Abstract. The vacuum (Casimir) energy of a quantized scalar field in a given geometrical situation is a certain moment of the eigenvalue density of an associated self-adjoint differential operator. For various classes of quantum graphs it has been calculated by several methods: (1) Direct calculation from the explicitly known spectrum is feasible only in simple cases. (2) Analysis of the secular equation determining the spectrum, as in the Kottos-Smilansky derivation of the trace formula, yields a sum over periodic orbits in the graph. (3) Construction of an associated integral kernel by the method of images yields a sum over closed (not necessarily periodic) orbits. We show that for the Kirchhoff and other scale-invariant boundary conditions the sum over nonperiodic orbits in fact makes no contribution to the total energy, whereas for more general (frequency-dependent) vertex scattering matrices it can make a nonvanishing contribution, which, however, is localized near vertices and hence can be "indexed" a posteriori by truly periodic orbits. For the scale-invariant cases complete calculations have been done by both methods (2) and (3), with identical results. Indeed, applying the image method to the resolvent kernel provides an alternative derivation of the trace formula.
Overview and acknowledgmentsVacuum energy [15,14,38,37,13] is a topic in physics (quantum field theory) that has turned out to have some resonance in mathematics. It provides an application of, and a new window on, the spectral theory of second-order differential operators. In particular, there are connections with the lore surrounding the relations between the operator's spectrum and the trajectories of an associated classical-mechanical system. Insofar as energy, pressure, etc. can be localized in space, the study of vacuum energy promotes attention to localized spectral information, a somewhat neglected topic. The present article is complementary to [21]; both are directed primarily to a mathematical audience.Much of the article is based on the undergraduate research [43] of J. Wilson under the direction of S. Fulling and G. Berkolaiko. Some results of that research have been published in [9,22,23], to which we refer for details. We are happy to acknowledge strong interactions with, and assistance from, the Texas A&M quantum graph research group