Supercooled Water Anomalies

Speculative If liquid water below its freezing point is really two liquids — a low-density form and a high-density form separated by a first-order line ending at a critical point nobody has reached — then a long list of otherwise unrelated oddities in the most measured substance on Earth collapses into one mechanism. The condition goes first because the critical point has never been observed in bulk liquid water; section 7 says what would locate or exclude it, and section 5 says why the measurement is hard.

Established The magnitude is best stated as the size of the hole in the phase diagram. Liquid water has a density maximum at 3.983035(670) °C, where its density is 0.99997495(84) g mL⁻¹; demineralised bulk water has been supercooled to −48.3 °C, or 224.8 K; and glassy water, whose transition sits near 136 K, can be warmed to about 150 K before it crystallises. Between roughly 150 K and the homogeneous freezing limit near 231 K, experiments on ordinary water find only crystal ice. That is an 80-kelvin band of the equilibrium behaviour of the best-characterised liquid in science in which no direct measurement of the liquid exists, and the whole argument is about what is inside it.

Frontier What a resolution buys, in the units the users actually work in, is a physical basis for numbers currently produced by extrapolation. Atmospheric models carry supercooled liquid water in mixed-phase clouds and need homogeneous and heterogeneous nucleation rates as functions of temperature and water activity; cryobiology needs the same rates to decide whether a sample vitrifies or crystallises; planetary science needs the equation of state of amorphous and supercooled water over ranges no laboratory has reached. Each of those is currently a fit. None of them is currently wrong in a way anyone has demonstrated, which bounds the prize honestly.

Handwave The gain accrues to modellers and to a small experimental community, and the condition is that a measurement can be made faster than ice forms, at more than one pressure. There is no product at the end of this page. What there is instead is the possibility that a textbook substance turns out to have a phase transition in it that nobody has seen, which would be an unusual thing to have been wrong about for a century. That possibility is worth stating precisely, and this brief was written under research conditions severe enough that a large part of its job is saying which parts of the standard account it could verify and which it could not.

Established Water is anomalous in the strict sense that it does several things the other simple liquids do not, and the best-measured of them is the density maximum. Liquid water is densest at 3.983035(670) °C rather than at its freezing point, which is why lakes freeze from the top and why the layer at the bottom of a deep temperate lake sits near 4 °C all winter. That single fact is the anchor of the whole subject: a liquid whose density rises on heating over some range is telling you that something in its structure is expanding as it cools.

Frontier The standard account attributes that to hydrogen bonding producing two competing local arrangements, and the anomalies then follow from their competition. One arrangement is open and tetrahedral, four neighbours held at arm’s length; the other is denser, with an interstitial neighbour pushed into the first shell. Cooling favours the open one, which is why the liquid expands; compression favours the dense one. Everything interesting in this subject is a consequence of asking whether those two arrangements are merely two ends of a continuum or two genuinely distinct liquids.

Speculative The strong version of that question is the second-critical-point hypothesis, and it is the reason this subject is in the pre-paradigmatic category. The proposal is that the low-density and high-density liquids are separated by a first-order transition line terminating at a critical point, and that the anomalies observed at ordinary temperatures are the tail of critical fluctuations emanating from a point deep in the supercooled region. The hypothesis is usually traced to a 1992 simulation study by Poole, Sciortino, Essmann and Stanley; this session could not retrieve that paper, so no number from it is quoted anywhere in this brief.

Speculative Competing accounts exist and have never been eliminated. The field’s own taxonomy names at least three rivals to the critical-point picture: a retracing spinodal in which the liquid’s limit of stability turns upward into the supercooled region, a singularity-free scenario in which the anomalies come from anticorrelated volume and entropy fluctuations with no transition at all, and a critical-point-free scenario in which the transition line runs to negative pressure without terminating. This brief states that taxonomy as the field’s own and flags it accordingly, because the primary sources for it were not retrievable in this session.

Established The reason the question has stayed open is a region of the phase diagram that experiments cannot hold open. Cool ordinary water and it freezes; warm glassy water and it crystallises. Between about 150 K, the highest temperature to which glassy water can be warmed without nucleation, and the homogeneous freezing limit near 231 K, experiments on bulk water find only crystal ice. The field calls that band no-man’s land, and the postulated critical point is inside it.

Established Solid amorphous water is real, comes in more than one density, and is the field’s proxy for the liquids it cannot reach. Low-density amorphous ice is made routinely by depositing water vapour onto a smooth cold metal surface below 120 K, and a high-density amorphous form was shown in 1994 to be produced by vapour deposition onto surfaces below 30 K. Heating high-density amorphous ice at 0.4 K per minute under 0.81 GPa crystallises it to ice IV at about 165 K. If the two amorphous solids are the glasses of two distinct liquids, the liquid-liquid transition exists; if they are not, the inference fails, and that inference is load-bearing for most of the evidence in section 3.

Handwave The count of anomalies that circulates in this field — seventy-odd, sometimes seventy-four — could not be verified for this brief and should be treated as rhetoric rather than as a measurement. The commission for this page gives it as “more than seventy measured anomalies”. No peer-reviewed source retrieved in this session states a count. The number appears to originate in a single privately maintained compilation, which this session could not reach; and a count of anomalies is in any case a count of how a list was partitioned, not a property of water.

Established This brief was written under a research pass that failed on most of its primary literature, and the first duty of this section is to say so precisely rather than to write around it. The session’s web-search budget was exhausted before the commission began, and the principal publishers of supercooled-water physics — the journals carrying the free-electron-laser droplet experiments and the simulation work on the critical point — returned permission or HTTP errors to every attempt. What follows separates what was verified from a retrieved source, what is named but unread, and what could not be checked at all. Every number printed below came from a document this session actually opened.

