Relaxed structures for the calculations reported in the manuscript are provided in "extended" XYZ format [1]. Q-TIP4P/F calculations have been carried out with the LAMMPS code [2] (interfaced via ASE [3]). MB-pol calculations are based on an in-house ASE-calculator building on the MB-pol plugin for OpenMM (Release 1.1.2), available from the Paesani group [4]. This implementation can only handle periodic boundary conditions for orthogonal cells of a certain minimum size. For this reason, a 3x3x3 (3x2x7) supercell of the structures provided below for ice Ih (ice II) has been constructed internally in our ASE MB-pol calculator. [1] https://wiki.fysik.dtu.dk/ase/dev/_modules/ase/io/extxyz.html [2] https://lammps.sandia.gov/doc/Howto_tip4p.html [3] https://wiki.fysik.dtu.dk/ase/index.html [4] https://github.com/paesanilab/mbpol_openmm_plugin - ice Ih: ==> ih-qtip4pf.xyz <== 288 Lattice="13.3973455228 0.000288218955933 -0.000168644185498 0.000333186810814 15.483704165 0.000699787747983 -0.000183692738196 0.000659521318457 14.5916671855" ==> ih-mbpol.xyz <== 288 Lattice="13.5401973366 0.0 0.0 0.0 15.6520245142 0.0 0.0 0.0 14.7539983072" - ice II: ==> ii-qtip4pf.xyz <== 36 Lattice="7.69183810264 -0.00515898305999 -0.0111177281685 -3.02253873772 7.07309406065 -0.0111177281666 -3.02253873772 -4.57522765162 5.39407778325" ==> ii-mbpol.xyz <== 216 Lattice="12.9751818454 0.0 0.0 -2.99255363976e-06 22.4702673861 0.0 0.0 0.0 6.25158943718" # This cell given by the lattice vectors a^T = [12.9751818454, 0.0, 0.0] b^T = [-2.99255363976e-06, 22.4702673861, 0.0] c^T = [0.0, 0.0, 6.25158943718] can be reduced to the primitive unit cell of ice II according to (a',b',c') = (a,b,c) * P = ( P_{1,1}a+P_{2,1}b+P_{3,1}c, P_{1,2}a+P_{2,2}b+P_{3,2}c, P_{1,3}a+P_{2,3}b+P_{3,3}c ) where ( 0.5 0.0 -0.5 ) P= (-0.16666 0.3333 -0.1666 ) ( 0.33333 0.3333 0.3333 ) Note that the resulting cell (described by a',b' and c') is rotated from the conventionally chosen reference frame (a' along x-axis, b' in xy-plane).