On the second day of the astrometry.net group meeting, Mierle, Lang, and I realized (or, more properly, bolstered our nascent view) that indexing all stellar quads (recall we are using quads of stars to get first guesses on image positions on the sky) on the entire sky is a very bad idea, for the two related reasons that (a) many stars tend to be involved in many quads, so the loss of a single star in the test image at test time can lead to a large loss in the number of expected matching quads, and (b) small n-tuples (like 5-tuples and 6-tuples) of stars can produce enormous numbers of stars, so we get false positive matches because there are many similar-looking 5-tuples on the sky. Yes, many! So we planned for Mierle and Lang to work tonight working on an index that contains no more than one quad for each star.
On the first day of the astrometry.net group meeting, Lang, Mierle, and I worked on the SDSS data demonstration project, with Mierle assigned to get post-mortem code (code for understanding failures and false positives) working again (after a re-write) and Lang assigned to re-run a subsample of SDSS fields with different settings of the tolerance and the number of stars to use in quads.
I discussed merger rate projects with Eyal Kazin, a possible new member of the NYU galaxy astrophysics group!
I spent an inordinate amount of time on the weekend tracking down a bug in the (hacked) astrometry.net prototype tweak code (take an image with a close-to-correct WCS and make it beautiful, using either SDSS or USNO-B1.0 stars). I got it, and Willman is now solving her astrometry on a bunch of images. This week is the astrometry.net group meeting, so I hope to get that code re-written by a competent programmer.
I downloaded the Bright Star Catalog, and wrote a reader for it, and plotted it—it contains all stars brighter than about V=6.5 mag—for use in some SDSS demos we make, including our famous zoom, which is very out of date. I am preparing to re-make it, but this time it will show all the constellations!
Sheldon previewed his job talk for UC Davis this morning; his results—which show the cross-correlation between mass and galaxy clusters—are awesome. They strongly confirm the NFW profile and what we know about intermediate-scale clustering.
Padmanabhan (Princeton) and I discussed averaging SDSS spectra, transparency, and future massive spectroscopic surveys. No real project yet, but many on the horizon and in planning.
I prepared slides for my talk at ISCAP at Columbia tomorrow. I am going to concentrate on constraining small-scale dynamics of the dark and light sectors using galaxy interactions, mergers, and accretion.
David Kaplan (JHU) gave the HET talk about imposing severe symmetries to cancel out cosmological-constant-like terms (ie, vacuum energy density) in the standard model. It involved duplicating the entire standard model with a whole negative-energy "ghost" standard model, and thus having enormous numbers of terms cancel out. A nice idea (because the small cosmological constant would be required by the symmetry, broken only by gravity), but the theory has super-strange properties, like a rapidly decaying vacuum (!) and possibly causality issues.
I spent the rest of my research time today hacking on the morphology–color–density figures and the post-starburst environment overlap-region analysis.
Worked on figures for the future paper on the morphology–color–density relation. We have been saying for years (eg, here and here) that indirect indicators of galaxy morphology do not correlate directly with environmental density, but that they correlate with galaxy star-formation history, which themselves correlate with density. I realized (very late, I must admit), that we can show this directly by simply showing true-color images of the damned galaxies; everyone can "do" morphology however they want on those images.
Maintaining 15 Tb of astronomical data takes a significant amount of human effort! If I were clever, I would have perl scripts take care of all contingencies. Blanton has written some great stuff that keeps our directory structure (built out of many cross-mounted partitions on many large servers) all consistent and unified even when machines go down or data are moved from one to another. But we still hit problems relating to the finite size of our directories and partitions (and the human-factors value in having related data close in hardware). I spent a good fraction of my research time this weekend cleaning up.
I started the process of investigating the effect of spectrograph constraints (the SDSS spectrograph can't simultaneously take spectra of two galaxies closer than 55 arcsec) on our galaxy environments results. I am investigating it empirically, using the fact that a large fraction of the sky is covered by the spectrograph twice (the
plates overlap), and hence many of the dropped targets get picked back up again.
Roweis and I talked a bit about the agenda for the full-week astrometry.net group meeting coming up in about 10 days. We are going to focus on getting our alpha system up on the WWW.
Dragan Huterer (Chicago) gave two talks, one about statistical anomalies in the CMB at large angular scales, which are very interesting (despite being very a posteriori). Like most cosmologists, I assume these will turn out to be either flukes or systematic errors. His other talk was about the future of constraining the dark energy with observations.
Jackie Chen (Chicago) spoke about using lensing and galaxy surveys to look at dark-matter substructure in collapsed dark-matter halos. She sowed quite a bit of doubt about the current lensing detections of substructure.
I continued downloading lots of public imaging data relevant to the PRIMUS project. There are a lot of public multi-wavelength deep imaging data sets.
Quintero provoked a nice discussion of filaments in group meeting. Evidently filaments do exist; the questions are as follows: (1) Are the galaxies in filaments different from other galaxies not in filaments but that live in the same local mass density environments? (2) Can we understand the evolution of clusters as the infall (and simultaneous evolution) of galaxies along the filaments? Not sure if we will pursue these questions, but there clearly are
first steps we could take.
Masjedi showed me brand-new results on the rate of mergers of LRGs (that's
luminous red galaxies) with smaller galaxies, as a function of color and magnitude of the smaller galaxy. Of course he really measures the cross-correlation functions, but we can use those to (severely) limit the merger rate. This project will give us the mass spectrum of mergers (or constrain it) and, we hope, limit the total mass accretion rate.
