Knots, BPS states, and algebraic curves (Q739567): Difference between revisions

From the polynomial invariants (in particular the HOMFLY superpolynomial) of a knot \(K\) many interesting objects can be extracted. This paper focuses on two: certain algebraic curves, which exhibit special geometric properties, and BPS invariants, which exhibit special integrality properties. The main general result of the paper is that the latter can be read off from the former. Specifically, for each knot \(K\), a certain generating function in variables \(x,a\) packaging the BPS invariants \(b_{r,i}\), \(r\in \mathbb N\), \(i \in \mathbb Z\), equals an expression of the form \(x (\partial_x y(x,a)) / y\), where remarkably \(y(x,a)\) is an algebraic function determined by a certain curve called the ``dual A-polynomial'' of \(K\). The authors use this result to compute all ``extremal'' BPS invariants (those with \(i\) extremal for a fixed \(r\); they depend only on the coefficients of maximal and minimal powers in \(a\) of the HOMFLY polynomial) for twist knots. The formulas are remarkable: for example, for \(p\geq 2\), the negative extremal BPS invariants of the \(p\)th hyperbolic twist knot \(K_p\) are \[ b_{K_p,r}^- = \frac1{r^2} \sum_{d|r} \mu(r/d) (-1)^d \binom{2d-1}{d-1}. \] Because the BPS invariants are integers, this sum is an integer, a fact that is hard to prove directly. The authors in fact conjecture an even stronger integrality condition for extremal BPS invariants, which they check for twist and torus knots and for several knots with up to 10 crossings.