光纖雷射輔助碳化矽晶圓加速研磨製程之表面改質研究

Abstract

隨著電動車與5G通訊需求增加，傳統矽材料在高頻高壓環境下已顯不足。碳化矽(Silicon carbide, SiC)具備寬能隙、高導熱與優異的熱穩定性，是第三代半導體的關鍵材料之一。然而，因其具備高硬度特性而難以加工，進而導致製程成本上升。為改善此問題，本研究採用奈秒級紅外光光纖雷射進行表面預處理，透過熱誘導氧化反應於表面形成軟化層，以降低材料表面硬度並提升後續加工性。相較於飛秒或皮秒雷射，所採用之光纖雷射成本低廉、維護簡便，其設備成本約低100至1000倍，大幅提升產業應用之可行性。研究中以有限元素分析模擬雷射移動熱源之熱場變化，並建構機器學習模型預測不同參數對移除深度與粗糙度之影響。結果顯示當雷射速度為1000 mm/s、重疊率為44%時，表面硬度由40.17 GPa 降至0.77 GPa，提升52.25%研磨效率，有效提升加工效率並降低耗材消耗。

With the growing demand for electric vehicles and 5G communications, conventional silicon materials are increasingly inadequate under high-frequency and high-voltage conditions. Silicon carbide (SiC), with its wide bandgap, high thermal conductivity, and excellent thermal stability, is a pivotal third-generation semiconductor material. Nevertheless, its high hardness imposes substantial challenges on wafer-level fabrication. The present study applies a nanosecond infrared fiber laser for surface pretreatment.Thermal oxidation is induced to form a softened surface layer, reducing hardness and improving subsequent machinability. Compared to femtosecond or picosecond lasers, fiber lasers offer significantly lower equipment and maintenance costs, approximately 100 to 1000 times cheaper, enhancing industrial feasibility. Finite element analysis was performed to simulate the transient thermal fields of the laser moving heat source, while machine learning models were developed to predict the influence of processing parameters on material removal depth and surface roughness. When processed under optimized conditions, specifically at a scanning speed of 1000 mm/s and a 44% overlap, the surface hardness was reduced from 40.17 GPa to 0.77 GPa, improving grinding efficiency by 52.25%. The proposed method effectively enhances processing performance and reduces consumable wear, offering a promising solution for precision machining of hard materials.