In order to analyze the interface quality, GIXRR measurements have been carried out, and the experimental data were simulated with the commercial software of reflectivity and fluorescence simulation [RFS] provided by Bede (Bede Scientific Incorporated). The results are shown in Figure . Figure shows the experimental curve and simulation curve of RTD-1 and RTD-2, respectively. A clear peak can be observed at the range of 1,000 to 2,000 arc sec in GIXRR curve of RTD-1, but nearly no satellite peak can be observed in Figure which is RTD sample with embedded InAs QD layer. The appearance of the satellite peaks in Figure suggests a better interfacial smoothness of RTD-1 than RTD-2. It is also shown in the same figure that the simulation curve is well-fitted with the experimental curve of the sample without InAs QD layer. Because the experimental curve of RTD-2 as shown in Figure had no obvious satellite peaks, the simulation curve did not fit with the experimental curve accurately. The root mean square [RMS] roughness of different layers for the two samples were obtained and listed in Table . Note that due to the sharp decrease of the X-ray intensity during the measurement of RTD-2, the simulated curve cannot match the experimental results well. The RMS roughness of this sample is only listed as a reference but does not accurately represent the real structural characteristics of this sample. It can be clearly found that the RMS roughness of AlAs UBL was higher than DBL for both samples. Considerably, the same growth conditions were adopted for the two samples. There are reasons to believe that deposited InAs QDs between double barriers worsened the interfacial flatness.
GIXRR measurements of the experimental samples. The GIXRR for (a) RTD-1, (b) RTD-2, and (c) RTD-3 are measured. GIXRR curve (black solid line) and corresponding simulation curve (red dash line)
Simulated result of RMS roughness of every layer of RTD-1, RTD-2, and RTD-3
The penetration depth of X-ray is very low in the test of GIXRR, and the reflection light intensity will show a power exponent downward trend with the increase of penetration depth [10
]. So, the GIXRR better reflects the quality of the interface close to the surface. Hence, it is believable that the decline of AlAs UBL/InGaAs interface flatness contributes to the deteriorative interfacial quality of RTD-2 more than AlAs DBL/InGaAs interface. Because the AlAs UBL/InGaAs hetero-junction was deposited after the growth of InAs QDs, it will definitely increase the difficulty to obtain a flat hetero-junction interface. Obviously, in order to put this type of QD-RTD into real use, the interfacial flatness must be improved.
To improve the interface quality, three methods may be feasible. The first one is depositing thicker InGaAs capping layer, but the increasing thickness of quantum well will introduce a serious degradation on resonant tunneling performance of RTD [13
], at the same time, this change of material structure will increase the total strain accumulation of InAs/InGaAs system [14
]. The second one is to raise the growth temperature of InGaAs capping layer. Higher temperature is conducive to increase the atom migration ability and, thus, improve the interface flatness. However, for the sake of the weaker In-As chemical bond, this approach may lead to the deviation of indium component of InGaAs capping layer from the setting value and may even cause the InAs QDs to dissolve. The last one is to increase the growth temperature of depositing AlAs UBL. Because the InGaAs layer is strained, the system tends to reduce the strain energy through segregating indium atoms onto surface [15
]. This phenomenon will increase the roughness of InGaAs surface. Raising temperature after InGaAs growth as an annealing treatment can evaporate excess indium atoms at InGaAs layer surface and will result a better growth of front flatness [17
]. So, we adopted the last method to grow simple RTD-3. Its structure was exactly the same as RTD-2; the only difference was the improved growth temperature of the AlAs UBL from 500°C to 610°C with an interruption of 2 min.
Figure shows the GIXRR curve of RTD-3. The appearance of multi-level satellite peaks indicates that the interface quality has been truly improved. The experimental data were simulated with RFS, and the RMS roughness of every layer was also listed in Table . According to the simulation of RTD-3, the RMS roughness for InAs QDs layer was 4.62 nm which was consistent with the average height of the Stranski-Krastanov growth mode QDs (4 to 7 nm) [19
], indicating that the simulating result was very close to real value; so, this simulation should be convincing. In Table , it is shown that the flatness of UBL has been greatly optimized via the improvement of growth condition. The RMS roughness of RTD-3 UBL decreased from 0.31 nm (of RTD-1) to 0.18 nm. In addition, it should also be noticed that UBL and DBL of RTD-3 have RMS roughness of 0.17 nm and 0.18 nm, respectively. Obviously, the interfacial flatness of UBL has been improved to be close to the level of DBL.
In order to obtain the structural characteristics at the atomic level, the cross-sectional HRTEM image, which was taken along the [1 1 0] direction of sample RTD-3, was obtained. The results are shown in Figure . The position of the AlAs barriers and the InAs QDs was marked in Figure , and shows the enlarged image of the part. It is shown in Figure that the roughness of AlAs UBL and DBL have no obvious difference and are both in the range of 1 to 3 ML. Another phenomenon could be observed from Figure ; both the interfaces of UBL and DBL have very similar 2 ML step at the same location. This feature suggests that, at least, parts of the steps in UBL were transferred from the DBL but did not form during the growth process of UBL. This phenomenon may also explain the same RMS of AlAs UBL and DBL measured by GIXRR.
HRTEM images for InAs/InGaAs/AlAs structure. (a) Low magnification, and (b) high magnification. Inset image is the Fourier-transformed image of selected area.
Considering the change of growth condition that may dissolve InAs QDs, PL spectra at 77 K for samples RTD-2 and RTD-3 were acquired and shown in Figure . It can be found that each PL spectrum shows a dominant peak for the two samples, respectively, which is related to the inter-band transitions of InAs QDs. This result indicated that using higher temperature for UBL cannot dissolve InAs QDs. In addition, the PL peak position of sample RTD-3 shifts to shorter wavelengths of 27 nm (from 1,051 nm to 1,024 nm). This behavior may be attributed to the lower indium composition of InAs QDs resulting from local inter-diffusion of In and Ga atoms in the InGaAs/InAs QDs layers when using higher growth temperature [21
PL spectra of RTD-2 (red dash line) and RTD-3 (black solid line) samples.