Established The anchor’s density figure is correct and can be stated more precisely than the commission stated it. The temperature of maximum density is 3.983035(670) °C and the density there is 0.99997495(84) g mL⁻¹. The commission’s 3.98 °C is right to the digits it gives.

Frontier The anchor’s no-man’s-land boundaries are wrong at the bottom by about thirty kelvin, and the error matters because it halves the size of the unknown region. The commission gives 180–235 K. The lower boundary of no-man’s land is set by the temperature to which glassy water can be warmed before it crystallises, which is about 150 K, and the retrieved statement of the inaccessible band is that between 150 and 231 K experiments find only crystal ice. The region is therefore roughly 80 K wide, not 55, and the postulated critical point sits inside the wider band. Nothing between 150 and 180 K is more accessible than the rest of it.

Frontier The upper boundary is not a single number either, and the sources disagree with each other in a way that is itself informative. One retrieved reference gives the homogeneous nucleation point of water at standard pressure as about 231 K and another, on the same site, gives almost 224.8 K, quoted as the deepest supercooling achieved in chemically demineralised water. A third retrieved measurement takes the bulk homogeneous limit as about −38 °C, or 235 K. These are not contradictions so much as different quantities: the temperature at which a given sample of a given size freezes within a given observation time. Homogeneous nucleation is a rate, and quoting it as a temperature hides the sample volume and the clock.

Established The clearest demonstration that the freezing limit is a rate and not a boundary came from droplets a few nanometres across. Water confined in pores as small as 2 nm, surrounded by octane, froze at about −41 °C, below the bulk homogeneous limit taken there as about −38 °C. The transformation temperature showed a pronounced length-scale dependence below 10 nm; Laplace pressures reached about 500 bar, reducing the chemical-potential difference between ice and water and raising the nucleation barrier; and the soft oil-water interface deformed into ripples that flattened the curvature contribution, suppressing nucleation rates by over two orders of magnitude at the smallest sizes.

Frontier That result cuts both ways for this brief and the honest reading is the awkward one. It shows that no-man’s land can be entered, by shrinking the sample. It also shows that the sample that enters it is under hundreds of bar of Laplace pressure with a deformable interface, which is not bulk water. Every technique that reaches deep supercooling buys the access with a confinement, a pressure or a clock, and the question of whether the thing being measured is still the substance of interest is not rhetorical.

Established Ice nucleation from the vapour has now been watched molecule by molecule, and at 102 K it has no barrier at all. In-situ cryogenic transmission electron microscopy at about 102 K and 10⁻⁶ Pa, with millisecond temporal and picometre spatial resolution, followed heterogeneous ice nucleation through an adsorption-mediated pathway in which amorphous ice forms first. The initial growth velocity was 4.7 × 10⁻² nm s⁻¹ and nucleation rates reached 4.6 × 10⁷ cm⁻² s⁻¹, three to four orders of magnitude above atmospheric supercooling studies; statistics on more than a hundred individual nuclei found no critical nucleus size; hexagonal and cubic ice competed, with crystallites reaching about 50 nm in roughly an hour.

Frontier Simulation has located the surface as the place where crystallisation starts, and has named a structure nobody expected. A 2024 study of water’s free surface reports that nucleation begins about 1.5 nm inside the surface rather than at the interface itself, driven by surface tension creating a negative-pressure zone that favours precursor structures containing five-, six- and seven-membered rings rather than the six-membered rings of ordinary ice. Nucleation rates in thin films exceeded bulk rates, with the enhancement increasing as the film thinned. The same work argues that the choice of order parameter used to identify crystal precursors significantly changes the estimated rate, which is a methodological warning aimed directly at the rest of the field.

Established The strongest recent evidence for two distinct amorphous states is a 2026 machine-learning analysis, and its result is the one the two-liquid picture needs. Applying a data-driven framework to low-density and high-density amorphous ice, the authors report that phase identity is encoded within the first coordination shell — specifically in the local hydrogen density — rather than in global descriptors, that the transition between them shows no evidence for intermediate structures and is consistent with a first-order-like transition, and that local hydrogen density reveals pronounced structural hysteresis between compression and decompression that orientational order parameters do not show. Two of the five authors are long-standing proponents of the critical-point hypothesis, which is worth stating rather than hiding.

Speculative That result is about solids, and the inference to liquids is the step where the argument becomes an argument. A first-order-like transition between two amorphous solids is not the same claim as a first-order transition between two liquids; it becomes one only if each amorphous solid is the glass of a distinct liquid, and no experiment has equilibrated either liquid. The hysteresis reported in the same study is consistent with a first-order transition and also with the kinetics of a strongly non-equilibrium compression, and the study does not discuss the liquid-liquid transition.

Established A September 2026 result adds a specific and underappreciated hazard to the whole ultrafast-heating method the field relies on. Amorphous solid water films about 100 nm thick, held at 100 K on a platinum substrate and heated by ultrafast laser pulses that drove the metal to peak temperatures of 700–1200 K, showed almost no structural change within 10 ns. The explanation offered is an interfacial vapour film about 6 nm thick: with the ice at 0.6 W m⁻¹ K⁻¹ against platinum at 72, and an interfacial thermal conductance of about 130 MW m⁻² K⁻¹, the substrate simply fails to deliver its heat into the ice.

