This is the most surprising and exciting result of my career: we were running simulations of NaCl with a neural network potential that implicitly accounts for the effect of the water, ie a continuum solvent model (trained on normal MD) when Junji noticed something strange: 1/n

🚨 BREAKING: Water’s coolest mystery, solved! 🌊 Our latest study, published in @NaturePhysics, provides the most realistic molecular picture of supercooled #water to date! 🚀 👉 Enabled by our MB-pol #datadriven #manybody potential, we have achieved, for the first time, CCSD(T)-level accuracy for microsecond-long simulations, allowing us to resolve one of the most debated questions in water research: Does supercooled #water have a liquid–liquid critical point?💧❄️ Main results: ✅ We predict a liquid–liquid critical point at ~200 K and ~1,250 atm, consistent with experimental extrapolations where direct measurements of water’s phase behavior are possible. ✅ Our simulations provide compelling evidence that water can exist in two distinct liquid phases at low temperatures and high pressures, resolving a three-decade-long scientific puzzle. ✅ These findings open the door for experimental validation, including potential measurements in nanodroplets. Why is this important? For over 30 years, the existence and precise location of this critical point have been hotly debated. Our MB-pol #datadriven #manybody simulations set a new benchmark for predicting water’s phase behavior across thermodynamic conditions and environments. 🏄‍♀️ Beyond fundamental science, our results pave the way for realistic simulations of water-mediated processes in: 🌧️ Atmospheric and environmental chemistry (e.g., cloud formation, climate modeling) 🌎 Geochemistry and planetary science (e.g., water in extreme conditions) 🧬 Biophysics and biochemistry (e.g., water’s role in biology) 🔬 Materials science (e.g., aqueous solutions in energy applications) We are very grateful to @AFOSR for funding this research through our "MURI: Unraveling the Mechanisms of Ice Nucleation and Anti-Icing through an Integrated Multiscale Approach", and to @DoD_HPCMP, @ACCESSforCI, and @SimonsFdn for providing #GPU and #CPU time! @UCSanDiego @UCSDPhySci @UCSDChemBiochem @HDSIUCSD @SDSC_UCSD

Overall very impressed compared to o1. Reminds me of talking to someone who knows a huge amount fairly shallowly, and is exceedingly overconfident. But clearly has passed the threshold of genuine usefulness.