SPRAY: A smoothed particle radiation hydrodynamics code for modeling high intensity laser-plasma interactions

This paper introduces SPRAY, the first massively parallel GPU-accelerated, mesh-free smoothed particle hydrodynamics code designed to accurately simulate high-intensity laser-plasma interactions by overcoming numerical challenges through tailored SPH formulations, flux-limited diffusion, and a mesh-free WKB-based laser coupling module.

The Gist

Imagine you are trying to predict how a drop of water behaves when a giant, high-powered fire hose blasts it. Now, imagine that "water" is actually a solid block of metal, and the "fire hose" is an incredibly intense laser beam. When they collide, the metal doesn't just melt; it explodes, turns into a super-hot gas (plasma), and gets crushed by shockwaves. This is the world of High-Energy Density Physics, and it's the key to understanding how stars work and how we might one day create unlimited clean energy (fusion).

For decades, scientists have tried to simulate these explosions on computers. But the old tools were like trying to track a swirling tornado using a rigid grid of boxes. As the gas swirls and stretches, the boxes get in the way, or the simulation crashes because the "tornado" doesn't fit the grid.

Enter SPRAY.

What is SPRAY?

Think of SPRAY not as a computer program, but as a digital swarm of bees.

Instead of using a rigid grid (like a chessboard), SPRAY treats the plasma as millions of tiny, individual particles (the bees). Each particle carries its own mass, temperature, and speed.

• The Old Way (Grid-based): Imagine trying to track a crowd of people by only looking at them through a window with square panes. If the crowd moves diagonally or swirls, you lose track of who is where.
• The SPRAY Way (Particle-based): Imagine the crowd is the simulation. Every person (particle) knows where their neighbors are and how they are moving. If the crowd stretches into a long line or swirls into a ball, the simulation adapts instantly because there are no fixed boxes to get in the way.

The Superpowers of SPRAY

1. The "Ghost" Laser Beam
Lasers don't just hit a target; they bend and twist as they pass through the hot, changing plasma (like light bending through a heat haze).

• The Innovation: SPRAY has a special module that treats the laser beam like a swarm of invisible "ghost particles." These ghosts fly through the plasma, bending naturally as the density changes, and drop off energy exactly where they get absorbed.
• Why it matters: Previous methods tried to force the laser to fit into a grid, which was inaccurate. SPRAY lets the laser flow freely, just like light does in real life.

2. The "Mirror" Trick for Edges
When a laser hits a solid target, the surface blows off (ablation), creating a rapidly expanding cloud. In computer simulations, the edge of this cloud is a nightmare because there are no particles "outside" the edge to help calculate the physics.

• The Innovation: SPRAY uses a clever trick called Particle Mirroring. It imagines "ghost particles" on the other side of the edge, mirroring the real ones. This fills in the missing information, allowing the simulation to calculate the pressure and density at the edge perfectly, without the numbers crashing or becoming nonsense.

3. The Three-Temperature Brain
In a normal fire, everything is roughly the same temperature. But in a laser-plasma explosion, the electrons (tiny particles) get super hot instantly, while the ions (heavier atoms) stay cooler for a moment. They are like two people in a room: one is running a marathon (electrons), and the other is sitting in a chair (ions).

• The Innovation: SPRAY tracks these three things separately: Electrons, Ions, and Radiation. It calculates how they talk to each other and exchange heat. This is crucial because if you assume they are all the same temperature, your prediction of the explosion will be wrong.

4. The GPU Powerhouse
Simulating millions of particles is incredibly heavy lifting.

• The Innovation: SPRAY is built to run on GPUs (the same powerful chips found in high-end video game computers). It splits the work among thousands of tiny processors working in parallel. It's like having a stadium full of workers instead of a single person trying to move a mountain. This allows the simulation to run fast enough to be useful for real-world experiments.

Why Do We Care?

