鎳鍺化物與N型鍺接面摻雜析離對於蕭基位障的影響:透過第一原理計算

Abstract

近年來隨著傳統的矽金氧半場效電晶體技術已經逐漸到達其微縮的極限。為了能夠持續追求元件性能的提升，鍺因為它的載子擁有優越的本質遷移率，被認為在不久的將來可以取代矽作為通道的材料而受到廣泛的研究。然而一般的Ｎ型摻雜物例如磷與砷在鍺基板不止固態溶解度比在矽基板還低而且具有更快的擴散速率，所以很難製作出深度很淺的接面。更糟的是，接面的費米能階被釘扎在很靠近鍺價帶的頂端，造成相當高的電子蕭基位障。實驗上發現透過摻雜析離技術可以使得摻雜物在界面析離進而達到更淺的接面深度與更濃的摻雜濃度。但是，摻雜物對於改變界面特性所扮演的角色仍然不是很清楚。 在本篇論文中，我們只用單一的鎳鍺化物（１１２）晶相去簡化真實的多晶鎳鍺化物並與鍺形成接面結構，然後透過第一原理計算使用LDA交換相關能泛函研究Ｎ型摻雜物在界面的行為以及使用HSE06泛函探討摻雜物析離在界面是否可以有效降低鎳鍺化物與鍺接面的物理蕭基位障。 對於一般的Ｎ型摻雜物例如磷與砷而言，我們的計算結果顯示這兩種摻雜物都會在界面析離，雖然磷較偏好析離在界面靠近鎳鍺化物的那一側而砷則是偏好在鍺那一端。這些結果建議我們假若要摻雜磷，在形成鎳鍺化物先植入磷會是個較好的製程選擇；對於摻雜砷，不管製程是在形成鎳鍺化物之前或是之後植入砷，砷都會移到鍺那一端並且堆積在界面。 然後，我們指出這兩種摻雜物析離在界面靠近鍺那一端造成鎳鍺化物與鍺接面的物理蕭基位障降低的量少於0.1eV，而這麼小的降幅是因為很強的費米能階釘扎效應所造成。順帶一提，當摻雜磷在界面靠近鎳鍺化物的那一側不具有任何降低物理蕭基位障的效果。總結以上的結果，我們認為摻雜析離造成鎳鍺化物與鍺接面特性的改善主要歸功於增加在界面附近的摻雜濃度，少部分是由於物理蕭基位障降低的貢獻。 最後我們探討當摻雜氮這種特殊的情況，計算的結果指出氮可以析離在界面，然而也產生大量的界面態散佈在整個鍺的能隙。雖然鍺的等效導帶邊界透過連續分布的界面態延伸而更接近費米能階，然而界面態在遠離界面約17埃即消失殆盡，所以鎳鍺化物與鍺接面的物理蕭基位障還是幾乎沒有任何改變。 以上這些第一原理的計算結果給予我們更深入的了解對於摻雜物在鎳鍺化物與鍺接面所扮演的角色，也與實驗所觀察到的結果相互呼應。將來，更進一步的計算希望可以協助新製程的開發。

Recently, traditional Si-based MOSFETs are approaching its fundamental scaling limits, and then Ge has been comprehensively explored as a potential channel material to replace Si due to its high intrinsic carrier mobility for further performance enhancement. Nevertheless, the shallow junction depth is hard to form since the conventional n-type dopants such as phosphorous and arsenic have not only lower solid solubility but also faster diffusion rate in Ge substrate than in Si. Moreover, strong Fermi-level pinning near the valence band edge of Ge leads to high electron Schottky barrier height. Dopant segregation technique has been proposed to achieve shallower junction depth and heavier dopant concentration experimentally due to dopant segregated around the interface. However, the role of dopants at the NiGe/Ge interface is not clear. In this thesis, we build the realistic polycrystalline phases NiGe/Ge contact by including NiGe (112) phase only, and then the first-principles calculations are employed to investigate the behaviors of the n-type dopant around the interface by LDA functional and whether the physical Schottky barrier height of the NiGe/Ge contact is reduced by dopant segregation or not is calculated by HSE06 hybrid functional. For the conventional n-type dopant such as phosphorous and arsenic, our calculations show that those two elements may be segregated at the interface, but the preferred segregated site of phosphorous and arsenic are on the NiGe and Ge side, respectively. These results suggest that phosphorous would be a better choice for implantation before germanide process, while arsenic can migrate into the Ge layer and pile up at the interface in both implantation before and after germanide processes. Then, we show that the physical Schottky barrier height of the NiGe/Ge contact modified by dopant segregation using those two elements on the Ge side is reduced by less than 0.1 eV. This small value is due to the strong Fermi-level pinning effect. By the way, there is no effect to modify the physical Schottky barrier height by doping phosphorous on the NiGe side. To sum up, the improvement of the NiGe/n-type Ge junction characteristics by dopant segregation using phosphorous and arsenic are mainly attributed to the increase of dopant concentration around the interface and partially attributed to the reduction of the physical Schottky barrier height. For the specific case using nitrogen dopant, the calculated results show that it can be segregated around the interface but yield a large number interface states spreading the Ge bandgap. Although the effective conduction band edge is closer to the Femi level due to continuous interface states, the physical Schottky barrier is almost unchanged since the interface states disappear at about 17 Å away from the interface. These first-principles calculations provide deep insight on the role of dopants near the NiGe/Ge interface and can explain the experimental observations very well. Further calculations can also help new process development.