Controlled induction of crystal nucleation is a highly desirable but elusive goal. Attempts to speed up crystallization, such as high super saturation or working near a liquid-liquid critical point, always lead to irregular and uncontrollable crystal growth. Here, we show that under highly nonequilibrium conditions of spinodal decomposition, water crystals grow as thin wires in a template-less formation of "Haareis". This suggests that such nonequilibrium conditions may be employed more widely as a mechanism for crystal growth control.The physical chemistry of crystal nucleation is of great fundamental and practical importance but is yet poorly understood. It is therefore one of the grand challenges on the border between physics, chemistry, and chemical engineering. Crystal nucleation in melt or solution is typically described by Gibbs's classical nucleation theory based on the competition between the free energy of solidification and the free energy due to the presence of the interface. 1, 2 The latter results in a barrier to crystallization and hence super-saturation and a metastable nonequilibrium state. Thermodynamic fluctuations then lead to pre-nucleation sites, the majority of which will redissolve. 3 Occasionally, a nucleus will grow big enough to overcome the barrier (a critical nucleus) and continue to grow. Only at considerable super-saturation will the energy barrier disappear, at which point homogeneous nucleation will occur.As a result, crystal nucleation is generally a rare process that is difficult to study either experimentally or even through computer simulation. In addition, Ostwald's rule of stages suggests that there are intermediate metastable states critical to the understanding of the path and thermodynamics of nucleation. Such metastable states are typically too rare or short-lived to be observed.However, recent work by Gebauer and others has shown that in some cases (such as the nucleation of carbonates from aqueous solution 4-7 ) solute clusters may form that aggregate into amorphous clusters, which then transform into crystal nuclei. 4, 7-9 Such non-classical nucleation theories do not require a "critical nucleus". These theories appear to, but may not necessarily, 10 be counter to thermodynamic theory. Interestingly, a number of light scattering studies of solutions have shown anomalous clustering in solution suggesting that the effect might be more general. 11,12 In the 1990s, Frenkel introduced the concept of the enhancement of crystal nucleation due to the presence of liquid-liquid critical points. 13 Such a critical point would induce concentration fluctuations that would give rise to droplets of so-called "dense fluid" in which the nucleation probability would be greatly enhanced. 7,[14][15][16] Thus, in this scheme the nucleation mechanism is not changed (it could be classical or non-classical) other than to provide an environment with an increased concentration. Although Frenkel's theory was developed for protein crystallization, it is now widely used in chemical-engineering d...