Crystallization is a fundamental and important physical process, which, though widely studied, remains very poorly understood.
In particular, the rate at which macroscopic crystals form is typically higher by many orders of magnitude, compared to theoretical predictions.
The conventional experimental techniques, such as x-ray scattering and electron microscopy, are unable to detect the detailed structure and
thermodynamical properties of early nuclei. Thus, the whole paradigm of crystal nucleation is based on untested assumptions.

Hard sphere colloids, micron-sized particles suspended in a molecular solvent,
are among the simplest systems which exhibit crystal nucleation; thus, this system provides
an important insight onto the basics of nucleation, which are very poorly understood.
Colloids are much larger than atoms or molecules, such that confocal microscopy allows to track
individual particles in a dense three dimensional system, in real time; yet colloids minimize their
free energy, akin to atoms and molecules.

We show that early nuclei are not compact, as most theories of nucleation assume. Instead, the nuclei adopt a wide
range of ramified morphologies, limited solely by the liquid-solid interfacial tension; the existence of these morphologies modifies
the free energy of the nuclei, resulting into an enormous increase in the rates of crystal nucleation.
Remarkably, recent computer simulations indicate that similar morphologies are present in early nuclei, in metallic Cu;
thus, ramified morphologies of nuclei are likely to play an important role in crystal nucleation in a wide range of materials.