Our sample cell appears schematically in Fig. 1. The suspension of volume fraction is confined to a thin horizontal layer between smooth clean glass walls separated by 2.5 m. This sample volume is hermetically sealed at the edges to minimize the invasion of air-borne contaminants. Reservoirs of ion exchange resin in diffusive contact with the observation volume further help to maintain chemical purity. Under these conditions the screening length is limited by the intrinsic concentration of counterions to roughly nm. Clean glass surfaces develop large negative surface charges in contact with water  which repel the negatively charged colloid and provide a smooth and featureless confining potential . They also prevent colloid from aggregating onto the walls through van der Waals interactions. The glass walls and their supports provide a transparent area of about 1 cm for observation.
We image the colloidal spheres using a conventional reflected light microscope with a 140 N.A. 1.3 oil immersion objective and a Xe arc illuminator filtered to avoid sample heating and degradation. The microscope's depth of focus is 0.2 m, which is comparable to a sphere diameter. Images are captured at a rate of 30 frames per second with a CCD video camera and recorded on video tape before being digitized and analyzed on frame-by-frame basis. The total system magnification is 68.2 nm/pixel at the CCD camera. Although the spheres are too small to form diffraction-limited images, they appear as distinct bright features on an otherwise dark background, as can be seen in Fig. 2.
The upper glass surface of the sample volume has a pair of parallel plate thin film electrodes patterned onto it in transparent chemically inert -InO. A potential applied across the 3 mm wide gap induces flow in the suspension through electrophoresis and electro-osmosis . This motion is resisted by the rigidity of the colloidal lattice and so the observed direction of flow depends also on the lattice's orientation, at least at low flow rates. Higher flow rates shear-align the lattice. Because the flowing lattice is damaged in an uncontrolled fashion at the edges of the cell, we drive the flow with a small amplitude (1Vpp) sinusoidal signal to prevent propagation of this damage into the observation region. The peak-to-peak amplitude of the response at 1 Hz is roughly 100 m.
The lattice deforms both elastically and plastically as it flows around obstructions such as the columnar feature shown schematically in Fig. 1(b). Measurements of the lattice's deformation can be used to estimate its elastic moduli . We will describe such measurements on geometrically confined suspensions elsewhere. Sufficiently vigorous shearing around obstructions introduces a sufficient density of defects to melt the lattice as can be seen in Figs. 2(c) and (d). The melting point is gauged by the disappearance of shear rigidity, at which point the driving signal is abruptly turned off. There is a brief coherent flow of particles as the pressure in the system equilibrates after the driving stops. Because imaging is difficult during this transient, we start taking data at sec after cessation of shearing.
Although shear melting releases the lattice's latent heat of crystallization, the heat capacity of the surrounding water is so large that the system's overall temperature does not change measurably. This isothermal isobaric process therefore produces in a supercooled fluid whose degree of supercooling is exactly the binding free energy of the initial crystal's unit cell. Following our earlier work , we use this observation to define an effective supercooling parameter:
where is the volume fraction at melting for this colloid at the experimental temperature and ionic environment. Eqn. (4) corresponds to the usual supercooling parameter for conventional materials under the assumption that and after linearizing about the typical separation . This is at best a rough estimate of the degree of supercooling for the present study since is measured for bulk unconfined colloid in a different sample cell. The sample in this study has while the melting point for the bulk fcc crystal is about . The value corresponds to a supercooling of roughly C in water.