We develop weakly compressible multiphase flow solvers for large-scale GPU computing. By avoiding the pressure Poisson equation, our approach achieves strong parallel scalability and remains stable for challenging flows, including high density ratios and rapidly moving interfaces. Our ongoing work improves momentum conservation, interface capturing, numerical stability and the treatment of complex boundaries.

Adaptive mesh refinement (AMR) concentrates fine grids where they are needed, substantially reducing the cost of resolving thin films and gas–liquid interfaces. We combine a block-structured AMR method with our weakly compressible solver and dynamically redistribute the computational domain across multiple GPUs as the interface moves. This enables efficient simulations with very high local resolution.

We use pore-scale geometries reconstructed from X-ray CT images to simulate multiphase flow in rocks and other porous materials. These simulations provide relative permeability, pressure and velocity fields, and insight into injection strategies. Our methods make it possible to study low-capillary-number flows and high density ratios that are difficult to resolve with conventional approaches.

At low injection rates, surface tension can drive abrupt movements of fluid interfaces within pores. We analyse the conversion of surface, kinetic and viscous energy during these events to understand flow conditions beyond the predictions of conventional Darcy-scale models and to inform more efficient injection strategies.

Storing CO₂ below the seabed under conditions of negative buoyancy requires understanding its interaction with unconsolidated sediments. We couple the discrete element method (DEM) with multiphase flow simulation to investigate the motion of sand particles and the surrounding fluids. Scalable simulations of these interactions also support the study of other particle-laden flows.

By concentrating fine grids at moving interfaces, AMR allows us to resolve flow inside thin liquid films surrounded by gas on both sides. The figure shows the dynamically refined block structure; each block contains 4 × 4 cells in this view. We use this capability to investigate film dynamics that have been difficult to simulate directly.

Surfactants undergo advection, diffusion, adsorption and desorption between the liquid and the gas–liquid interface. We develop concentration transport models and couple them to multiphase flow solvers to capture their effects on bubble and film dynamics. Resolving these effects in thin films requires both accurate transport models and high spatial resolution.

Printing technologies are expanding to new substrates and inks with a wide range of physical properties. High-fidelity multiphase flow simulations allow us to examine droplet flight and impact across conditions that are difficult to explore with a single experimental apparatus.

We investigate complex fluids whose viscosity depends on shear rate and whose behaviour includes viscoelastic effects. Improving constitutive models and validating simulations against experiments helps us understand these fluids and their applications in printing and other fluid-based devices.

Compact solid oxide fuel cells (SOFCs) use fluid-mechanical effects in microchannels and porous fuel electrodes. In collaboration with Prof. Tetsuya Yamada at Institute of Science Tokyo, we study microchannel design through topology optimisation and perform pore-resolved simulations of gas transport to support more efficient devices.

Our in-house AMR solver dynamically distributes refined blocks across multiple GPUs. This concentrates resolution near gas–liquid interfaces and solid boundaries, allowing us to reproduce the onset and development of interfacial disturbances in oil jets discharged from bent pipes. The colours indicate the blocks assigned to each GPU.

Gas–liquid separators are used in reactor systems to separate gases from liquids such as molten salts. Their strongly disturbed multiphase flows require both high resolution and numerical stability. We develop scalable simulation methods to explore separator geometries and operating conditions that are difficult to investigate experimentally.