Cl2/Ar電漿廣泛應用於各類半導體之電漿蝕刻製程,然而目前之模型研究均針對矽基材之電漿製程技術,至今除了電漿模型外,仍未有化合物半導體之蝕刻模型出現。本研究建立Cl2 /Ar電漿蝕刻GaN化合物之數學模型,其中包含Cl2 /Ar電漿模型、GaN蝕刻速率模型和深寬比效應模型,討論在不同操作參數下,電漿中各物種濃度及GaN材料蝕刻速率之變化。 藉由與實驗量測值之比較,建立零維電漿模型,並探討Cl2 /Ar電漿之主要反應機制,再利用二維電漿模型求解,探討操作參數對電漿均勻性之影響。研究結果顯示,操作壓力是主要影響電漿均勻性之主因,操作壓力高時,電漿均勻性較差。 利用電漿模型求得之離子及自由基濃度,代入GaN蝕刻速率模型中,配合實驗量測GaN之蝕刻速率值,迴歸求得Cl原子於GaN材料表面之黏滯係數(r=0.3)及離子之濺擊產率(Yi=0.25 (Vbias0.5-Vth0.5)),其中Vbias為基板偏壓,Eth為材料之特性係數。經由蝕刻速率模型計算,亦可獲得不同操作參數下,GaN材料表面蝕刻先驅物Clx-GaN(s) (x=1~3) 之組成。 結合上述兩模型求得之Cl2 /Ar電漿蝕刻GaN相關資訊,建立深寬比效應模型,並探討不同操作參數下,孔徑大小相異之引洞蝕刻之深寬比效應,研究顯示在高吸附/蝕刻速率比(Rn0/Ri)環境下,可有效拉進大小引洞之蝕刻速率差,並減低深寬比效應。
Cl2/Ar plasmas are widely used in the dry etching of semiconductors. However, current modeling studies still focus on plasma processing of silicon-based materials. Very few modeling studies on compound semiconductors exist. In order to investigate the plasma etching of compound semiconductors using Cl2/Ar chemistry, we developed an integrated model, which includes Cl2/Ar plasmas model, GaN etch rate model, and aspect ratio dependent etch (ARDE) model. First, a zero dimensional (0D) model was developed to discuss the Cl2/Ar reaction mechanism by comparing with experimental data from literature. A two dimensional (2D) was also developed to evaluate the influences of various operating parameters on species density uniformity across the reactor. It showed that plasma uniformity was mainly determined by gas pressure. Plasma uniformity is better as the pressure gets lower. Then, an etch-rate model was developed to study the interaction between plasma species and the surface of compound semiconductor. By comparing with experimental data from literature, we found that the sticking coefficient of Cl radical is about 0.3 and the sputtering yield of ionic species is about 0.25 (Vbias0.5-Vth0.5), where Vbias and Vth represent self-bias voltage and threshold voltage. Etch rate constants Ri and adsorption rate constants Rn0 can also be derived in this model under various operating conditions. Finally, Ri and Rn0 calculated from etch rate model were put into ARDE model. It was found that ARDE effect can be predicted by Rn0/Ri ratio under various operating conditions. Higher Rn0/Ri ratio resulted in less ARDE effect.