Punctures and \(p\)-spin curves from matrix models. III: \(D_l\) type and logarithmic potential (Q2149619)

There is a long history since Witten's conjecture in 1991 of connecting intersection numbers on the moduli space of curves to partition functions of matrix models. This paper, the third in a series, considers one particular generalisation, namely that of \(p\)-spin curves, where the matrix model is associated to the Lie algebras of type \(D_l\). In this case, \(p\) is restricted to be given by the Coxeter number of the Lie algebra, namely \(2l-2\). References to the origins of these models are provided within. \S2, ``One Point Function for \(A_l\) Type'', considers work previously done by the authors and others in the case of models associated to \(A_l\) Lie algebras as a demonstration of methods to be used later. Here \(p\) is restricted to be \(l+1\). The one-point functions for genus up to 6 are written explicitly for general \(p\), and for specific values \(p=3, 4\) the generating function is written explicitly in terms of special functions. Moreover, periodic vanishing relations for one-point functions are discussed briefly. \S 3, ``One Point Function for \(D_l\) Type'', completes many of the same calculations demonstrated in \S 2. Using algebraic manipulations to simplify the process of expanding the generating function, the authors are able to give the one-point functions for genus up to \(11\) in terms of general \(p\). The function for generic \(g\) is given for \(p=2\), and periodic vanishing is again discussed. \S 4, ``One Point Function of \(D_l\) Type for the Non-positive Integer cases'', considers the analytic continuation of solutions away from the restriction \(p=2l-2\) to the case of \(p\) a negative integer or half-integer value, relations to other matrix models, and proves some results from earlier. \begin{itemize} \item \S4.1 gives a change-of-variables trick to be used in later computations of the generating function. \item \S4.2 demonstrates, by comparing one-point functions, the equivalence of the \(D_l\) model to that of the generalised Kontsevish-Penner model for specific parameter values. \item \S4.3 demonstrates that all one-point functions vanish in the \(A_l\) case when \(p\) is allowed to equal \(1\). \item \S4.4 proves the formula for the \(p=2\) one-point functions given in \S3. \item \S4.5 considers \(p=-1\), obtaining Euler characteristics in terms of Bernoulli numbers. \item \S4.6 considers \(p=-1/2\), and shows that the one-point functions vanish for both \(A_l\) and \(D_l\) when \(g>1\). \item \S4.7 considers \(p=-2\), and shows that the \(D_l\) one-point functions vanish for all \(g\). \item \S4.8 gives the large-argument limit of the generating function in the \(D_l\) case with \(p=-2\). \item \S4.9 considers \(p=1/2\) and \(p=3/2\), obtaining exact computations of the generating function for \(A_l\), as well as \(D_l\) for the former. \end{itemize} \S 5, ``Large \(p\), Large \(g\) Limits and Integrality'', obtains asymptotic results for the one-point functions by using methods of stationary phase for the corresponding integrals. An expression for the one-point functions asymptotic in either of \(g\) and \(p\) is obtained, and these are related to special functions. Subsequent sections \S6 and \S7 give some cases of extensions to the work above, including two-, three-, and four-point functions for some specific values of \(p\), as well as the `strong-coupling' (large argument) limit for these cases.