed Poincaré polynomial by \(W_r(z,t)\), which is independent of \(e\).NEWLINENEWLINEOn the other hand, there is the \textit{Hitchin moduli space} of stable Higgs bundles \(\mathcal{H}_r^e(X)\) of rank \(r\) and degree \(e\). It is a smooth quasi-projective variety and its rational cohomology carries a natural perverse filtration. Denote by \(P_r(z,t)\) its mixed Poincaré polynomial.NEWLINENEWLINEAs smooth real manifolds, \(\mathcal{C}_r^e(X) \cong \mathcal{H}_r^e(X)\), but there is no algebraic isomorphism. The \textit{P=W} conjecture states that the two above filtrations agree. At the level of Poincare polynomials, this reads as NEWLINE\[NEWLINE W_r(z,t)=P_r(z,t). NEWLINE\]NEWLINE The \textit{P=W} conjecture is due to de Cataldo, Hausel and Migliorini, who proved it in rank 2.NEWLINENEWLINEBy weighting the \(W_r(z,t)\)'s with some polynomials into a generating function \(F_{\mathrm{HRV}}(z,t,T)\), Hausel and Rodriguez-Villegas conjecture that NEWLINE\[NEWLINE F_{\mathrm{HRV}}(z,t,T) = \ln Z_{\mathrm{HRV}}(z,t,T), NEWLINE\]NEWLINE where \(Z_{\mathrm{HRV}}(z,t,T)\) is an explicit sum of rational functions associated to Young diagrams. This is known as the HRV conjecture. It was generalized to punctured curves by Hausel, Letellier and Rodriguez-Villegas and is known in that case as the HLRV conjecture.NEWLINENEWLINEConsider now the local Calabi-Yau threefold \(Y\), which is given as is the total space of \(\mathcal{O}_X(-D)\oplus K_X(D)\), where \(D\) is an effective divisor on \(X\). To \(Y\) are associated refined Gopakumar-Vafa (GV) invariants (also called BPS state counts). One of the contributions of the present paper is that the authors give a direct cohomological definition of the refined GV invariants for \(Y\). Namely, consider the moduli space \(\mathcal{M}(Y,\beta,n)\) of slope semistable pure dimensional sheaves \(F\) on \(Y\) such that \(ch_2(F)=\beta\) and \(\chi(F)=n\). For \(Y\), the moduli spaces \(\mathcal{M}(Y,\beta,n)\) are identified with some Hitchin moduli spaces and consequently the refined GV invariants are perverse Betti numbers of Hitchin moduli spaces. They are packaged into a generating function NEWLINE\[NEWLINE F_{GV,Y}(q,Q,y). NEWLINE\]NEWLINE On the other hand, one considers refined stable pair invariants, which are packaged into a generating function NEWLINE\[NEWLINE Z^{\mathrm{ref}}_{PT,Y}(q,Q,y). NEWLINE\]NEWLINE The expectation is that NEWLINE\[NEWLINE Z^{\mathrm{ref}}_{PT,Y}(q,Q,y)=\exp(F_{GV,Y}(q,Q,y)). NEWLINE\]NEWLINE This is an extension to refined invariants of the M-theory/IIA duality in string theory.NEWLINENEWLINEConsider now the change of variables NEWLINE\[NEWLINE (q,Q,y)=((zt)^{-1},(zt^2)^{g-1}T,t). NEWLINE\]NEWLINE On one side, the authors of the paper under review calculate \(F_{GV,Y}\) for the new variables. They use the cohomological identification of the refined Gopakumar-Vafa invariants as perverse Betti numbers of the Hitchin moduli space. Via the \(P=W\) conjecture then, these Betti numbers correspond to Betti numbers of character varieties. Assuming the \(P=W\) conjecture, they compute that NEWLINE\[NEWLINE F_{GV,Y}((zt)^{-1},(zt^2)^{g-1}T,t)=\sum_{k\geq 1}\sum_{r\geq 1}\frac{T^{kr}}{k}B_r(z^k,t^k)W_r(z^k,t^k), NEWLINE\]NEWLINE where NEWLINE\[NEWLINE B_r(z,t):=\frac{(zt^2)^{(1-g)r(r-1)}}{(1-z)(1-zt^2)}. NEWLINE\]NEWLINE On the other side, the authors compute \(Z^{\mathrm{ref}}_{PT,Y}\) for the same change of variable using geometric engineering and instanton sums, yielding that NEWLINE\[NEWLINE Z^{\mathrm{ref}}_{PT,Y}((zt)^{-1},(zt^2)^{g-1}T,t)=Z_{\mathrm{HRV}}(z,t,T), NEWLINE\]NEWLINE where \(Z_{\mathrm{HRV}}(z,t,T)\) is a sum of rational functions associated to Young diagrams.NEWLINENEWLINEWith these identifications, and assuming the \(P=W\) conjecture, the above expected relationship between the generating functions \(F_{GV,Y}\) and \(Z^{\mathrm{ref}}_{PT,Y}\) becomes NEWLINE\[NEWLINE F_{\mathrm{HRV}}(z,t,T) = \ln Z_{\mathrm{HRV}}(z,t,T). NEWLINE\]NEWLINE Remarkably, this is the HRV formula.NEWLINENEWLINERecently, Chuang, Diaconescu, Donagi and Pantev [\textit{W.-y Chuang} et al., Commun. Math. Phys. 335, No. 3, 1323--1379 (2015; Zbl 1367.14019)] extended the results of the present paper to the case of punctured curves and showed that they obtain the HLRV formula. The interested reader is in addition referred to their paper for a more detailed study of the objects and constructions relevant to the present paper under review.NEWLINENEWLINEFor the entire collection see [Zbl 1286.14002].