Matthias Ihme

CloudBench Dataset This dataset contains the post-processed statistical output from 10,000 Large-Eddy Simulations (LES) driven by a General Circulation Model (GCM), as part of the CloudBench dataset produced by Google Research. The data was generated by performing an ensemble of high-resolution large-eddy simulations (LES) covering diverse meteorological conditions across the tropical Pacific. The LES are forced with the large-scale conditions derived from a single GCM model (the GFDL CM4) for 500 locations, 4 representative months (January, April, July, and October), and 5 different climate scenarios. The simulations were conducted in Swirl-LM, a computational fluid dynamics (CFD) simulation framework that is accelerated by the Tensor Processing Unit (TPU). Refer to the documentation in Swirl-LM for information about installation, required hardware, and instructions on how to run large simulations on TPU nodes in Google Cloud.Further details on the dataset are available in the README file.

Understanding the dynamics of chemical reactions in solutions is vital, as their rates and kinetics are significantly affected by the solvent environment. Supercritical solvents offer extensive applications in chemical reactions by enabling the manipulation of the solution environment. In this study, we investigate the geminate recombination of iodine in a supercritical xenon solvent by using ReaxFF-based molecular dynamics simulations. Our findings reveal that the highest iodine recombination rate occurs near supercritical conditions, while lower-pressure conditions lead to reduced collision rates and unstable recombination, and higher-pressure conditions hinder iodine diffusion, resulting in a lower recombination rate. Our analysis shows that the xenon local density at the time of recombination is at least 2.5 times higher than the global density, confirming the presence of xenon clusters surrounding the Iodine atoms. This observation is further supported by coordination number analysis, which confirms an elevated xenon local density during recombination. In addition, the correlation between the total energy of xenon atoms within a cluster and recombined iodine atoms underscores the kinetic energy transfer process, validating the occurrence of geminate recombination. The excess kinetic energy from the recombining iodine atoms is transferred to the surrounding xenon atoms. Our examination of geminate recombination demonstrates that iodine atoms confined within xenon clusters-whether through manual insertion of atoms or the fast dissociation of an iodine molecule within xenon clusters-are more likely to recombine as primary geminate recombination. However, extending the iodine molecule dissociation time allows iodine atoms to diffuse out of the cluster, and the recombination to shift toward secondary geminate recombination.

Supercritical fluids are characterized by unique thermodynamic properties. One of these properties is the existence of two-component dynamics that is associated with distinct low-frequency and high-frequency vibrational responses of the fluid. However, the origin of this behavior remains unknown. By combining inelastic X-ray scattering and molecular dynamics simulations, we show that this behavior can be connected to density heterogeneities arising from molecular clusters. Analyses of measurements and molecular trajectories suggest that the two-component dynamics emerges due to distinct momentum fluctuations of clustered and unbound molecules. This connection between clusters and two-component dynamics highlights the importance of molecular-structural heterogeneities in supercritical fluids, colloids, and condensed-matter systems.

Discontinuous Galerkin (DG) methods combine features of finite element (FE) and finite volume (FV) solvers to solve partial differential equations in an element-local framework. This offers inherent advantages for massively-parallel computing architectures due to the high arithmetic intensity and lower memory requirements of element-local schemes. This paper presents a DG framework built using JAX, which targets several hardware platforms using the XLA compiler. The framework achieves an arbitrarily high order of accuracy and correctly simulates canonical fluid dynamics test cases using single-precision, but the associated convergence rate degrades very quickly using single-precision. Concurrently, single- and double-precision compute time is compared across a variety of hardware architecture including graphics processing units (GPUs), tensor processing units (TPUs), and both x86_64 and ARM computers. It is shown that single-precision produces as much as 2x speedup even on traditional architectures. This motivates the development of a mixed-precision framework that combines the speedup from using single-precision with the improved accuracy from double-precision representations. To this end, this paper demonstrates that selectively increasing the precision of certain operations in a DG solver can lower the machine-precision error floor. Some observations on the contributions of the precision of different operations in a DG solver to its overall accuracy and convergence rate are also presented.

