Wang et al. reply

REPLYING TO W. Zhou et al. Nature 528, 10.1038/nature16145 (2015)

In the accompanying Comment¹, Zhou et al. reproduced our² molecular dynamics (MD) results and pointed out that the simulated 2D ice is slightly rhomboidal, in contrast to the square lattice seen in the transmission electron microscope (TEM) images². We were aware of this disagreement, but did not discuss it in ref. 2 for the following reasons. First, previous MD simulations^3,4 have reported ‘square ice’, although it remains unclear whether this ice is different to the distorted lattice we found². Second, and more importantly, we were convinced that the simulated, slightly rhomboidal structures should be observed experimentally as square ice.

Indeed, our MD snapshots² (and those presented in ref. 1) show substantial disorder. Each realization is metastable, and the finite temperature is expected to move such defects through the crystal lattice. Our simulations show that this happens on a timescale of tens of nanoseconds for nanometre-sized ice crystals, much longer than the time used by Zhou et al.¹, but much shorter than the time needed to obtain experimental images (about 1 s). To simulate this time-averaging effect, we created a number of intermittent states (such as that shown in figure 2d in ref. 2) and superimposed them, keeping the positions of only the edge molecules fixed to simulate the confinement. We found that the slightly rhomboidal lattice averaged out into one that is indistinguishable from a perfect square (not shown in ref. 2).

Finally, perfectly square ice discussed in ref. 2 was subsequently found to be the most stable configuration using first-principle analyses^5,6. Therefore, we maintain that square ice can theoretically occur in hydrophobic nanocapillaries, in agreement with the experiment².

R. R. Nair and I. V. Grigorieva support this Reply, but did not contribute to the part of research that was addressed in the accompanying Comment.