The paper proves that SPRAY works by testing it against known problems:

• The Shockwave Test: It successfully simulated a classic "shock tube" problem, proving it can handle violent explosions without breaking.
• The Laser Test: It simulated a laser hitting aluminum and matched the results of a famous, established code called MULTI-IFE almost perfectly.
• The Instability Test: It simulated the "Rayleigh-Taylor instability" (think of oil floating on water and mixing chaotically). SPRAY predicted how the mixing would happen, matching the results of other super-computers.

The Bottom Line

SPRAY is a new, super-fast, and flexible tool for scientists. It uses a "swarm of particles" approach instead of a rigid grid to simulate how lasers blast through matter. By handling the messy, chaotic edges of explosions and tracking the complex dance of heat and light, it helps scientists design better fusion reactors and understand the universe's most violent events.

It's essentially the difference between trying to model a fluid with a stiff net versus modeling it with a swarm of intelligent, interacting bees. The bees win every time.

Technical Summary

1. Problem Statement

The interaction between high-intensity lasers and plasmas is central to High Energy Density Physics (HEDP) and Inertial Confinement Fusion (ICF) research. These interactions involve complex physical phenomena, including:

• Fluid Deformation: Highly non-linear fluid dynamics driven by instabilities (e.g., Ablative Rayleigh-Taylor Instabilities) and shock waves.
• Multi-Scale Physics: Coupling between hydrodynamics, radiation transport, and thermal conduction across vastly different time and spatial scales.
• Boundary Complexity: The formation of rapidly expanding coronal plasmas and free surfaces where density gradients are steep ($10^3 - 10^4$).

Limitations of Existing Methods:

• Eulerian (Grid-based) Methods: Struggle to track moving multi-fluid interfaces and plasma boundaries without complex Adaptive Mesh Refinement (AMR), which is computationally expensive.
• Lagrangian (Cell-based) Methods: While they track boundaries better, they are limited in handling large fluid deformations common in hydrodynamic instabilities.
• Existing SPH (Smoothed Particle Hydrodynamics): Historically lacked detailed physical modeling for radiation transport and laser coupling specific to HEDP, often relying on simplified approximations or mesh-based ray tracing.

2. Methodology: The SPRAY Code

The authors developed SPRAY (Smoothed Particle RAdiation hYdrodynamics), a massively parallel, GPU-accelerated, mesh-free, Lagrangian code.

A. Core SPH Formulations

• Kernel: Uses the Wendland C2 kernel for stability against pairing instabilities.
• Conservation Laws:
  • Momentum: Formulated using a conjugate pair approach ($\phi=1$) to ensure energy conservation.
  • Energy: Implements a single-fluid, three-temperature model (Ion, Electron, Radiation). This is crucial for high-intensity laser interactions where ion-electron equilibrium is not instantaneous, and radiation temperature differs significantly from electron temperature.
  • Mass: Uses the continuity equation with Kernel Gradient Correction (KGC) to improve density estimation near free surfaces.
• Numerical Dissipation: Implements an energy-conserving artificial viscosity scheme (Price's method) with a dynamic dissipation switch to capture shock waves without generating unphysical oscillations.

B. Radiation Transport

• Approach: Uses the flux-limited diffusion approximation derived from the zeroth moment of the radiation transport equation. This is justified by the generally thick optical depth of laser-produced plasmas.
• Coupling: Solves radiation transport implicitly (using a backward Euler scheme with Jacobi iteration) while hydrodynamics is solved explicitly (operator splitting).
• Non-linearity: Retains the quartic dependence of Planck energy density on electron temperature ($T_e^4$) rather than linearizing it, preventing significant errors in energy extrapolation.
• Solver: Uses a Krylov subspace iterative method (BiCGSTAB with incomplete LU preconditioning) on GPUs to solve the resulting linear systems efficiently.

C. Laser Energy Deposition

• Physics: Models inverse bremsstrahlung absorption using the Wentzel-Kramers-Brillouin (WKB) approximation.
• Ray Tracing: A novel mesh-free ray-tracing scheme where laser rays are treated as hypothetical SPH particles.
  • Rays traverse the electron density gradient field.
  • Refraction is calculated based on the electron density gradient.
  • Energy deposition is distributed to fluid particles via kernel weighting.
  • This avoids the grid-dependency of traditional ray-tracing methods (e.g., in FLASH or HYDRA).