Flammable liquids leaking on hot surfaces are an ignition hazard for fires in aircraft, vehicles, and heavy machinery. Engineering analysis to assess fire risk in these devices, must account for the flammability properties of fuels and other liquids. The autoignition temperature (AIT) is most often used for this purpose. AIT is determined following a standard procedure such as ASTM E659, but it is well-established that the minimum ignition temperature in practical industrial configurations is often significantly higher than the AIT. In this work, we introduce a canonical experimental apparatus that addresses this gap. This proposed apparatus enables the characterization of the probabilistic hot surface ignition behavior of flammable liquids in configurations featuring thermal stratification and reduced residence times compared to ASTM E659. It is highly flexible, enabling testing with various surface materials, droplet diameter, impact velocities, and repetition rate. The introduction of automated optical diagnostic enables rapid testing of multiple fuels. The main focus of the present work is to demonstrate the repeatability of the measurements conducted in this apparatus: In repeated experiments with n-hexane, we find that the standard deviation of the hot surface ignition temperature is 3.8 K. A thorough analysis of the experimental uncertainties is also conducted. We demonstrate the capabilities of the apparatus by examining fuel effects on the ignition behavior of eight fuels: n-hexane, a conventionally-derived Jet-A fuel, and six synthetic aviation turbine fuels and blending compounds (SAFs). We find that the 50 % ignition probability temperature of the most ignitable and least ignitable of our six SAFs differed by 70 K, which motivates future efforts to better account for fuel effects in ignition risk analysis as new SAFs are rapidly entering the aviation fuel market. This deposit contains the supplementary materials for this article: composition of the fuel samples and experimental data.

In January 2025, a series of severe wildfires in Los Angeles County burned over 50000 acres, destroyed over 16000 structures, and caused 30 fatalities. These fires, driven by intense Santa Ana winds, rank among the most destructive and deadliest fires in California’s history. Beyond the destruction and loss of lives, these fires also contributed to unhealthy levels of pollutants, such as fine particulate matter (PM 2.5 ), which cause increased respiratory and cardiopulmonary morbidity, yet this smoke-specific mortality burden is not routinely documented in official reports of wildfire damage and losses. By employing high-resolution air-quality forecast data, we quantify the excess mortality from smoke exposure during a wildfire episode in Los Angeles County in 2025. We found unhealthy air quality in the most populous census tracts immediately after the fire outbreak that persisted for several days. We estimated a total of 14 unaccounted excess deaths from acute wildfire smoke exposure, which is equivalent to 47% of direct fire fatalities. Our study shows that excess mortality from smoke exposure, even for a few days, causes increased mortality, which requires consideration for a comprehensive assessment of wildfires.

Supersonic retropropulsion (SRP) is an enabling technology for human missions to Mars and other planetary bodies, as existing parachute systems cannot provide sufficient deceleration for human-scale payloads. SRP is the use of rockets to decelerate a spacecraft by exhausting in the opposite direction of its descent. We present a physical model that predicts the boundary of a high-thrust and underexpanded single-jet plume in blunt penetration mode at zero angle of attack, which is then used in a control-volume analysis to predict the deceleration force caused by the impingement of the freestream flow on the spacecraft heatshield. The jet plume model is based on the axisymmetric method of characteristics and accounts for the post-bow-shock pressure and Mach disk location on the jet plume boundary. This model considers effects of stagnation pressure in the recirculation region for which we add a pressure deficit term in the force computation. Results demonstrate improved accuracy in predicting jet plume boundaries and force, under various operating conditions compared to existing analytical methods. Our method offers an improved low-order physics-based approach that aids in performing rapid analyses of SRP quantities of interest, thus adding to the engineering methods for designing entry, descent, and landing architectures.

