Large-scale visualization with machine learning of dislocation networks in colloidal single crystals

Understanding the formation and evolution of dislocation networks and their effect on the mechanical properties of solids is challenging, as it spans a vast hierarchy of length and time scales. Hard-sphere colloidal suspensions provide a unique model system to address these difficulties.

To visualize the large-scale collective dislocation dynamics in colloidal single crystals, we developed a laser diffraction imaging technique inspired by the classical TEM imaging methods in atomic systems. We harness convolutional neural networks to address the inverse problem of inferring the structure of the dislocation network from the multiple complex diffraction images. The training process, however, imposes a significant problem as experimental data is sparse. This difficulty is bypassed by generating a training dataset composed of multiple simulated random dislocation structures. We will show that the trained neural network reliably reconstructs the spatial positions of the dislocations and their Burgers vectors by testing the results with in-situ confocal microscopy data on a smaller portion of the sample. We demonstrate the success of our method by applying it to the network of misfit dislocations formed during particle sedimentation on a strained substrate.