Cosmic rays are constantly crashing into the Earth's atmosphere, and when they do they produce a large shower of particles. There are various experiments to look at these showers, and the similar ones produced when very high energy gamma rays hit the atmosphere. But among the things that are produced in these showers are muons. (A muon is a heavier, short-lived cousin of an electron.) These muons, some of them, are moving fast enough to make it to ground level. They mostly come from above, but there's a fairly wide range of angles they come sleeting down from — in fact the distribution is roughly the cosine of the zenith angle. So they come from roughly the whole sky.
Not only do these muons pass through the atmosphere, they pass through solid rock; up to about twenty or thirty meters of solid rock, in fact, you can still detect appreciable numbers of muons. So the idea is to plant a detector somewhere nearby and below an interesting object and look how many muons it blocks. This is roughly equivalent to using an X-ray to look inside people. In fact, because the main limitation is collecting enough muons, the idea is to place several detectors, each of which collects muons and maps the space around it. Then you combine these several images to reconstruct the masses and voids inside the object of interest. It's a little like a CAT scan, though with a rather limited set of scans.
The detector is based on plastic scintillators: a muon that passes through one leaves behind a flash of Cherenkov light, which is channelled through a ("wavelength-changing") optical fibre to a photomultiplier tube. The detector is a cylinder made of scintillator tubes interwoven in such a way that a muon entering the cylinder will cause a flash in three tubes, localizing it. The muon then does the same thing on the way out, so the detector can reconstruct the straight-line track of the muon through the detector. It doesn't get any kind of energy resolution, which is kind of annoying because the lower-energy muons are more likely to have been scattered, but because the detector is fairly large, you get some degree of stereo vision even from a single detector.
Scattering, it turns out, is the limiting issue for the detector's resolution. As the muon passes through the solid object, a few of the muons will be slightly deflected by passing near atoms (more probably atomic nuclei). This small amount of scattering serves to blur the image slightly; in practice the blurring is on the order of a milliradian — a millimeter at arm's length, so not fantastically sharp. But you can't change the incoming muons, so the detector was built to have only about a milliradian resolution, which limits the number of photomultipliers needed.
To test the detector, they set up a detector at Sandia National Laboratories, which "has lots of tunnels for rent". It's about the size of a hot water heater, so they loaded it on a cart that rolls back and forth along a track. The whole setup was run remotely; the detector team logged in over the Net, simply asking the Sandia folks to roll the detector back and forth along the track every now and then. The Sandia folks then set up some test objects at ground level above the tunnel, and the detector team formed an image with which they tried to figure out the nature of the test objects.
I found the results kind of interesting; they presented them as analogous to "light field" camera images, that is, two-dimensional images that can be refocused at various depths. So the presentation showed a video that looked very like what you see when you use a microscope: you adjust the focus up and down through the object to try to build up a three-dimensional image in your mind.
The results of the test run were fairly positive. The easiest test object, a cube of lead bricks (I guess if you work at Sandia you just have these lying around) about a meter on a side, was unmissable, though the shape was a little vague. The reconstruction also managed to produce convincing-looking if blurry images of a piece of Jersey barrier and a culvert buried in sandbags. Some of the smaller objects, half-meter sandstone cubes apparently intended for lawn ornaments were harder to distinguish. But overall the test showed that given plenty of detectors or room and time to move one around, you really can do decent reconstructions.
(Now I'm wondering about those sandstone-cube lawn ornaments. Really? I guess I can imagine building a miniature Stonehenge on my lawn. If I had a lawn. And very tolerant neighbours. Maybe they'd fit in better on some of those Socorro lawns that just have a few cacti?)
So what about actual archaeology? Well, there was a group led by Luis Alvarez in the late 60s that used an older detector design to image the interior of the Great Pyramid, though unfortunately what they found was that there were no additional voids. But their system involved many tons of iron to get energy selectivity, and delicate and expensive spark-chamber detectors. So, not ideal for use in jungle-covered unexcavated Mayan ruins. This group's detectors, on the other hand, are water-heater-sized cylinders with one input (12 V) and one output (Ethernet) and they're fairly robust. So the idea seems fairly promising. But they still need to build a few more detectors — they have five, two of which are still underground at a test site of unspecified location — and they still need to sort out some issues of muon calibration and image formation. There are other groups working on this technique too.
Not this year, then, but maybe next year or the year after, we'll have people imaging the interiors of Mayan pyramids using the cosmic muons streaming through them. And hopefully no sudden visitations by aliens popping through ancient artifacts.