矽基板穿孔之氮化鎵高電子遷移率電晶體之高頻特性分析及其小訊號模型

Abstract

氮化鋁鎵/氮化鎵高電子遷移率電晶體在高頻應用上很有前景，而矽基板則是在商業化下最吸引人的選擇。但是在運用在高頻及高功率放大器時，低阻值的矽基板會造成寄生效應而降低其高頻表現。而矽基板移除是個被期待的辦法去跨越這個寄生效應衍生的瓶頸並保有低阻值矽基板的優點。在這個論文中，我們探討氮化鋁鎵/氮化鎵高電子遷移率電晶體在完全及局部移除矽基板之後的特性，並建立相對應的小訊號模型。 在第二章，我們探討氮化鋁鎵/氮化鎵高電子遷移率電晶體完全移除源極和汲極間矽基板之後的特性。元件在移除基板後在低頻下有較差的順向增益電壓，但在高頻下卻有較好的順向增益電壓，因為低頻時挖除基板的電晶體受到較嚴重的熱效應導致增益降低，但在高頻時因為移除矽基板降低漏流效果高於熱效應的影響，而可取得較好的順向增益電壓。我們觀察到挖除基板的電晶體有較高的電流增益截止頻率和功率增益截止頻率，雜訊的表現也被提升。 在先前的探討中，我們得知矽基板移除能有效的改善元件的高頻特性，在第三章，我們再更深入的研究在不同的操作偏壓下，選擇性移除矽基板時的高頻特性。矽基板背後製程的前後之間，直流特性並沒有太大的改變。磨薄及不同的矽穿孔對電流增益截止頻率並無太大影響。在矽基板背後製程前，功率增益截止頻率相差很大因為受非線性缺陷影響，而在汲極偏壓提升時，其變化也會被非線性缺陷屏蔽住。而在矽基板背後製程後，非線性缺陷被抑制，且寄生效應減少越多功率增益截止頻率有越大的提升。選擇性移除矽基板在汲極偏壓提升時，其寄生效應會跟著提升，導致其回到矽基板背後製程的前的狀態，而完全移除寄生效應相關部件則不會有隨著汲極偏壓提升而提升的寄生效應。最後包括寄生效應相關部件的小訊號模型被建立且模擬出上述探討的理論機制。結果再度顯示矽基板的移除能改善高頻特性，並且在不同的偏壓下，完全移除寄生效應相關部件有較穩定的高頻特性提升。

AlGaN/GaN HEMTs are very promising devices for high frequency applications and Si substrate is the most attractive choice for commercialization. For high-frequency and high-power amplifiers, the parasitic loss by the conductive Si substrate can degrade the performance of the devices. Si substrate removal is a promising approach to overcome the obstacles and sustain the advantages of the low-resistivity(LR) Si. In this thesis, we fabricated AlGaN/GaN/Si HEMTs with full and selective Si substrate removal. DC electrical properties and RF characterizations are investigated. Small signal models with parasitic elements are also developed. In chapter 2, the full Si substrate removal between source and drain pad on microwave performance is investigated. Our results show a lower transconductance, gm, due to heat effect which degrades forward voltage gain, S21, at low frequency. Nevertheless, the 2DEG in the channel couples with conductive substrate at high frequency which has larger impact on S21 than heat effect and results in a larger S21 at high frequency after Si removal. The current gain cut-off frequency, fT, and power gain cut-off frequency, fmax, are also improved since higher power efficiency being achieved. Moreover, substrate leakage current has a longer response time comparing to channel current and plays a role as a noise current. The measurement results show the noise figure can be improved after Si removal. From earlier discussion, we conclude that Si removal is an effective method to increase microwave performance of HEMT on LR Si. In chapter 3, the impact of selective substrate removal at different bias conditions on microwave performance is studied. The DC electrical properties remain consistent after backside process. Thinning down the low-resistivity Si substrate and the removal of the Si substrate underneath the mesa region have little impact on fT. Before backside process, fmax differs a lot because of the non-linear defects. The tendency of fmax is also screened by these defects as increasing VD. After backside process, the non-linear effect is diminished and the more the decrease of the coupling effect by the backside process, the higher the fmax will be. Without full removal of the Si substrate under all the mesa area, fmax before and after backside process shrink with a higher VD due to increasing coupling effect. Nevertheless, fmax of the device with full removal of the parasitic elements remain consistent with increasing VD. Finally, a small signal circuit model with parasitic elements is employed to elucidate the idea. The results can conclude that Si removal can improve RF performance and full removal of the parasitic elements can have more stable improvement at different bias condition.