Frontier The significance of that for this brief is direct: experiments that infer a transient liquid by heating amorphous ice must know what temperature the ice actually reached. A vapour layer at the interface is a thermometer error of unknown size in exactly the experiments that are used to argue for a liquid-liquid transition. This brief could not retrieve those experiments and therefore cannot say whether the hazard applies to any of them; it can say that the hazard is newly documented and that nothing in the retrieved literature rules it out.

Established Meanwhile the crystalline side of water’s phase diagram is still producing new phases, which is a useful measure of how much is left to find. A first-order transition to ice XXII, an orthorhombic Pbcm structure described as the densest H₂O polymorph experimentally observed, was reported above 308(5) GPa at room temperature in 2026. In the same year a commentary described five different crystallisation pathways of high-density ice at room temperature and a previously unknown metastable phase, ice XXI, with a body-centred tetragonal structure. Water in 2026 is a substance whose ordered phases are still being counted.

Established The commission’s second anchor — a 2020 X-ray scattering experiment read by its authors as a liquid-liquid transition — could not be verified in this session, and the associated decisive experiment appears to have been attached to the wrong result. The commission specifies the decisive experiment as femtosecond X-ray probing of evaporatively cooled droplets. From a retrieved 2023 Nature Reviews Physics review of atmospheric ice nucleation, the droplet line of work is dated to a 2015 paper by Laksmono and colleagues, whose title reports anomalous behaviour of the homogeneous ice nucleation rate in no-man’s land, and to a 2017 paper by Kim and colleagues in Science 358, 1589–1593, whose title reports maxima in the thermodynamic response and correlation functions of deeply supercooled water. Those are bibliographic facts read from a retrieved bibliography; neither paper itself was retrieved, and no finding from either is asserted here beyond what its title states.

Frontier The consequence is that the commission’s anchor and its decisive experiment are probably two different experiments. A droplet cooled by evaporation is a droplet at essentially one pressure; a 2020 result described as reading a transition, rather than a response-function maximum, is doing something the single-isobar droplet technique cannot do. This brief cannot confirm which experiment the commission meant, and rather than guess it records the mismatch and rebuilds the decisive experiment from first principles in section 7.

Handwave What this section cannot do is adjudicate the interpretation, and it will not pretend otherwise. The 2020 experiment, the simulation literature on the critical point in realistic water models, the long dispute over whether the transition seen in one model was a liquid-solid transition misread as liquid-liquid, and the 2023 phase-separation experiments on low-density amorphous ice are all named in the reading list and none was retrieved. A reader wanting the state of that argument should go to those papers directly. What this brief offers instead is a corrected map of the region, an audit of the anchor, and a decisive experiment that does not depend on trusting any of the contested results.

Speculative Question one: is there a second critical point in liquid water, and if so where? This is the question the subject exists for. It has a sharp form — a location in the temperature-pressure plane — and the honest position from the retrieved evidence is that no coordinates have been established by experiment on bulk liquid water, because no bulk liquid has been equilibrated inside no-man’s land.

Speculative Question two: are the two amorphous solids the glasses of two liquids? The 2026 finding that phase identity between low-density and high-density amorphous ice is encoded in the first coordination shell, with no intermediate structures and a transition consistent with first-order behaviour, is the strongest available support for a two-state picture. It is a statement about solids. Whether each solid is the arrested form of a distinct liquid is an inference that the experiment does not make and that the field has argued about for thirty years.

Frontier Question three: how much of the observed anomaly is critical and how much is just hydrogen bonding? The singularity-free scenario says the anomalies follow from anticorrelated fluctuations in volume and entropy with no transition anywhere. This is not a fringe position; it reproduces the qualitative behaviour, and distinguishing it from a critical scenario requires quantitative response functions along more than one isobar, which is the measurement section 7 asks for.

Frontier Question four: does confinement preserve the physics or replace it? Nanoconfinement is the main route into no-man’s land, and the 2021 nanodroplet work shows exactly what it costs: a 2 nm droplet carries Laplace pressures of order 500 bar and sits against a deformable interface that suppresses nucleation by more than two orders of magnitude. Results from confined water are evidence about confined water; the extrapolation to bulk is an assumption the field makes routinely and tests rarely.

Speculative Question five: is the nucleation problem separable from the structure problem? The 2024 surface study locates nucleation 1.5 nm below the free surface and attributes it to negative pressure favouring precursors with five-, six- and seven-membered rings, and the 2025 cryo-TEM work finds no critical nucleus at 102 K. If the structures that nucleate ice are the same structures the two-liquid picture invokes, then nucleation and the liquid-liquid question are one problem and the field has been treating them as two.

Handwave Question six: how much of the experimental record is thermometry? The September 2026 vapour-film result shows that an amorphous ice film on a metal substrate can stay structurally unchanged for 10 ns while the substrate is driven to 700–1200 K. Any experiment that infers a transient liquid state from a heating protocol depends on knowing the sample temperature during the measurement. This is a question about the field’s own record that the field has not obviously asked.

Handwave Question seven: does any of this generalise, or is water sui generis? Polyamorphism and two-state behaviour have been claimed for other network formers — silicon, silica, phosphorus. If water’s second critical point is real, it is a member of a class; if it is not, the class may be empty. The sibling brief on Glass Transition Theory reaches an adjacent version of the same worry from the other direction: whether one theory can cover glass-formers that share no chemistry.

Established The binding constraint is a clock, and it belongs to ice rather than to the experimenter. Inside no-man’s land liquid water crystallises faster than any conventional measurement can equilibrate it, which is what makes the band a no-man’s land rather than a hard region. Every technique in this subject is a way of buying time: shrink the sample so nucleation is rare, cool so fast that the liquid outruns the nucleus, or heat a glass so fast that it becomes liquid before it becomes ice.

