Scientists claim to have found the two types of water that explain the liquid’s oddness

Despite being the most familiar liquid, water is weird. It breaks many of the usual rules that govern the liquid state. In 1992 a team of researchers suggested why that is. Perhaps, they said, there are two types of liquid water, which become distinct only at temperatures well below freezing point, where it’s all but impossible to keep water liquid.¹ Researchers have sought evidence for this bold conjecture ever since – and now an international team claims to have found it.

Anders Nilsson of Stockholm University and his co-workers have used ultrafast laser pulses to rapidly melt ice at temperatures and pressures close to those at which the two deeply supercooled liquid phases of water are thought to exist. They then used x-ray scattering to see a signature of the two liquids and the liquid–liquid phase transition between them – an abrupt (first-order) transition that, like the transition between a normal liquid and gas phase, ends in a critical point where the two phases become indistinguishable.²

‘The new results are the most persuasive evidence to date of a liquid–liquid critical point [LLCP] in water,’ says physicist Greg Kimmel of Pacific Northwest National Laboratory in Richland, Washington.

‘The LLCP is important because it is the source of the water anomalies,’ says Nilsson. It means that ordinary water at ambient conditions is in fact a supercritical liquid, existing at temperatures above the critical point of the two metastable (provisionally stable) supercooled liquid phases. One of these phases is denser than ordinary water because it lacks many of the hydrogen bonds between H₂O molecules that keep the molecules at ‘arm’s length’ in the other, lower-density liquid. The competition between a hydrogen-bonded, more open structure – which makes normal ice less dense than water itself – and the closer-packed structure is ultimately what creates anomalies such as water’s density maximum being 4°C above freezing.

The liquid–liquid transition (LLT) is predicted at temperatures below a LLCP at about -50°C and at a pressure of around 1000 atmospheres. But it’s virtually impossible to probe this region of the phase diagram without water freezing to ice. Several previous experiments have offered indirect hints of the LLT and LLCP^3-5, but nothing definitive. In collaboration with Kyung Hwan Kim of the Pohang University of Science and Technology in Korea and others, Nilsson’s group has been using the combination of rapid laser heating of amorphous ice – a disorderly form of ice that is like an ‘arrested liquid’ – and x-ray scattering to see signs of a LLT in the few microseconds before ice crystals began to nucleate and grow.⁴

‘If you want to move the state conditions of water close to the LLCP and make a measurement, you have to beat the crystallisation timescale,’ Nilsson explains. In the new experiments they can melt the ice in less than a nanosecond. Then the sample, held in a vacuum chamber, begins to expand and decompress. ‘If you probe it at various times during decompression, you can study the liquid at different pressures,’ says Nilsson. ‘All of this has to be performed within a few microseconds, otherwise you will only see ice.’ Physicist Francesco Sciortino of the Sapienza University of Rome, one of the team that predicted water’s LLCP in 1992, says that experiments like this ‘could not have been imagined even 10 years ago’.

Nilsson and Kim have now tweaked the experiments to get a more complete view of the phase diagram in this region. ‘The new results provide strong support for the LLT,’ says Kimmel. The researchers could also measure the heat capacity of their sample and found that this increased sharply, seemingly without limit, as the temperature approached about -63°C – exactly the kind of ‘divergence’ that signals a critical point.

‘This behaviour can only be interpreted by invoking an underlying free-energy landscape shaped by a critical point,’ says Sciortino. He cautions that there are still some experimental uncertainties – for example, the laser pulse ‘heats the sample in a way that is not fully controlled, and temperature inhomogeneities may develop’. But despite these issues, he says, ‘these beautiful experiments are fully consistent with the LLCP scenario’.