Scientists use deep neural networks to map elusive liquid-liquid transition points

New natural physics studies shed light on long-standing liquid-liquid critical critical points where water is simultaneously present in two different liquid forms, opening new possibilities for experimental verification.

水は異常な特性で知られています。ほとんどの物質ではなく、水は液体状態で最も密なものであり、固体ではありません。 This leads to unique behaviors such as ice floating in the water.

However, research on liquid-liquid phase transitions (LLPT), which are assumed to occur in ultra-cooled regimes, faces challenges that researchers want to address.

Phys.org spoke with co-authors of the study, Professor Francesco Sialtino at Sapienza, Rome, and Professor Francesco Paesani at the University of California, San Diego.

“One of the long-standing hypotheses is that water is present in two different liquid phases, high density liquids and low density liquids, under extreme conditions, especially under extremely low temperatures and high pressures. It suggests that it can be done.”

Professor Sciortino said, “The point in which these two phases are indistinguishable is known as the liquid-liquid critical point. However, its experimental confirmation is that before these conditions are reached, the water is noted. It remains elusive due to the challenge of preventing it from freezing.”

Cooling pure water to -38°C will result in liquid form despite passing the freezing point to 0°C. This is known as a super cooled state.

In 1992, researchers first proposed that there could be a water liquid phase transition (LLPT) under -38°C at the supercooling point where water is present in two different liquid states or phases .

Professor Sciortino worked on the issue in 1992 as a postdoctoral student at Boston University.

The difficulty comes from what researchers call “No Man’s” land. This is an area where liquid water typically crystallizes instantly into ice before measurements can be made. This occurs below the -38°C supercooled critical point.

The inability to perform measurements in real time forced researchers to rely heavily on computer simulations to predict water behavior.

Previous studies have provided wide-ranging predictions for the location of the proposed liquid-liquid critical critical point (LLCP), with estimated critical pressures ranging from 36-270 MPa to -123°C to -23°C (or from 150-250 to k).

They pursued this study through a mix of curiosity and skepticism surrounding whether MB-POL can rigorously investigate the validity of two-liquid scenarios in deep cooled water.

Using deep neural networks

“Despite its accuracy, MB-POL is more computationally demanding than empirical models. To overcome this limitation, Sigbjørn Bore, the third author of this paper, has been founded in MB-PoL data. We developed the potential for trained deep neural networks (DNN@MB-PoL). “Professor Paesani explained the involvement of neural networks in their research.

Unlike previous water models, this approach stems from first-principles chemistry at the binding cluster level, which is considered the gold standard for molecular interactions.

Using the DNN@MB-POL model, the researchers performed microsecond long molecular dynamics simulations.

“These are important for studying water in deep cooling as molecular diffusion slows dramatically as temperatures drop. This slowdown will increase the system to reach metastable equilibrium.困難になり、関連するダイナミクスをキャプチャするために非常に長いシミュレーションが必要になります」とProfは説明しました。 .

Simulations were conducted at 280 different state points ranging from 20 temperatures (188–368 k or -85°C to 95°C) and 14 pressures (0.1–131.7 MPa).

All simulations were performed on a system of 256 water molecules under periodic boundary conditions.

Identifying phase transitions

When studying water at -85°C (188 k), the researchers observed dramatic density variations occurring on a microsecond timescale, with water reaching high and low density states at about 101.3 MPa.自然に切り替えました。

These observations confirmed the presence of a first-order phase transition between two liquid forms of water. This is a clear signature of such a transition, a barrier of increased free energy upon cooling.

Researchers were also able to construct a comprehensive phase diagram showing liquid-liquid coexistence curves.

“We are very confident in our estimated liquid-liquid critical critical points, as it is developed from first-principles quantum chemistry in coupled cluster-level theory (the gold standard for electronic structure calculations). ”

The results provide the strongest computational evidence for the presence of LLPT in water and help to solve scientific problems that have been persisting for over 30 years.

Researchers believe that water nanovertical (hanging in a nanometer wide water droplet or medium present in a confined space) can experimentally verify the results of LLPT.

「直径数ナノメートルのナノ増殖の場合、内圧は液体液体臨界圧力（〜1,250 ATM）に匹敵する値に達する可能性があります。これは、慎重に制御されたナノ樹液がLLCPをプローブするためのIt suggests that we can provide an experimental route,” the professor said. Pesani.

Professor Sciortino added, “Neutron and X-ray scattering experiments can be used to detect the structural signatures of two liquid states within these limited droplets.”

“Specifically, scattering techniques may reveal the density variations and correlations characteristic of important phenomena. Furthermore, time-resolved spectroscopy provides the dynamics of interconversion between two liquid phases.捉えるのに役立ちます。」

LLPT discoveries have a wide range of impacts on multiple scientific disciplines.

Understanding the behavior of water dual states improves climate modeling and weather forecasting, provides insight into the oceans of distant moons and planets, and enhances understanding of cellular processes driven by phase separation,エネルギー貯蔵と水処理における前進技術を高めます。

©2025 Science X Network