Dr. Chang was recently awarded a personal fellowship – a Newton Advanced Fellowship from the UK Royal Society. This fellowship provides funding support for Dr. Chang and his group for collaborating with Prof Richard Harrison group at the Department of Earth Sciences, Cambridge University, on a project entitled ‘A machine-learning approach to multiscale environmental magnetism’. This project will employ sets of cutting-edge nanoscale 2D/3D imaging techniques, machine-learning based data analyses, and computational modelling to develop new multiscale environmental magnetic analytical tools. The new tools will enable extracting key environmental signals from natural samples at an unprecedent level. Additional funding support for this project is provided by the National Natural Science Foundation of China (NSFC).
Below is a glimpse of some relevant work set out in this project:
Figure 1. Transmission electron microscopy (TEM) tomography of magnetic mineral assemblages (from Midgley and Dunin-Borkowski, Nat. Mater., 2009).
Figure 2. Focused ion beam nano-tomorgaphy (FIB tomographic reconstruction of a dusty olivine grain in a chondritic meteorite, image courtesy to Josh Elnsle, University of Glasgow).
Figure 3. (a) TEM 3D tomography of a single silicate (grey) extracted from deep-sea sediment containing titanomagnetite nano-inclusions (colors) using a tilt series of TEM images, (b) results of micromagnetic simulation of the yellow region in (a). Elnsle, Chang, Harrison, unpublished data acquired at Cambridge.
Figure 4. Machine learning-based image processing of magnetic mineral nanoparticle assemblages: (left) A bright-filed TEM image, (middle) processed image using Canny Edge Detection, (right) processed image using Holisticaly-Nested Edge Detection (Pei Chang, unpublished data).
Figure 5. Illustrations of a work flow on extracting magnetic signals from marine sediments. (a) TEM image of large number of magnetic mineral grains, (b) identification and statistics of the particle sizes extracted from image data, (c) building of the 3D micromagnetic model that may contain hundreds and thousands of particles, (d) results supporting the paleoenvironmental proxy (modified from Chang et al., Nat. Commun. 2018; Chang et al., EPSL, 2019).