Machine Learning Nucleation Collective Variables with Graph Neural Networks DOI Creative Commons
Florian Dietrich, Xavier R. Advincula, Gianpaolo Gobbo

et al.

Published: June 30, 2023

The efficient calculation of nucleation collective variables (CVs) is one the main limitations to ap- plication enhanced sampling methods investigation processes in realistic environments. Here we discuss development a graph-based model for approximation CVs, which en- ables orders-of-magnitude gains computational efficiency on-the-fly evaluation CVs. By performing simulations on nucleating colloidal system, assess model’s both postprocessing and biasing trajectories, thereby mimicking multistep process from solu- tion. Moreover, probe transferability graph approximations across systems by using CV based sixth-order Steinhardt parameters. was trained with data collected system used drive crystalline copper its melt. Our approach general fully transferable more complex as well different

Language: Английский

Developing a Vial-Scale Methodology for the Measurement of Nucleation Kinetics Using Evaporative Crystallization: A Case Study with Sodium Chloride DOI Creative Commons
Michele Chen, Leif-Thore Deck, Luca Bosetti

et al.

Crystal Growth & Design, Journal Year: 2025, Volume and Issue: unknown

Published: April 4, 2025

Understanding nucleation kinetics is vital for designing crystallization processes, yet traditional measurement methods based on cooling are unsuitable compounds with temperature-independent solubility. This study introduces an experimental procedure to measure the evaporative and applies it sodium chloride (NaCl) in water. By systematically varying conditions such as temperature evaporation gas flow rate, we obtained a comprehensive data set of NaCl crystals that allowed estimating kinetic parameters using rate expression derived from classical theory (CNT). work demonstrates robustness method measuring applicable regardless how solubility compound depends temperature.

Language: Английский

Citations

1
Machine Learning Nucleation Collective Variables with Graph Neural Networks DOI Creative Commons
Florian M. Dietrich, Xavier R. Advincula, Gianpaolo Gobbo

et al.

Journal of Chemical Theory and Computation, Journal Year: 2023, Volume and Issue: 20(4), P. 1600 - 1611

Published: Oct. 25, 2023

The efficient calculation of nucleation collective variables (CVs) is one the main limitations to application enhanced sampling methods investigation processes in realistic environments. Here we discuss development a graph-based model for approximation CVs that enables orders-of-magnitude gains computational efficiency on-the-fly evaluation CVs. By performing simulations on nucleating colloidal system mimicking multistep process from solution, assess model's both postprocessing and biasing trajectories with pulling, umbrella sampling, metadynamics simulations. Moreover, probe transferability models across systems using CV based sixth-order Steinhardt parameters trained drive crystalline copper its melt. Our approach general potentially transferable more complex as well different

Language: Английский

Citations

16
Exploring Carbamazepine Polymorph Crystal Growth in Water by Enhanced Sampling Simulations DOI Creative Commons
Radost Herboth, Alexander P. Lyubartsev

ACS Omega, Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 15, 2024

In this work, the polymorphism of active pharmaceutical ingredient carbamazepine (CBZ) was investigated by using molecular dynamics simulations with an enhanced sampling scheme. A single molecule CBZ attaching to flat surfaces different polymorphs used as a model for secondary nucleation in water. novel approach developed compute free energy profile characterizing adsorption molecules orientation aligned crystal structure surface. The distribution states that showed alignment rescale include only contribution is consistent growth. resulting favorable thermodynamics most stable form, Form III and second I. primary crystallization product, dihydrate, found be less favorable, implying nonclassical pathway. We suggest major determining energetics hydrophobicity This thermodynamic ranking provides valuable information about pathways polymorph growth will further contribute understanding process CBZ, which imperative since formation can alter physical properties drug significantly.

Language: Английский

Citations

1
Machine Learning Nucleation Collective Variables with Graph Neural Networks DOI Creative Commons
Florian Dietrich, Xavier R. Advincula, Gianpaolo Gobbo

et al.

Published: Oct. 4, 2023

The efficient calculation of nucleation collective variables (CVs) is one the main limitations to application enhanced sampling methods investigation processes in realistic environments. Here we discuss development a graph-based model for approximation CVs, which enables orders-of-magnitude gains computational efficiency on-the-fly evaluation CVs. By performing simulations on nucleating colloidal system mimicking multistep process from solution, assess model's both postprocessing and biasing trajectories with pulling, umbrella metadynamics simulations. Moreover, probe transferability models CVs across systems by using CV based sixth-order Steinhardt parameters trained drive crystalline copper its melt. Our approach general potentially transferable more complex as well different

Language: Английский

Citations

1
Machine Learning Nucleation Collective Variables with Graph Neural Networks DOI Creative Commons
Florian Dietrich, Xavier R. Advincula, Gianpaolo Gobbo

et al.

Published: Oct. 6, 2023

The efficient calculation of nucleation collective variables (CVs) is one the main limitations to application enhanced sampling methods investigation processes in realistic environments. Here we discuss development a graph-based model for approximation CVs, which enables orders-of-magnitude gains computational efficiency on-the-fly evaluation CVs. By performing simulations on nucleating colloidal system mimicking multistep process from solution, assess model's both postprocessing and biasing trajectories with pulling, umbrella metadynamics simulations. Moreover, probe transferability models CVs across systems by using CV based sixth-order Steinhardt parameters trained drive crystalline copper its melt. Our approach general potentially transferable more complex as well different

Language: Английский

Citations

0
Machine Learning Nucleation Collective Variables with Graph Neural Networks DOI Creative Commons
Florian Dietrich, Xavier R. Advincula, Gianpaolo Gobbo

et al.

Published: June 30, 2023

The efficient calculation of nucleation collective variables (CVs) is one the main limitations to ap- plication enhanced sampling methods investigation processes in realistic environments. Here we discuss development a graph-based model for approximation CVs, which en- ables orders-of-magnitude gains computational efficiency on-the-fly evaluation CVs. By performing simulations on nucleating colloidal system, assess model’s both postprocessing and biasing trajectories, thereby mimicking multistep process from solu- tion. Moreover, probe transferability graph approximations across systems by using CV based sixth-order Steinhardt parameters. was trained with data collected system used drive crystalline copper its melt. Our approach general fully transferable more complex as well different

Language: Английский

Citations

0