publications

Due to recent technological advances, the creation of 3D objects is becoming more important. It is critical to offer appropriate and accessible tools to users from diverse backgrounds. Unfortunately, the tools traditionally used in 3D creation are designed for professionals, require complex and time-consuming training, and are unsuitable for inexperienced users who form the vast majority of potential users. We propose a simplified creation tool that uses sketch-based methods. First, the desired mesh is created from a traced outline. The interior is inflated following the method of Dvoroznak et al. Second, the height (displacement) of the mesh is achieved by altering the strips created by shading. Shading is the result of a Lambertian reflection model for a given directional light. Our method consists of using the strips from the method of characteristics applied to solve Shape-From-Shading. Using the strips, we identify the regions affected by the change in shading. One of these regions will be used to interpolate the shading according to the new isophote. From this new shading, the characteristic strips will be used to find the new height, ensuring that only the region affected by the change in shading is altered. The meshes created can then be combined, inspired by the method proposed by Dvoroznak et al. to form a single, complex mesh. Our tool is designed to be more intuitive than the ones provided by professional 3D software. Our results illustrate its potential.

In the last few years, much effort has gone into developing general machine-learning potentials capable of describing interactions for a wide range of structures and phases. Yet, as attention turns to more complex materials, including alloys and disordered and heterogeneous systems, the challenge of providing reliable descriptions for all possible environments becomes ever more costly. In this work, we evaluate the benefits of using specific vs general potentials for the study of activated mechanisms in solid-state materials. More specifically, we test three machine-learning fitting approaches using the moment-tensor potential to reproduce a reference potential when exploring the energy landscape around a vacancy in Stillinger–Weber silicon crystal and silicon–germanium zincblende structures using the activation-relaxation technique nouveau (ARTn). We find that a targeted on-the-fly approach specific to and integrated into ARTn generates the highest precision on the energetics and geometry of activated barriers while remaining cost-effective. This approach expands the types of problems that can be addressed with high-accuracy ML potential.