Have you ever felt that you have spoken plainly to a partner about a material only to realize that you don't exactly mean the same thing? This is really common when industrialists, experimentalists, and modelers communicate. The first step to bridging the gap is to become aware of the communication challenge - something I started thinking about when I within two weeks attended the "Europe Materials Modeling Council workshop on interoperability" and the "Future of Materials Summit" in Luxembourg.
- What is the problem you say? Let us define a material based on the ideal - "diamond" is a perfect carbon lattice extending to infinity in all directions. But what is it is the imperfections, the defects, that give the material the properties we are looking for? If the defects are included in a statistical model then what about contaminants? Any atom from the periodic table could in principle be present in the crystal. Without knowing how we created the material how do we know what contaminants it is even relevant to include?
- Ok, you say - let us instead use the process. If we can describe exactly the input and the processes used, then we have defined the material. But even small imperfections in modeling combined with the "butterfly effect" for highly non-linear processes could translate inputs and processes to end products extremely difficult. Not to mention that this definition could lead to the weird situation that two samples with the same configuration of atoms should by called by different names.
- Well then, maybe we should just double down on sensing and define a material by the results of experimental characterization. The downside is that we will always be blind to what we do not measure, and there will still be something we do not measure - especially in non-destructive measurements. We cannot get a full desciption this way.
- Or you could argue we should just forget about the atomic arrangement - what is essential for applications is what the material can do - the performance. A performance definition of a material is a compelling idea, since one material with, for instance, a specific minimum yield strength could be a perfect substitute for another. But the problem is that this definition leaves too many things undefined. Imagine that we change a process and the material behaves the same in all the basic tests - yet suddenly handlers get sick from arsenic poisoning. Your definition, focusing only on the metrics you thought off, can lead you to ignore that a new process might have introduced poisonous contaminants.
That was just four ways! The last, which I recommend, is all-of-the-above. None of these descriptions paint the full picture.
Use the ideal to describe how your desired properties arise from the configuration of atoms. Use the process to explain how you arrive from feed materials to result. Materials modeling can then predict the structure. You can then test simulations of the performance and characterization data against empirical reality.
If you can't quite make the pieces fit together you still have some work ahead of you - but you can see where the discrepancy is. On the other hand, if after all your hard work, everything fits together snugly you have can tell a compelling story to standards authorities, collaborators, clients, and the general public.
Thanks to employees of ModuMetal, Siemens, Citrine, NanoMatch, Ocsial, Datastories and Materialise for inspiring talks and discussions over the last few weeks.