We describe recent developments to the Castro astrophysics simulation code, focusing on new features that enable our simulations of X-ray bursts. Two highlights of Castro's ongoing development are the new integration technique to couple hydrodynamics and reactions to high order and GPU offloading. We discuss how these features will help offset some of the computational expense in X-ray burst models.

Modelling long-time convective flows in the interiors of stars is extremely challenging using conventional compressible hydrodynamics codes due to the acoustic timestep limitation. Many of these flows are in the low Mach number regime, which allows us to exploit the relationship between acoustic and advective time scales to develop a more computationally efficient approach. MAESTROeX is an open source low Mach number stellar hydrodynamics code that allows much larger timesteps to be taken, therefore enabling systems to be modelled for much longer periods of time. This is particularly important for the problem of convection in the cores of rotating massive stars prior to core collapse. To fully capture the dynamics, it is necessary to model these systems in three dimensions at high resolution over many rotational periods. We present an overview of MAESTROeX's current capabilities, describe ongoing work to incorporate the effects of rotation and discuss how we are optimising the code to run on GPUs.

Simulations in stellar astrophysics involve the coupling of hydrodynamics and nuclear reactions under a wide variety of conditions, from simmering convective flows to explosive nucleosynthesis. Numerical techniques such as operator splitting (most notably Strang splitting) are usually employed to couple the physical processes, but this can affect the accuracy of the simulation, particularly when the burning is vigorous. Furthermore, Strang splitting does not have a straightforward extension to higher-order integration in time. We present a new temporal integration strategy based on spectral deferred corrections and describe the second- and fourth-order implementations in the open-source, finite-volume, compressible hydrodynamics code Castro. One notable advantage to these schemes is that they combine standard low-order discretizations for individual physical processes in a way that achieves an arbitrarily high order of accuracy. We demonstrate the improved accuracy of the new methods on several test problems of increasing complexity.

We present MAESTROeX, a massively parallel solver for low Mach number astrophysical flows. The underlying low Mach number equation set allows for efficient, long-time integration for highly subsonic flows compared to compressible approaches. MAESTROeX is suitable for modeling full spherical stars as well as well as planar simulations of dynamics within localized regions of a star, and can robustly handle several orders of magnitude of density and pressure stratification. Previously, we have described the development of the predecessor of MAESTROeX, called MAESTRO, in a series of papers. Here, we present a new, greatly simplified temporal integration scheme that retains the same order of accuracy as our previous approaches. We also explore the use of alternative spatial mapping of the one-dimensional base state onto the full Cartesian grid. The code leverages the new AMReX software framework for block-structured adaptive mesh refinement (AMR) applications, allowing for scalability to large fractions of leadership-class machines. Using our previous studies on the convective phase of single-degenerate progenitor models of Type Ia supernovae as a guide, we characterize the performance of the code and validate the new algorithmic features. Like MAESTRO, MAESTROeX is fully open source.

We discuss the challenges of modeling X-ray bursts in multi-dimensions, review the different calculations done to date, and discuss our new set of ongoing simulations. We also describe algorithmic improvements that may help in the future to offset some of the expense of these simulations, and describe what may be possible with exascale computing.

We describe the AMReX suite of astrophysics codes and their application to modeling problems in stellar astrophysics. Maestro is tuned to efficiently model subsonic convective flows while Castro models the highly compressible flows associated with stellar explosions. Both are built on the block-structured adaptive mesh refinement library AMReX. Together, these codes enable a thorough investigation of stellar phenomena, including Type Ia supernovae and X-ray bursts. We describe these science applications and the approach we are taking to make these codes performant on current and future many-core and GPU-based architectures.

The dynamics of helium shell convection driven by nuclear burning establish the conditions for runaway in the sub-Chandrasekhar mass, double detonation model for Type Ia supernovae, as well as for a variety of other explosive phenomena. We explore these convection dynamics for a range of white dwarf core and helium shell masses in three dimensions using the low Mach number hydrodynamics code Maestro. We present calculations of the bulk properties of this evolution, including time-series evolution of global diagnostics, lateral averages of the 3D state, and the global 3D state. We find a variety of outcomes including quasi-equilibrium, localized runaway, and nova-like runaway. Our results suggest the double detonation progenitor model is promising, that 3D, dynamic convection plays a key role, and that these systems warrant further study.

