removed while the etching of GaN. Therefore, compared with the sample without voids, openings above the voids in AlN interlayer may introduce additional dislocations into overlying GaN which leads to quality degradation.
Moreover, voids brought more significant effect to the optical properties of GaN/AlN multilayer structure on Si than to crystal quality. The cross-section view and plane view of voids in GaN layers were shown in Fig. B.8a and b. From the plane view it was found that etching of GaN occurred along the regular facets of GaN. Such voids with crystal facets are good light scattering structure. When light comes to the facets, it would be reflected or scattered to various directions which it not normal to the plane of (0001) of GaN anymore.
Only a part of the light goes back to the direction which it comes along, as represented in Fig.
B.8c. As a result, as plotted in Fig. B.8d, as the GaN thickness increased and the density of voids accumulated, the intensity of reflected incident light with wavelength of 950 nm from the curvature and temperature monitor head was decreasing. The same thing happened with the reflectivity measurement of these two samples. Depending on the wavelength, the reflectivity of GaN with voids was about 0.05 lower than that without voids. More significantly, in the photoluminescence (PL) measurement, no PL light could be collected from the sample with voids. The normal peak around 550 nm from GaN was obtained. These experimental results proved that the assumption of scattering by the voids was reasonable.
There are some potential applications of the light scattering effect of void structure in devices. Such light scattering structure is favorable in solar cells to trap lights and increase the optical path length in the device to enhance its light-electricity conversion efficiency.
Another application is in LEDs. Due to scattering of the voids, the intensity of emitted light from LED would be less angle-dependent. Especially, it may be helpful to improve the external quantum efficiency of LESs on GaN-on-Si wafer if the influence of voids on the stress control and GaN quality is ignorable. Because on GaN-on-Si wafer, a portion of the light emitted from the LED active area may be absorbed by Si substrate. The presence of void structure reduces the intensity of the light goes to the Si substrate side.
points is ignorable, the light scattering effect of the voids has potential application in some devices like solar cells and LEDs.
References
[1] S. Arulkumaran, G.I. Ng, S. Vicknesh, H. Wang, K.S. Ang, J.P.Y. Tan, V.K. Lin, S. Todd, G.Q. Lo, and S.
Tripathy,Direct Current and Microwave Characteristics of Sub-micron AlGaN/GaN High-Electron-Mobility Transistors on 8-Inch Si(111) Substrate, Jpn. J. Appl. Phys. 51, (2012).
[2] D.D. Koleske, A.E. Wickenden, R.L. Henry, J.C. Culbertson, and M.E. Twigg,GaN decomposition in H2 and N2 at MOVPE temperatures and pressures, J. Cryst. Growth 223, 466(2001).
[3] M.A. Mastro, O.M. Kryliouk, M.D. Reed, T.J. Anderson, A. Davydov, and A. Shapiro,Thermal Stability of MOCVD and HVPE GaN Layers in H2_HCl_NH3 and N2, Phys. Status Solidi (a) 188, 467(2001).
[4] E.E. Zavarin, D.S. Sizov, W.V. Lundin, A.F. Tsatsulnikov, R.A. Talalaev, A.V. Kondratyev, and O.V.
Bord,In-situ investigatioins of GaN chemical unstability during MOCVD, Proceedings of the Fifteenth International European Conference on Chemical Vapor Deposition (EUROCVD-15) 2005-09, 299(2005).
[5] D.D. Koleske, M.E. Coltrin, A.A. Allerman, K.C. Cross, C.C. Mitchell, and J.J. Figiel,In situ measurements of GaN nucleation layer decompostion, Appl. Phys. Lett. 82, 1170(2003).
[6] E.V. Yakovlev, R.A. Talalaev, A.S. Segal, A.V. Lobanova, W.V. Lundin, E.E. Zavarin, M.A. Sinitsyn, A.F. Tsatsulnikov, and A.E. Nikolaev,Hydrogen effects in III-nitride MOVPE, J. Cryst. Growth 310, 4862(2008).