Supercritical mixtures of ethanol (EtOH) and carbon dioxide (CO 2 ) are classified as type-I mixtures, with complete macroscopic miscibility. However, differences in molecular polarity and interactions suggest a distinct phase behavior at the microscopic level. Here, we combine small angle X-ray scattering experiments and molecular dynamics (MD) simulations to investigate the microscopic structure of EtOH–CO 2 mixtures under supercritical conditions. The structure factor exhibits nonlinear composition-dependent behavior, revealing pronounced local density fluctuations. The complementary MD simulations, using optimized force field parameters, provide atomistic insight, showing that EtOH forms self-associated, hydrogen-bonded aggregates, while CO 2 remains more uniformly distributed. Cluster analysis identifies a preferential EtOH-rich composition exceeding the bulk average, governed by a balance between energetic and entropic competition. These results demonstrate that, contrary to macroscopic expectations, the mixture exhibits significant microscopic heterogeneity and immiscibility, which may influence solubility, reactivity, transport properties, and thermodynamic response functions. These findings challenge the conventional views of type-I fluids and emphasize the necessity of revising mixture states and considering molecular polarity.

Understanding and controlling microscopic dynamics across spatial and temporal scales has driven major progress in science and technology over the past several decades. While ultrafast laser-based techniques have enabled probing nanoscale dynamics at their intrinsic temporal scales down to femto- and attoseconds, the long wavelengths of optical lasers have prevented the interrogation and manipulation of such dynamics with nanoscale spatial specificity. With advances in hard X-ray free electron lasers (FELs), significant progress has been made developing X-ray transient grating (XTG) spectroscopy, aiming at the coherent control of elementary excitations with nanoscale X-ray standing waves. So far, XTGs have been probed only at optical wavelengths, thus intrinsically limiting the achievable periodicities to several hundreds of nm. By achieving sub-femtosecond synchronization of two hard X-ray pulses at a controlled crossing angle, we demonstrate the generation of an XTG with spatial periods of 10 nm. The XTG excitation drives a thermal grating that drives coherent monochromatic longitudinal acoustic phonons in the cubic perovskite, SrTiO3 (STO). With a third X-ray pulse with the same photon energy, time-and-momentum resolved measurement of the XTG-induced scattering intensity modulation provides evidence of ballistic thermal transport at nanometer scale in STO. These results highlight the great potential of XTG for studying high-wave-vector excitations and nanoscale transport in condensed matter, and establish XTG as a powerful platform for the coherent control and study of nanoscale dynamics.

High-resolution spatiotemporal simulations effectively capture the complexities of atmospheric plume dispersion in complex terrain. However, their high computational cost makes them impractical for applications requiring rapid responses or iterative processes, such as optimization, uncertainty quantification, or inverse modeling. To address this challenge, this work introduces the Dual-Stage Temporal Three-dimensional UNet Super-resolution (DST3D-UNet-SR) model, a highly efficient deep learning model for plume dispersion predictions. DST3D-UNet-SR is composed of two sequential modules: the temporal module (TM), which predicts the transient evolution of a plume in complex terrain from low-resolution temporal data, and the spatial refinement module (SRM), which subsequently enhances the spatial resolution of the TM predictions. We train DST3D-UNet-SR using a comprehensive dataset derived from high-resolution large eddy simulations (LES) of plume transport. We propose the DST3D-UNet-SR model to significantly accelerate LES of three-dimensional (3D) plume dispersion by three orders of magnitude. Additionally, the model demonstrates the ability to dynamically adapt to evolving conditions through the incorporation of new observational data, substantially improving prediction accuracy in high-concentration regions near the source.

