The physics of colloidal glass

Work done by Eric Weeks while I was at Harvard University.

Lab Home -- People -- Experimental facilities -- Lab pictures -- Experimental pictures

Brief explanation: When molten glass is cooled, it flows slower and slower, and at some point it's essentially solid. Unlike water, which is either water or ice, glass smoothly changes from a fluid to a really slow fluid to a glass. Why? We use small colloidal particles to model atoms in a glass; we look at them with a microscope. As we pack the particles together, if one of them wants to move, its neighbors have to cooperate. As they are packed even tighter, more of the particles have to cooperate for any to move. Perhaps when all of the particles in the sample have to cooperate, the sample is essentially a solid -- thus explaining what is happening with glass as it's cooled. This is a classic theory, and for the first time we can look at a real physical experiment and directly see cooperative motion.

Most solid materials are crystals; the atoms are arranged in regular patterns, for example, stacked like cannonballs. The structure of crystals has been carefully studied for quite a long time, and is related to properties such as the strength of materials, their conductivity, how they break, and how they form.

Glass has no underlying regular structure. Instead, the atoms or molecules making up a glass are jumbled together, like the picture at left. They may be packed in so tightly they cannot move, but they are not packed in a regular way. Some materials naturally form glasses when they are cooled, such as silica (SiO2) (the primary chemical component of normal glass). What's weird is that structurally, glasses are the same as liquids -- if you just look at a microscopic snapshot of the position of the atoms, you can't tell the difference (whereas it's really obvious that a crystal is something different from a liquid). So why is a glass a glass, and not a liquid? Or is it a liquid...?

When a glass-forming liquid is cooled, its viscosity increases -- it flows slower and slower. A simple definition of a glass is a liquid with a viscosity that is 10,000,000,000,000 times larger than the viscosity of water. This is somewhat arbitrary. In fact, a big question is, do glasses actually flow, or are they completely solid (infinite viscosity)? My best answer is that I believe glasses do not flow any more than any other solid object flows, although perhaps even rocks and glasses and crystals will flow in some fashion if we wait eons.

A colloid is simply a fluid filled with lots of very small solid particles; this includes black ink, blood (filled with blood cells), and paint (filled with particles which stick to surfaces when they dry). Typically these particles are very small, between 1 nm and 1000 nm (one-millionth to one-thousandth of a millimeter). We use particles that are 0.002 mm in diameter, made from PMMA (the same material which is Plexiglas, if you have a much larger hunk of it). These particles basically act like marbles, that is, they ignore each other unless they bump into each other -- they don't have electric charges on them, for example.

Sometimes, our colloidal particles form crystalline arrays, like the picture at left, or the hexagons shown above. These are similar in many ways to regular atomic crystals. In other cases, the colloidal particles pack close together in a random way, and form a colloidal glass; this is what is shown in some of the other pictures on this page. While normal atomic solids are formed by cooling, colloidal crystals and glasses are formed by cramming the particles together, usually by centrifuging them.

(animated gif)
We use a confocal microscope to take 3D pictures of our samples, to see what the individual particles are doing as they move around in a colloidal glass. We can follow several thousand particles simultaneously, and watch their motion for several hours (sometimes several days). We look at how their motion changes when they are packed closer together, as the sample becomes a glass. Hopefully by understanding what occurs in a colloidal glass system, we can learn general properties of all glasses. The larger size of colloidal particles (as compared to atoms in regular glasses) make them possible to see, and it also means they move slower -- thus we don't need really fast electronics or techniques to see what they're doing, unlike people who study molecular glasses, who have to be extremely clever. (We make no such claims.)

We find that particles have to cooperate to move: if one particle can move a little ways, then one of its neighboring particles can move into the space left behind by the first particle; and then perhaps a third particle can follow the second particle, and so on. The more glass-like the sample is, the more particles cooperate at the same time. However, it takes longer and longer times before we see one of these cooperative events. Thus, it is possible that when all of the particles have to cooperate in order for any of them to move, you have a glass. Perhaps the time between cooperative rearrangements diverges, as well as the number of particles needed to cooperate, and it is the divergence of these two dyanamical quantities that distinguishes a glass from a liquid. Our data isn't inconsistent with this hypothesis -- which is a weasly scientist way of saying we have no idea if this really happens, but it's an intriguing idea that could be possible, so I mention it here on this web page. At least, the possibility is one reason why I find this interesting.

This is from the experimental data, showing cooperative clusters of particles (the largest in this image is highlighted). All of the particles are actually the same size (the size of the large ones), I have just made the slower particles smaller so that the fast ones stand out.

This work was conducted at Harvard University, before I came to Emory University.

I also have two movies you can download, below. Each of them is 10 MB in size. They are animated GIFs, so they should display in any graphical browser without any additional work or plugins.

The movies are similar to the picture above: fast particles are drawn big (to scale) and slow particles are drawn small (not to scale). The scale bar on the left indicates the time scale, as noted below. The scale bar on the right indicates the fraction of the sample that is "fast" according to our somewhat arbitrary definition that over time, on average 5% of the sample will be fast (thus this scale bar will fluctuate around 5%).

supercooled fluid
volume fraction = 56%
dt* = 1000 s

For a supercooled fluid, we have very large clusters of cooperative fast particles. This movie is from the sample that is closest to the glass transition, without actually being a glass.

volume fraction = 60%
dt* = 3000 s

For glasses, there are no large cooperative clusters. Since we're defining the fastest 5% of the particles as "fast", there is always some activity going on. However, it does not appear to involve many particles moving in a cooperative fashion as occurs for the liquid-like samples. Instead, there are small groups of particles which move slowly side-to-side, in somewhat of a cooperative fashion. This motion never results in large-scale rearrangements, unlike the supercooled fluids.

If you have questions or comments send me email: weeks(at)


An article in Science by C. A. Angell (see the references below) lists a few other intriguing facts about glasses. While glasses occur naturally (such as volcanic glass), it has recently been discovered that another natural glass is in comets. Water in comets exists in a glass state (unlike water here on Earth!). It has also been found that the glass transition is related to the suspension of insect life in a desert during a drought.

Eric R. Weeks