J. Phys. Chem. 1994,98, 8096-8100
8096
Extrinsic Premelting at the Ice-Glass Interface D. Beaglehole' and P. Wilson Physics Department, Victoria University, P.O. Box 600, Wellington, New Zealand Received: February 8, 1994; In Final Form: May 7, 1994'
This paper describes an investigation into premelting a t the ice-glass interface. Using ellipsometry to monitor the thickness of the iceglass interface, we show that no intrinsic premelting occurs a t the interface between ice and smooth clean glass. However, premelting is observed a t interfaces with roughened glass surfaces and with the introduction of surface impurities. Roughened glass which was made hydrophobic showed no premelting. The implications for theories of premelting are discussed.
Introduction Premelting at the ice-solid interface region is thought to be a component of such diverse phenomena as frost heave in soils and glacier sliding. There are many reports of unfrozen water in subzero soils,' but there is no general agreement as to the cause of the premelting or the nature of the water a t the interfaces.* Kuroda3 has proposed a theory to explain frost heave of soils based on the thermodynamical properties of a thin water layer between ice and soil particles: more recently Dash and colleagues have considered layers of water between ice and solids to explain frost heave,4 regelation,5 and in relation to the ice-vapor surface thunderstorm activity.6 A recent paper of theirs analyzes water layers between ice and porous solids.7 Beagleholes has discussed the wetting of small particles and has considered impurity effects in more detail. Nenow and Trayanovg review surface premelting phenomena. Kroeslo has made molecular dynamic simulations of surface melting on basal faces of ice. It is now understood that therecan bevarious causes for surface melting. An intrinsic lowering of the excess interfacial energy may occur for flat surfaces by the formation of a thin liquid or quasiliquid layer, Extrinsic mechanismssuch as surface roughness (interface curvature) or impurities soluble only in a liquid phase may also produce an interfacial liquid layer. The thickness of a liquid layer on the surface of the ice will generally depend upon temperature, and if complete wetting occurs will diverge as AT 0 as A P where AT = T , - T , with T , the bulk melting temperature. When the excess surface energy is due to van der Waals interaction, the exponent a for intrinsic premelting is predicted to be Measurements of exponents greater than have been attributed to the curvature effects of small particle size, and impurities also lead to larger values. Dash5 has analyzed Gilpin'sll regelation data and has found some evidence for the van der Waals exponent. The dynamics of water layers on small particles has been studied by nuclear magnetic resonance by Mizuno and Hanafusa.12 Maruyama et al.13 used neutron diffraction to look at ice premelting in powders of graphite or talc and reported unfrozen water to temperatures below -30 OC. Gay et al.14 have studied exfoliated graphiteflakes. In these works exponents have been found to lie in the range
-
0.5-2.
Powders, porous solids, and soils are not well characterized, and the interpretation of measurements on these materials makes assumptions about size distributions, purity, ice crystallinity, etc. In an attempt to provide more direct evidence of premelting, Furukawa et al.IsJ6 made an ellipsometric study of the ice-glass interface for flat glass surfaces and single ice crystals. They found evidence for a surface layer with a thickness of about 100 A and refractive index midway between glass and ice at ~
e Abstract
published in Advance ACS Abstracts, July 1, 1994.
