In collaboration with Diamond Light Source we run the EPSRC funded CONEXS network for X-ray Spectroscopy. CONEXS will bring together experimentalists and theoreticians working in the area of X-ray spectroscopy to achieve new levels of understanding, especially for the interpretation of experimental data.
Progress in the theory of X-ray spectroscopy is rapidly developing. We are developing a new approach to speed up the analysis of X-ray spectra based upon a deep neural network [paper1, paper2]. A general schematic of the DNN used is given below. It takes the local environment around an atomic absorption site as input.
Progress in the theory of X-ray spectroscopy is rapidly developing. We are developing a new approach to speed up the analysis of X-ray spectra based upon a deep neural network [paper1, paper2]. A general schematic of the DNN used is given below. It takes the local environment around an atomic absorption site as input.
We have used this to estimate Fe K-edge X-ray absorption near-edge structure spectra in less than a second with no input beyond geometric information about the local environment of the absorption site. Our DNN, which is still in the development phase, can predict peak positions with sub-eV accuracy and peak intensities with errors over an order of magnitude smaller than the spectral variations that the model is engineered to capture (Examples are shown below).
Estimations drawn from the first centile of performance (upper 3 panels) cannot be distinguished from the target XANES spectra illustrating the performance of the network. Impressively, even the worst estimations, i.e. those drawn from the ninety-nineth centile, reproduce faithfully the spectral shapes of their target.
The performance of the DNN can also be illustrated by its application to the structural refinement of iron(II)tris(bipyridine) and nitrosylmyoglobin (below).
The performance of the DNN can also be illustrated by its application to the structural refinement of iron(II)tris(bipyridine) and nitrosylmyoglobin (below).
An important consideration when developing a deep neural network (DNN) for the prediction of molecular properties is the representation of the chemical space. In paper2 we have explored the effect of the representation on the performance of our DNN engineered and addressed the question: How important is the choice of representation for the local environment around an arbitrary absorption site? While Coulomb matrix (CM) and pair-distribution/radial distribution curve (RDC) featurisation are demonstrably robust descriptors, it is possible to obtain a smaller mean squared error (MSE) between the target and estimated XANES spectra when using RDC featurisation, and converge to this state a) faster and b) using fewer data samples.
A presentation some of our preliminary work in area can be found below:
A presentation some of our preliminary work in area can be found below:
A more general presentation on the theory of X-ray spectroscopy can be found: