What I did on my Summer Vacation, part I

OK, not a vacation in the true sense of the word: I’ve been in the US, attending meetings (in Berkeley), workshops (in Santa Fe), conferences (in Pasadena) and, because I can’t seem to escape them, teleconferences everywhere and all the time.


In Berkeley, I attended the first all-hands collaboration meeting for PolarBear, an experiment that will measure the polarization of the CMB from a telescope that will eventually be situated on the Atacama desert plain in Chile — one of the highest, driest, least accessible places on the earth, and one of the least contaminated with light or radio interference. (Despite the name of the experiment, there are no polar bears there.) First, we’ll test it at the somewhat less remote White Mountain facility in California, shake out all the bugs. PolarBear is one of a new generation of experiments that will measure the CMB using not just a few tens of detectors, but a few thousand, which brings with it all sorts of technical challenges. In hardware, the first challenge is simply making so many detectors and keeping their properties uniform each to each — these are among the most sensitive microwave detectors ever built, essentially as good as the constraints of quantum mechanics and thermodynamics allow. The second, related to the first, is to pack as many of these into a small space — the focal plane of the telescope — as possible. Traditionally, microwave detectors have used horns to guide the electromagnetic waves from the sky onto the detectors, but those horns are much wider than the detector hardware. For experiments like PolarBear, we put the detectors themselves right at the focus of the telescope and make each of them into a little antenna, receiving directly the focused light after passing through a hemispherical lens. The final hardware challenge is to get the information from these thousands of detectors off of the telescope and into our computers, which the PolarBear designers have solved with a new technique called “frequency-domain multiplexing“. Sort of like the way FM radio manages to convey the full spectrum of sound by modulating at a particular frequency, the very high-tech SQUIDs (Superconducting QUantum Interference Devices) can then amplify these tiny CMB signals into data we can analyze.

In fact, the data analysis and computing challenges are almost as significant as those faced in hardware. With thousands of detectors and a telescope that will run for the better part of several, we have many orders of magnitude more CMB data than we’ve ever dealt with before, combined with a sensitivity goal better than a millionth of a degree. By adding more and more detectors, we can make the raw experiment itself sensitive enough to do this. What we don’t know is whether we can eliminate everything else that can possibly contaminate our results: light may spill over our shield from the 300 degree ground or directly from the atmosphere; dust in our solar system or our galaxy also glows in the bands we want to measure, as do external galaxies millions of light-years away. So our task is to compress the terabytes of data into a few interesting numbers (like the energy scale of inflation) and to simultaneously separate the cosmic signal from the that produced by instrument and from the rest of the Universe (which may be much brighter!). Suffice to say, we have some good ideas but until we’re confronted with real data we won’t know how successful we’ll be.

Plus, I ate bagels (better than London; not as good as New York) and burritos, and bought shoes at cheap American prices (at least when I think in British Pounds).

Next up, Santa Fe