鉑鍺化物與N型鍺接面對於蕭特基位障的影響−藉由結合實驗與第一原理計算

Abstract

本論文結合第一原理模擬與實驗對「鉑鍺合金/鍺」蕭特基接面進行研究。我們首先在實驗上探討不同退火溫度成長的鉑鍺合金對於電子蕭特基能障的影響。由電性量測結果，發現電子蕭特基能障會隨著退火溫度增加先行微幅下降然後趨於穩定，經由低掠角X-射線繞射檢測，發現鉑鍺合金隨退火溫度增加其組成與晶面方向亦先隨之變化而後趨於穩定。此外，為了確認接面原子尺度結構並縮小後續理論計算探索分析的範疇，我們使用X-射線光電子能譜檢測，排除了在退火過程中，接面周邊材料擴散進來的可能性。其次，我們從實驗樣本的低掠角X-射線繞射檢測訊號中，確認鉑鍺合金元素組成比例與晶面，並在模擬計算上初步考慮數種訊號強度較強的晶面。經由考量每一種晶面匹配所需的計算資源後，我們最終選定並建構出PtGe(121)/Ge、Pt2Ge3(800)/Ge和PtGe2(120)/Ge三種適合計算的接面結構，來對應實驗上各自退火溫度下的接面組成。透過第一原理計算，我們首先研究這三種具有代表性的單晶面「鉑鍺合金/鍺」接面對電子蕭特基能障的影響，發現其等效電子蕭特基能障對不同接面組成結構的變化趨勢與實驗吻合。而且這些接面組成結構之間的能障差異量少於 0.1 eV，此趨勢與實驗不謀而合，其原因可能來自於很強的費米能階釘扎效應。最後，我們探討數種摻雜元素析離於界面時，其各自對降低電子蕭特基能障的效果。計算結果顯示其堆積於界面頗為穩定，而典型Ｎ型摻雜元素磷與砷皆有助於降低其能障高度。此外，我們發現在所研究的摻雜原子種類中，以鈧離子摻雜將等效能障高度從摻雜前的0.671 eV降低至0.388 eV，調整幅度最大。未來可將其摻雜應用至實際製程以檢驗此特性的合理性。綜合以上結果，第一原理模擬計算可幫助我們初步探究不同接面合金晶面與摻雜物對等效電子蕭特基能障高度的改善效果。我們預期這樣的理論計算能夠為實際製程提供明確的改善方向，以大幅縮短製程開發的時程與成本。

In this thesis, we combine the first-principles simulations and experiments to study the Pt-germanide/Ge Schottky junction. In experiments, we first investigate how various PDA treatments affect the SBHs of a Pt-germanide/Ge junction. The I-V measurements show that SBHs are slightly reduced and gradually saturated as the annealing temperature increases. As being inspected by GIXRD analysis, the Pt-germanide phase and crystalline plane is found to vary slightly and gradually saturated as the annealing temperature increases. In order to realize the Pt-germanide/Ge interface structure in atomic scale and also to reduce possible computational trials, we use the XPS analysis to eliminate the possibility that the materials surrounding the junctions diffuse in during annealing. Next, we figure out the Pt-germanide compositions and crystalline planes from the GIXRD analysis, and do an initial screening to keep only several crystalline planes of the stronger peak signals for simulation modeling. After further considering the lattice mismatches and their desire computational resources, we finally choose and construct the three interfaces: PtGe(121)/Ge, PtGe(800)/Ge, and PtGe(120)/Ge, corresponding to the junctions formation under different annealing temperatures. By the first-principles calculations, we first investigate how SBH changes with these three single-crystalline PtmGen/Ge interfaces. We find that their SBHs depend on different Pt-germanide phases in a way in agreement with the experiments. In addition, their values differ by less than 0.1eV, consistent with the experiments. This can result from the strong Fermi level pinning. Finally, we compare the SBH-lowering effects under segregation of various dopants. Calculations reveal that they tend to segregate to the interface, and the conventional N-type dopants P and As can both reduce SBH. In addition, we find that the Sc dopant can reduce SBH from 0.671 to 0.388eV. It is the most effective among these dopants, which can be testified in the further experiments. In summary, first-principles calculations can help us explore SBHs of different crystalline planes and dopants, and pursue its reduction for device purposes. We expect that such calculations can provide a better direction for realistic fabrication, and can further shorten process development cycles and lower the cost.