Machine learning-based design of anti-electrostatic integrated semiconductor bridge initiators

Herein, we propose a machine learning (ML)-driven design paradigm for anti-electrostatic integrated semiconductor bridge (ISCB) chips. Electro-explosive performance experiments and electrostatic discharge (ESD) experiments were conducted on SCB chips protected by discrete devices. Based on these experiments, a database containing over 2000 valid trials was established. Subsequently, six machine learning algorithms were systematically evaluated and compared using this database. Finally, based on ML model, monolithic ISCB chips with the optimal parameters were fabricated. The electro-explosive performance and ESD experiments were subsequently conducted on monolithic ISCB chips. The ML modeling results indicate that the random forest model was identified as the optimal predictive model for electrical explosion characteristics and post-ESD resistance change, achieving coefficients of determination ( R 2 ) of 92.45% and 96.02%, and mean absolute errors of 0.4872 μs and 3.98%, respectively. The ML model quantitatively reveals fundamental physical mechanisms: increased bridge area ( A ) prolongs burst duration through strong linear correlation while reducing resistance variation rate, whereas higher breakdown voltage( V BR ) shortens ignition time nonlinearly but compromises ESD safety. Leveraging response surface methodology, we designed two optimized parameter combinations: A = 20,720 μm 2 with V BR = 6.13-12.72 V, and A = 38,000 μm 2 with V BR = 8.34-15.36 V. Experiments of the monolithic ISCB demonstrates that ML-optimized ISCB initiators maintain excellent electro-explosive performance (burst time deviation < 2 μs, burst energy variation rate < 5%, similar plasma cloud volumes) while achieving 95% reduction in peak power dissipation and zero structural damage under 25 kV/500 pF/500 Ω ESD conditions. This work successfully resolves the longstanding contradiction between electro-explosive performance and ESD reliability, providing a robust ML framework for advanced initiator design. This study introduces a novel machine learning framework to co-optimize the design of Integrated Semiconductor Bridge chips. For the first time, predictive models are trained on discrete device test data, establishing quantitative links between key parameters and critical performance metrics. These models directly guide the rational design and fabrication of ISCBs. Final chip validation confirms the reliability of this data-driven approach, significantly accelerating development compared to traditional methods. • Pioneering a machine learning co-design framework for ISCB chips, establishing a new paradigm for AI-optimized electro-explosive devices. • Developing high-precision predictive models based on a database of 2,000+ experiments, achieving over 92% accuracy in forecasting electro-explosive performance . • The ML-model-designed ISCB chip maintains precise ignition (< 2 μs deviation) while cutting ESD peak power by 95% and sustaining no damage under 25 kV ESD.