Simulations of complex aeroelastic configurations of practical interest
The design of helicopter rotors is a complex task as it requires a coupled analysis of the aerodynamic and mechanical behaviour of the elastic rotor blades in various operating conditions.
Our activity as members of the MultiXscale Center of Excellence is fully focused on the development of simulation-workflows aimed at increasing the physical accuracy and the computational efficiency of these coupled aero-structural calculations.
Within these workflows, the open-source Lattice Boltzmann solver waLBerla is in charge of evaluating the aerodynamic forces acting on the rotating blades. The computed forces are then transferred to a structure-mechanic solver which evaluates the dynamics of the elastic blades as a result of inertial and aerodynamic forces.
waLBerla’s remarkable scalability properties on GPU-based hardware is expected to dramatically reduce the time needed for this type of simulations while offering the necessary high level of physical accuracy required in the rotor design process.
The Use of Electrostatics: Unlocking Research in Materials Science, Battery Development, Biomedics, and Astrophysics
In the goal to create more efficient batteries, develop novel materials with tailored properties, and elucidate the intricate mechanisms of biological systems, scientists rely on modeling the forces that act on particles and systems. One fundamental force that underlies all charged systems is based on electrostatics. The complex interplay of electrostatic forces governs the behavior of ions, molecules, and electrons in diverse systems, from lithium-ion batteries to protein-ligand interactions. By accurately modeling these electrostatic interactions, researchers can uncover new insights into the underlying systems, driving innovation in fields like energy storage, materials science, and biomedicine.
However, simulating electrostatic interactions poses a significant computational challenge, requiring both massive resources and up-scaling to thousands of processors. To overcome this, we need simulation codes that can efficiently harness the capabilities of modern computing architectures, from single GPUs to powerful massive parallel supercomputers. The key to achieving this is performance portability – the ability of a code to run efficiently on a wide range of hardware platforms without requiring extensive rewriting or optimization.
With the intention to overcome these limitations, our research team is developing novel, scalable algorithms for electrostatics that can leverage the power of diverse computing architectures. By utilizing libraries like Kokkos and Cabana, we are creating performance-portable electrostatic solvers that can seamlessly make the transition between different platforms, ensuring maximum efficiency by minimizing the need for platform-specific optimization. This innovative approach is set to enable new discoveries and advancements in fields where electrostatic interactions are crucial.
Simulations of large-scale carbon electrode materials for supercapacitors
Electrochemical double layer capacitors, often called supercapacitors, are energy storage systems which accumulate and release energy through reversible ion adsorption at electrode/electrolyte interfaces. Porous carbons are commonly used as electrode materials due to their relatively low cost and good electronic conductivity. Over the past decade, most of the simulations of supercapacitors were performed at the microscopic scale, using molecular dynamics software such as LAMMPS. However it is well known from experiments that commercial materials are highly inhomogeneous, while molecular simulations, where electrode sizes are a few nanometers, only allow for the inclusion of a few pores. It is therefore necessary to simulate electrodes and supercapacitors at larger scales.
To this end, as part of the MultiXscale Center of Excellence, we are developing a software, LPC3D, designed for mesoscopic simulations of capacitive properties of carbon-carbon capacitors, based on a lattice-gas model. The code calculates properties such as quantities of adsorbed ions, diffusion coefficients and NMR spectra for ions adsorbed in porous carbons. To increase the accuracy of the model without sacrificing its effectiveness, we are coupling LPC3D with molecular density functional theory (MDFT), which allows a better description of the electrolyte properties in each pore. In particular, MDFT provides a fast and accurate computation of the free energies in the solvents of interest for supercapacitors, such as water or acetonitrile, which is used as the main ingredient for the LPC3D simulations. The availability of the code will be useful not only for supercapacitor applications, but it can also be used in other problems that involve nanoporous materials such as capacitive deionization.
Coupling ESPResSo with waLBerla
We address a key challenge in soft matter science and process engineering: simulating systems where objects like polymers or nanoparticles are immersed in a solvent. While understanding the dynamics of these systems requires accounting for the solvent’s flow, resolving every tiny solvent molecule (such as water) isn’t feasible.
Instead, we couple two simulation methods. Molecular dynamics models the larger structures, while a lattice Boltzmann solver handles the fluid flow. This approach, for example, means that when an object moves faster than the surrounding fluid, it both slows down and pushes the liquid, accurately capturing the dynamic interactions.
Within MultiXscale, we integrate ESPResSo, a simulation package for coarse-grained models with the highly scalable waLBerla framework for lattice-Boltzmann an other lattice-based algorithms. We will thereby provide a tool which is fast and scalable, while still being accessible for domain scientists — even those with limited high-performance computing experience.
In practice, ESPResso uses waLBerla as a library. The package is controlled using Python, a language well known and widely used in the research community.
Lastly, we also provide a second coupling between the lattice Boltzmann method and a solver for diffusion, advection, and reaction equations, enabling studies such as the behavior of small ions in a flowing fluid. We ship lattice-Boltzmann implementations on both CPUs and GPUs for enhanced performance. The diffusion-advection-reaction solver will also be ported to GPU in the near future.
Ultrasound for biomedical applications
Understanding complex systems is like zooming in and out with a camera—sometimes you need a close-up to see fine details, and other times you need a wide-angle view to capture the whole scene. In science, this means studying both tiny atoms and larger structures like cells or materials. Multiscale simulations help bridge these levels, letting us see the big picture without losing important small-scale details.
As part of the MultiXscale Center of Excellence, we are implementing efficient multiscale coupling techniques into widely used simulation packages like LAMMPS and waLBerla. This makes it easier for researchers to combine detailed molecular dynamics (MD) with large-scale fluid simulations using the Lattice Boltzmann Method (LBM).
This is especially exciting for ultrasound research. By linking LBM for large-scale fluid flow with MD for atomic-level interactions, we can accurately simulate how ultrasound waves interact with proteins in our cells. The availability of efficient codes capable of handling such multiscale systems has the potential to enhance medical treatments, improve drug delivery, and advance ultrasound-based therapies, pushing the boundaries of scientific discovery.
