A MIMO Decoupling Algorithm for Rapid Heating Systems of Wafers

In the semiconductor industry supply chain, there are complex processing steps from wafer to chip. Rapid Thermal Processing (RTP) is one of the key technologies, using lasers or high-temperature lamps to rapidly heat the wafer to over 1000°C for several seconds. This technique can produce high-quality silicides and oxides on the wafer while reducing the thermal budget in terms of temperature and time. RTP systems must be able to change the spatial energy distribution radiated onto the wafer to maintain temperature uniformity under different processing conditions, including varying gases, pressures, and wafer structures. However, the non-equilibrium conditions in RTP make modeling and prediction challenging, with temperature coupling issues in different wafer regions. This paper models the wafer heating system as a 5th-order time-delayed first-order inertia matrix and proposes a corresponding feedforward decoupling method. Traditional incremental PID control, when ramping up temperatures above 1000°C, leads to prolonged oscillations near the target temperature due to temperature coupling between regions, resulting in slow temperature convergence. The method proposed in this paper can effectively eliminate inter-regional coupling effects, unify PID parameters for better determination, and deploy them in practical RTP systems. Experimental results demonstrate that temperature convergence can be achieved rapidly, enhancing system heating efficiency. This study provides an effective solution to temperature control issues in wafer processing, with important practical implications for the semiconductor equipment manufacturing industry.