Frontier The second bottleneck is that the interesting axis is pressure and the accessible techniques are isobaric. A critical point is a point in a two-dimensional plane. An evaporatively cooled droplet in vacuum is at one pressure. A measurement along a single isobar can find a maximum in a response function; it cannot by itself distinguish a critical point from a smooth crossover, because a maximum along one line is what both scenarios predict. Locating or excluding a critical point requires varying pressure at fixed deep supercooling, and the techniques that give deep supercooling mostly do not give pressure control.

Established The third bottleneck is that confinement changes the answer by a measurable amount. The 2 nm droplet result puts numbers on this: length-scale dependence of the freezing temperature below 10 nm, Laplace pressures of order 500 bar, and over two orders of magnitude of nucleation suppression traced to interfacial deformation. Confinement is not a neutral window onto bulk behaviour; it is a different thermodynamic system whose relationship to the bulk has to be argued case by case.

Frontier The fourth bottleneck is thermometry during ultrafast heating, and it is newly documented. A 100 nm amorphous solid water film on platinum, with the metal driven to 700–1200 K, showed minimal structural change within 10 ns because an interfacial vapour film about 6 nm thick limited the interfacial thermal conductance to about 130 MW m⁻² K⁻¹. Where the sample temperature is inferred rather than measured, the inference now has a named failure mode.

Speculative The fifth bottleneck is that the simulation evidence is model-dependent and the models are fitted to ambient water. Classical water models are parameterised against properties measured near room temperature and then run at 180 K and kilobar pressures, far outside their fitting range. Machine-learned potentials trained on electronic-structure data relax that objection and introduce a different one, since the reference method itself has to be right about hydrogen bonding in a regime where it has not been tested. This brief could not retrieve the primary literature on either, and says so rather than characterising it.

Handwave The sixth bottleneck is the order parameter. The 2024 surface-nucleation study reports that the choice of order parameter for identifying crystal precursors significantly changes the estimated nucleation rate. The same problem afflicts the liquid-liquid question in a sharper form: two liquids are distinguished by a structural criterion, the criterion is chosen by the analyst, and the 2026 amorphous-ice work found that the discriminating information sits in the first coordination shell where several conventional parameters do not look. A dispute about whether two states exist is partly a dispute about what counts as a state.

Handwave The seventh bottleneck is access, and this brief is an instance of it. The experimental record of this subject lives behind publisher paywalls at a handful of journals, and its key facilities are a small number of free-electron lasers with competitive beamtime. A research pass that cannot open the papers cannot audit the field, and a field whose central claims rest on a few unreplicated facility experiments is unusually exposed to that.

Frontier The first dependency is a pressure-resolved measurement inside no-man’s land. Everything else is secondary. Until response functions can be measured at more than one pressure at temperatures between 150 and 231 K, on a sample large enough that Laplace pressure and interfacial deformation are not doing the work, the critical point can be argued for and against indefinitely.

Established The second dependency is a temperature measurement during the experiment rather than a temperature inference after it. The 2026 vapour-film result makes this concrete: in one configuration a substrate at up to 1200 K left a 100 nm ice film structurally unchanged over 10 ns. Any protocol that reconstructs a thermal history from a model needs a measured check, and the field has one available in principle through the temperature dependence of the scattering itself.

Speculative The third dependency is an agreed structural criterion for what makes two liquids two. The 2026 amorphous-ice analysis found the discriminating information in local hydrogen density within the first coordination shell, and found that orientational order parameters hid the hysteresis entirely. A field that cannot agree on the order parameter cannot agree on the count of phases, and this is a convention problem before it is a physics problem.

Frontier The fourth dependency is nucleation rates that hold across sample size. The measured rates now span an enormous range — 4.6 × 10⁷ cm⁻² s⁻¹ for adsorption-mediated nucleation at 102 K, three to four orders above atmospheric supercooling studies, with no critical nucleus observed — and the atmospheric community needs rates as a function of temperature and water activity for use in models. Reconciling those is a prerequisite for treating any deep-supercooling experiment as a measurement of the liquid rather than of its destruction.

Handwave The fifth dependency is bibliographic and this brief is the evidence for it. A subject whose central experimental claims sit behind paywalls at three or four journals, and whose most-cited anomaly count originates in a privately maintained web compilation, is a subject whose public record is weaker than its scientific record. Open deposition of the facility data would cost the field nothing and would let an audit like this one be worth more than it is.

Frontier The commissioned decisive experiment has to be rebuilt before it can be ranked, because as stated it cannot return a decisive answer. The commission asks to locate or exclude a second critical point in no-man’s land by femtosecond X-ray probing of evaporatively cooled droplets. Evaporative cooling in vacuum fixes the pressure. A critical point is a point in the temperature-pressure plane, and a single-isobar scan through a region containing one will show a maximum in the compressibility whether the point is there or not, because a maximum along a line through a critical region and a maximum along a line through a smooth crossover look alike. The droplet technique is how the region became measurable; it is not by itself how the question gets settled.

Speculative The decisive experiment for this brief is the two-dimensional version: measure the isothermal compressibility and the isobaric heat capacity of bulk supercooled water at several distinct pressures, at temperatures inside the 150–231 K band, faster than ice nucleates. The signature that discriminates is not a maximum but the behaviour of the locus of maxima as pressure is varied: a critical point requires the maxima in different response functions to converge onto a single line that terminates, with the peak height growing as the point is approached; a singularity-free scenario requires the loci to remain separate and the peaks bounded. One isobar cannot see the difference. Three or four can.