We perform the first detailed three-dimensional simulation of low Mach number convection preceding thermonuclear ignition in a mixed H/He X-ray burst. Our simulations include a moderate-sized, approximate network that captures hydrogen and helium burning up through rp-process breakout. We look in detail at the difference between two- and three-dimensional convective fields, including the details of the turbulent convection.

Type I X-ray Bursts (XRBs) are thermonuclear explosions of accreted material on the surfaces of a neutron stars in low mass X-ray binaries. Prior to the ignition of a subsonic burning front, runaway burning at the base of the accreted layer drives convection that mixes fuel and heavy-element ashes. In this second paper in a series, we explore the behavior of this low Mach number convection in mixed hydrogen/helium layers on the surface of a neutron star using two-dimensional simulations with the Maestro code. Maestro takes advantage of the highly subsonic flow field by filtering dynamically unimportant sound waves while retaining local compressibility effects, such as those due to stratification and energy release from nuclear reactions. In these preliminary calculations, we find that the rp-process approximate network creates a convective region that is split into two layers. While this splitting appears artificial due to the approximations of the network regarding nuclear flow out of the breakout reaction 18Ne(a,p)21Na, these calculations hint at further simplifications and improvements of the burning treatment for use in subsequent calculations in three dimensions for a future paper.

We present high-resolution, full-star simulations of the post-ignition phase of Type Ia supernovae using the compressible hydrodynamics code Castro. Initial conditions, including the turbulent velocity field and ignition site, are imported directly from a simulation of the last few hours of presupernova convection using a low Mach number code, Maestro. Adaptive mesh refinement allows the initial burning front to be modeled with an effective resolution of 36,864^3 zones (~136 m/zone). The initial rise and expansion of the deflagration front are tracked until burning reaches the star's edge and the role of the background turbulence on the flame is investigated. The effect of artificially moving the ignition location closer to the star's center is explored. The degree to which turbulence affects the burning front decreases with increasing ignition radius since the buoyancy force is stronger at larger radii. Even central ignition --- in the presence of a background convective flow field --- is rapidly carried off-center as the flame is carried by the flow field. We compare our results to analytic models for burning thermals, and find that they reproduce the general trends of the bubble's size and mass, but underpredict the amount of buoyant acceleration due to simplifying assumptions of the bubble's properties. Overall, we find that the amount of mass that burns prior to flame break out is small, consistent with a "gravitationally confined detonation" occurring at a later epoch, but additional burning will occur following breakout that may modify this conclusion.

We assess the robustness of a low Mach number hydrodynamics algorithm for modeling helium shell convection on the surface of a white dwarf in the context of the sub-Chandrasekhar model for Type Ia supernovae. We use the low Mach number stellar hydrodynamics code, MAESTRO, to perform three-dimensional, spatially-adaptive simulations of convection leading up to the point of the ignition of a burning front. We show that the low Mach number hydrodynamics model provides a robust description of the system.

We extend our previous three-dimensional, full-star simulations of the final hours of convection preceding ignition in Type Ia supernovae to higher resolution using the adaptive mesh refinement capability of our low Mach number code, MAESTRO. We report the statistics of the ignition of the first flame at an effective 4.34 km resolution, and general flow field properties at an effective 2.17 km resolution. We find that off-center ignition is likely, with radius of 50 km most favored and a likely range of 40 to 75 km. This is consistent with our previous coarser (8.68 km resolution) simulations, implying that we have achieved sufficient resolution in our determination of likely ignition radii. The dynamics of the last few hot spots preceding ignition suggest that a multiple ignition scenario is not likely. With improved resolution, we can more clearly see the general flow pattern in the convective region, characterized by a strong outward plume with a lower speed recirculation. We show that the convective core is turbulent with a Kolmogorov spectrum and has a lower turbulent intensity and larger integral length scale than previously thought (on the order of 16 km s$^{-1}$ and 200 km, respectively), and we discuss the potential consequences for the first flames.

