鍺/鍺錫光激發光及其光電元件之研究

Photoluminescence Characterization and Photonic Applications of Ge/GeSn

林宗毅(Chung-Yi Lin)

指導教授 ： 劉致為

國立臺灣大學/電機資訊學院/光電工程學研究所/博士(2020年)

https://doi.org/10.6342/NTU202001712

摘要

由於鍺為間接能隙材料故其發光的效率及其吸收光的效率都有加強的空間，而我們發現摻入7%-10%錫可以使其Gamma valley下降甚至低於Gamma valley使其變成直接能隙材料，目前鍺錫材料已被用來利用光來產生雷射也用來製作電晶體因為其電洞之遷移率高於鍺，但是其材料特性、發光元件、以及光偵測器皆需要更進一步的研究，尤其是光偵測器的部分不管其吸收材料是矽、鍺、甚至鍺錫皆需要進一步的研究，因為目前在物聯網、自駕車或是在光聯結上在接受端的發展皆是非常需要的。
表面鈍化是可用來提升鍺錫材料之光電元件後的效率，表面鈍化可以利用兩個方式達成一個是降低表面斷鍵產生的缺陷但是這種方式需要高溫來產生，而在鍺錫材料上熱預算約為400度所以其降低表面斷鍵的效果並不好，第二種達到表面鈍化的方式是利用帶電的氧化物將表面載子與缺陷複合的機率降低，而這種方式也應用在高效率的太陽電池當中，在論文的第二章原子層堆積被利用來成長了二氧化矽以及氧化鋁在矽鍺表面利用其帶負電的特性來鈍化表面，金氧半的電容也被製作來萃取其帶電量，而光激發光的頻譜強度被用來判斷其鈍化的效果，當光打入鍺錫材料中會產生電子電洞對如果鈍化成功的話表面的載子就不會產生非放光式的複合，而在製程的優化下20-cycle的二氧化矽搭配上60-cycle的氧化鋁能產生最好的鈍化效果。
再來由於鍺錫材料在成長時需成長在鍺的虛擬基板上面因為錫跟矽的晶格大小差距約為20%，而錫跟鍺也具有14%的晶格差距所以在成長上會產生應力，而應力會影響載子的分佈也會影響其能隙之大小，在論文的第三章我們利用治具來施加額外的應力來了解經過應力以後載子的分佈及能隙之變化，利用經驗公式以及固態理論來fitting光激發光的頻譜可以得知鍺錫材料的能隙，再來我們發現經過外加拉伸應力鍺的直接能隙發光是十分敏感的但是鍺錫材料的直接能隙發光卻是非常不敏感的。
由於發光元件在目前的四族光電元件當中是十分缺乏的所以在第四章我們利用金氧半結構來製作電激發光的元件，目前的發光元件皆需要額外的摻雜利如p-n結構或是p-i-n的結構而摻雜皆有可能產生額外的缺陷，利用金氧半結構也可以偵測到真正的薄膜品質，我們利用原子層堆積極薄的3nm氧化鋁來達成金氧半的結構，由於為了不讓錫產生析出鍺錫材料的表面皆覆蓋了一層鍺的覆蓋層以提升磊晶後薄膜之品質，但是這層錫會影響載子的分佈，我們發現當有覆蓋鍺以後電激發光的頻譜會以鍺的直接能隙發光為主，那我們利用熱氧法去除覆蓋錫的以後就看到了以鍺錫的直接能隙發光為主。
研究完發光元件以後我們就將研究光接收端，目前窄頻的光偵測器尤其是紫外光的波段是非常重要的，在第五章我們也一樣利用金氧半結構來製作光偵測器並且利用上面的金屬來當作濾波器，那之前的研究發現銀可以當作紫外光的濾光片所以我們的金屬電極就利用銀再來我們利用原子層堆積成長了不同的氧化物在矽的表面，我們發現氧化鋁可以有效的鈍化表面所以金氧半結構的氧化物就利用氧化鋁，製作出來其窄頻的光偵測器可以達到0.174mW/A的光響應然後利用印刷電路板搭配上電路設計我們製作出了符合物聯網之窄頻光偵測器。但是光偵測器最重要的還是在光連結上面，在第六章我們製作了不同厚度的p型鍺虛擬基板的垂直式光偵測器並量測其直流跟交流的特性，我們發現p型鍺虛擬基板其直流跟交流式越厚越好的，那我們將虛擬基板抽掉直接成長鍺在矽基板上發現其直流特性會下降許多但是交流特性會提升，為了增加偵測的波段我們也將錫摻入鍺當中製作p-n二極體，我們利用了多層的磊晶方式來達到更高濃度的錫摻雜，那我們發現當最上層的鍺錫厚度會影響到元件的特性，越厚的最上層的鍺錫層可以達到越高的光響應但是相對的其暗電流密度也是最高的，暗電流密度之中我發現其主要是產生自空乏區而這樣證明了其較高光響應的原因。

