What is Computational Materials Science?

Computational materials science is an interdisciplinary field that enables more efficient materials discovery, materials design, failure characterization, and materials modeling in both fundamental research and product design. Computational materials science consists of a set of methodologies that enables engineers to investigate the material behavior and the properties of materials, such as mechanical, thermal, and electromagnetic properties.

Aside from the ability to understand new material design, computational materials science has enabled knowledge transfer through interdisciplinary research. Advances in computational materials science are now enabling more industrial sectors to design more efficient material systems and better-performing products without the need for multiple prototyping rounds.

Computational methods are used in many areas of chemistry and materials science research and development, including advanced materials, composites (ceramic, carbon, and polymer composites), and other solid-state materials. Computational materials science also extends to many technological applications, including energy production and semiconductors.

Today, computational materials science is continuously evolving with advancements in computing power and simulation software.

One of the most prominent trends in the past few years (that has now become an added feature) is the link to process modeling and how the manufacturing process affects the properties of the material. As more companies adopt digital manufacturing, computational materials science methodologies provide more capabilities to understand and improve these processes.

Another area that has seen a lot of development and use in recent years is multiscale modeling, which combines computational and engineering techniques to predict material properties and material behavior to be optimized across multiple length scales, from the atomic level to the macroscopic level. This is becoming a popular simulation approach when designing materials. It’s being combined with experimental characterization techniques, such as transmission electron microscopy (TEM) and scanning electron microscopy (SEM), to further material design from the nanoscale upward.

On the software side, one of the biggest current trends, like many areas today, is artificial intelligence (AI). Different AI algorithms can be used to provide better prediction and optimization of material properties and processes when fed with the relevant simulation (and experimental) data. This is a tool that is starting to mature in computational materials science and material informatics, where machine learning algorithms are used to predict material properties and improve the efficiency of materials development.

Computational materials science can be used to understand the structure and properties of materials at low length scales, including the atomistic and nanoscale levels.

Aside from modeling the material itself, high-throughput computational screening can take the known properties of yet-to-be-made materials ― based on their composition and crystal structure ― from large databases to identify the material/material structure with the ideal properties for a specific application. Because a potential material of interest often has multiple sought-after properties, this approach focuses on one property at a time, narrowing the number of potentially suitable materials over time as multiple properties are explored. This saves time and energy by not having to perform trial-and-error experimental approaches to find the best fit and can accelerate material development. However, these simulations are computationally expensive and require a lot of time or a computer with much processing power (or both).

Many computational modeling methods simulate at the atomistic level, and many variations of core methods have been adapted for specific materials and applications. Out of all the approaches, density functional theory (DFT), molecular dynamics (MD) simulations, and Monte Carlo simulations are most common:

• Density Functional Theory (DFT): DFT is a quantum mechanical model that simulates how electrons behave in a material based on their density. It is the most widely used electronic structure method for deducing chemical and energy properties — predicting the ground-state properties (lowest energy state of an atom) and mechanical properties of a material.
• Molecular Dynamics (MD): MD is a physics-based model that predicts how every atom in a simulation will behave and interact with other atoms, as well as give insight into how atoms physically move over time. The positions and stresses of atoms can be used to predict material properties, and machine learning potential is being integrated into MD simulations to improve their accuracy and reduce the computational cost of simulations.
• Monte Carlo: Monte Carlo simulations are computational algorithms that predict the chance of a result using repeated random sampling. It’s a probability-based tool that simulates particle interactions and complex systems.

Many computational tools also are used to predict material properties at much larger scales — throughout the many layers of a material rather than just at the atomic scale. These multiscale modeling approaches look at the macro properties of the material (mechanical, electromagnetic, etc.), investigate the microstructure, and see how the material behaves under extreme conditions (especially for demanding applications).

Modeling methods at larger scales fall under the umbrella of continuum level modeling and take the information gained at the smaller molecular scales to establish a connection with the larger material system. This sequential approach permits more precise but computationally expensive modeling to be done at the atomistic level, followed by less computationally intensive microscale modeling to be performed once the fundamentals are in place.

While many computational tools are available for full material system modeling, some common ones include finite element method (FEM), phase field method, and computer coupling of phase diagrams and thermochemistry (CALPHAD).

