An issue first raised privately to us by some colleagues and later in [101], [102] concerns the nature of the matter distribution in our galactic model. They have noted that given the existence of the discontinuity of the function N z that we had pointed to in [63], a significant surface tensor S k i can be constructed with a surface density component given by to first order in G, i.e. G 1 . The square bracket notation [..] denotes the jump over a discontinuity of the given function, here at z = 0. Using (9.7), this becomesIt was claimed that this necessarily implied the existence of a singular physical surface of mass in the galactic plane above and beyond the continuous mass distribution that we had found, thus rendering our model unphysical.Having received this challenge, we calculated the surface mass that was said to be present in the four galaxies that we had studied by integrating (A.2) over the surface without paying heed to the actual sign of the result. Suspicions were aroused from the discovery that (A.2) in each case gave a numerical value slightly less than the mass that we had derived from the volume integral of our entire continuous mass density distribution using (9.18), (9.15) and (9.12). a With echoes from undergraduate mathematics courses, this pointed to a plausible explanation: in our case, with our choice of model, there is no physical mass layer present on the z = 0 plane. The surface integral of this singular layer is merely a mathematical construct that indirectly describes most of the continuously distributed mass by means of the Gauss divergence theorem. To see this, consider the vector F defined as bas a first option. We choose A(r, z) so thatwhere M is the total mass. As a more transparent second option, we choosewhere we definea It should be noted that the two terms in (A.2) were found to contribute equally.b er and ez are unit vectors in the r and z directions.
A. CRITICAL CHALLENGES AND OUR REPLIES
197From these definitions, we deduce the form of A(r, z) in order to produce the density as expressed through N in (9.12). We calculate the mass over the cylindrical volume defined by −∞ < z < ∞, 0 < r < r galaxy . By the Gauss divergence theorem, the volume integral of ρ, via (A.7) is equal to the integral of the normal component of F over the bounding surfaces. However, the integration must be over a continuous domain and since the e z component is discontinuous over the z = 0 plane, the volume integral must be split into an upper and a lower half (see Figure A.1). The two new surface integrals together would constitute the jump integral of (A.2) in the first option if one were to be cavalier about the directions of unit outward normals, as we shall discuss in what follows. The surfaces above and below the galaxy give zero because of the exponential factors in z and the final small contribution comes from the cylinder wall via the A function. In our solution, the actual physical distribution of mass is not in concentrated layers over bounding surfaces: the Gauss theorem gives the va...