Speculative This experiment does not exist and its hardest requirement is a sample that is simultaneously deeply supercooled, under controlled pressure, and large enough to be bulk. Each pair of those is achievable today and the triple is not, which is a more useful statement of the problem than any restatement of the theory. It is also why this brief gives the experiment no schedule: nobody has proposed it as a programme, and it needs an instrument that has not been built.

Frontier Second in rank, and cheap: a thermometry audit of the ultrafast-heating experiments. Repeat the heating protocols used to infer transient liquid water from amorphous ice, on films of varied thickness and on substrates of varied thermal conductance, with the sample temperature determined from the scattering rather than from a model of the heat flow. The September 2026 result gives the specific thing to look for: an interfacial vapour film of order 6 nm limiting conductance to about 130 MW m⁻² K⁻¹, which in one configuration left a 100 nm film unchanged for 10 ns under a substrate at up to 1200 K. If the hazard is present in the liquid-liquid experiments, it is the most consequential correction available in this subject; if it is absent, that is worth establishing and takes one campaign.

Speculative Third: the confinement ladder, run as a controlled series rather than as a set of separate results. Measure the same observable in droplets from 2 nm to micrometres, correcting explicitly for Laplace pressure, which reaches about 500 bar at the small end, and for interfacial compliance, which the 2021 work shows suppresses nucleation by over two orders of magnitude. The output is the function that lets confined measurements be read as statements about bulk water, and its absence is why the confined-water literature has not settled anything.

Frontier Fourth: an order-parameter round robin on the amorphous ices. The 2026 analysis found that phase identity between low-density and high-density amorphous ice sits in local hydrogen density within the first coordination shell, that the transition shows no intermediate structures, and that orientational order parameters missed the compression-decompression hysteresis entirely. Distributing the same configurations to groups using different criteria, and comparing what each concludes about the number of phases, would separate the physics from the convention. This is the water analogue of the viscometry round robin that Glass Transition Theory asks for, and it has the same justification: when a claim depends on an analysis choice, the analysis choice has to be measured.

Speculative Fifth: nucleation-rate reconciliation across nine orders of magnitude. Adsorption-mediated nucleation at 102 K was measured at 4.6 × 10⁷ cm⁻² s⁻¹ with no critical nucleus size, three to four orders above atmospheric studies; classical nucleation theory is the framework atmospheric models use. A joint programme measuring rate against temperature, substrate and water activity with one analysis pipeline would tell the atmospheric community whether its parameterisations survive, which is the largest practical stake in this brief.

Handwave Sixth: the negative-pressure route, which is the one genuinely different idea in circulation. If the anomalies are driven by a transition line that runs to negative pressure without terminating, then stretched water — water under tension in inclusions or in acoustically cavitated samples — is the place to look, and the region is as inaccessible as no-man’s land for different reasons. This brief could not retrieve the primary literature on stretched water and therefore neither ranks this route nor dismisses it.

Handwave Seventh, and the one that would improve this page fastest: open the record. Deposit the raw scattering data from the free-electron-laser water experiments, with the thermal-history models used to assign temperatures, in a public archive. The interpretive dispute at the centre of this subject is about what a set of scattering patterns means; it is being conducted by people who have the patterns against people who have the papers.

Speculative Eighth, for completeness: the crystalline phase diagram is still yielding, and that is a reminder about priors. Ice XXII was reported above 308(5) GPa in 2026 and ice XXI, body-centred tetragonal, was described the same year among five crystallisation pathways of high-density ice at room temperature. A substance still producing new ordered phases in 2026 is not a substance whose metastable liquid states can be assumed to be understood.

Frontier The instrument this subject is built on is the X-ray free electron laser, and the requirement it imposes is a sample that is destroyed after the measurement rather than before. Femtosecond pulses allow a scattering pattern to be collected from a droplet that will crystallise microseconds later, which is the only reason the upper part of no-man’s land is accessible at all. The engineering constraint is that the droplet source, the vacuum, the cooling trajectory and the pulse must all be synchronised, and the resulting temperature is inferred from the evaporation model rather than measured.

Established Ultrafast heating of amorphous ice is the complementary instrument and it now has a documented failure mode. The requirement is a film thin enough to heat uniformly on the timescale of the probe and a substrate whose thermal contact is characterised rather than assumed: 100 nm of amorphous solid water at 100 K on platinum, with an interfacial vapour film of about 6 nm and an interfacial conductance near 130 MW m⁻² K⁻¹, did not follow a substrate driven to 700–1200 K within 10 ns. Designing that interface is now part of the experiment.

Established Cryogenic electron microscopy has become a quantitative instrument for this problem rather than an imaging one. In-situ cryo-TEM at about 102 K and 10⁻⁶ Pa, with millisecond temporal and picometre spatial resolution, produced growth velocities of 4.7 × 10⁻² nm s⁻¹ and statistics over a hundred individual nuclei. That is a measurement of nucleation one event at a time, which no bulk technique provides.

Speculative The instrument that does not exist is a pressure-controlled deep-supercooling cell with a fast probe. Diamond anvil cells give pressure; droplet sources give supercooling; neither gives both with a femtosecond probe and a bulk sample. Building one is the engineering statement of the decisive experiment in section 7, and it is the reason that experiment carries no date here.

Frontier Computation is a first-class instrument in this subject and its requirements are unusual. The systems are small and the timescales long, the interatomic model has to be right about hydrogen bonding far outside where it was fitted, and the analysis depends on order parameters whose choice has been shown to change the answer. The 2026 amorphous-ice study is a machine-learning analysis of structure rather than a simulation of dynamics, which is a different and more auditable use of the same tooling; the same distinction is drawn in Glass Transition Theory, where predictive models outperform interpretable ones and the field says so.

