Explicit formula for the hyperbolic scattering determinant (Q2641555)

The authors consider the scattering determinant for a general co-finite Fuchsian group \(\Gamma\) with \(\kappa\geq1\) cusps (i.e.~\(\Gamma\) is not co-compact). The scattering matrix for \(\Gamma\), \(\Phi(s)=(\varphi_{j,k}(s))_{1\leq j,k\leq\kappa}\), is given in terms of the constant terms of the Eisenstein series corresponding to the different cusps. More precisely, \(a_{j,k}(y,s)=\delta_{jk}y^{s}+\varphi_{j,k}(s)y^{1-s}\) is the constant term at the cusp \(k\) of the Eisenstein series associated to the cusp \(j\). The scattering determinant for \(\Gamma\) is simply the determinant of \(\Phi\): \(\varphi(s)=\det\Phi(s)\). It is well-known that \(\varphi(s)\) can be expressed as \[ \varphi(s)=K(s)H(s) \] where \[ K(s)=\pi^{\frac{\kappa}{2}}(\frac{\Gamma(s-\frac{1}{2})}{\Gamma(s)}) ^{\kappa}e^{c_{1}s+c_{2}} \] for some constants \(c_{1}\) and \(c_{2}\), and \(H(s)\) can be expressed as a Dirichlet series [cf. e.g. \textit{D. A. Hejhal}, The Selberg trace formula for \(\text{PSL}(2,\mathbb{R})\). Vol. 2. Lect. Notes Math. 1001. Berlin etc.: Springer-Verlag. (1983; Zbl 0543.10020), p. 299]. For a general Fuchsian group it is difficult to say anything explicit about the Dirichlet coefficients in \(H\). Explicit formulas has been worked out for certain congruence subgroups by Hejhal [loc. cit., Chapter 11] and for more general congruence subgroups by [\textit{M. N. Huxley}, Scattering Matrices for Congruence Subgroups. Modular forms, Symp. Durham/Engl. 1983, 141--156 (1984; Zbl 0554.10016)]. In addition, numerical investigations has also been carried out in a few non-congruence cases. See for example \textit{P. J. Howard} and \textit{P. F. Mahoney} [J.~Phys.~A, Math. Theor. 40, No.~31, 9275--9295 (2007; Zbl 1123.81027)] where they in addition to reporting on numerical experiments also obtained explicit formulas for the Hecke triangle groups \(G_{4}\) and \(G_{6}\) (these are unfortunately conjugate to the congruence subgroups \(\Gamma_{0}(2)\) and \(\Gamma_{0}(3)\), respectively) and \textit{H. Avelin} [Math.~Comput. 76, No.~257, 361--384 (2007; Zbl 1114.11049)] who tracked the zeros of the scattering determinant through a perturbation of the underlying surface in Teichmueller space. In the present paper, the authors considers instead the logarithmic derivative of \(H(s)\): \(\frac{H'}{H}(s)=\sum_{q}\frac{b(q)}{q^{s}}\) where the numbers \(q\) and \(b(q)\) depends on \(\Gamma\), \(0<q\rightarrow\infty\). The explicit formula mentioned in the title is not a formula for the individual Dirichlet coefficients but rather a trace formula relating the coefficients \(b(q)\) to the zeros of \(\varphi\): \[ \sum_{\rho}v(\rho)M_{\frac{1}{2}}g(\rho) =-\sum_{q}\frac{b(q)}{q^{\frac{1}{2}}}G(\log q)+\kappa\int_{0}^{\infty}\frac{e^{\frac{x}{2}}G(x)}{1+e^{\frac{x}{2}}}\,dx \] where \(g(x)=G(-\ln x)\) and \(G(x)=F(x)-F(-x)\) for a test function \(F\) satisfying certain conditions. The formula above generalizes a result of \textit{J. Jorgenson} and \textit{S. Lang} [Math.~Ann. 306, No.~1, 75--124 (1996; Zbl 0871.11058)]. The most important generalizations are that the test function \(F\) (and hence \(G\)) need not be continuous and there is no restriction on its order of vanishing at \(0\). It is however clear that \(G\) is odd by construction and since \(G\) has taken the role of \(F\) in Jorgenson and Lang [loc. cit.], the part about ``need not be odd'', in the remark on how the present result differs from that of Jorgenson and Lang [loc. cit.] should be disregarded. One advantage of the more general conditions in the present paper is that they allow the authors to use the test function \(F\) defined by: \(F(x)=e^{-(\frac{1}{2}+x)}\) for \(x>0,\) \(F(0)=\frac{1}{2}\) and \(F(x)=0\) for \(x<0\) which leads to a nice explicit formula (akin to the classical Riemann--von Mangoldt formula for \(\zeta(s)\)) for the scattering determinant \(\varphi\): \[ \frac{\varphi'}{\varphi}(s)-c_{1}=\sum_{\rho}\frac{m(\rho)}{s-\rho} \] where the sum is over all zeros and poles of \(\varphi\) with \(m(\rho)\) equal to the order of the zero if \(\rho\) is a zero and minus the order of the pole if \(\rho\) is a pole.