HgTe/HgCdTe Superlattices Grown on CdTe/Si by Molecular Beam Epitaxy for Infrared Detection

Journal of Electronic Materials, Jun 2004 by Selamet, Y, Zhou, Y D, Zhao, J, Chang, Y, Et al

We closely monitored the reflection, high-energy electron diffraction patterns at the beginning and during the growth. They were all streaky from the start of the bottom n^sup ^ layer and throughout the SL and top undoped layers. We concluded that the temperature stabilization was optimal, and the crystalline quality of the layers was good.

The cell temperatures were controlled to �0.1�C over the entire growth procedure to ensure stable fluxes. During growth, the steady-state values of the CdTe, Te, and Hg fluxes were selected to give a HgTe/CdTe SL growth rate of about 2-3 [Angstrom]/see, which is much lower than the HgCdTe alloy growth rate of 6-8 [Angstrom]/see. The lower growth rate provides us with better control over the thickness of HgTe wells and CdTe barriers. The difference in growth rates for the HgTe/CdTe SL and HgCdTe alloy layers leads to different growth conditions. We reduced the CdTe and Te fluxes to lower the growth rate, and, at the same time, the Hg flux was reduced according to the corresponding Te flux necessary to maintain good crystal quality. all these changes were made after completing the bottom HgCdTe alloy growth and were made again after completing the HgTe/CdTe SL before starting the top HgCdTe-alloy growth. The CdTe, Te, and In cell temperatures and Hg fluxes during the structure growth are shown in Fig. 2.

In-situ ellipsometry data was also collected during HgTe/CdTe SL growth and is shown in Fig. 3. It shows a periodic change in the SL optical properties alternating growth of HgTe and CdTe layers. We can observe two superimposed periodicities. The slower oscillation is due to the progressive increase in total thickness. This oscillation dies out once the layer is optically thick. The shorter-period oscillation is due to the successive deposition of materials characterized by different indices of refraction. The period of the shorter oscillation can be used to estimate the growth rate of HgTe layers. The calibration of the CdTe growth rate is not easily done by ellipsometry because it cannot distinguish between the CdTe SL layers and the CdTe in the CdTe/Si substrates. Therefore, we resorted to step profiling. We also checked the results of our ellipsometric growth rates against those obtained with step profiling and obtained good agreement between the two.

The crystalline quality of the material was studied using double-crystal rocking curves (DCRCs). Figure 4 shows the DCRCs of the detector structure. X-ray diffraction shows two peaks. Peak 1 located at - 657.5 arcsec is less intense and is attributed to the SL because the top HgCdTe alloy layer weakens and broadens the SL diffraction signal. The DCRC fullwidth at half-maximum (FWHM) for peak 1 is 110 arcsec. Peak 2, located at -8.9 arcsec, is attributed to the HgCdTe-alloy layers. The DCRC FWHM for peak 2 is 122 arcsec, implying good crystal quality. These FWHM values are comparable to those of HgCdTe alloys grown on CdTe/Si substrates.

Experimental Fourier transform infrared (FTIR) spectra show features caused by both HgCdTe-alloy layers and the HgTe/CdTe SL. The representative FTIR spectrum at 295 K of Fig. 5 shows absorption caused by different subband transitions. From the calibration runs, we expect the HgCdTe-alloy cutoff to be around 3,400 cm^sup -1^ or 2.94 �m in the wavelength scale because the targeted layer composition was around x = 0.38. This is in good agreement with the experimental feature at about 2.9 �m. The cutoff for the HgTe/CdTe SL is seen to be at 5 �m. The FTIR data of single HgTe/CdTe SL layers (not device structures) display a similar feature near 5 �m. The absorption coefficient has been calculated for an alloy with x = 0.38 andoa SL with well and barrier widths of 34 [Angstrom] and 40 [Angstrom], respectively, as shown in Fig. 6. Theoretical details can be found elsewhere.2 Agreement with experiment is good at 295 K. From the theoretical absorption coefficient for 80 K, the absorption edge of the alloy only shifts from 2.9 �m to 3.05 �m, whereas the absorption edge caused by the SL shifts from about 5 �m to 7 �m, which is in very good agreement with the spectral response discussed later.