Until lately, the field of material science has largely relied on trial-and-error lab experiments to determine a material’s properties. But this approach can’t handle problems such as predicting how long the steel in truck axles will hold up. Lab experiments in this area are impossible to complete in anyone’s lifetime. To solve the problems, an engineer must understand how carbon atoms are spaced among iron atoms in the steel, and how hydrogen atoms penetrate the structure as the material degrades. Fortunately, a new approach called computational materials science, developed at the Massachusetts Institute of Technology, generates numerical methods that use physics and chemistry principles to evaluate thousands of different variations in material composition.
In one example, MIT researchers applied the idea to axles on trucks that travel rough Siberian roads. An unexpected bearing failure on a remote stretch could literally endanger a driver’s life. Under the weight of the truck, the axles deform over time at an ever-increasing rate. “According to scientific theory, the exponent in the equations governing deformation should be three, but experiments conducted over decades always found it was actually four or five,” says MIT materials scientist Krystyn Van Vliet. “Computational techniques resolved the discrepancy, letting the truck company precisely analyze how steel degrades and find ways to slow or stop it.”
The approach can be used to put other materials under new scrutiny as well. MIT recently started a project to improve concrete’s properties and slash its carbon footprint. Researchers used computational techniques to decode, for the first time, the 3D structure of the basic unit of calcium-silicate-hydrate — the paste that forms and quickly hardens when cement powder and water mix.
Researchers caution that understanding detailed material properties still requires lab experiments — no computer models are ever perfect. But the guidance provided by modeling lets material engineers work more efficiently.