Adaptive Self-Tuning of Process Parameters for High-Voltage Power Supplies in Etching Equipment

In semiconductor manufacturing, the precision of plasma etching directly determines the quality of microstructural formation in chips. As the energy source for plasma, the real-time self-tuning capability of process parameters in high-voltage power supplies has become a core technology for breaking through the bottlenecks of nanoscale pattern transfer. Traditional etching equipment relies on manual parameter adjustments, which struggle to address process fluctuations and new material requirements. In contrast, intelligent algorithm-based self-tuning technologies are driving etching processes toward higher stability and adaptability. 
1. Key Challenges in Parameter Tuning
Plasma etching requires coordinated control of two critical parameters: Ion Energy (Ei) and Ion Flux (Γi): 
• Ion Energy determines etch directionality. For example, high-energy ions (>500 eV) enable vertical etching for high-aspect-ratio structures but risk mask carbonization damage, while low-energy ions (<100 eV) favor isotropic etching. 
• Ion Flux affects etch rates. Increased excitation power boosts plasma density, but excessive power (e.g., >600 W) reduces efficiency due to intensified collisions. 
Dual-frequency power architectures (e.g., 60 MHz high-frequency + 2 MHz low-frequency) decouple Ei and Γi control through independent frequency and power adjustments. However, coupling effects from gas chemistry, pressure (0.1–2.0 Pa), and wafer temperature make manual tuning inadequate for full-wafer uniformity. 
2. Advances in Intelligent Self-Tuning Technologies
2.1 Dynamic Impedance Matching & Closed-Loop Control 
High-voltage power supplies must respond to plasma impedance changes within microseconds. High-compute controllers (e.g., 300 MHz dual-core MCUs) track impedance drift in real time, automatically tuning frequencies to suppress power reflection and limit forward power fluctuations to ±0.5%. For instance, in diamond etching, dynamic matching networks compensate for impedance mismatch caused by gas ionization fluctuations, minimizing energy loss. 
2.2 Pulse Modulation & Customized Waveforms 
High-frequency pulsing (e.g., nanosecond-level pulse-width modulation) compresses ion energy distribution bandwidth to <5 eV full-width-at-half-maximum (FWHM) by adjusting duty cycles, reducing sidewall erosion from low-energy ions. Customized waveforms (e.g., square waves superimposed with RF) optimize electron collision paths, enhancing the generation efficiency of specific radicals (e.g., F radicals) and suppressing byproducts. 
2.3 Wide-Bandgap Semiconductors for Efficiency 
Third-generation semiconductors like silicon carbide (SiC) significantly improve dynamic response. SiC MOSFETs reduce switching losses by 70% compared to silicon-based devices and tolerate temperatures >200°C, ensuring thermal stability for 11 kW-level power supplies under pulsed operation—effectively curbing plasma density drift. 
3. Sheath Control in High-Aspect-Ratio Etching
As etch structures exceed 100:1 aspect ratios, nonlinear sheath oscillations in narrow trenches cause ion incident angle divergence. Solutions include: 
• Focus Ring Thermal Compensation: Real-time monitoring of focus ring temperature changes (e.g., time-variant curves from 90°C to 20°C) dynamically adjusts bias voltage to offset sheath height differences caused by ring consumption. 
• 3D Power Topology: Asymmetric field control algorithms for 3D structures (e.g., gate-all-around transistors) enable differentiated etching at trench bottoms and sidewalls. 
4. AI-Driven Parameter Optimization
Deep learning models trained on real-time plasma emission spectra predict optimal power-frequency combinations: 
• Data Input Layer: Voltage, current, and spectral intensity from process sensors. 
• Decision Layer: Long short-term memory (LSTM) networks model parameter couplings to output optimized Ei/Γi solutions. 
Experiments show this approach improves edge-to-center etch uniformity by >30% and reduces recipe-switching time by 40%. 
5. Future Directions
The essence of self-tuning lies in precise control of energy conversion from electricity to chemistry to kinetics. Future trends focus on: 
1. Multi-Parameter Synergy: Integrating gas chemistry models with power response algorithms for cross-dimensional optimization of pressure, temperature, and power. 
2. Quantum Computing-Assisted Design: Quantum annealing algorithms solve high-frequency parameter combinations, overcoming local optima limitations of conventional methods. 
Conclusion
Self-tuning technology for high-voltage power supplies in etching equipment is evolving from experience-driven to data-and-intelligence-driven. Through the deep integration of wide-bandgap devices, dynamic impedance matching, and artificial intelligence, high-voltage power supplies establish controlled energy transfer boundaries at the atomic scale. This advancement will ultimately propel semiconductor manufacturing toward intelligent self-aware, self-deciding, self-optimizing fabs, unlocking the industrial potential of sub-3nm processes.