temperatures below -1 O C . They attributed this to the surface roughness of the glass causing a region of intermediate dielectric constant. However, they also found a sharp increase in the thickness of that layer at temperatures above -1 O C which they described as a premelting layer of liquid water. They did not consider optical anisotropy in the calculations. To our knowledge, this study provides the only experimental evidence for premelting on flat ice-solid surfaces with single crystals of ice. A premelting layer of liquid water has been shown to exist at the ice-vapor interface using a variety of experimental techniques such as surface electrical conductivity, surface NMR, and ellipsometry. Single crystals have been studied using ellipsometry by Beaglehole and Nason," who found layers on the prism but not the basal face. At -1.5 O C they estimated the prism face thickness to be 130 A, reducing to a monolayer at -10 OC. Furukawa et a1.18 found somewhat similar results but thicker layers. Elbaum et al.,I9using reflectivity measurements,observed partial wetting of the basal and prism faces when the gas phase was pure water vapor but complete wetting on the basal face and on vicinal surfaces around both the basal and prism faces when air was added to the gas phase. Prediction of the intrinsic liquid layer is straightforward in principle but in practice uncertain due to uncertainty in the surface free energy parameters and, in the case of the ice-water system, due to the close similarity of the ice and water high-frequency electromagnetic properties. It is usually assumed that the liquid layer is stabilized by van der Waals dispersion interactions. In their simplest form, the strength and sign of this interaction are measured by the Hamaker constant W a (e - q)(e - ez), where el and e2 are the optical dielectric constants of the bounding media and c is the dielectric constant of the intruding layer. For the ice-vapor interface, e > e z , q and the sign of W will be positive. For the ice-solid interface where the solid (glass) has a dielectric constantof -2.25, Wwillbenegative. Thus,ifthereispremelting at the ice-vapor interface, there should not be premelting at the ice-solid interface. Elbaum and Schickz0have performed a more thorough calculation of the dispersion interaction for the i c e vapor system and have suggested that partial wetting occurs at the ice-vapor interface-the thickness of the layer should not diverge but rather remain finite (at around 30 A) at T,. The calculation has not been made for the ice-solid interface. In view of the uncertainty in the mechanism for wetting at the ice-solid interface, we have repeated the experiments of Furukawa,I5 studying the ice-glass interface using ellipsometry. In particular, we have used glass surfaces with differing surface roughness, and with thin layers of salt impurity. We have not found wetting layers on smooth impurity-free glass surfaces, though impurities and roughness both induce wetting. In the theoretical work7J it is always supposed that the wetting condition on the surface free energies is satisfied, namely that ub > uiw + uws,where uis, uiw,and uwsare the surface energies of the
0022-3654/94/2098-8096$04.50/0 0 1994 American Chemical Society
The Journal of Physical Chemisiry. Vol. 98, No. 33, 1994 8097
Extrinsic Premelting at the Ice-Glass Interface icesolid, ice-water, and water-solid interfaces, respectively. Since the experiments reported here have shown that wetting does not occur between ice and smooth clean glass surfaces, we have to consider alternatives to the theoriesmentioned above. In thelast section we give some consideration to the consequences of this finding for theories of premelting.
reflected
,'
light
J
h
incident light
Ellipsometry Ellipsometry is an optical technique which is sensitive to the surface properties with thickness resolutions of molecular dimensions. Transition layers lying between ice and glass induce small amounts of circular polarization in the reflection of a linearly polarized incident beam, and the amount of circular polarization, which is measured by ellipsometry, depends upon the thickness and dielectric properties of the transition layer. Microscopic roughness of a surface also induces circular polarization, and we are able to use ellipsometry to monitor the roughness on the glass surface during etching. When the transition layer, or in the case of the rough surface the height fluctuation, is much thinner than the wavelength of light, the ellipticity is linearly dependent upon its thickness i. However, as thelayer grows, theellipticity undergoes interference oscillations, with the first peak in ellipticity occurring a t typically 120 nm for the case of a water layer between ice and glass. We will see later peaks in the ellipticity ofsamples showing intruding wetting layers. The height fluctuations even of the most heavily etched glass surfaces are always quite small compared with the wavelength oflight, and roughness therefore produces a background ellipticity independent of sample temperature. That this background is due to roughness rather than an intervening liquid layer can he checked by measuring theellipticity of theglass-water interface, since the dielectric constants of ice and water have almost the same value, and in the latter case there will he no intervening layer. We measure the coefficient of ellipticity at the Brewster angle. Themethod is fullydescrihedelsewhere.".22 Iceis weakly optically anisotropic, and since we use single crystals of ice with known optic axis orientation, specific dielectric constants must he assigned to each orientation, and we use the full expressions for the ellipticity and dielectric constants given by Lekner" and Beaglehole and Wils0n.2~However, since the dielectric constant of glass (ei = 2.25) is not close to that of ice (ezx ch 1.72), theanisotropy has littleeffect here (though it played an important role in the study of the ice-water interface2').
- -
The lee-Glass Samples
Ice crystals were grown from distilled, deionized water (>18 MR cm) by a modified Bridgemann method and then divided into single crystals by irradiation from a 1000 W halogen lamp. This irradiation also produced Tyndall figures which indicated theorientationof thecrystals-the crystals werecut so that either basal or prism faces were in contact with the glass surface. The samples, 3.5 X 3.5 X 5 cm in size, were mounted on a cold stage (Figure I ) with the top surface melted to he as flat as possible. The glass, either a rectangular slab of float soda glass or a half cylinder of specially polished borosilicate glass, was placed on the ice with a thermocouple (