Current practice in stellar evolution is to employ one-dimensional calculations that quantitatively apply only to a minority of the observed stars (single non-rotating stars, or well detached binaries). Even in these systems, astrophysicists are dependent on approximations to handle complex three-dimensional processes like convection. Understanding the structure of binary stars, like those that lead to the Type Ia supernovae used to measure the expansion of the universe, are grossly non-spherical and await a 3D treatment. To approach very large problems like multi-dimensional modeling of stars, the Lawrence Livermore National Laboratory has invested in massively parallel computers and invested even more in developing the algorithms to utilize them on complex physics problems. We have leveraged skills from across the lab to develop a 3D stellar evolution code, Djehuty (after the Egyptian god for writing and calculation) that operates efficiently on platforms with thousands of nodes, with the best available physical data (opacities, EOS, etc.). Djehuty has incorporated all basic physics for modeling stars including an accurate equation of state, radiation transport by diffusion, thermonuclear reaction rates, and hydrodynamics, and we have begun testing it in a number of applications.

The DJEHUTY project is an intensive effort at the Lawrence Livermore National Laboratory (LLNL) to produce a general purpose 3-D stellar structure and evolution code to study dynamic processes in whole stars.

We present the results of self-consistent models of Compton-heated accretion disk coronae. The models are calculated using a new method for computing monochromatic radiative transfer in two-dimensions. The method splits the radiation into direct and scattered components. The direct radiation is computed by calculating the optical depth along rays, while transfer of the scattered radiation is approximated by flux-limited diffusion. The resulting code agrees with more accurate treatments to within 50\%, and is highly efficient, making it practical for use in large hydrodynamic simulations. The coronal models are used to confirm the results of earlier work, and to extend it to higher luminosities. In contrast to earlier work, which found the outer disks to be shadowed by the inner corona at high luminosities, we find our results to form an almost continuous extension of the models at lower luminosities. This is due to the presence of multiply-scattered radiation, which acts to partially offset the loss of direct radiation from the central source. Although the analytic methods derived at lower luminosities cannot be used to derive the coronal structure for $L/L\sr{Edd}\gapprox0.1$, the results of the models are amenable to semiempirical fits. We also discuss possible observational consequences of the results for coronal veiling and line fluorescence from the disk.