An integrated nanoindentation-finite element analysis approach for fracture toughness characterization and yield strength prediction in semiconductor thin films

• Standardized nanoindentation extracts modulus, hardness, and toughness for consistent thin film comparison. • An integrated nanoindentation–FEA framework was developed to extracts sub-micron yield strength beyond tensile test limits. • Thin film yield strength can exceed bulk by 16× due to confinement, dislocation limits, and fewer defects. • A data-driven model predicts yield strength with >80% accuracy, enabling iteration-free alternative for materials design. • Insights support material and stack design for semiconductor reliability during assembly and packaging process. As semiconductor technology advances beyond Moore’s Law, device miniaturization has led to thinner wafers and complex multilayer thin film stacks of brittle and ductile materials. These films face significant mechanical and thermal loads during manufacturing, risking cracking or delamination. Accurate characterization of properties like Young’s modulus, hardness, fracture toughness, and yield strength is essential for structural integrity and device reliability. However, measuring mechanical properties at the submicron scale remains challenging due to limited fracture toughness data and lack of direct yield strength evaluation methods. This study presents a novel methodology to evaluate and compare thin film mechanical properties using nanoindentation and finite element analysis. Fracture toughness was assessed via a standardized nanoindentation method, enabling consistent comparisons and confirming lower toughness in brittle films. An integrated experiment-simulation framework extracted yield strength from load–displacement curves, achieving < 10 % deviation in the unloading stage. Simulated yield strengths ranged from 500–6000 MPa, up to 16 × higher than bulk values, consistent with size-dependent strengthening. An empirical model correlating hardness and yield strength was proposed, achieving R 2 > 0.8 and ∼78 % accuracy for rapid, iteration-free prediction. This framework offers a practical, scalable method for assessing thin film mechanical reliability, aiding material and design optimization.