Established Metrology is a requirement here in the ordinary sense as well. The temperature of maximum density is known to 3.983035(670) °C and the density there to 0.99997495(84) g mL⁻¹; nothing in no-man’s land is known to anything like that precision, and the gap between those two standards of evidence is the subject. Metrology Infrastructure carries the general argument about reference data and traceability.

Established Atmospheric science is the largest consumer of this physics and the least interested in its metaphysics. Supercooled liquid droplets exist routinely in stratus and cumulus clouds, and a 2023 review treats homogeneous ice nucleation as affected by pressure, temperature and solute strength and predictable through water activity. Mixed-phase clouds are a first-order term in radiative transfer, which makes the nucleation rate a climate parameter. AI Weather Prediction inherits those parameterisations wholesale, and Weather Modification owns the deliberate-nucleation question.

Frontier Cryobiology is where the competition between vitrification and crystallisation is the entire engineering problem. Preserving a cell, a tissue or an organ means getting through the same temperature band this brief is about without letting ice form, and the margin is set by nucleation rates and by how much the sample can be confined or loaded with cryoprotectant. Cryonics and Brain Preservation carry the claims; Glass Transition Theory carries the glass side of the physics; this page carries the liquid.

Speculative Planetary and astrophysical ice is the largest reservoir of amorphous water in the universe and the least accessible. Low-density amorphous ice is made in the laboratory by vapour deposition below 120 K, which is also how it forms on interstellar grains and on outer solar system surfaces; the 2025 cryo-TEM work at 102 K is effectively laboratory astrophysics. Laboratory Astrophysics owns the instrumentation argument and Planetary Protection the icy-moon question; what belongs here is that the amorphous phases whose relationship is disputed are the common forms of water off Earth.

Speculative Freeze concentration and ice-slurry processing are the industrial neighbours, and this brief cannot quantify them. Separating water from a solution by freezing rather than by boiling is thermodynamically attractive and operationally awkward, and where it is used the binding constraint is nucleation control. Water Infrastructure Megaprojects is the nearest existing brief on this map and owns the economics of making and moving water at national scale — desalination as a purchase of electricity, transfer as a purchase of sovereign consent. It is the economics of the substance; this page is the physics of it, and the two do not overlap.

Handwave Biology’s interest is in whether water’s anomalies are load-bearing for life, and the claim is easy to overstate. That liquid water’s density maximum at 3.983 °C keeps deep lakes liquid under ice is a real and consequential fact. Arguments that extend from there to water being uniquely necessary for life are a different kind of claim, and Xenobiology carries them. Nothing in the retrieved evidence for this brief bears on them.

Speculative The sibling commissioned alongside this brief shares its central methodological problem. Glass Transition Theory is about a slowdown with two incompatible explanations that has survived every decisive experiment; this page is about a phase transition with two incompatible readings that has survived every decisive experiment. Both fields reach for amorphous solids as proxies for liquids they cannot equilibrate, both have order-parameter conventions doing load-bearing work, and both would be improved by the same round robin.

Frontier The institutional fact that shapes this subject is that its decisive instruments are a handful of free-electron lasers with competitive beamtime. An experiment that needs a facility campaign is an experiment that happens when a committee agrees it should, and a null result is a harder proposal to write than a positive one. That is the ordinary economics of large-facility science; what is unusual here is that the claim under test is a claim about a substance available in every laboratory.

Established The second institutional fact is that this subject’s public record is unusually closed, and this brief is direct evidence. A research pass that retrieved 2021, 2023, 2024, 2025 and 2026 primary papers on ice nucleation and amorphous ice could not open a single primary paper on the liquid-liquid transition itself. Whatever the cause in any individual case, the effect is that the central claim of the field is harder to audit from outside than its periphery.

Handwave The third is that the field’s most-repeated summary statistic has no institutional owner. The count of water anomalies — seventy, seventy-four, more than seventy — is repeated in commissioning documents, review introductions and popular accounts, and traces to a privately maintained compilation rather than to a standards body or a review series. A community that wants a number quoted should own the number.

Speculative The fourth is that the reference-data institutions stop at the edge of the interesting region. The equation of state of water is among the most carefully maintained reference datasets in physical science above the freezing point. Below it, and especially below 231 K, there is nothing to maintain because there is nothing measured. Extending a reference standard into a region where the phase behaviour is disputed is not a formality, and it is the institutional form of the decisive experiment.

Handwave The fifth is data deposition, which would cost this field nothing and change what outsiders can check. Scattering patterns, thermal-history models and the analysis code that converts one into a temperature are the whole content of the dispute. Depositing them would convert an argument about interpretation into an argument that a third party can adjudicate, which is the outcome the two camps have not produced in thirty years of publishing.

Established There is no direct harm surface here and the indirect one runs through weather and climate models. Ice nucleation rates enter mixed-phase cloud parameterisations, which enter radiative transfer, which enters projections that governments spend money on. A systematic error in a nucleation rate is not a laboratory curiosity; it is a bias in a number that is used. That is a reason for the reconciliation programme in section 7 and not a reason for alarm.

Speculative The claims-discipline problem in this subject is specific and worth naming. Water attracts more pseudoscience than any other substance — structured water, memory of water, exclusion-zone water — and it attracts it partly because the genuine open question at the centre of this brief is real, unresolved and describable in language that sounds like the fake version. A field with a real anomaly has a duty to be precise about its boundaries, because imprecision is exploited.

