radiative heat transfer

This work addresses the challenge of high-fidelity thermal modeling for lunar rovers operating under extreme thermal environments, where large temperature gradients, radiative heat transfer, and complex surface conditions hinder accurate simulation. To overcome this, the authors propose an adaptive coarse-grid thermal simulation framework that synergistically integrates physics-informed machine learning with numerical solvers. Specifically, the method combines a transferable neural network with a differentiable finite difference solver to dynamically refine mesh resolution and reconstruct high-resolution temperature fields while enforcing physical consistency. Evaluated against conventional coarse-grid models and purely data-driven artificial neural networks, the proposed approach improves prediction accuracy by 50% and 39%, respectively, while achieving a computational speedup of threefold compared to high-fidelity simulations—effectively balancing efficiency and accuracy.

This study addresses thermal accumulation in high-density orbital AI clusters caused by convective and radiative thermal crosstalk, which leads to performance throttling and reduced hardware lifespan. The work proposes a novel paradigm—thermal-aware heterogeneity—that treats spatial cooling disparities as a primary resource dimension, thereby transcending the limitations of conventional uniform load distribution. By developing a thermal-aware scheduling framework (TLB) grounded in real-time fluid temperature and radiation absorption metrics, the approach dynamically migrates large language model workloads to nodes with optimal cooling conditions, achieving thermal load balancing. This strategy effectively mitigates thermal bottlenecks, restores model floating-point utilization (MFU), reduces thermal stress, enhances computational efficiency, and extends the operational lifetime of orbital hardware, thus enabling sustainable AI deployment in space environments.

Existing topology optimization methods struggle to accurately model multidirectional mutual radiation effects, limiting performance improvements in radiative heat dissipation devices. This work proposes a density-based topology optimization framework that couples finite element heat conduction analysis with a radiative heat transfer model combining the zonal method and ray tracing. The approach explicitly accounts for mutual radiation during optimization and treats intermediate-density materials as participating media, thereby physically consistent modeling of radiative behavior at implicit boundaries. Notably, this is the first topology optimization method to employ ray tracing for precise modeling of mutual radiation, successfully generating high-performance radiative heat sinks and multilayer insulation structures unattainable by conventional approaches. The results demonstrate the critical influence of the balance between conduction and radiation on optimal configurations, with designs significantly outperforming existing solutions.

This work addresses the limitations of conventional on-orbit computing satellites, which struggle to support large-scale AI inference due to low computational power per unit mass, poor thermal dissipation, and bulky power systems. The authors propose an integrated distributed satellite architecture that unifies structural, power, computing, and thermal management subsystems. By incorporating large-area vapor chamber radiators, custom space-grade chips, high-density photovoltaics, and modular subarray communications, the design achieves exceptional thermal regulation and specific power. The resulting system delivers over 100 kW/ton of computational capacity and a specific power of 500 W/kg. A single 150 kg satellite can thus provide 16 MW of compute power, concurrently supporting 31 subarrays and enabling more than 7,900 large-model inference sessions, each achieving a throughput of 553 tokens per second.

Thermal control engineers lack efficient tools for extracting and visualizing thermal flow data from Thermal Desktop (TD), resulting in inefficient post-processing. To address this, we propose a MATLAB/C++ hybrid GUI system that integrates the OpenTD API with a custom CSR file parser. Leveraging an implicit node–path–submodel ID mapping embedded in CSR files—exploited via a “side-effect” mechanism—the system enables millisecond-level association and loading of thermal flows, temperatures, admittances, and submodel metrics. This approach improves data-matching efficiency by two to three orders of magnitude, substantially reducing post-processing time. The system bridges a critical gap in the TD ecosystem by enabling deep, interactive visualization of thermal flow metrics. Moreover, it provides a reproducible technical pathway and empirical foundation for enhancing thermal analysis capabilities in future OpenTD releases.

This work addresses the thermal constraints of in-orbit data centers, where limited radiative heat dissipation restricts the performance of conventional GPUs due to their high heat density and resulting hotspots that necessitate frequency throttling. To overcome this challenge, the authors propose a “radiator-in-the-loop” co-design framework that, for the first time, jointly optimizes radiative cooling capacity with computational architecture energy efficiency. Through thermal simulations and multi-workload evaluations, they demonstrate that compute-in-memory (CIM) architectures exhibit significantly more uniform thermal distribution and higher TOPS/W efficiency compared to GPUs. Experimental results under realistic orbital thermal constraints show that CIM consistently outperforms GPUs across varying thermal budgets, thereby validating its feasibility and superiority as an AI accelerator for space applications.

This study investigates the scattering and sputtering mechanisms of solar wind ions interacting with lunar regolith, with a particular focus on negative ion emission. A physically motivated semi-analytical model, applicable to particles of arbitrary charge states impinging on homogeneous multi-component surfaces, is developed and validated against in situ negative ion measurements from the NILS instrument aboard the Chang’e-6 mission using Bayesian inference for parameter inversion. The results demonstrate excellent agreement between model predictions and observations, yielding the first quantitative estimates of proton scattering (22%) and hydrogen atom sputtering (8%) probabilities. The surface binding energy of lunar regolith is inferred to be 5.5 eV, and 7–20% of sputtered hydrogen atoms are found to be emitted as negative ions, significantly advancing the understanding of ion–surface interactions at the lunar surface.

This study addresses the lack of an efficient, open-source, and physically credible multidisciplinary trade-off tool for lunar micro-rovers under 50 kg during their preliminary design phase. The authors present an open-source analytical framework that integrates terramechanics, mass estimation, power consumption, thermal survivability, and path planning. Capable of evaluating a single mission scenario in just 30 milliseconds, the tool supports NSGA-II-based multi-objective optimization and enables, for the first time, high-speed physics-driven trade-off analyses tailored to micro lunar rovers. Pareto fronts generated across diverse terrain scenarios demonstrate a median absolute error of 13.3% in mass prediction, with real-world designs closely approaching the optimization boundary—validating the method’s robustness. The results also reveal mission-dependent dominant constraints and challenge the conventional assumption that six-wheel configurations are inherently superior.

This work addresses the challenge of constrained data transmission for large-scale AI tasks in space due to limited ground-to-orbit communication bandwidth. To overcome this bottleneck, the study proposes a novel orbital data center architecture that integrates semantic communication with a multi-layer heterogeneous satellite network comprising relay and in-orbit computing nodes. By transmitting semantic information instead of raw data, the approach substantially reduces uplink traffic. A coupled energy–thermal management model is introduced to evaluate system feasibility. The first comprehensive analysis demonstrates that, under gigabit-class ground-to-orbit links, the proposed architecture can support petabyte-scale internal data exchange, thereby establishing a foundation for scalable and energy-efficient orbital AI systems.

This work addresses the longstanding reliance on inefficient and error-prone manual workflows in infrared radiative transfer calculations, which hinders the demand for fast and reliable computation in climate science and remote sensing. The authors propose InfEngine, an autonomous intelligent computing engine that automates task execution, self-validation, and self-optimization through multi-agent collaboration. Scientific correctness is ensured via a solver-evaluator co-debugging mechanism, while workflow optimization is achieved through an evolutionary algorithm guided by a self-discovered fitness function. Integrated with four specialized agents, 270 tools, and a dedicated benchmark suite (InfBench), InfEngine achieves a 92.7% pass rate across 200 tasks and generates code 21 times faster than expert manual implementation. This system pioneers the transformation of reusable, verified code into persistent scientific assets, advancing research workflows toward a human–AI collaborative paradigm.