[7] Y.-H. Yeh, K.-M. Chen, Y.-H. Wu, Y.-C. Hsu, and W.-I. Lee,Hydrogen etching on the surface of GaN for producing patterned structures, J. Cryst. Growth 314, 9(2011).
[8] Y.-H. Yeh, K.-M. Chen, Y.-H. Wu, Y.-C. Hsu, T.-Y. Yu, and W.-I. Lee,Hydrogen etching of GaN and its application to produce free-standing GaN thick films, J. Cryst. Growth 333, 16(2011).
[9] D.D. Koleske, A.E. Wickenden, R.L. Henry, J.C. Culbertson, and M.E. Twigg,<GaN decomposition in H2 and N2 at MOVPE temperatures and pressures.pdf>, J. Cryst. Growth 223, 466(2001).
171
172
Appendix C Suggestions on further research
This section is for the people who want to do further research in the field of GaN-on-Si.
Although this field has been studied more than a decade, there are still some public problems which should be solved or improved further. The key topic is how to control the stress and improve the crystal quality more efficiently in smaller thickness. The flow of this part is the same as the growth sequence of the sample structure.
(1) First of all is the buffer layer. Buffer layer plays a critical role in the system of GaN-on-Si, it influence both of the stress and quality of the 1st GaN layer as well as the whole system. As discussed in previous chapters, the quality of AlN buffer layer should be as high as possible. So far, the growth of AlN on Si is far from being optimized. The quality of AlN on Si substrate is much lower than that on sapphire even employing the same growth conditions. The microstructural growth mechanism of AlN on Si is different from that on sapphire which is yet to be clarified. The growth technologies of AlN on sapphire like 3D growth mode and pulse-injection method can be transplanted to it on Si.
(2) The quality improvement of the 1st GaN. Since the interlayers may be not helpful to improve the GaN quality, it should be enhanced in the 1st GaN as much as possible.
There are some other strategies to improve the quality of the 1st GaN besides optimizing the growth of AlN buffer and applying 3D growth of itself. Masking and patterning on GaN are promising methods for it, by SiNx or MgNx for example.
(3) Most importantly, the capability of AlN interlayers of inducing compressive stress in overlying GaN layers can be elevated further. Employing more effective interlayers, the number of interlayers and total GaN thickness can be reduced. The growth conditions of new prototypes of interlayers in chapter 6 can be optimized and better performance of them is highly expectable. The micro-structures of them are highly worth investigating. Moreover, another type of two-step interlayer which consists of lower AlGaN and upper AlN is also attractive since it is capable of producing high-quality upper interface of it with small AlN lattice constant.
(4) More attention should be paid to the device on the template of GaN-on-Si, such as LED. In the end of chapter 5, there is an interesting phenomena that the peak position of blue LEDs through the wafer was uniform but not for the case of PL intensity.
Surprisingly, PL intensity in the edge was seven times of that in the center in spite of generally better template quality in the center. Topics including the uniformity of wafer bow, temperature and stress distribution, InGaN composition, thickness and morpholgy of InGaN/GaN layers in the MQW are quite critical to the performance of LEDs on GaN-on-Si. The influence of wafer bow which can affect the stress state and bandgap profile in the MQW on the LED performance is also very important and interesting. There are little reports about the topics listed above.
Moreover, there is some comment on the production efficiency of MOVPE of GaN-on-Si.
Up to date, almost all buffer layers were grown by MOVPE for GaN-on-Si. This requires
173
cleaning of Si wafer and reactor dry cleaning or AlN coating every time prior to the growth of GaN-on-Si. This is not time efficient. If the AlN buffer layer is grown by physical vapor deposition (PVD), such cleaning processes can be simplified to save time. MOVPE of GaN-on-Si can start directly from the growth of the 1st GaN layer. More importantly is, according to the work of Hongbo Wang, quality of GaN on PVD-AlN buffer is much better than that on MOVPE-AlN buffer, which would lead to more efficient stress control in GaN-on-Si and better crystal quality. In conclusion, PVD-AlN buffer layer on Si is a promising routine to improve the growth of GaN on Si.
174