跳转至

文章背景与核心概要

随着半导体工业向更复杂的计算器件和更严格的工艺容差推进，先进过程控制（APC）变得至关重要。本文介绍了一种新颖的框架，能够直接从原始设备日志文件的时间序列数据中推断半导体工艺的底层动力学。

该研究将机台动力学建模为一个结合了确定性与随机性组件的随机动力系统，并利用 Liang-Kleeman 和 Pires 形式主义来估计熵转移率。初步评估表明，在推断出的依赖关系中，7.5% 是已知的，36.0% 是合理的，17.5% 代表了未表征的新关系，而 39.0% 与当前的工艺认知不一致。这些发现在凸显该框架发掘新型因果洞察强大能力的同时，也指明了未来需要进一步改进以减少不一致结果的方向。

Information-Theoretic Causal Modelling of Semiconductor Process Dynamics

Summary

As the semiconductor industry pushes toward increasingly complex compute devices and tighter process tolerances, advanced process control (APC) becomes critical. This paper introduces a novel framework for inferring the underlying dynamics of semiconductor processes directly from raw equipment log-file time-series data. By modeling tool dynamics as a stochastic dynamical system—combining deterministic and stochastic components—the authors estimate entropy transfer rates using the Liang-Kleeman and Pires formalism. Preliminary evaluations reveal that 7.5% of the inferred dependencies were previously known, 36.0% were plausible, 17.5% represented uncharacterized relationships, and 39.0% were inconsistent with current process knowledge. These findings highlight the framework's strong capability to uncover novel causal insights while pointing to areas for future refinement to reduce inconsistencies.

As the semiconductor industry pushes toward increasingly complex compute devices and tighter process tolerances, advanced process control (APC) becomes critical. This paper introduces a novel framework for inferring the underlying dynamics of semiconductor processes directly from raw equipment log-file time-series data. By modeling tool dynamics as a stochastic dynamical system—combining deterministic and stochastic components—the authors estimate entropy transfer rates using the Liang-Kleeman and Pires formalism. Preliminary evaluations reveal that 7.5% of the inferred dependencies were previously known, 36.0% were plausible, 17.5% represented uncharacterized relationships, and 39.0% were inconsistent with current process knowledge. These findings highlight the framework's strong capability to uncover novel causal insights while pointing to areas for future refinement to reduce inconsistencies.

Document Details

  • arXiv Identifier: arXiv:2608.14678
  • Primary Subject: Signal Processing (eess.SP)
  • Secondary Subjects: Artificial Intelligence (cs.AI), Information Theory (cs.IT)
  • Submission Date: August 4, 2026
  • Conference Context: To be presented at the 2026 IEEE 33rd International Conference on Electronics, Circuits and Systems (ICECS) and published in the IEEE conference proceedings.
  • arXiv Identifier: arXiv:2608.14678
  • Primary Subject: Signal Processing (eess.SP)
  • Secondary Subjects: Artificial Intelligence (cs.AI), Information Theory (cs.IT)
  • Submission Date: August 4, 2026
  • Conference Context: To be presented at the 2026 IEEE 33rd International Conference on Electronics, Circuits and Systems (ICECS) and published in the IEEE conference proceedings.

Authors

  • Daniel Sørensen
  • Giorgio Melchiorre
  • Sudip Bandyopadhyay
  • Sandip Halder
  • Roel Wuyts
  • Bappaditya Dey
  • Daniel Sørensen
  • Giorgio Melchiorre
  • Sudip Bandyopadhyay
  • Sandip Halder
  • Roel Wuyts
  • Bappaditya Dey

Abstract

With the progress of the semiconductor industry toward increasingly complex compute devices and tighter process tolerances, advanced process control has become crucial. This work explores a novel framework to infer the underlying dynamics of semiconductor processes, directly from raw equipment log-file time-series data. By modelling the tool dynamics as a stochastic dynamical system comprising (a) a deterministic component and (b) a stochastic component, we estimate entropy transfer rates between variables through the Liang-Kleeman and Pires formalism. Preliminary results indicated that 7.5% of the inferred dependencies were known, 36.0% were plausible, 17.5% represented previously uncharacterised relationships, and 39.0% were inconsistent with established process knowledge. These findings demonstrate the framework's capability to uncover novel causal insights, while motivating further improvements to reduce inconsistent findings.

With the progress of the semiconductor industry toward increasingly complex compute devices and tighter process tolerances, advanced process control has become crucial. This work explores a novel framework to infer the underlying dynamics of semiconductor processes, directly from raw equipment log-file time-series data. By modelling the tool dynamics as a stochastic dynamical system comprising (a) a deterministic component and (b) a stochastic component, we estimate entropy transfer rates between variables through the Liang-Kleeman and Pires formalism. Preliminary results indicated that 7.5% of the inferred dependencies were known, 36.0% were plausible, 17.5% represented previously uncharacterised relationships, and 39.0% were inconsistent with established process knowledge. These findings demonstrate the framework's capability to uncover novel causal insights, while motivating further improvements to reduce inconsistent findings.

Full-Text & Access Links

References & External Tools