SearcharxivSearch

arXiv subjects

Lars Kürten

Publications and source records attributed to Lars Kürten.

2 recordsLinked to original sources

The Free-Energy Barrier of Precritical Nuclei in Hard Spheres is Consistent with Predictions

Predicting crystal nucleation is among the most significant long--standing challenges in condensed matter. In the system most studied (hard sphere colloids), the comparison between experiments performed using static light scattering and computer simulations is woeful, with a discrepancy of over 10 orders of magnitude. The situation with other well-studied materials (such as water and sodium chloride) is no better. It has thus far proven impossible to access the regime of this discrepancy with particle-resolved techniques which might shed light on its origins, due to the relatively sluggish dynamics of the larger colloids required for confocal microscopy. Here we address this challenge with two developments. Our work is a marked improvement in the precision of mapping the state point of experiments to simulation. For this, we employed a combination of novel machine-learning methods for particle tracking and higher-order correlation functions. Our second innovation is to consider the free energy of pre-critical nuclei which can be detected in the discrepancy regime. These are in agreement with computer simulation. This is the first time that such free energies have been successfully compared between experiment and simulation in any material as far as we are aware. The agreement provides important validation of rare event sampling techniques which are used very widely in simulation, but which can seldom be directly compared with experiment.

cond-mat.soft

Colloidoscope: Detecting Dense Colloids in 3d with Deep Learning

Colloidoscope is a deep learning pipeline employing a 3D residual Unet architecture, designed to enhance the tracking of dense colloidal suspensions through confocal microscopy. This methodology uses a simulated training dataset that reflects a wide array of real-world imaging conditions, specifically targeting high colloid volume fraction and low-contrast scenarios where traditional detection methods struggle. Central to our approach is the use of experimental signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and point-spread-functions (PSFs) to accurately quantify and simulate the experimental data. Our findings reveal that Colloidoscope achieves superior recall in particle detection (finds more particles) compared to conventional heuristic methods. Simultaneously, high precision is maintained (high fraction of true positives.) The model demonstrates a notable robustness to photobleached samples, thereby prolonging the imaging time and number of frames than may be acquired. Furthermore, Colloidoscope maintains small scale resolution sufficient to classify local structural motifs. Evaluated across both simulated and experimental datasets, Colloidoscope brings the advancements in computer vision offered by deep learning to particle tracking at high volume fractions. We offer a promising tool for researchers in the soft matter community, this model is deployed and available to use pretrained: https://github.com/wahabk/colloidoscope.

cond-mat.soft