Research introduction
Everyday materials such as rubber, paint, clay, yoghurt, foam, detergent, creams, gels and plastics are well-known examples of soft-condensed materials.
What they have in common is the fact that the material properties vary between simple elastic solids and simple viscous liquids. These complex material properties are not directly caused by their atomic or molecular constituents, but rather result from mesoscopic structures that emerge spontaneously due to the systems interactions and thermal fluctuations.
Hence, a common theme for studying soft-matter with computers is to build models that describe the geometry of the relevant structures and their dynamics. We take great care to make computationally effective models that disregard irrelevant details, but capture the relevant physics of the systems. This process is called coarse-graining and is a tool that emerges out of statistical mechanics.
Experimentally, it is relatively easy to measure macroscopic material properties, but the molecular structure of the samples are almost never known. This makes it difficult to test theories based on experimental data. To address this problem we develop computer models. For instance, we build rubber models molecule by molecule and deform these using a supercomputer. Such simulations provide details insights into how the elastic properties of rubber emerges from a knotted molecular network formed by the cross-linked polymers.
Simulations produce data, but data in itself are not providing unique knowledge. Based on fundamental theoretical physics, we develop analysis methods and implement these as software. This allows us to analyse the simulation data, thereby converting data into valuable knowledge and insights into the physics of these materials. These insights allow us to critically test and improve theories. We also generate data corresponding to a virtual experiment performed on a model. Analysing virtual experimental data using the same methods as real experimental data, we can learn whether the experimental analysis methods can be trusted, and how to improve them.
From a practical perspective, we typically use different simulation techniques such as Molecular Dynamics or Monte Carlo simulations depending on what is computationally most efficient. Typically, simulation codes are written in C++, and analysis code uses scripting languages such as Python, Perl, Matlab or Mathematica. Some simulations run on a single server, others require high-performance computational facilities and we run these on various supercomputers. Hence in practice, most of the time we implement new algorithms and write code as a tool to do physics.