Is there a temperature at which ice is denser than water?

Normally ice would float on water because its density is less compared to that of water as a liquid. But is it possible that its density will increase due to a very low temperature or is ice in any case lighter than water?

• The density of the solids (and liquids) doesn't vary with temperature (and even with pressure) so much and we usually assume that their density is constant. – lucas Jul 1 '16 at 13:38
• @lucas: We do?… – Williham Totland Jul 1 '16 at 16:48
• @lucas: Because if we do, we're in for a rude awakening: Water, as an example, varies by almost 10% over the all things considered fairly narrow range 4°C to 100°C. That's a big variation to straight up ignore. – Williham Totland Jul 1 '16 at 16:52
• @WillihamTotland You are right. Thanks for teaching. – lucas Jul 1 '16 at 16:58
• @lucas: Moreover, one of the major reasons for the development of the constants-based SI system (e.g. where the meter is defined as the distance traveled by light in 1/299 792 458 s instead of the length of a meter-sized rod) is that the rod contracts and expands according to temperature and pressure; although, yes, for practical purposes that seems like a degenerate case. But when building railroads, you have to leave lengths of rail a certain distance apart, or they'll bunch up on one another in hot weather, causing all sorts of badness. (The genesis of the kthunk-kthunk of trains, btw.) – Williham Totland Jul 1 '16 at 17:18

Ice can be denser than water for certain values of $P,T$. Look at these two pictures taken from here:  The darker areas in the second picture denotes areas of greater density. So you can clearly see that when pressure is increased, ice becomes denser than water along the coexistence line.

For example at $T=400$ K ice VII is clearly denser than water along the coexistence line ($P \simeq 2$ GPa).

Quoting from the page:

As pressure increases, the ice phases become denser. They achieve this by initially bending bonds, forming tighter ring or helical networks, and finally including greater amounts of network inter-penetration. This is particularly evident when comparing ice-five with the metastable ices (ice-four and ice-twelve) that may exist in its phase space.

At atmospheric pressure, $P_{atm} \simeq 100$ kPa $=10^5$ Pa, ordinary ice is always less dense than water.

When I posted this answer, I may have taken for granted that everybody was able to read this kind of phase diagram, but since it looks like I was wrong, I will try to explain them better.

The first diagram shows the various phases of water as a function of the two parameters $P,T$. The first thing that must be noticed is that the pressure axis is logarithmic while the temperature axis is linear. This means that the plot is "compressed" in the vertical direction (you can see that the $T$ axis goes from $1$ to $800$ (almost 3 orders of magnitude) while the $P$ axis goes from $0.1$ to $10^{12}$ (13 orders of magnitude!)).

The black lines are coexistence lines. This means that along those lines two phases can coexist. If we want to compare water and ice, I think that the only meaningful way to do so is to compare them along the coexistence line, because only there it will be possible to have both of them. For example, you can see that ice VIII can never coexist with liquid water.

Our world is located at $P= 1$bar$\simeq 10^5$ Pa (red line): You can in fact see that, at the red line, the solid-liquid transition is at $273$ K ($0$°C) and the liquid-vapor is at $373$ K ($100$°C) - as expected.

But things get different at different pressures. For example, at $10^6$ Pa ($10\times$atm.pressure), the liquid-vapor transition is at $450$K, and at $10^2$ Pa ($1/1000$ of atm.pressure) ice sublimates directly into vapor (there is no liquid state!).

Now, the density. You have to look at the second plot to see the density. For example, let's take the $400$ K-$2\cdot10^9$Pa point (yellow arrow in the first plot). To see the density, look at the corresponding point in the second plot. You can see that the area corresponding to ice (ice VII) is darker than the area corresponding to water, so you can tell that ice is denser than water there, and so on.

If you take P=$10^5$ Pa (atm.pressure), you can see that ice is always less dense than water (lighter shading) there.

• So it looks like only with pressure it would be possible? – Marijn Jul 1 '16 at 15:40
• @Marijn At atmospheric pressure it is not possible for ice to be denser than water, if that is what you mean. – valerio Jul 1 '16 at 15:53
• Do you know why the vapor curve begins at around 200K, isn't that around -73C? Isn't that too low for vaporising? And am I wright that ice form XI is only created at temperatures which are close to the absolute zero point 0K? – Marijn Jul 1 '16 at 18:09
• @Marijn "too low for vaporizing" is just a subjective statement because we are used to atmospheric conditions. As pressure lowers, so does evaporation temperature and far below atmospheric conditions, things seem odd. What feels normal here, is not normal under other conditions. – Steeven Jul 1 '16 at 21:15
• +1, even though I don't fully understand the graphs, having never seen such graphs. The important takeaways are that ice is always less dense than water at normal atmospheric temperature, but could be more dense at certain pressures. The obvious followup is: at what pressure is the ice more dense? – DCShannon Jul 1 '16 at 21:49