• Finite Element Method (FEM): FEM is a numerical-based simulation that splits up the complex material system into a mesh of smaller elements using differential equations, and each “element” can be investigated individually. Understanding how each element behaves enables engineers to understand how the whole material behaves under certain conditions. FEM is used for performing structural analyses of a material, as well as investigating mass transport, fluid flow, and the electromagnetic properties of a material. The results of an FEM can be interpreted by a finite element analysis (FEA).
• Phase Field Simulations: Phase field models are mathematical models that solve problems at material interfaces. They often are used to model the solidification and interfacial dynamics of materials and can be used to study mechanical properties, such as fracture mechanics and brittleness. They can also be used to build multiphase models that examine a material’s microstructure based on certain parameters, such as a crystallographic orientation or specific phase, and can investigate the properties of materials with multiple thermodynamic phases.
• CALPHAD: A model that produces phase diagrams to predict phase stability of a material at different temperatures and chemical compositions. These models use the thermodynamic properties of each phase in a material to perform the simulation and are often used to better understand how a material behaves under different conditions.

Like any simulation or computational method, there are advantages and disadvantages to using a computational materials science approach. The advantages and disadvantages can also differ based on which set of tools is chosen, because the right simulation approach for one material is not always a good fit for a different material. Despite the different computational tool sets that can be used, there are general advantages and disadvantages.

Opportunities of Computational Materials Science

• It can reduce the need for extensive experimental development — both material synthesis and testing — in the early stages of the product cycle by replacing trial-and-error approaches with targeted simulations.
• It can be used to probe highly specific properties, processes, fundamental mechanisms, material environments, and application scenarios that are difficult to achieve through experimentation.
• It is helping to bridge the gap between fundamental materials science and other industries and application areas.
• It enables many industries to have a better understanding of materials.
• It is helping end users to think outside the box and to consider materials as design parameters, leading to more complex materials being developed.

Limitations of Computational Materials Science

• Information gained about the structure of materials and their property relationships can sometimes be too generalized, whereas most properties and material parameters in experiments are often quantifiable to specific values.
• Due to the complexity of some of the mechanisms, there will be a need to perform experiments to have a proper understanding of material behavior. Phenomena such as failure and damage are very hard to capture only with computation.
• Simulations can be limited by the computational power available, leading to long simulation times.

Challenges of Computational Materials Science

• Return on investment (ROI) is unclear because it’s difficult to develop computational science methodology for a specific case.
• It is an expensive, time-consuming process that is advantageous only if the methodology is reused.
• It often needs specific interdisciplinary expertise to leverage the full benefits of the computational models.

As mentioned earlier, the use of AI algorithms is becoming commonplace in computational materials science, but how exactly does AI help engineers in the materials engineering and mechanical engineering fields?

AI is finding more use at the atomistic-to-microscopic level and is helping to improve the prediction capabilities of simulation software so more accurate material properties can be identified. Machine learning is also having a big impact on MD simulations by creating a similar level of accuracy as DFT. (DFT accuracy is typically higher.) Other AI applications are emerging — for instance, automated characterization analysis, self-driving labs, process optimization, and multiscale modeling.

Rapid advancements in commercial AI technologies mean that there are lots of options available, and the AI in use today is improving the accessibility to material sciences information. The use of AI in computational materials science is still in a transition phase, with a wider integration across more tools likely to be established when one of two things happens:

• There’s technical need to use it.
• It becomes more cost-effective for the materials science community to use in more areas of material design.

New materials are becoming inherently more complex to achieve advanced functionalities and characteristics, and it’s more important than ever to maximize all available tools during the various material design phases.

While Ansys, part of Synopsys, can help at the macroscale level, a range of simulation tools for material scientists can help with multiscale simulation of your material. It should be noted that when it comes to simulating the structure and properties of a material system, there is not a one-size-fits-all approach. Depending on the material and application in question, a wide combination of simulation tools can be used, from the internal tools used at Ansys to Python-based approaches.

Here are examples of Ansys tools that can be used during the material design stages:

• Ansys LS-DYNA nonlinear dynamics structural simulation software: LS-DYNA software can be used to provide multiscale simulations that combine the microscale with the macroscale.
• The Ansys Materials Designer tool: The Materials Designer tool can be used to explore material networks and look at the microstructure. The tool also provides a native add-on and a user-friendly front end for homogenization studies.
• The Ansys Granta materials information, selection, and data management product collection: You can use Granta software to manage material data and support various models during the design process. Granta AI plug-ins can also be used to improve data management and support.
• Ansys Minerva simulation process and data management software: Minerva software provides secure simulation and process data to ensure that there is traceability for all the simulations and data that is created throughout the material design process.
• Ansys Mechanical structural finite element analysis software: With Mechanical software, crystal plasticity and MAPDL can be used for multiscale simulations that combine the microscale with the macroscale. While testing representative elementary volumes (RVEs) with different phases, periodic boundaries can be emulated through multipoint constraints (MPCs).

Ansys will tailor the approach to meet the needs of your material design process using the expertise of our material and simulation engineers. To find out what the optimal mix of tools would be for your material systems, contact our engineering team to get a customized simulation solution.

Related Resources