Leaving no stone unturned in the intense search for dark matter: focusing on the low-mass region with CDMS

Nov 19, 2014

By: Kristi Schneck

Over the past few years, several dark matter direct detection experiments have released results implying that dark matter may be hiding in an unexpected place.  These results suggest that dark matter particles may be ten times lighter than many physicists originally believed.  The Cryogenic Dark Matter Search (CDMS) collaboration, which includes many scientists from Stanford and KIPAC, has recently published several papers weighing in on the existence of this interesting type of dark matter particle.

Detecting dark matter in the lab is a bit like a game of billiards.  As the Earth moves through the Milky Way’s dark matter halo, it is possible that a dark matter particle will collide with an atom in our detector, as in the figure above.  This will cause the atom’s nucleus to recoil, much like how you send the eight ball careening towards the pocket at the end of a game of pool.  Unlike in billiards, though, the masses of the dark matter particle (the cue ball in this analogy) and the target nucleus can be wildly different.  As you can imagine, using a ping pong ball instead of the cue ball won’t cause the target ball to travel very far, since the ping pong ball cannot transfer much energy to the target.  This means that dark matter detectors need to be sensitive to extremely small energy deposits if we want to search for low-mass dark matter.  In addition, we would like to use a lighter target element if we’re looking for low-mass dark matter:  a tennis ball hit by a high-speed ping-pong ball is going to recoil more than a bowling ball will.

CDMS is currently operating 15 germanium detectors 2341 feet below ground in the Soudan Underground Lab in northern Minnesota.  (Visitors to the lab are reminded of this fact by the sign in the photo above, complete with one of the lab’s ubiquitous bats.)  The large rock overburden serves to shield the experiment from cosmic rays that hit the earth’s surface.  Additional shielding protects the detectors from ambient radioactivity and the few cosmic rays that make it deep into the earth.  

Even with all this shielding, a few non-dark-matter particles occasionally manage to deposit energy in the detectors, especially at the energies where a low-mass dark matter particle would show up.  When that happens, we can use statistical techniques to discriminate between these background events and the expected signature of dark matter.  Background events from photons and electrons tend to collide with the electrons of the detector, as shown in the first figure.  As a result, they produce a larger number of electrons than a dark matter particle, which is expected to interact with the nuclei.  We can use the ratio of charges collected to the recoil energy of the collision to discriminate between this type of background and a dark matter signal.  In addition, backgrounds from radioactive contamination tend to deposit energy near the surfaces of the detector, whereas a dark matter collision can occur anywhere in the detector.  We can therefore use the position of the event in the detector as a second way of separating signal from background.

One recent CDMS paper used machine learning to teach the computer to recognize the differences between signal and background and remove background events from data taken using the detectors currently operating at Soudan.  A second paper developed a sophisticated model of the expected backgrounds in CDMS-II, an earlier version of the CDMS experiment, and performed a likelihood analysis to look for dark matter.  Both analyses concluded that what was seen in the detectors was consistent with the expected backgrounds.

This figure shows dark matter mass and cross section, the two main characteristics of interest to experiments like CDMS, focusing on the low-mass region.  The mass of the dark matter particle is on the x-axis (in these units, 1 GeV/c^2 is about the mass of one proton).  The cross section, which measures how often dark matter interacts with normal matter, is on the y-axis.  Larger cross sections indicate more frequent interactions.  The solid black line and the dashed maroon line show the recent CDMS results.  Since both these results were consistent with the expected backgrounds, we set limits which rule out the existence of dark matter particles with certain characteristics.  In this plot, everything above a given line is excluded at the 90% confidence level by that experiment.  This means that, if we performed ten identical experiments, we expect that nine of them would see evidence of a dark matter particle with a given mass and cross section if such a particle existed.  We acknowledge the possibility that our real experiment could be the one out of ten that did not see dark matter by quoting the confidence level (as is the standard statistical procedure in physics).  For dark matter particles with mass and cross section further above and to the right of the curve, there’s an even smaller chance that our experiment would not find them.  

Compare these exclusion limits with the closed contours, which show the characteristics of dark matter particles that would be consistent with the possible signals seen by other experiments.  Under the simplest dark matter models, the new CDMS results (along with the LUX result in green) exclude the earlier possible dark matter signals with high confidence.  However, it may be possible to reduce the tension between experiments by changing the assumptions that go into calculating the exclusion limits.  Since we know next to nothing about how dark matter interacts, it’s possible that some experiments may be more sensitive due to the different properties of the target elements they use.  This is why it’s important to use several different targets in the search for dark matter.  

The planned upgrade to SuperCDMS, called SuperCDMS SNOLAB, will include both silicon and germanium detectors, whereas the currently running experiment uses only germanium.  Silicon is particularly sensitive to low-mass dark matter since it has a relatively light nucleus. A low-mass dark matter particle will deposit more energy in a silicon detector than a heavier germanium or xenon detector, as in the ping pong ball/tennis ball/bowling ball analogy.  In addition, one of the possible dark matter signals was seen in silicon detectors, so we would like to test this observation with the same type of target.  Combined with the xenon target used by LUX and its planned upgrade LZ, SuperCDMS SNOLAB will be capable of probing even more of the large region of dark matter parameter space -- watch for many more interesting results in the future!