Ice III is a hydrogen-disordered phase of ice
that is stable between about 0.2 and 0.35 GPa. Upon cooling, it transforms to
its hydrogen-ordered counterpart ice IX within the stability region of ice II.
Because of this metastability, detailed studies of the ice III to ice IX phase
transition have so far not been carried out. Using ammonium fluoride doping to
prevent the formation of ice II, we now present a detailed study on this phase
transition using in-situ powder neutron diffraction. The <i>a</i> and <i>c</i>
lattice constants are found to expand and contract, respectively, upon hydrogen
ordering yielding an overall negative volume change. Interestingly, the
anisotropy in the lattice constants persists when ice IX is fully formed and
negative thermal expansion is observed. Analogous to the isostructural keatite
and <i>b</i>-spodumenes, the
negative thermal expansion can be explained through the build-up of torsional
strain within in the <i>a</i>-<i>b</i> plane as the helical ‘springs’ within
the structure expand upon heating. The reversibility of the phase transition
was demonstrated for the first time upon heating. The ammonium fluoride doping
induces additional residual hydrogen disorder in ice IX and is suggested to be
a chemical way for ‘excitation’ of the ice-rules configurational manifold.
Compared to ices II and VIII, the induced hydrogen disorder in ice IX is
smaller which suggests a higher density of configurational states close to the
ground state. This study highlights the importance of dopants for exploring water’s
phase diagram and underpins the highly complex solid-state chemistry of ice.