Towards predictive atomistic simulations of SiC crystal growth

Alexander Reichmann a, Zahra Rajabzadeh a, Sebastian Hofer b, René Hammer b,c, Lorenz Romaner a · 2025

Read Towards predictive atomistic simulations of SiC crystal growth by Alexander Reichmann a, Zahra Rajabzadeh a, Sebastian Hofer b, René Hammer b,c, Lorenz Romaner a (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-20804.txt.

Introduction

Alexander Reichmann and four co-authors work on growing silicon carbide crystals inside a computer. Growers need these crystals for electronics, but atom-by-atom simulation forces unrealistically fast deposition. Atoms slammed down that quickly form messy layers and many defects, more like sputtering than real growth.

The team instead use a method that places each atom directly at its lowest-energy spot, stretching the effective time scale. They grew crystals on stepped silicon carbide surfaces cut at two angles, testing settings and atom-interaction models. Their tuned setup reproduced stable step-flow growth that ordinary simulation could not. They also measured dislocations, surface roughness, and atom movement in the results.

Catalog record

CREATORAlexander Reichmann a, Zahra Rajabzadeh a, Sebastian Hofer b, René Hammer b,c, Lorenz Romaner a
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE28 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-20804

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Towards predictive atomistic simulations of SiC crystal growth Preprint: APS/123-QED Alexander Reichmann a, Zahra Rajabzadeh a, Sebastian Hofer b, René Hammer b,c, Lorenz Romaner a Abstract Simulations of SiC crystal growth using molecular dynamics (MD) have become popular in recent years. They, however, simulate very fast deposition rates, to reduce computational costs. Therefore, they are more akin to surface sputtering, leading to abnormal growth effects, including thick amorphous layers and large defect densities. A recently developed method, called the minimum energy atomic deposition (MEAD), tries to overcome this problem by depositing the atoms directly at the minimum energy positions, increasing the time scale. We apply the MEAD method to simulate SiC crystal growth on stepped C-terminated 4H substrates with 4° and 8° off-cut angle. We explore relevant calculations settings, such as amount of equilibration steps between depositions and influence of simulation cell sizes and bench mark different interatomic potentials. The carefully calibrated methodology is able to replicate the stable step-flow growth, which was so far not possible using conventional MD simulations. Furthermore, the simulated crystals are evaluated in terms of their dislocations, surface roughness and atom mobility. Our methodology paves the way for future high fidelity investigations of surface phenomena in crystal growth. I Introduction The wide band gap material silicon carbide (SiC) has gained more and more importance for the realization of electronic devices, due to its unique properties, which enable operation at elevated frequencies, voltages, and temperatures. These properties allow SiC to experience a wide range of use-cases, which range from power devices 48 , the automotive industry 55 , to quantum computing 49 and quantum sensing 4 . While SiC has several advantages over Si and is believed to replace it in many applications, its complex and time consuming crystal growth process leads to much higher production costs. SiC crystals are usually grown at very high temperatures, between 2200-2800 K, and mainly via the physical vapor transport (PVT) method, which has much smaller growth rates compared to growth from the melt. It is difficult to track the thermodynamic processes that occur in the furnace during growth 10 , and therefore our understanding of the detailed conditions of growth and the formation of defects is still limited 31 . One type of defect, the stacking fault in particular, is of great relevance, since SiC can occur in more than 200 different polytypes 19 ; 57 , with different electrical and mechanical properties. Especially the band gap of the most commonly occurring polytypes 3C, 6H and 4H is 2.4, 3 and 3.2 eV, respectively 54 , showing considerable variation. Understanding the conditions that favor the growth of a certain polytype is crucial, and computational studies can provide deep insights. Polytype stability has been investigated to a large extent using ab-initio methods 21 ; 41 ; 34 ; 18 ; 30 and compared to experimental results 17 ; 14 ; 59 . The experimental investigations show that the polytype stability depends heavily on the temperature. At lower temperatures 3C is generally formed 14 , while 4H and 6H are favored at elevated temperatures 59 . In addition, the polytype grown is also influenced by the surface’s characteristics 25 , the addition of doping elements 46 or other factors, such as the C/Si ratio in the growing surroundings 52 . DFT investigations at 0K, also don’t paint a clear picture, since the free energy difference between the polytypes is in the range of a few meV/atom, which is small compared to the energy difference induced by the choice of exchange-correlation functional. For example, using the generalized gradient approximation (GGA), 4H and 6H are lower in energy than 3C at 0K. Adding van der Waals corrections can lead to 3C being the most stable polytype, but not for every implementation 23 . Dou et al. 8 have reported that, looking at the phonon interactions, 4H becomes the most stable polytype, while 3C is metastable. Scalise et al. 45 have analyzed the stability at elevated temperatures via calculations of vibrational entropy and found that 4H and 6H are more stable than 3C, in agreement with experiment. Similar to the experiment, it has also shown in DFT calculations that doping changes the polytype stability 36 ; 43 , especially for donors since the conduction band of the polytypes is energetically strongly shifted. Of great importance are surface phenomena that may allow a polytype to grow, even if its bulk free energy is unfavorable. Recently Ramakers et al. 44 have calculated surface energetics and concluded that, while 4H and 6H are favored in the bulk phase, 3C has a lower surface energy which may favor it during growth at lower temperatures. Direct simulations of crystal growth phenoman on the surface are, however, computationally too demanding for DFT. In the past, such simulations have mostly been carried out using kinetic Monte Carlo (kMC) methods 5 , including super lattice kMC 2 ; 3 ; 9 or competitive lattice kMC 13 ; 16 . While these methods capture important processes such as step-bunching 29 ; 6 or island growth 1 , they rely on predefined deposition sites, that remain fixed during the growth simulation. This makes it difficult to realistically simulate dislocation behaviour during growth. Furthermore, the important processes taking place on the surface and their reaction rates must be input to the calculation and the corresponding choices may drastically determine the outcome of the simulation. In recent years atomistic investigations of SiC crystal growth have increasingly been carried out via molecular dynamics (MD) simulations 20 ; 58 ; 56 ; 24 , using interatomic potentials (IPs). While no constraints on crystal structure or defects are imposed in such simulations, they have substantial computational costs compared to kMC, result in smaller length scales (a few nm) and shorter timescales (a few ns). Therefore, deposition rates in the simulations exceed the experimental values by far ( \sim 10^{7} - 10^{9} times higher). This may be the reason why previous MD simulations 20 ; 56 have revealed thick amorphous layers and predominant growth of 3C on top of a [0001] oriented 4H substrate, even at high temperatures. So far MD methods were not able to replicate the kinetic stabilization of a 4H polytype using a stepped substrate, as it is commonly used in experiments 33 ; 25 . Therefore, the real growth mechanism is not captured by such simulations. Promising approaches to combine the advantages of MD and MC have been proposed. Most of these are hybrid approaches 35 , such as time-stamped force biased MC (tfMC), which combines the force calculation of MD with the hopping probability of MC. Another method was recently proposed by Karewar et al. 22 and is called the minimum energy atomic deposition (MEAD) method. It works by first scanning over the potential energy landscape (PEL) of a system and finding minimum energy positions (MEPs). Afterwards, atoms are deposited in these MEPs. Before the next scanning of the PEL is commenced, the system can be equilibrated by tfMC or, as it was proposed in the original work, using the fast inertial relaxation engine (FIRE) 12 . The high versatility of the method makes it ideally suited to explore growth phenomena for many systems. We

Read the complete text