The Cryogenic Dark Matter Search (CDMS) is a series of experiments designed to directly detect particle dark matter in the form of WIMPs. Using an array of semiconductor detectors at millikelvin temperatures, CDMS has set the most sensitive limits to date on the interactions of WIMP dark matter with terrestrial materials. The first experiment, CDMS I, was run in a tunnel under the Stanford University campus. The current experiment, SuperCDMS, is located deep underground in the Soudan Mine in northern Minnesota.
Observations of the large-scale structure of the universe show that matter is aggregated into very large structures that have not had time to form under the force of their own self-gravitation. It is generally believed that some form of missing mass is responsible for increasing the gravitational force at these scales, although this mass has not been directly observed. This is a problem; normal matter in space will heat up until it gives off light, so if this missing mass exists, it is generally assumed to be in a form that is not commonly observed on earth.
A number of proposed candidates for the missing mass have been put forward over time. Early candidates included heavy baryons that would have had to be created in the big bang, but more recent work on nucleosynthesis seems to have ruled most of these out. Another candidate are new types of particles known as weakly interacting massive particles, or "WIMP"s. As the name implies, WIMPs interact weakly with normal matter, which explains why they are not easily visible.
Detecting WIMPs thus presents a problem; if the WIMPs are very weakly interacting, detecting them will be extremely difficult. Detectors like CDMS and similar experiments measure huge numbers of interactions within their detector volume in order to find the extremely rare WIMP events.
The CDMS detectors measure the ionization and phonons produced by every particle interaction in their germanium and silicon crystal substrates. These two measurements determine the energy deposited in the crystal in each interaction, but also give information about what kind of particle caused the event. The ratio of ionization signal to phonon signal differs for particle interactions with atomic electrons ("electron recoils") and atomic nuclei ("nuclear recoils"). The vast majority of background particle interactions are electron recoils, while WIMPs (and neutrons) are expected to produce nuclear recoils. This allows WIMP-scattering events to be identified even though they are rare compared to the vast majority of unwanted background interactions.