Shape transitions of metastable surface nanostructures D. J. Vine, D. E. Jesson,* and M. J. Morgan School of Physics, Monash University, Victoria 3800, Australia V. A. Shchukin and D. Bimberg Institut für Festkörperphysik, Technische Universität Berlin, D-10623 Berlin, Germany Received 15 August 2005; revised manuscript received 21 September 2005; published 9 December 2005 A shape transition between surface nanostructures which, as a function of island size, are associated with minima in formation energy per atom is modeled using a Fokker-Planck equation. We find that metastable states, associated with positive gradients in island chemical potential, can dominate the dynamics of the transition. The resulting bimodal island size distribution function is metastable to Ostwald ripening which has important implications for the self-organization of quantum dots. DOI: 10.1103/PhysRevB.72.241304 PACS numbers: 81.07.Ta, 68.35.Md, 68.65.-k, 81.16.-c The self-assembly and self-organization of nanostructures on surfaces can be utilized to produce quantum dot arrays for device applications. 1,2 This can be readily achieved, for ex- ample, by depositing thin films using deposition techniques such as molecular beam epitaxy. 3–5 The resulting islands, or dots, can be overgrown by appropriate layers to form the basis for devices such as semiconductor lasers. However, size uniformity is critical for many applications which has led to significant efforts to understand the key factors gov- erning the coarsening of quantum dot arrays. Surface nanostructures that possess a minimum in forma- tion energy per atom MEAas a function of island size MEA systemsare particularly attractive candidates for de- vice applications because they are associated with a thermo- dynamically favored size. By simply annealing such struc- tures, one might anticipate the creation of arrays with good size uniformity. Although it is not possible to identify MEA systems a priori, theoretical studies have shown that coher- ently strained two-dimensional 2Dislands, 6–8 three- dimensional 3Dislands with surface stress discontinuities at their edges 9,10 or 3D islands with strain renormalized sur- face energy 9,11 are potential candidates for MEA systems. A feature of particular interest in the case of 3D nano- structures is the possibility that surface islands can undergo shape transitions. 12–23 This can result in a multimodal island size distribution function during the self-organization of quantum dot systems that can deleteriously influence device performance. Understanding the dynamics of shape transi- tions is therefore of critical importance to control island size distributions and obtain good size uniformity. Theoretical descriptions of quantum dot systems undergo- ing shape transitions can be broadly classified as being ther- modynamic or kinetic in nature. Kinetic models have empha- sized the discontinuity in island chemical potential as islands attain a critical transition size. 12 Upon transformation to a new shape, the lower chemical potential islands grow rap- idly, resulting in a bimodal size distribution. In contrast, ther- modynamic models associate peaks in the island size distri- bution function with minima in formation energy per atom for different island shapes. 13,15,17 However, as discussed by Rudd et al., 15 the dynamics of the transition between “stable” states in thermodynamic models and the consequences for self-organization are relatively unexplored. In this Rapid Communication, we therefore develop a the- oretical description of shape transition dynamics in MEA systems. Surprisingly, metastable states, associated with positive gradients in chemical potential, are found to domi- nate the dynamics of the transition. These states are distinct from energy per atom minima and have important implica- tions for self-organization of MEA systems and the applica- tion of thermodynamic models to interpret experimental data. 13,15,17 To model a shape transition between nanostructures ex- hibiting MEA behavior we consider the specific case of 3D semiconductor islands with strain-renormalized surface en- ergy which are assumed to exhibit MEA properties. 9,11 The dimensionless formation energy E s Nof a faceted quantum dot as a function of the number of atoms N it contains is given by 9,11 E s N=- s N + s N 2/3 -2N 1/3 ln e 1/2 N 1/3 . 1 The first term is the island relaxation energy, the second term incorporates the change in renormalized surface energy due to island formation, and the third term combines the positive short range energy of the island edges with the nega- tive surface stress induced elastic relaxation energy at the edges, E elastic edges . The parameter s is the ratio of the volume relaxation energy to |E elastic edges | and s is the ratio of the renor- malized surface energy to |E elastic edges |. The subscript s =1, 2 of the parameters s , s , refers to two different nanostructure shapes. We assume that the island array is sufficiently dilute so that the elastic interaction between islands can be ne- glected. Quantitative values of the parameters s , s are presently unknown. For the purpose of this simulation we assume the arbitrary but physically reasonable values of 1 = 4.2, 1 =-0.8, 2 =5.2 and 2 =0.0. In Fig. 1awe plot the island formation energy per atom s N= E s N/ N as a function of island size for two shapes 1 and 2. Each shape is associated with a minimum in s Nat 1 N 1 E and 2 N 2 E . For small island sizes, shape 1 is energetically favorable. However, PHYSICAL REVIEW B 72, 241304R2005 RAPID COMMUNICATIONS 1098-0121/2005/7224/2413044/$23.00 ©2005 The American Physical Society 241304-1