Statistics, damned statistics and nanoscience – using data science to meet the challenge of nanomaterial complexity
For many years dealing with the complexity of nanoscale materials, the polydispersivity of individual samples, and the persistent imperfection of individual nanostructures has been secondary to our search for novel properties and promising applications. For our science to translate into technology, however, we will inevitably need to deal with the issue of structural diversity and integrate this feature into the next generation of more realistic structure/property predictions. This is challenging in the field of nanoscience where atomic level precision is typically inaccessible (experimentally), but properties can depend on structural variations at the atomic scale. Fortunately there exists a range of reliable statistical methods that are entirely applicable to nanoscale materials; ideal for navigating and analysing enormous amount of information required to accurately describe realistic samples. Combined with advances in automation and information technology the field of data science can assist us in dealing with our big data, characterising our uncertainties, and more rapidly identifying useful structure/property relationships. Taking greater advantage of data-driven methods involves thinking differently about our research, but applied appropriately these methods can accelerate the discovery of nanomaterials that are optimised to make the transition from science to technology.