In high school, my Earth Science teacher held up a globe and announced that the world was spinning itself apart. He spun the globe, orienting the north pole towards the class to show its counter-clockwise west-to-east rotation. Then he turned the globe around revealing, as viewed from above the south pole, that the earth was simultaneously spinning clockwise. The Earth must be spinning itself apart! Q.E.D. A nice little demonstration about rotation and asymmetry.
So, when I became a card-carrying member of the Galaxy Zoo community, I was more than a little surprised to discover that they were interested in having users distinguish between clockwise and counter-clockwise spiral galaxies. For those of you who haven’t heard of Galaxy Zoo, it is an effort to classify the morphology (spiral, elliptical, merger) of galaxies by enlisting the aid of Internet users. Galaxy Zoo makes use of images millions of galaxies catalogued from the Sloan Digital Sky Survey (SDSS). The classification of galaxies has been a topic of machine learning research for many years. Not surprisingly, computer-based classification is less and less accurate as we look deeper into space at smaller fainter galaxies. On the other hand, manual inspection and classification of the individual images can be both time-consuming and mind-numbing. But Galaxy Zoo, with its simple interface and active user forums succeeds in engaging users with the opportunity to make a real and valuable contribution to science. It is particularly exciting to realize that most of the galaxies being examined are extracted from such a tiny portion of the sky that they are in effect being examined for the first time, opening up the opportunity for fundamental discoveries – most notably Hanny’s Voorwerp, discovered last year by a dutch school teacher.
The Galaxy Zoo interface – one of the 1,700+ galaxies I’ve classified – hopefully correctly!
But why bother to distinguish clockwise and counter-clockwise spiral galaxies? Afterall, if spiral galaxies are strewn around with random orientations, then the rotational directions isn’t telling you anything physical about the galaxy itself, only about your relative perspective. Could this underlying assumption about randomness of galaxy orientations be wrong? Dr. Michael Longo of the University of Michigan, in a recent paper with the intriguing title Does the Universe Have a Handedness, writes:
Symmetry has a strong appeal to the human psyche. Nature, however, exhibits some surprising asymmetries. On the smallest scales, an asymmetry (parity violation) was found in the angular distribution of electrons in the beta decay of spin oriented 60 Co, confirming the proposal by Lee and Yang (1956) that parity was violated in weak decays. On the molecular scale, there is a large predominance of left-handed amino acids over right-handed ones in organisms, the origin of which is still not well understood. It is reasonable to ask if nature exhibits such an asymmetry on the largest scales.
Longo’s analysis of 2,817 visually characterized spiral galaxies gave R (clockwise) = 1256, L (counterclockwise) = 1360, and U (unknown) = 201, a small (8.3%) but statistically significant enrichment in the observed number of counter-clockwise spirals. As an example of a potential violation of the cosmological principle, this observation has enormous implications for understanding the origin and evolution of the universe. Longo has proposed a large-scale magnetic field as a source for this apparent anisotropy.
Galaxy Zoo, with its 85,000+ users (recently reported to be now over 150,000 users) generated 36 million classifications for 893,000 relatively bright galaxies – a significant sample-size increase! So what did the Galaxy Zoo team find? The results are described in a 2008 paper by Land et al. entitled Galaxy Zoo: The large-scale spin statistics of spiral galaxies in the Sloan Digital Sky Survey. They extracted two subsets of galaxies for which their was greater classification consensus among users. (On average, each galaxy is classified by about 40 different people.) One “clean” subset consisted of galaxies for which there was 80% agreement. A second, superclean subsample, achieved 95% agreement. Sure enough, there are more counter-clockwise spirals in both subsamples. In the clean subsample, the results were clockwise=17,100 / counterclockwise=18,471, an 8.0% increase in the number of counter-clockwise spiral galaxies – similar to the Longo result mentioned above. For the smaller superclean subset, the % difference is even greater (clockwise=6106 / counterclockwise=7034). So at first glance it seems that the rotational assymetry of spiral galaxies has been confirmed. But that is not the end of the story.
Land et al. conducted a bias study to determine if users might be sub-consciously skewing the results. For example, the authors point out that the human visual system might be better adept at detecting one orientation versus another, or there might be something about the Galaxy Zoo interface that causes users to err on the side of counter-clockwise classifications. The way they check for bias is quite simple: they generated another subset of galaxies and flipped the images! In the absence of bias, you would expect a reversal in the percentages of each classification – that is, more clockwise spirals. But in fact, even when the images are reversed, counter-clockwise spirals occur more frequently. When you adjust the counts for this inherent bias, the number of clockwise and counter-clockwise galaxies are almost identical. The conclusion (quoting from Land et al):
These results aid our understanding of the formation of large-scale structure. They limit the possibility of coherent large-scale magnetic fields that assist in the formation of galactic angular momentum (Longo 2007b). Similarly they constrain scenarios of galaxy formation – providing support for tidal torque theory (Barnes & Efstathiou 1987) as opposed to scenarios in which the angular momentum of galaxies are coherent over large scales. Such scenarios include those where the angular momentum of galaxies originates from primordial vorticity, or from collapsing inflows (Sugai & Iye 1995).
In fairness, I should note that Longo also attempted to account for possible human bias by randomly reversing the images prior to classification, so the discrepancy in the two results is still something of a mystery.
Galaxy Zoo highlights the remarkable potential for everyday users to contribute to astrophysical research in this new age of survey astronomy. It also highlights the additional steps required to account for inherent bias, whenever humans are a fundamental part of the data acquisition process. Importantly, as the Galaxy Zoo team demonstrates, these intrinsic sources of bias can be rigorously accounted for, and I suspect that an ever-growing database of galaxy morphology classifications will not only enable research requiring increasining sensitive statistical analysis, but also aid in the development of fully-automated maching learning algorithms capable of one day exceeding human performance. The mere fact that Galaxy Zoo continues to inspire both general interest in Astronomy as well as a greater appreciation for the immensity and diversity of the Universe is, in itself, a worthy achievement.
Posted by John Rachlin 