Rethinking the evidence for a liquid–liquid transition in water: What decompression experiments reveal

Abstract

The possibility of a liquid–liquid transition (LLT) in supercooled water has sparked decades of debate. Recent pump–probe experiments interpret two peaks in the structure factor S(q) during and after decompression of high-density liquid (HDL) as evidence of coexistence with low-density liquid (LDL). However, this interpretation presents a fundamental puzzle: such coexistence is implausible at ambient pressure, below the estimated location of the liquid–liquid critical point (LLCP). Here, we use decompression simulations with ML-BOP to reconcile this contradiction. Even when water decompresses along the LLT, S(q) retains a single peak because HDL and LDL domains remain nanoscopic. We explain the two-peak S(q) observed experimentally as a single evolving liquid peak superimposed on a slower to respond, colder HDL arising from the temperature gradient across the sample. The simulations reveal that the decisive LLT signature is a transient growth and decay of the apparent correlation length ξ at low q, which emerges only when decompression proceeds along the LLT, with maximum ξ near the LLCP. Importantly, ξ remains low when decompressing from T ≥ Tc, or too rapidly. The experimental signatures could be explained by an exponential pressure drop to the LLT in ∼10 ns, the growth of ξ as LDL domains develop, peaking near the LLCP at ∼50 ns, and subsequent entry into the single-phase regime, from which crystallization proceeds. Our findings resolve the contradiction between the LLCP location and structural signatures, identifying the low q region of S(q) evolution—not peak splitting—as the key structural marker of the LLT in water.