Frontier The anomaly count is the clearest instance of that imprecision and it is coming from inside the house. Repeating “more than seventy anomalies” without a source, a definition of anomaly or a reference substance is the kind of statement that reads as rigour and functions as advertising. The density maximum at 3.983035(670) °C is a fact with an uncertainty. Seventy is not.

Speculative The cryopreservation adjacency carries the field’s only real consent question, and it is not this page’s. Whether a vitrified sample retains the information a person cares about is a claim examined in Brain Preservation and Cryonics. What this page supplies is the physical margin: the band between the homogeneous freezing limit and the glass transition is where that margin lives, and the liquid inside it has never been directly measured.

Handwave The distributional question is thin and honest to state. Nobody is disadvantaged by the current state of ignorance about supercooled water, and nobody will be enriched by resolving it. It is a case where the value is almost entirely epistemic, which is a legitimate thing for a research portfolio to contain and a reason to be sceptical of any page that dresses it as an application.

Established Water’s anomalies are already load-bearing for the biosphere, and resolving their cause would not change that. The density maximum at 3.983 °C is why deep temperate lakes hold liquid water beneath winter ice; the supercooled droplets in clouds are why precipitation works the way it does. These facts operate regardless of which theoretical account turns out to be right, which bounds the civilisational stake of the resolution more than the subject’s rhetoric usually admits.

Speculative The consequence with real magnitude runs through climate projection, and it is a consequence of getting the rates right rather than the theory. Mixed-phase clouds are among the larger uncertainties in radiative transfer, and the nucleation rates that set their ice fraction are parameterised from measurements spanning a range that a 2025 experiment extended by three to four orders of magnitude in a regime the atmosphere does not sample. Whether the theoretical question is settled matters less to that than whether the rate measurements are reconciled.

Speculative The second consequence is preservation and it is shared with the sibling brief. Every scheme for holding biological material outside time depends on crossing this temperature band without nucleating ice, and every estimate of how reliably that can be done is an extrapolation of rates measured elsewhere. A predictive account of nucleation in deeply supercooled water would change the confidence interval on a family of proposals that currently rest on demonstrations rather than on theory.

Handwave The third is epistemic and is the honest headline. If liquid water turns out to contain a phase transition that was missed for a century in the most-measured substance on Earth, that is a fact about how confidently a mature science can be wrong about something ordinary. If instead the anomalies dissolve into hydrogen-bond statistics with no transition anywhere, that is a fact about how long a well-posed hypothesis can survive without being decidable. Both outcomes are informative and neither builds anything.

Handwave The counterfactual closes the section. If nobody resolves this in fifty years, cloud models will still run, organs will still be vitrified, and lakes will still freeze from the top. The cost of continued ignorance is a permanent uncertainty in a few parameterisations and an unfinished chapter in the physics of liquids, and a page claiming more than that is overselling.

These horizons track measurement access to the 150–231 K band, because nothing else in this subject can move without it:

• 10 yr: Frontier Expect continued incremental entry into no-man’s land from both ends — smaller droplets from above, faster heating of amorphous ice from below — with each result contested on whether the sample was still bulk water. Speculative Expect the thermometry audit of section 7 to happen, because it is cheap and because the September 2026 vapour-film result makes it obvious. Speculative Expect machine-learning structural analyses of the kind published in 2026 to multiply, and expect the number of claimed amorphous phases to depend on the order parameter chosen.
• 25 yr: Speculative This is the window in which a pressure-resolved measurement inside the band becomes plausible, and it is the only development that would settle anything. Speculative A positive result — converging loci of response maxima terminating at a point — would make water the first substance with an experimentally located second critical point and would immediately reopen the question for silicon, silica and phosphorus. Handwave A null across the accessible pressures would not exclude the hypothesis, because the point could always be placed somewhere not yet reached, and that asymmetry is the subject’s central methodological weakness.
• 50 yr: Handwave On this horizon the plausible outcome is not a verdict but a reclassification: the anomalies get an agreed quantitative account, the word “transition” is quietly dropped or quietly kept, and the question stops being asked in the form it is asked today. Speculative The atmospheric and cryobiological users will have had usable nucleation rates for decades by then, obtained empirically, and will not have waited.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The defensible statement is about shape: a hypothesis that places its decisive object in a region no instrument can hold open either gets an instrument or becomes a convention. Handwave Any specific claim about what a located second critical point in water would be used for is an assertion, and this page does not make one.

