We examine the underlying physics of the marine controlledsource electromagnetic ͑CSEM͒ method through the use of cross-sectional plots of the vector-current density. A systematic comparison of the cross-sectional current-density distribution within uniform and reservoir-bearing seafloor models reveals that the method induces detectable reservoir responses at the seafloor for source-receiver offsets that are frequency dependent. Higher frequencies generally result in larger anomalous differences between the two models at shorter offsets up to a frequency where induced currents no longer effectively interact with the reservoir due to electromagnetic ͑EM͒ attenuation. At zero and low frequencies, the less-attenuated background EM fields mask the reservoir response, although large induced currents are normally incident upon the reservoir. The reservoir response is also masked at larger offsets and/or in shallow environments by the airwave that can be thought of as energy diffusing up and down through the seawater column and propagating along the air-seawater interface. As the background EM fields and airwave are nearly horizontal because of the air-seawater boundary, vertical electric-field ͑E͒ measurements are free of the masking effect and, thus, might serve as additional useful information, especially in shallow environments where horizontal E measurements lose resolution of the reservoir. Horizontal magnetic field ͑B͒ measurements supplement the horizontal E measurements as E and B decay at different rates away from a source position in the near-field zone of a horizontal electric-dipole source. Comparison of CSEM responses for 1D and 3D reservoirs indicates that the 3D reservoir response is substantially smaller because of the limited surface area of the localized 3D reservoir, and that the pattern of the 3D reservoir response radically varies with a source position relative to reservoir boundaries.

We present a finite-element time-domain ͑FETD͒ approach for the simulation of 3D electromagnetic ͑EM͒ diffusion phenomena. The finite-element algorithm efficiently simulates transient electric fields and the time derivatives of magnetic fields in general anisotropic earth media excited by multiple arbitrarily configured electric dipoles with various signal waveforms. To compute transient electromagnetic fields, the electric field diffusion equation is transformed into a system of differential equations via Galerkin's method with homogeneous Dirichlet boundary conditions. To ensure numerical stability and an efficient time step, the system of the differential equations is discretized in time using an implicit backward Euler scheme. The resultant FETD matrix-vector equation is solved using a sparse direct solver along with a fill-in reduced ordering technique. When advancing the solution in time, the FETD algorithm adjusts the time step by examining whether or not the current step size can be doubled without unacceptably affecting the accuracy of the solution. To simulate a step-off source waveform, the 3D FETD algorithm also incorporates a 3D finite-element direct current ͑FEDC͒ algorithm that solves Poisson's equation using a secondary potential method for a general anisotropic earth model. Examples of controlled-source FETD simulations are compared with analytic and/or 3D finite-difference time-domain solutions and are used to confirm the accuracy and efficiency of the 3D FETD algorithm.