Supercritical fluids (SCFs) have attracted significant attention as solvents for chemical reactions due to their unique properties, such as high diffusivity, low viscosity, and tunable solvation properties. These properties profoundly influence reaction kinetics and are often attributed to the formation of molecular clusters within SCFs. To study the effect of supercritical solvent on chemical reactivity and dynamics of reactions, one needs to understand the dynamics of clusters in supercritical fluid. Extensive experiments on the photodissociation and recombination of iodine in supercritical fluids served as a model system for understanding these effects. Experimental studies have been complemented by theoretical and computational investigations, which mostly employ Monte Carlo or empirical molecular dynamics simulations. However, computational studies using non-reactive force fields and ab initio approaches present challenges in capturing reactive processes at larger scales within supercritical fluids. In this work, we developed the ReaxFF parameters by training against quantum mechanics data. ReaxFF reactive force field based molecular dynamics simulations were performed, studying the dynamics of a xenon solvent and cage effect at different thermodynamic conditions for the iodine recombination reaction. We show that the conditions near the critical point are the optimal conditions to study the cage effect. We show that the average lifetime of xenon clusters ranging between 5 and 11 ps is comparable to iodine geminate recombination. Our simulation results of iodine recombination in xenon solvent demonstrate the higher probability of iodine molecule formation in the presence of xenon clusters. Finally, we show that the supercritical condition exhibits the highest recombination rate for iodine atoms.

Planar entropy waves are commonly assumed for predicting indirect combustion noise. However, the non-planar and turbulent nature of flows found in most practical combustors challenges this assumption. In the present paper, we examine the indirect noise generated by non-planar and turbulent entropy fields through subsonic nozzles. Firstly, we introduce a new transfer function framework that accounts for the contribution of non-planar Fourier modes of the entropy field to the indirect noise spectra. When applied to a turbulent flow field, this method demonstrates a significant improvement in spectral predictions compared with a conventional approach that only considers the planar mode. Secondly, simulations show that non-planar Fourier modes become significant above a threshold frequency $f_{thresh}$ , found in the mid- to high-frequency range. This contribution of non-planar modes is explained by two-dimensional shear effects that distort the entropy waves. A scaling relation that uses residence times along streamlines is developed for $f_{thresh}$ , showing good agreement with simulation results. Finally, we show that the indirect noise from non-planar entropy modes found in aviation combustors can be significant at frequencies below 1 kHz, which might be relevant in situations of thermo-acoustic instabilities coupled to indirect noise.

When a flammable liquid is put in contact with a very hot surface, thermal ignition of fuel vapors can occur. In the event of a leak, this process, called hot surface ignition, can lead to fires in aircrafts, spacecrafts, vehicles, and machinery. In the aerospace industry, design practices and certification processes must ensure that this fire hazard is mitigated. In the present work, we performed experiments to assess whether sustainable aviation fuels have a different hot surface ignition behavior compared to petroleum-derived jet fuel. In a canonical configuration, 2.5mm fuel droplets were released onto a high temperature optically accessible surface. After a 150mm fall representative of a typical aircraft engine compartment, the droplets broke up upon impact and ignition occured if the surface temperature was sufficiently high. The temperature of ignition was quantified and we found that all investigated fuels had a similar or higher hot surface ignition temperature to jet-A in this configuration. The temperature of ignition spanned a 70K range for the 4 fuels investigated. High speed shadowgraphy revealed the effect of fuel surface tension on the droplet break-up process, and high speed chemiluminescence imaging revealed the multi-step ignition process that the droplet underwent.

Novel combustion systems in aerospace propulsion systems as scramjets and rotating detonation engines utilize higher pressure conditions to achieve optimal thermodynamic efficiency. As a result, the combustion chamber pressure may affect the interface stability of water, one of the main combustion products. Although the immiscibility between the liquid water droplet and the ambiance guarantees the formation of the subcritical interface, its dynamics and impact is typically neglected in many simulations. These are even more prominent with interfacial dynamics due to transcritical effects that arise due to strong shocks inside the combustion chamber, which is typical in modern detonation engines. To elucidate this, we consider a shock of M=1.36 for a water/nitrogen system, where the post-shock liquid- and vapor-phase conditions (T_1=600 K and T_2=730 K) are fully transcritical with respect to the critical mixing point. In this work, we employ the Regularized Interface Method (RIM), since it is able to resolve both subcritical interfacial dynamics along with supercritical mixing dynamics. Vanishing and spontaneous emergence of the interface is observed, indicating the need of resolving the interfacial dynamics and effects.

