Stranski-Krastanow growth mode
The layer thickness for the transition from two-dimensional to three-dimensional growth as characteristic feature of the SK growth mode depends on lattice mismatch and therefore on In composition for the ternary InKGai_KN material. In situ growth mode analysis during MBE using Reflection High-Energy Diffraction (RHEED) is used to investigate the 2D/3D critical layer thickness. The results for MBE using ammonia are shown in Fig. 5.12 for 0 < xIn< 0.50. By investigation of the QD formation at 550° C growth temperature, the minimum In composition for SK-QD formation is around 12%, below which only surface roughening occurs . A different onset of the transition at 18% indium content is measured for plasma-assisted MBE using nitrogen. For such an MBE method, SK-QD require stoichiometric or slightly nitrogen-rich conditions. For an Ino.35Gao.65N layer, the critical layer thickness is 1.7 MLs, which is also lower than the value taken from Fig. 5.12. In terms of compositional dependence of the critical layer thickness and lower boundary for the 2D/3D transition, the InGaN material systems resembles very much the behavior of the well-studied InGaAs/GaAs system .
Proper MOVPE growth conditions for QD formation are different from those reported for MBE. Foremost, growth temperatures ranging from 600 to 800° C
Fig. 5.12. Phase diagram for 2D/3D transition during growth of InGaN by MBE. (Reprinted with permission from , © 1998 American Institute of Physics.)
Fig. 5.13. Left) Surface morphologies taken by AFM after MOVPE growth of InGaN on GaN(0001) surfaces for InGaN coverages of a) 5.3, b) 10.4, and c) 18.4 MLs. Right) Reduction of the QD density with increasing temperature for an nominal coverage of 19 ML InGaN. (Reprinted with permission from , © 1999 American Institute of Physics.)
and V/III ratios of several thousands are applied owing to the low decomposition rate of ammonia (NH3). These values are more alike quantum-well growth parameters, which is in contrast to InGaAs/GaAs(001), where lower growth temperatures and low V/III ratios are favored. Also, the critical layer thicknesses for QD formation of 5-10 MLs for In compositions of around 20-40% are significantly larger than for MBE (Fig. 5.13). In part, the latter deviations can be attributed to uncertainties due to ex situ calibration of growth rate and composition of InGaN layers for MOVPE. The right panel of Fig. 5.13 shows an example of the decrease of the QD density with increasing temperature. An exponential relation of the QD density with inverse temperature is measured [43, 44].
Typical dimensions of InGaN QDs are between 1 and 5 nm height and 5 and 30 nm base length both for MOVPE and MBE growth. The electronic confinement in the lateral direction may not be sufficient for discretization of energy levels because of the much smaller excitonic Bohr radius of only about 3.4 nm. This may also explain that the 2D/3D transition is not easily detected in the emission spectra as a transition from a spectrally narrow quantum-well emission to an inhomogeneously size-broadened QD ensemble luminescence.
Relatively high ammonia partial pressures are needed to stabilize InGaN surfaces against In droplet formation. The investigation of the impact of different V/III ratios on QD formation at growth temperatures of 755° C using InGaN layers with approximately 20-30% indium indicates that an optimum for the NH3 flow rate (and hence for the V/III ratio) exists . At very high ammonia flow rates, large three-dimensional objects can show up on the surface which are either relaxed clusters or do not provide three-dimensional quantum confinement for charge carriers. In contrast, very low ammonia flow rates may lead to suppression of the QD formation as In atoms may evaporate from the surface.
Growth interruptions (GRI) after deposition of the wetting layer are commonly used in MOVPE of InGaAs/GaAs(001) QDs to control average QD size and density of QD ensembles. Material transfer from the WL into QDs and from small to large QDs leads to red-shifted emission spectra with increasing growth interruption times. Simultaneously, the QD density gradually decreases due to the dissolution of smaller QDs. The evolution of size and density of InGaN QDs is affected by indium desorption effects as illustrated by Fig. 5.14. During ammonia-stabilized growth interruptions the size distributions at 60s GRI shows a reduction in height and width of the QDs and an increase in QD density as compared to 30s GRI. Indium desorption competes with the QD nucleation during the initial stage of the GRI. Afterwards, the QD density decreases and the QD sizes increase, indicating material transfer from small to large QDs. The continued blue-shift of the emission spectra indicate the desorption of In atoms into the gas phase also at this stage.
In order to provide sufficient optical gain for laser diode devices, stacking of QD planes need to be considered. Thus, overgrowth of QDs with GaN barrier layers is necessary which, due to the volatility of the In-N bond, cannot be performed at temperatures far above the QD growth temperatures. By choosing
Fig. 5.14. Evolution of InGaN a) QD size and b) QD density for different duration of growth interruption during MOVPE. (Reprinted from  with permission from IOP.) the same temperature for QDs and barrier layers, overgrowth of QDs is achieved for x/„<0.5 [42, 48, 49]. At very high In concentrations in the QDs, dissolution during overgrowth takes place both for MBE and for MOVPE . Insight into vertically correlated growth of stacked InGaN quantum dot layers has not yet been gained.