We present our end-to-end capability for computing the convective phase through the explosion phase of Type Ia supernovae. We compute the convective phase up to the time of ignition using our low Mach number code, MAESTRO, and the subsequent explosion phase using our compressible code, CASTRO. Both codes share the same BoxLib software framework and use finite-volume, block-structured adaptive mesh refinement (AMR) to enable high-resolution, three-dimensional full-star simulations that scale to 100,000+ cores. We present preliminary results from the first-ever simulations of convection preceding ignition using MAESTRO with AMR. We also demonstrate our ability to initialize a compressible simulation of the explosion phase in CASTRO using data obtained directly from MAESTRO just before ignition. Some care must be taken during this initialization procedure when interpreting the size and distribution of hot spots.

We present multidimensional simulations of the early convective phase preceding ignition in a Type I X-ray burst using the low Mach number hydrodynamics code, MAESTRO. A low Mach number approach is necessary in order to perform long-time integration required to study such phenomena. Using MAESTRO, we are able to capture the expansion of the atmosphere due to large-scale heating while capturing local compressibility effects such as those due to reactions and thermal diffusion. We also discuss the preparation of one-dimensional initial models and the subsequent mapping into our multidimensional framework. Our method of initial model generation differs from that used in previous multidimensional studies, which evolved a system through multiple bursts in one dimension before mapping onto a multidimensional grid. In our multidimensional simulations, we find that the resolution necessary to properly resolve the burning layer is an order of magnitude greater than that used in the earlier studies mentioned above. We characterize the convective patterns that form and discuss their resulting influence on the state of the convective region, which is important in modeling the outburst itself.

Performing high-resolution, high-fidelity, three-dimensional simulations of Type Ia supernovae (SNe Ia) requires not only algorithms that accurately represent the correct physics, but also codes that effectively harness the resources of the most powerful supercomputers. We are developing a suite of codes that provide the capability to perform end-to-end simulations of SNe Ia, from the early convective phase leading up to ignition to the explosion phase in which deflagration/detonation waves explode the star to the computation of the light curves resulting from the explosion. In this paper we discuss these codes with an emphasis on the techniques needed to scale them to petascale architectures. We also demonstrate our ability to map data from a low Mach number formulation to a compressible solver.

Many astrophysical phenomena are highly subsonic, requiring specialized numerical methods suitable for long-time integration. In a series of earlier papers we described the development of MAESTRO, a low Mach number stellar hydrodynamics code that can be used to simulate long-time, low-speed flows that would be prohibitively expensive to model using traditional compressible codes. MAESTRO is based on an equation set derived using low Mach number asymptotics; this equation set does not explicitly track acoustic waves and thus allows a significant increase in the time step. MAESTRO is suitable for two- and three-dimensional local atmospheric flows as well as three-dimensional full-star flows. Here, we continue the development of MAESTRO by incorporating adaptive mesh refinement (AMR). The primary difference between MAESTRO and other structured grid AMR approaches for incompressible and low Mach number flows is the presence of the time-dependent base state, whose evolution is coupled to the evolution of the full solution. We also describe how to incorporate the expansion of the base state for full-star flows, which involves a novel mapping technique between the one-dimensional base state and the Cartesian grid, as well as a number of overall improvements to the algorithm. We examine the efficiency and accuracy of our adaptive code, and demonstrate that it is suitable for further study of our initial scientific application, the convective phase of Type Ia supernovae.

We present a new code, CASTRO, that solves the multicomponent compressible hydrodynamic equations for astrophysical flows including self-gravity, nuclear reactions and radiation. CASTRO uses an Eulerian grid and incorporates adaptive mesh refinement (AMR). Our approach to AMR uses a nested hierarchy of logically-rectangular grids with simultaneous refinement in both space and time. The radiation component of CASTRO will be described in detail in the next paper, Part II, of this series.

We present the first three-dimensional, full-star simulations of convection in a white dwarf preceding a Type Ia supernova, specifically the last few hours before ignition. For these long-time calculations we use our low Mach number hydrodynamics code, MAESTRO, which we have further developed to treat spherical stars centered in a three-dimensional Cartesian geometry. The main change required is a procedure to map the one-dimensional radial base state to and from the Cartesian grid. Our models recover the dipole structure of the flow seen in previous calculations, but our long-time integration shows that the orientation of the dipole changes with time. Furthermore, we show the development of gravity waves in the outer, stable portion of the star. Finally, we evolve several calculations to the point of ignition and discuss the range of ignition radii.