We have observed what appears to be a first-order phase change from deeply supercooled liquid triphenyl phosphite at 1 atm to a rigid, "apparently" amorphous phase which we denote as the "glacial phase". This is a new, crisper, and rather different addition to the examples of polyamorphism that have recently been studied. In order to "deeply" supercool the liquid, it must be quick-quenched to a low temperature: if heated slowly, but immediately, it crystallizes; if allowed to stand for several hours at low temperature, it transforms to the glacial phase; and if subsequently heated it, too, crystallizes, but at a higher temperature than that for liquid crystallization. The glacial phase can be clearly distinguished from both the normal crystal and the ordinary glass. We propose a model for the formation of this "apparently" amorphous glacial phase.

We have discovered a solid, apparently amorphous phase of triphenyl phosphite to which the supercooled liquid converts, a phase distinct from both the glass and the crystal. To date, this is the clearest and best identified case of a first-order transition from a liquid to another apparently amorphous condensed phase. We discuss this phenomenon in terms of a recently formulated theory of supercooled liquids that predicts and naturally incorporates the existence of such low-temperature phases, thereby suggesting that its existence is a general phenomenon intimately connected with the existence and properties of supercooled liquids. In accord with the theory, we also suggest that although the X-ray scans do not indicate any lattice structure, these apparently amorphous phases may in fact be defect-ordered structures with large unit cells.