D. Boundary and Search Algorithms

• Free Surface Tracking: A novel scheme combining free surface tracking with a particle mirroring method. It creates virtual "mirrored" particles across the boundary to compensate for particle deficiency, ensuring accurate gradient estimation for pressure and velocity at the expanding plasma edge.
• Nearest Neighbor Particle Search (NNPS): Uses a cell-based approach optimized for GPUs. It employs a dual-smoothing-length strategy (one for resolution, one for search range) to handle the extreme density variations caused by ablation and compression.
• Anisotropic Kernels: Implements ellipsoidal kernels that evolve based on the velocity field gradient to handle anisotropic expansion (e.g., laser ablation direction).

E. Parallelization

• Architecture: Written in CUDA C/C++ for NVIDIA GPUs.
• Scaling: Utilizes multi-GPU support with optimized inter-GPU particle exchange routines to minimize memory transfer overhead. The code achieves near-linear scaling ($O(N)$) up to 10 million particles.

3. Key Contributions

1. First SPH Application in HEDP: To the authors' knowledge, this is the first application of SPH methods to laser-plasma interactions in HEDP with detailed physical modeling (EOS, opacity, multi-temperature).
2. Mesh-Free Laser Ray Tracing: Development of a fully mesh-free ray-tracing module coupled directly to the SPH particle distribution, capable of handling arbitrary geometries and refraction.
3. Robust Free Surface Handling: A new particle mirroring and tracking algorithm that accurately resolves steep density gradients at the ablation front, a known failure point for standard SPH.
4. Three-Temperature RHD: Implementation of a coupled ion-electron-radiation model with implicit radiation transport and explicit hydrodynamics, specifically tailored for nanosecond-pulse laser interactions.
5. GPU Optimization: A highly optimized, massively parallel implementation demonstrating strong scaling on modern GPU architectures.

4. Results and Verification

The code was validated against analytical solutions and established reference codes (MULTI-IFE and ATHENA):

• Sod Shock Tube: Successfully captured shock waves and rarefaction waves, matching analytical solutions and demonstrating the effectiveness of the artificial viscosity scheme.
• Laser Irradiation of Aluminum: Compared against the cell-based code MULTI-IFE. SPRAY results for density, velocity, ion/electron temperatures, and energy deposition were nearly identical to MULTI-IFE, validating the laser coupling and radiation transport modules.
• Rayleigh-Taylor Instability: Benchmarked against the MHD code ATHENA. SPRAY successfully reproduced the growth rates of spikes and bubbles and the qualitative evolution of the instability across various resolutions and Atwood numbers.
• Implosion Benchmark: Simulated a cylindrical shell implosion (2.8 million particles) on 4 NVIDIA A100 GPUs, resolving high-mode instabilities and turbulent mixing.
• ICF Target Compression: Simulated a laser-driven implosion of a DT/CH target. The results matched the 1D reference code MULTI-IFE regarding implosion velocity, confirming the accuracy of the mesh-free ray-tracing and energy coupling.

5. Significance and Future Work

Significance:
SPRAY bridges the gap between the robustness of SPH in handling complex deformations and the rigorous physics required for HEDP. It offers a promising alternative to grid-based codes for problems involving large fluid deformations and complex boundaries, leveraging GPU acceleration for efficiency.

Limitations & Future Directions:

• Phase Changes: Modeling the solid-to-plasma phase transition remains a challenge for SPH.
• Radiation Transport: Currently uses a gray approximation. Future work aims to implement multi-group diffusion to better model X-ray driven ICF (indirect drive).
• Beam Interactions: The current ray tracer assumes independent rays. Future versions will include Cross-Beam Energy Transfer (CBET) and beam-beam interactions.
• 3D Extension: The current study is limited to 2D; 3D implementation is planned.

In conclusion, SPRAY represents a significant step forward in computational HEDP, demonstrating that particle-based, mesh-free methods can achieve accuracy comparable to established grid-based codes while offering superior handling of complex fluid boundaries.