The mixing dynamics of injected propellants is a key factor in determining the ignition performance and combustion-instability response of liquid rocket engines. A key source of uncertainty in the prediction of LOX/GH2 mixing dynamics is the immiscibility of the injected propellants at supercritical pressure conditions, which induces liquid/vapor phase separation and surface-tension effects. However, modeling of LOX/GH2 mixing layers typically employs the Diffuse Interface Method (DIM), where interfacial dynamics are neglected. Recently, the Regularized Interface Method (RIM) was developed to consistently model the interfacial flows of immiscible mixtures at high pressure. Thus, the objective of this paper is to provide quantification of the impact of resolving interfacial effects on modeling of LOX/GH2 injection dynamics in both cryogenic temperature and transcritical conditions, simulating cold-start and stable operation, respectively. To accomplish this, we analyze simulation results of a benchmark high-pressure LOX/GH2 mixing layer configuration using RIM and DIM formulations at these respective conditions. Results of the study will quantify the need for interface-capturing simulations of propellant injection, even at supercritical pressure, as well as shed light on the complex transition dynamics from subcritical interfacial flows to supercritical mixing as the surface temperature is heated beyond the mixture's critical point.

High-resolution spatiotemporal simulations effectively capture the complexities of atmospheric plume dispersion in complex terrain. However, their high computational cost makes them impractical for applications requiring rapid responses or iterative processes, such as optimization, uncertainty quantification, or inverse modeling. To address this challenge, this work introduces the Dual-Stage Temporal Three-dimensional UNet Super-resolution (DST3D-UNet-SR) model, a highly efficient deep learning model for plume dispersion prediction. DST3D-UNet-SR is composed of two sequential modules: the temporal module (TM), which predicts the transient evolution of a plume in complex terrain from low-resolution temporal data, and the spatial refinement module (SRM), which subsequently enhances the spatial resolution of the TM predictions. We train DST3DUNet- SR using a comprehensive dataset derived from high-resolution large eddy simulations (LES) of plume transport. We propose the DST3D-UNet-SR model to significantly accelerate LES simulations of three-dimensional plume dispersion by three orders of magnitude. Additionally, the model demonstrates the ability to dynamically adapt to evolving conditions through the incorporation of new observational data, substantially improving prediction accuracy in high-concentration regions near the source. Keywords: Atmospheric sciences, Geosciences, Plume transport,3D temporal sequences, Artificial intelligence, CNN, LSTM, Autoencoder, Autoregressive model, U-Net, Super-resolution, Spatial Refinement.

Microstructural heterogeneities arising from molecular clusters directly affect the nonlinear thermodynamic properties of supercritical fluids. We present a physical model to elucidate the relation between energy exchange and heterogeneous cluster dynamics during the transition from liquidlike to gaslike conditions. By analyzing molecular-dynamics data and employing physical principles, the model considers contributions from three key processes, namely, changing cluster density, cluster separation, and transfer of molecules between clusters. We show that the proposed model is consistent with the energetics at subcritical conditions and can be used to explain the nonlinear behavior of thermodynamic response functions, including the peak in the isobaric heat capacity.

Supercritical fluids exhibit distinct thermodynamic and transport properties, making them of particular interest for a wide range of scientific and engineering applications. These anomalous properties emerge from structural heterogeneities due to the formation of molecular clusters at conditions above the critical point. While the static behavior of these clusters and their effects on the thermodynamic response functions have been recognized, the relation between the ultrafast cluster dynamics and transport properties remains elusive. By measuring the intermediate scattering function in carbon dioxide at conditions near the critical point with X-ray photon correlation spectroscopy, we directly capture the cross-over dynamics between 4 and 13 picoseconds, revealing the transition between ballistic and diffusive motion. Complementary analysis using large-scale molecular dynamics simulations reveals that this behavior arises from collisions between unbound molecules and clusters. This study provides direct evidence of the ultrafast momentum exchange between clusters, which has significant impact on transport properties, solvation processes, and reaction kinetics in supercritical fluids. A transition from ballistic to diffusive motion within 10 ps is observed in supercritical carbon dioxide with X-ray photon correlation spectroscopy. Collisions of unbound molecules with clusters are responsible for the ultrafast momentum exchange.

