First-Principles-Based Kinetic Modeling of Surface Growth During Plasma Deposition of Silicon Thin Films

Hydrogenated amorphous silicon (a-Si:H) thin films are used extensively in the fabrication of large-area electronic and optoelectronic devices. Some of their important applications are as thin-film transistors for active-matrix displays, multiple-junction photovoltaic solar cells, active pixels in X-ray imaging, photoreceptors, and sensing devices. Growth of these a-Si:H thin films by plasma enhanced chemical vapor deposition (PECVD) has been an active area of research for over two decades. In spite of the substantial amount of information that has been generated from experiments, the growth models proposed so far have been able to neither explain fully the complex growth process nor account for the surface composition and its temperature dependence, as well as the relative roles of coordination defects in the growth of a-Si:H thin films.

Predictive modeling of surface growth, as determined by plasma-surface interactions during PECVD, requires properly developed coarse-grained dynamical simulations capable of capturing effectively surface length scales and deposition time scales. Toward this end, we have carried out a computational study of the growth mechanisms of plasma deposited a-Si:H thin films based on kinetic Monte Carlo (KMC) simulations according to a transition probability database constructed by first-principles density functional theory (DFT) calculations. Based on the results of our study, we propose a comprehensive model of a-Si:H thin-film growth by plasma deposition under conditions that make the silyl radical the dominant deposition precursor. The transition probabilities for the various kinetic events accounted for in the KMC simulations are based on DFT calculations on the H-terminated Si(001)-(2x1) surface; this surface is considered as a representative model of the local chemical environment and atomic coordination on the growing film surfaces. The relevant surface transport and reaction processes were identified in molecular-dynamics simulations of SiH₃ radical impingement on the growth surface and include SiH₃ surface diffusion, SiH₃ chemisorption and insertion into Si-Si bonds, surface H abstraction reactions, surface hydride dissociation reactions, as well as SiH₄ and Si₂H₆ desorption into the gas phase. The DFT calculations have been conducted within the generalized gradient approximation (GGA) and employed plane-wave basis sets, ultra-soft pseudopotentials, slab supercells, and the nudged elastic band method for determining optimal surface reaction/diffusion pathways and the corresponding activation energy barriers. Results are presented for two types of KMC simulations. The first one employs a fully ab initio database of activation energy barriers for the surface rate processes involved and is appropriate for modeling the very early stages of growth. The second one uses approximate rates for all the relevant processes to account properly for the effects on the activation energetics of interactions between species adsorbed at neighboring surface sites; it is appropriate for modeling later stages of growth toward a steady state of the surface composition.

We find that the relative roles of surface coordination defects are crucial in determining the surface composition of plasma deposited a-Si:H films and should be properly accounted for. Furthermore, we explain adequately the critical roles of both dangling bonds and floating bonds during a-Si:H film growth and their relative contributions to the growth process over a broad range of substrate temperatures. From the set of possible surface rate processes, of particular importance is the radical dissociative adsorption mediated by Si floating bonds along the reaction pathway that forms silicon dihydrides on the surface. The KMC predictions for the temperature dependence (over the range from 300 K to 700 K) of the surface concentration of SiH_x(s) (x = 1,2,3) species, the surface hydrogen content, and the surface dangling-bond coverage are compared with experimental measurements on a-Si:H films deposited under operating conditions that render the SiH₃ radical the dominant deposition precursor. The predictions of both KMC simulation types are consistent with the reported experimental data, which are based on in situ attenuated total reflection Fourier transformed infrared (ATR-FTIR) spectroscopy. The agreement of our computational modeling predictions with the experimental measurements is good quantitatively and excellent in terms of the qualitative trends captured.