Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →In a 2016 computational study, Tobias Morawietz and colleagues modeled the melting point of ice using quantum-mechanical calculations accelerated by a neural network. The work was not a new experiment measuring water’s everyday freezing temperature, and the Chemistry World report did not give the calculation’s exact numerical result. Its significance was showing how a more efficient model could capture molecular behavior tied to water’s unusual density.
What “from scratch” means in this study
The phrase refers to a calculation grounded in quantum-mechanical modeling, rather than a measurement made by cooling a sample of water. Chemistry World’s 7 July 2016 report describes the research as calculating ice’s melting point using ab initio molecular dynamics. Melting and freezing are opposite directions across the same equilibrium phase boundary, though the report frames the modeled quantity as ice’s melting point.
As an Amazon Associate I earn from qualifying purchases.
The calculation was not an approximation-free simulation of every electron and molecule. The team, from the University of Vienna and Ruhr-University Bochum, trained a neural network to reproduce results from density functional theory (DFT), then used a van der Waals correction to represent weak intermolecular attractions that matter for water’s structure. The report characterized the approach as computationally more efficient than running DFT directly.
Why a faster model was needed
According to Chemistry World, conventional DFT-based molecular dynamics was costly enough to limit simulations to a few picoseconds, while studying the relevant behavior called for nanosecond-duration periods. DFT also did not accurately reproduce small van der Waals forces in this context. Those weak attractions can still affect how water molecules organize, so longer simulations alone would not solve the modeling problem if those interactions were represented poorly.
#1 Best Overall
The neural network offered a way to reproduce DFT results at lower computational cost, and the van der Waals correction addressed the interactions the report says were consequential. The trade-off is resolution: David Keffer of the University of Tennessee cautioned that the method exchanged a fine-grained approach for computational efficiency, which he described as “a soundly-based improvement.” That makes the model a useful approximation, not an exact or assumption-free account of water.
How the molecular picture relates to water’s density
Ice has an open hydrogen-bonded structure
Hydrogen bonds hold molecules in ice in a relatively open three-dimensional arrangement. When ice melts, some of those bonds weaken and molecules can pack closer together. In liquid water, this competition helps explain why density rises as ice melts and cools toward roughly 4°C, where the report says liquid water reaches its maximum density.
Liquid water’s molecular shells can rearrange
The report describes a balance between contraction of the nearest shell of molecules around a molecule and movement of molecules from the second shell into the first. Cooling strengthens the hydrogen-bond network and pulls the nearest shell closer, but the liquid can still contain “intruder” molecules in that shell. At lower temperatures, a more rigid network rejects them.
In the study’s account, correctly treating van der Waals forces gives the hydrogen-bond network enough flexibility for molecules to move between the second and first shells. Morawietz summarized that point by saying the network has the right flexibility only when those forces are correctly included. This molecular rearrangement is part of the explanation for water’s density behavior; it is not, by itself, a complete explanation of every property of water.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the report does—and does not—establish
Chemistry World identifies the original paper as T. Morawietz and colleagues, published in the Proceedings of the National Academy of Sciences in 2016 (DOI 10.1073/pnas.1602375113). The report does not state the exact computed melting point or a numerical uncertainty, so neither should be inferred from its headline. The study’s reported contribution is methodological: using a neural-network model trained on DFT results, together with a van der Waals correction, to make longer simulations of water’s molecular behavior more practical.
Quick Recap
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
