Growth Mode Transfer

Figure 3.2 shows the plane view of AFM for samples 2, 4, 5, 6, and 7. At both ends of the axis, the black and white represent the lowest point at 0 nm and the highest point at 16 nm. In Fig. 3.2(a), the weak black and white contrast indicates the smooth surface of the ZnSe buffer layer. The average roughness of the ZnSe surface is less than 0.28 nm. When the average CdSe coverage was 2.0 MLs (sample 5), the average surface roughness changed from 0.28 to 0.42 nm. No clear 3D island formation is observed in Fig. 3.2(b). When the average coverage was further increased to 2.5 MLs (sample 6), the QD structures still could not be observed but the surface roughness increased slightly to 0.45 nm. The only change associated with the increase in the CdSe coverage from 2.0 to 2.5 MLs is the roughing of the sample surface. As the average coverage increased further to 2.7 MLs (sample 7), as shown in Fig. 3.2(d), the 3D dots began to appear. The dot density of sample 7 is 2.0×10⁸ cm^-2. The average dot height and diameter are 2.95 and 68.3 nm, respectively. And the average aspect ratio, α= H/D, is 4.3×10^-2. This aspect ratio is much smaller than that (α= 0.25) measured using the amplitude scan of the taping mode by K. Kitamura et al. [3] The dot density is ten times lower too. However, the above observation indicates that the critical thickness is between 2.5 to 2.7 MLs for the transition of 2D to 3D growth, i.e.

the thickness of the wetting layer is about 2.5 to 2.7 MLs. Reference 15 reported the similar result.

In Fig. 3.2(e), the average coverage is 3.0 MLs (sample 2) and two groups of QD were observed. The average diameter and height of the larger (smaller) dots are 86 (66) and 5.35 (2.21) nm, respectively. The dot density of the larger (smaller) dots is 15.2×10⁸ (4.8×10⁸) cm^-2. The dot size distribution (the number of dots versus the dot volume) of sample 2 was summaried in Fig. 3.3. The dot volume was calculated based

on the assumption that the dots are conical. The two group size distributions were observed. The coexistence of small and large dots is similar to that in the second ripen (R2) growth mode, which is theoretically defined by I. Daruka and A.L. Barabasi [4, 5], as shown in the equilibrium phase diagram of Fig. 3.4. Figure 3.4 refers to phase diagram II in Ref. 5. It was obtained by minimizing the sum of the energy of the strained overlayer, the free energy of the QDs and the total energy density of the ripened islands. The x-axis represents the misfit between the buffer layer and the material grown above. While the y-axis represents the average coverage of QD. For a very small misfit and low coverage, only the 2D layer-by-layer growth mode, i.e. the FM mode, is allowed. For a misfit that is less than but close to 0.065, the growth mode starts with the FM mode. If the average coverage is above 3.0 MLs, a ripened (R1) growth mode begins. For misfit between 0.069 and 0.075, the growth starts with the FM mode and proceeds to the SK1 mode and then to the R2 mode as the coverage increases. In the SK1 growth mode, the 3D QDs were grown on the wetting layer. In the R2 growth mode, the larger ripened QDs and the smaller coherent SK QDs coexist.

In current study, the misfit between the ZnSe buffer layer and the CdSe QD is about 0.07, as indicated by the dotted line in Fig. 3.4. The equilibrium phases of the Volmer-Weber (VW), SK2 and R3 growth modes are not discussed herein, since in this work, the misfit is less than 0.1.

To summarize the above investigation, a schematic growth dynamics is shown in Fig. 3.5. Figures 3.5(a) to (c), which correspond to CdSe coverages of 0, 2.5 and 2.7 MLs, respectively, indicate the SK growth mode of CdSe QDs. In Fig. 3.5(d), the CdSe coverage was 3.0 MLs; the original smaller QDs grow into larger ripened QDs, while some smaller SK dots appeared in the flat spaces to relax the remaining stress.

As a result, the growth mode changed form the SK mode to the R2 growth mode, as represented by the dotted line in Fig. 3.4. The coherent SK growth mode was

observed at TG = 260 ^oC with the average CdSe coverage from 2.5 to 3.0 MLs. Figure 3.2(e) clearly shows the ripen growth state for a CdSe coverage from 3.0 to 3.5 MLs, which is also schematically demonstrated in Fig. 3.5(d).

In addition to the above characterization by AFM, PL studies were also performed. The PL spectra for several samples with the CdSe coverage of 1.0 MLs to 3.3 MLs were obtained at 10 K, as shown in Fig. 3.6(a). The peak position clearly shifts toward lower energy as the CdSe coverage increases. Figure 3.6(b) presents the dependence of the PL peak position on the CdSe coverage. Two linear dependences were found. The abrupt change in the slope is about 2.5 MLs, which value is consistent with the thickness of the CdSe QD’s wetting layer, observed by the AFM studies. Restated, when the CdSe coverage was less than the thickness of the wetting layer, no 3D dots were formed. The dependence of PL peak position on CdSe coverage follows a linear behavior contributed by the 2D quantum well. The high PL peak energy, which is very close to the 2.812 eV value of the bulk ZnSe energy gap, can be attributed to the effect of inter-diffusion, energy shift due to strain and the confinement energy of the 2D quantum well. When the CdSe coverage exceeded the thickness of the wetting layer (2.5 MLs), the QD structure formed. The dependence of PL peak position on CdSe coverage obeys another linear relation. The PL transition energy of QDs decreases as the dot size increases. This statement was verified by temperature-dependent PL spectra. Figure 3.7 plots the activation energy (Ea) versus the CdSe coverage. Ea was obtained form the following equation

)] respectively. D is a fitting parameter, and k is the Boltzmann constant. In Fig. 3.8, curves A and B are fitting using equation (3.1) for 2.0 and 2.7 MLs samples, respectively. Ea increased steeply when the CdSe coverage was near 2.5 MLs. This

finding supports the QD morphology obtained by AFM and the study of the dependence of the PL peak position on the CdSe coverage mentioned above. Below 2.5 MLs, the exciton transition energy is close to the band gap of the ZnSe barrier. As a result, the activation energy which to dissociate the exciton from the CdSe 2D quantum well into the ZnSe barrier is small. As the CdSe coverage increases, the increased thickness of 2D quantum well will reduce the confinement of carriers.

Therefore, the activation energy decreases with the CdSe coverage increases as the coverage less than 2.5 MLs. In addition, 2.5 MLs, 3D QDs were formed. Ea increases abruptly because the exciton binding energy in QDs is much strong. A further increase in CdSe coverage reduces Ea because the exciton binding energy decreases as the dot size increases. The fitting curve B initially increases and then decrease as the temperature rises. This behavior is typical of QDs [6-8]. Fitting curve A decreases rapidly as the temperature ruses, indicating 2D quantum well behavior.