Droplets of a liquid alloy on a silicon surface can rearrange the surface atoms so that they mimic the short-range ordering of atoms in the alloy. Remarkably, this effect inhibits freezing of the droplets.
Promoting freezing in a liquid is conceptually straightforward — you simply need to add suitable templates. The templates can be either 'seeds' of the crystalline phase that would form from the liquid, or small crystals of another material whose atomic-level surface structure in some way matches that of such seeds.
What is much more difficult to conceive of is a solid surface that inhibits freezing by acting as a template for the liquid. However, on page 1174 of this issue, Schülli et al.1 describe evidence suggesting that such a template is possible. Their results have wide implications not only for fundamental studies of freezing, but also for the practical control of this phase transition.
When a liquid is cooled, there is a thermodynamically defined temperature — the freezing point, or liquidus temperature — at which it should start to crystallize. But the crystal nucleation that initiates freezing requires a driving force, and occurs only at temperatures below the ideal freezing temperature. The cooling of a liquid to below the ideal freezing temperature, known as supercooling (Fig. 1), is of great interest in diverse areas ranging from the control of microstructure in metallic welds and castings2 to the inhibition (or promotion) of ice formation necessary for the survival of living systems3. Although in some situations it is desirable to nucleate crystals at the highest possible temperature by minimizing supercooling, in others the challenge is to avoid nucleation so that liquids and their freezing processes can be studied at the lowest temperature possible (greatest possible supercooling).
Schülli et al.1 studied the freezing of a liquid gold–silicon alloy near its eutectic (lowest freezing point) composition. Not only is this alloy a useful model system, but it also has practical significance — it is used in the vapour–liquid–solid process for growing silicon nanowires4. The authors formed microscopic islands of gold on a single-crystal silicon substrate and heated them until they melted, whereupon some of the silicon dissolved into the gold to form liquid-alloy droplets. On cooling, these droplets froze at a temperature that was reproducible over repeated heating–cooling cycles, but that depended greatly on the upper temperature limit of the heating.
Using an X-ray scattering technique under ultra-high vacuum conditions, Schülli et al.1 characterized the structure of the liquid–substrate interfaces in their system in situ. They found that on crystallographically defined silicon substrates (for which particular planes within the lattice were exposed at the surface) the freezing point of the gold–silicon alloy was generally 563 kelvin. By performing a detailed analysis of extensive X-ray data acquired at a wide range of angles to the substrate, the authors showed that this freezing point corresponds to a slight, gold-induced reconstruction of the atomic arrangement of the silicon surface. But when the team had previously heated the liquid to temperatures above 673 K, the onset of freezing was remarkably depressed to 513 K. In this case, the authors observed that the silicon had undergone a more radical reconstruction to yield what is known as a 6 × 6 superstructure.
The freezing point of the alloy in contact with the 6 × 6 silicon surface is about 120 K below that of eutectic gold–silicon, but Schülli et al. point out that the eutectic freezing temperature is not the most appropriate reference point from which supercooling of the system should be measured. As the liquid alloy droplets cool, silicon comes out of solution and redeposits on the substrate, enriching the droplets in gold. The observed freezing point of 513 K therefore represents a supercooling of about 360 K below the liquidus of the resulting composition. This is more than 40% of the expected freezing temperature, an exceptionally high value for a metallic system5.
This supercooling is all the more remarkable because of the ordering of the liquid at the solid–liquid interface. It is accepted that a liquid in contact with a planar solid shows out-of-plane ordering — the liquid atoms form layers parallel to the surface of the solid6. This might be thought to favour crystallization of the liquid, but whether or not this is so depends on the nature of the atomic ordering within each layer (in-plane ordering), the characterization of which has proved challenging7. In-plane ordering is exactly what would be expected when a liquid comes into contact with a substrate that acts as a template for crystal nucleation in freezing.
Schülli et al.1 have succeeded in the difficult task of characterizing in-plane order in the liquid gold–silicon alloy adjacent to the 6 × 6 silicon surface. They found that the liquid is anisotropic, with atomic positions strongly correlated with the structure on the underlying surface. It is remarkable that this correlation with a periodic surface pattern impedes the crystallization of the liquid, rather than inducing it.
The researchers1 also characterized the structure of the gold–silicon liquid away from the silicon substrate and found that, in common with most metallic liquids8, it shows icosahedral short-range order that becomes more pronounced on cooling. Crucially, the authors observed that the pentagonal clusters of atoms typical of the icosahedral order seem to be stabilized by similar pentagonal arrangements in the 6 × 6 silicon superstructure (see Fig. 4 on page 1177). However, when the alloy droplets freeze on the 6 × 6 silicon surface, the resulting gold crystals form in random orientations. This suggests that the substrate has no orienting role in freezing; the actual site and mechanism of crystal nucleation remain undetermined.
For several years, there has been intense interest in attaining a large supercooling effect of liquids before the onset of freezing. To avoid crystal nucleation caused by contact with a solid, the favoured strategy is to process the liquid without using a container. This can be done in various ways9, including by levitating liquid drops electromagnetically, electrostatically or acoustically, or by studying drops as they fall through a tube or tower. Because the surface energies of most liquids are lower than those of the crystals that form from them10, free surfaces — in these experiments, the liquid surfaces at air–liquid interfaces — also stabilize the liquid state of levitating or free-falling drops. The work of Schülli et al.1, however, opens up an attractive alternative to dispensing with the container: why not just disguise the container's surface as a liquid?
Schülli, T. U. et al. Nature 464, 1174–1177 (2010).
Spittle, J. A. Int. Mater. Rev. 51, 247–269 (2006).
Zachariassen, K. E. & Kristiansen, E. Cryobiology 41, 257–279 (2000).
Hannon, J. B., Kodambaka, S., Ross, F. M. & Tromp, R. M. Nature 440, 69–71 (2006).
Kelton, K. F. Solid State Phys. 45, 75–177 (1991).
Oh, S. H., Kauffmann, Y., Scheu, C., Kaplan, W. D. & Rühle, M. Science 310, 661–663 (2005).
Greer, A. L. Nature Mater. 5, 13–14 (2006).
Lee, G. W. et al. Phys. Rev. B 77, 184102 (2008).
Herlach, D. M., Cochrane, R. F., Egry, I., Fecht, H. J. & Greer, A. L. Int. Mater. Rev. 38, 273–347 (1993).
Oxtoby, D. W. Nature 347, 725–730 (1990).
About this article
Physical Chemistry Chemical Physics (2020)
Chemical ordering and crystal nucleation at the liquid surface: A comparison of Cu50Zr50 and Ni50Al50 alloys
The Journal of Chemical Physics (2018)
Modeling and simulation of microstructural evolution in Zr based Bulk Metallic Glass Matrix Composites during solidification
MRS Advances (2017)
Atomic study on the ordered structure in Al melts induced by liquid/substrate interface with Ti solute
Applied Physics Letters (2015)
Atomic dynamics in molten AlCu alloys of different compositions and at different temperatures by cold neutron scattering
Physica B: Condensed Matter (2013)