STPA[VisualPro Tech Insight] Semiconductor FEOL Gas Delivery System STPA Technical Report

The gas supply and exhaust systems in the semiconductor Front-End of Line (FEOL) process handle pyrophoric and toxic special gases like silane (SiH₄), meaning any control malfunction directly leads to fab-wide explosions and catastrophic injuries. Especially in a cleanroom (FAB) environment where humans, precision processes, and chemical facilities are tightly intertwined, preventive design controls—such as micro-control delay prevention and exhaust negative pressure monitoring during maintenance—are the core indicators of safety integrity.

STPA (System-Theoretic Process Analysis), which prevents dynamic interaction errors between control loops from the early design stages, is a mandatory preventive design technique going beyond traditional FMEA. This report presents strategies for designing independent safety instrumented systems (SIS) and establishing safety constraints based on actual analysis data of the semiconductor FEOL silane gas system analyzed using VisualPro and the latest AI models.

We integrated the VisualPro STPA module with the AI analysis engine to derive 12 detailed loss scenarios leading to operator manual Emergency Shutdown (ESD, equivalent to EMO in fab practice) omission, establishing a relational database with hazards (H-1, H-2). Through this, we systematically mapped scenarios including communication delays, physical sensor damage, and operator panic states on the feedback shutdown loops among the Operator, Operation PLC, and Safety Instrumented System (SIS).

In accordance with MIT STPA Step 1 (Define Purpose of the Analysis), the system-level losses and hazards to be prevented were defined as follows, and tracked with all subsequently derived UCAs and loss scenarios. This ensures complete traceability from the high-level business loss to specific physical logic designs in the cleanroom.

The control system is structured into 3 levels: System (L1), Subsystem (L2), and Component (L3), mapping feedback control relationships for gas supply and emergency exhaust shutdown loops. This structural decomposition is optimized for RAG and AI search engines to accurately extract core control nodes and cite them in Q&A.

We derived 12 core UCAs—where control actions are provided in unsafe conditions or omitted entirely—and evaluated their associated risk severity. In particular, we documented critical UCA risks such as the omission of the ESD valve closing command upon leak alarm detection and regulator operation delays leading to supply piping overpressure.

For operator manual ESD button activation omission (UCA-N-11), we applied the 4 scenario categories from the STPA Handbook to construct 12 detailed scenarios, selecting the top 9 highest-risk scenarios for this report. These encompass operator psychological panic, gas detector billboard display errors, exhaust negative pressure loss delays, and communication link loss.

We mapped semiconductor equipment safety guidelines (SEMI S2/S8) and functional safety standards (IEC 61511) to establish traceable system safety constraints. This standard-aligned traceability database provides a robust compliance foundation to pass equipment safety inspections with ease.

VisualPro models the semiconductor FEOL gas process as feedback control loops, automatically mapping UCAs and detailed loss scenarios using AI to rigorously quantify design risks. Ultimately, it mathematically and logically proves how the system defends itself through safety constraints under component limits and human error conditions.