關鍵字

鍺錫合金 ； 原子層堆積 ； 表面鈍化 ； 光激發光頻譜 ； 鍺錫合金能隙 ； 金氧半發光二極體 ； 光偵測器

並列摘要

The bandgap tuning and direct bandgap emission of GeSn can enhance the performance of group IV photonic devices on Si platform. The indirect-to-direct transition of relaxed Ge1-xSnx was reported at the Sn content of 6-11%. GeSn was proposed to be an active layer in electrically pumped laser with the advantage of wavelength tuning by adjusting Sn content and the tensile strain generated by underneath GeSn layers. GeSn high performace and low dark current photodetectors were also demonstrated due to GeSn can have higher absorption coefficient at optical interconnect wavelength and also extent the detection range to FIR regime.
Surface passivation is a critical technique because it can enhance the photonic devices performance. There two methods can achieve surface passivation. The first one is lower surface dangling bond which can treat as defect and it can create non-radiation recombination. But this method usually needs thermal treatment which is not suitable for GeSn because GeSn has low thermal budget. The second scheme is using oxide charge to propel the carriers away from the surface. This method is widely used in high efficiency solar cell which also prove it can be used in the industry. In the second chapter of this dissertation, plasma-enhanced atomic layer deposition is applied to grow SiO2 and Al2O3. By using metal-oxide-semiconductor capacitors, the flat band voltage is extracted to estimate the oxide charge density. The SiO2 and Al2O3 has negative oxide charge around 2x1012 cm-3. The photoluminescence (PL) spectra intensity is also used to distinguish the success of surface passivation. After process optimization, 60-cycle Al2O3 on 20-cycle SiO2 can have the best surface passivation.
Due to the 14% lattice mismatch between Sn and Ge, the GeSn needs Ge virtual substrate. But if the thickness is above critical thickness, the strain will relaxed. The strain will affect the carrier distribution and also the bandgap energies. In the chapter three, the external tensile strain is applied to the Ge/GeSn/Ge quantum well. By using EPM and model solid theory to fit the PL spectra, the bandgap energies are extracted. It is also found that Ge direct emission is sensitive to tensile strain while GeSn direct emission is not.
Due to low emission efficiency for the group four material, light emitting devices is needed. In chapter four, a MIS structure light emitting diode (LED) is fabricated. Currently, the light emitting diode requires doping which can be made into p-n or p-i-n structure but doping will introduce more defect which will lower device performance. By utilizing MIS structure with ultrathin oxide layer, the MISLED can probe the epitaxy film quality without further processes. During epitaxial grown of GeSn layer, a thin Ge cap layer is deposited to prevent Sn segregation but the Ge cap layer will affect carrier distribution. It is found that with Ge cap layer the spectra of GeSn MISLED will mainly attribute by Ge direct emission while without Ge cap layer the spectra of GeSn MISLED will mainly attribute by GeSn direct emission.
After discussing the transmit end devices, in chapter five and six the photodetectors are studied. For the narrow band photodetectors, UV regime detection is very popular due to outdoor activities. By using MIS structure, the metal layer can be treated as optical filter especially Ag. It is also found that Al2O3 can passivate the Si surface. By using ultrathin Al2O3 and Ag, the narrow band Si photodetector can achieve 0.174mA/W at -2V. With proper package and circuit, the device can be integrated into IoT circuit. For the silicon photonics, high performance Ge photodetector is required. It is found that Ge virtual substrate will affect the device DC and AC characteristic. With increasing Ge virtual substrate, the device DC and AC characteristic will also enhanced. In order to have larger 3dB bandwidth, Ge virtual substrate was removed to increase the bandwidth but the DC properties are worse than the devices with Ge virtual substrate. In order to enable extend the detection range, Sn was alloyed into Ge to decrease the bandgap to form a p-n diode by in-situ doping. To achieve high Sn content, three GeSn layer is grown. By increasing top layer GeSn the p-n diode will have better responsivity but the dark current density will also increase. By fabricated different size photodetectors, the dark current was mainly attributed by bulk defect.

並列關鍵字

GeSn ； atomic layer deposition ； surface passivation ； photoluminescence ； GeSn nandgap energies ； light emitting diodes ； photodetectors

參考文獻

[1] Suyog Gupta, Blanka Magyari-Köpe, Yoshio Nishi, and Krishna C. Saraswat, “Achieving direct band gap in germanium through integration of Sn alloying and external strain.” J. Appl. Phys. 113, 073707 (2013).

[2] H.-S. Lan and C. W. Liu, “Ballistic electron transport calculation of strained germanium-tin fin field-effect transistors.” Appl. Phys. Lett. 104, 192101 (2014).

[3] S. Wirths, R. Geiger, N. von den Driesch, G. Mussler, T. Stoica, S.Mantl, Z. Ikonic, M. Luysberg, S. Chiussi, J. M. Hartmann, H. Sigg, J. Faist, D. Buca and D. Grützmacher, “Lasing in direct-bandgap GeSn alloy grown on Si.” Nat. Photonics 9, 88 (2015).

[4] Richard Soref,” Direct-bandgap compositions of the CSiGeSn group-IV alloy” Opt. Mater. Express 4, 836-842 (2014)

[5] Yuan Dong, Wei Wang, Xin Xu, Xiao Gong, Dian Lei, Qian Zhou, Zhe Xu, Soon-Fatt Yoon, Gengchiau Liang, and Yee-Chia Yeo, “Germanium-Tin on Silicon Avalanche Photodiode for Short-Wave Infrared Imaging.” in Proceedings of IEEE Symposia on VLSI Technology and Circuits (IEEE, 2014), pp. 21.2