• Depends on Nothing on this map blocks this subject. Water Infrastructure Megaprojects is the nearest existing brief and shares only the substance: it owns desalination energy, inter-basin transfer and the politics of moving water, and nothing in it bears on the phase behaviour of the supercooled liquid. Glass Transition Theory, commissioned alongside this page, supplies the vocabulary of fictive temperature, fragility and amorphous states that section 3 uses, and shares the order-parameter problem rather than causing it. Laboratory Astrophysics supplies the low-temperature deposition craft that makes amorphous ice available at all.
• Requires (not on this map) Five conditions, in the order of the tokens above. The measurement. Isothermal compressibility and isobaric heat capacity at several distinct pressures at temperatures between 150 and 231 K, in a sample large enough to be bulk, taken faster than ice nucleates; the discriminating signature is the behaviour of the locus of maxima with pressure, which a single isobar cannot show. The cell. An instrument combining pressure control, deep supercooling and a femtosecond probe; each pair exists today and the triple does not, which is why section 7 gives its decisive experiment no date. The criterion. An agreed structural definition of what makes two liquid states two, after a 2026 analysis found the discriminating information in local hydrogen density within the first coordination shell and found orientational order parameters blind to the hysteresis. The archive. Public deposition of the facility scattering data and the thermal-history models used to assign temperatures to it, without which the central interpretive dispute cannot be adjudicated by anyone who was not on the beamline. The rate standards. Reference nucleation rates traceable across sample size, after measurements at 102 K came in three to four orders of magnitude above atmospheric studies and nanodroplet confinement suppressed rates by over two orders of magnitude.
• Enables A resolved account of nucleation in deeply supercooled water would tighten the mixed-phase cloud parameterisations that AI Weather Prediction inherits and that Weather Modification acts on, and would narrow the vitrification margin that Cryonics and Brain Preservation depend on. No typed enabling edge is claimed: those fields are already operating on empirical rates and will continue to, which is the honest statement of this subject’s leverage.
• Adjacent Atmospheric ice nucleation, cryobiology, planetary ice and high-pressure crystallography supply every number on this page. Within this map: Glass Transition Theory for the amorphous side, Planetary Protection for the icy moons, Xenobiology for the claims about water and life, Metrology Infrastructure for the reference-data argument, and Objective Collapse Models as the sibling case of a subject whose decisive parameter sits where no instrument reaches.

Handwave “Water has more than seventy measured anomalies.” This is the most-repeated claim in the subject and this brief could not source it. No peer-reviewed document retrieved for this page states a count; the figure appears to originate in a single privately maintained compilation that was not reachable in this session. A count of anomalies depends entirely on how finely the list is subdivided and on which reference substance is used for comparison, so even a sourced count would be a fact about a list. The individual anomalies are real; the number is not a measurement.

Established “No-man’s land runs from about 180 to 235 K.” The commission for this brief says exactly this and the lower bound is wrong by roughly thirty kelvin. Glassy water can be warmed to about 150 K before nucleation occurs, and the retrieved statement of the inaccessible region is that between 150 and 231 K experiments find only crystal ice. The band is about 80 K wide. Anyone using the narrower figure will underestimate the region a decisive experiment has to cross.

Frontier “Water freezes at 0 °C, and the homogeneous limit is −38 °C.” Both numbers are conventions attached to a rate. Demineralised bulk water has been supercooled to −48.3 °C without special technique; 2 nm droplets in octane froze at about −41 °C against a bulk limit taken as about −38 °C; and one widely used reference gives the standard-pressure homogeneous nucleation point as about 231 K while another on the same site gives 224.8 K. A freezing temperature without a sample volume and an observation time is incomplete.

Speculative “The second critical point has been observed.” Not in bulk liquid water, on the evidence this session could retrieve. What has been reported is a first-order-like transition between two amorphous solids with no intermediate structures, a set of response maxima along a single isobar in droplets, and simulation results in models. Each of those is consistent with the hypothesis and none of them is the observation. The distinction matters because the hypothesis is falsifiable and the weaker claims are not.

Speculative “The critical point has been ruled out.” Equally unsupported, and structurally harder: a critical point can always be relocated to a temperature and pressure not yet examined. That asymmetry is the subject’s methodological weakness and is the reason section 7 asks for the locus of response maxima rather than for a single measurement, since a locus constrains where the point can be even when it is not reached.

Frontier “Amorphous ice is just water frozen very fast.” It is at least two distinct materials. Low-density amorphous ice is made by vapour deposition onto smooth metal below 120 K; a high-density form was shown in 1994 to arise from deposition below 30 K; and a 2026 analysis reports that the two are distinguished by local hydrogen density within the first coordination shell, with no intermediate structures between them and pronounced hysteresis between compression and decompression. Whether that makes them the glasses of two liquids is the open question, not a settled one.

Established “Ice nucleation is understood; it is just classical nucleation theory.” In-situ cryo-TEM at about 102 K found an adsorption-mediated, barrierless pathway with no critical nucleus size across more than a hundred individual nuclei, and nucleation rates three to four orders of magnitude above those in atmospheric supercooling studies. Separately, simulation of the free surface places nucleation about 1.5 nm inside the liquid, driven by negative pressure favouring precursors with five-, six- and seven-membered rings. Classical nucleation theory is a parameterisation that works where it has been fitted.

Frontier “If you heat amorphous ice quickly enough you get the liquid, and you know its temperature.” The second half is now doubtful. A 100 nm amorphous solid water film on platinum, with the metal driven to 700–1200 K, showed almost no structural change within 10 ns, attributed to an interfacial vapour film about 6 nm thick limiting interfacial conductance to about 130 MW m⁻² K⁻¹. Where sample temperature is inferred from a heat-flow model rather than measured, this is a documented failure mode and not a hypothetical one.

Handwave “The anomalies prove water is special in a way that explains life, or health, or structure.” Water attracts a large pseudoscientific literature precisely because it has a real unresolved anomaly at its centre, and the two are not connected. The density maximum keeps deep lakes liquid under ice; that is a consequential and completely ordinary piece of physics. Nothing in the retrieved evidence for this brief supports any claim about structured water, water memory, or biological specialness, and the existence of a genuine open question is not evidence for a fake one.

Handwave “This is a solved problem that the field keeps arguing about for funding reasons.” The cynical reading is available and this brief does not endorse it. The region in dispute is genuinely inaccessible: liquid water inside it crystallises faster than a conventional measurement equilibrates, the techniques that reach it buy access with confinement, pressure or a clock, and the crystalline phase diagram of the same substance was still producing new phases in 2026. A dispute can be unresolved because the measurement is hard, and here it is.