In the present work, a first-of-its-kind enrichment wall-model is developed within the spectral element method (SEM) framework for large-eddy simulations (LES) of wall-bounded turbulent flows. The method augments the polynomial solution in the wall-adjacent elements with an analytical law-of-the-wall enrichment function representing the mean velocity near the wall. In the solution representation, this enrichment function captures the large gradients in the boundary layer, which allows the polynomial modes to represent the turbulent fluctuations. The enriched solution is able to resolve the shear stress at the wall without any modification to the no-slip wall boundary conditions, which allows for greater accuracy in the near-wall region compared to traditional methods. The enrichment wall modeling approach is implemented in a high-order SEM computational fluid dynamics solver, Nek5000, and its performance is assessed in turbulent channel flow wall-modeled LES for a range of Reynolds numbers. It is demonstrated that the enrichment wall-model improves solution accuracy on under-resolved near-wall grids as compared to traditional shear stress wall-models.

Uncertainties in wildfire simulations pose a major challenge for making decisions about fire management, mitigation, and evacuations. However, ensemble calculations to quantify uncertainties are prohibitively expensive with high-fidelity models that are needed to capture today's ever-more intense and severe wildfires. This work shows that surrogate models trained on related data enable scaling multifidelity uncertainty quantification to high-fidelity wildfire simulations of unprecedented scale with billions of degrees of freedom. The key insight is that correlation is all that matters while bias is irrelevant for speeding up uncertainty quantification when surrogate models are combined with high-fidelity models in multifidelity approaches. This allows the surrogate models to be trained on abundantly available or cheaply generated related data samples that can be strongly biased as long as they are correlated to predictions of high-fidelity simulations. Numerical results with scenarios of the Tubbs 2017 wildfire demonstrate that surrogate models trained on related data make multifidelity uncertainty quantification in large-scale wildfire simulations practical by reducing the training time by several orders of magnitude from 3 months to under 3 h and predicting the burned area at least twice as accurately compared with using high-fidelity simulations alone for a fixed computational budget. More generally, the results suggest that leveraging related data can greatly extend the scope of surrogate modeling, potentially benefiting other fields that require uncertainty quantification in computationally expensive high-fidelity simulations.

Background Wildfire research uses ensemble methods to analyse fire behaviours and assess uncertainties. Nonetheless, current research methods are either confined to simple models or complex simulations with limitations. Modern computing tools could allow for efficient, high-fidelity ensemble simulations. Aims This study proposes a high-fidelity ensemble wildfire simulation framework for studying wildfire behaviour, assessing fire risks, analysing uncertainties, and training machine learning (ML) models. Methods We present a simulation framework that integrates the Swirl-Fire large-eddy simulation tool for wildfire predictions with the Vizier optimisation platform for automated run-time management of ensemble simulations and large-scale batch processing. All simulations are executed on tensor-processing units to enhance computational efficiency. Key results A dataset of 117 simulations is created, each with 1.35 billion mesh points. The simulations are compared to existing experimental data and show good agreement in terms of fire rate of spread. Analysis is performed for fire acceleration, mean rate of spread, and fireline intensity. Conclusions Strong coupling between wind speed and slope is observed for fire-spread rate and intermittency. A critical Froude number that delineates fires from plume-dominated to wind-dominated is identified and confirmed with literature observations. Implications The ensemble simulation framework is efficient in facilitating large-scale parametric wildfire studies.