OpenTM: An open-source, single-GPU, large-scale thermal microstructure design framework
Yuchen Quan, Xiaoya Zhai, Xiao‐Ming Fu · Smart Materials in Manufacturing · 2026
Thermal microstructures are engineered to manipulate heat flow. We present OpenTM, an open-source GPU-based framework for designing periodic 3D high-resolution thermal microstructures via inverse optimization of target conductivity. To ensure numerical stability without incurring large memory overheads, we employ an adaptive volume-fraction strategy within the Optimality Criteria (OC) method. Practical demonstrations at a resolution of 512 × 512 × 512 achieve runtimes under 12 s per iteration on an NVIDIA Tesla P40 GPU, with a peak memory footprint of 10.78 GB. Our open-source, high-performance implementation is publicly available at https://github.com/quanyuchen2000/OPENTM and is easily installed via Anaconda. A Python interface ensures accessibility for non-C/C++ users. This work highlights the potential of OpenTM as a powerful tool for the rapid structural optimization of thermal microstructures, fostering deeper exploration of their exotic heat-management capabilities.