We develop laser-based manufacturing processes that sense, predict, and correct themselves. Our group combines laser–matter interaction, multi-modal in-situ monitoring, and physics-informed AI to make processes for hard-to-process materials and next-generation semiconductor packages reliable and scalable.
Multi-Material Metal Additive Manufacturing
High-reflectivity metals such as copper absorb little infrared laser energy, which makes them difficult to print and even harder to join with dissimilar alloys. We use a 1 kW blue-laser (445 nm) directed energy deposition (DED) platform to process pure copper and build copper–Inconel functionally graded structures.
Blue-laser DED of copper and copper alloys
Cu–Inconel 718 functionally graded and bimetallic structures
Multi-sensor melt pool monitoring (two-color pyrometry, OES, multispectral, and LWIR imaging)
Real-time closed-loop control of melt pool temperature and size
Advanced packages demand finer pitch, thinner wafers, and tighter thermal budgets than conventional processes can deliver. We develop femtosecond-laser processes and optical thermometry tools that measure what happens inside the package during bonding.
Laser-assisted bonding for fine-pitch microbump and hybrid bonding
Femtosecond laser slicing of SiC wafers and microLED chip transfer
Through-silicon thermoreflectance for in-situ temperature and bond-quality measurement
Thermal conductivity and interfacial thermal conductance characterization (SSTR/FDTR)
Physics-Informed AI for Manufacturing
High-fidelity simulation is too slow for process planning and control, and purely data-driven models fail outside their training range. We build physics-informed neural networks that predict thermal fields across process parameters in near real time.
Parametric PINNs for thermal modeling of metal AM
Adaptive sampling and training strategies for PINNs
Sensor-data fusion and ML for process state prediction