As my loyal readers know (both of them, I hope), the astrometry.net project, among other things, will use an index of quadrangles of stars to identify fields
blind, ie, even when the user doesn't
remember the image pointing, rotation, or scale. I spent quite a bit of time this weekend working out (and recalling) the statistics of such quadrangles: How many are there on the sky? How many unique quads can we index, given finite positional errors? How many quads do we expect in a finite, rectangular image of given dimensions? What fraction of fields will we fail to solve, given the need for redundant successes? Etc. The goal is to re-solve all of the SDSS images blind.
Joe Hennawi (Berkeley) gave a nice talk on the clustering of quasars, and on what can be learned about quasar impact on environment from quasar pairs close on the sky (but not in redshift). There is a big mystery about the proximity effect—in the transverse direction there is lots of absorption by neutral gas, but along the line of sight the quasar ionizes everything—which either means that things are very anisotropic or quasar lifetimes are super-short, I think. We spent time afterwards discussing clustering at small scales, because his work has a lot of overlap with that of Masjedi.
I failed to mention that Dave Monet (USNO) sent me all the data upon which the USNO-B1.0 catalog was constructed. This is yet another respect in which Monet is a great scientist, contributing always to the public good! Now we'll see if the astrometry.net crowd is crazy enough to do anything serious with these data...
On the bus (yes, the flammable Chinatown bus) to Boston and in Burles's office, I spent time thinking about fundamental observed properties of our Universe, inspired by our transparency discussions of the last few weeks. I was also inspired by the non-observation that the Universe is accelerating. The world seems to take as gospel that acceleration has been demonstrated, but the data are in fact fully consistent with positive q0 at the present day. What really has been shown is that if you restrict yourself to the two-dimensional space of matter plus cosmological constant, the best-fit model has present-day acceleration. That does not say that there is sufficient weight in the low-redshift data to see the acceleration locally; indeed there is not. It may be a subtle distinction, but it is an important one, because there can easily be new components emerging at the present day that are important but don't have much impact on the current observations.
That said, I was trying to list the truly fundamental observations, the ones that will never go away. Here are some:
- The night sky is dark (Olbers's paradox).
- We are not being shredded by gravitational radiation (the gravitational radiation equivalent of Olbers's paradox).
- Spacetime is locally flat (related to the above but stronger, really).
- The Universe is transparent at
- The Universe is isotropic.
- The Universe is expanding.
- The laws of physics look very similar at enormous distances from us (and therefore in the distant past).
- The Universe was hotter in the past.
Here's a window into my research soul: The figure below shows three white iso-failure-rate contour lines (for 90 percent success, 99 percent, and 99.9 percent) as a function of the density of standard stars and the angular scale of our quad index, assuming Poisson statistics, no problems/issues and that we need three quads in an image to declare a match
good. The cyan lines are lines of constant index size (ie, constant RAM usage). The plot shows that we are better going with a higher density of stars and a smaller angular scale. However, what the plot does not show is that if we make arbitrarily small quads, we get less good astrometric solutions (ie, we have to increase the quad–quad match tolerance).
In my few minutes of research time today, I attended an informal talk by Gabadadze (NYU) on modifications to gravity that involve a light scalar interacting with the graviton. This was inspired by last week's talk by van Acoleyen (mentioned here), and clarified some issues about these kinds of theories. Gabadadze's position seems to be that if you give such a theory enough freedom to explain both acceleration and dark-matter halos, you have to do enormous amounts of fine tuning to get things right. No symmetries protect you.
Maller, Quintero, and I discussed Quintero's qualitative mini-project to look at the filament paradigm (ie, the idea that galaxies flow into clusters along filaments). Maller shared my feeling that this effect would not be startlingly obvious in the data, despite the fact that it is widely accepted lore in the 'munity.
I worked out the (theoretical) blind astrometry failure rate as a function of field size, standard-star density, and angular scale of our quad indexing (remember that?). The failure rate is a very strong function of all variables, for the main reason that we depend on quadrangles of stars, and their abundance goes as the fourth power of number density, fourth power of field size, and sixth power of index angular scale. My calculation is crude, because it assumes all data are good, but it won't be off by many orders of magnitude.
Roweis and I had a long conversation about the
next step in the astrometry.net project after Lang and Mierle solve all of the SDSS fields. One possibility would be to build a next-generation astrometric standards catalog, using existing data but better (or more interesting) statistical methods.
Rafelski (Arizona) spoke about the quark-gluon plasma, in RHIC and in the early Universe. Evidently RHIC has some awesome results, although he did not focus on them. Also, I was reminded that even after all these years on the lattice, we still don't know the u, d, or s quark masses!
Quintero questioned the existence of filaments in the galaxy distribution. I told him he was in good company (eg, Jim Peebles). We came up with a simple project to look for them. Very simple.
Padmanabhan (Princeton), Blanton, Weiner, Masjedi, and I discussed transparency (again). It really seems to me that this is fundamental and observably constrainable, whatever the theories predict; of course Padmanabhan and Weiner think we should rule out specific models, but I am happy to say what we can, but say it in such a way that it will rule out future models, even those not yet worked-out.
We found five clear and possible tests: Average spectrum of old galaxies as a function of redshift (most non-trasparent models predict wavelength-dependent changes that are not attributable to age or chemical composition), average spectrum of galaxies as a function of angle at the same redshift (most models have anisotropies in absorption/conversion of photons), average spectrum of galaxies as a function of foreground structure (most models have absorption/conversion depend on line-of-sight details), angular size of high redshift sources (most models blur point sources at high redshift), and the Tolman test (described earlier). The first three are trivial with SDSS data, the fourth is trivial in the literature and/or with HST imaging of quasars, and the fifth will be possible in the next five years or so.
Golding (Princeton) gave a nice talk on measuring rates and variability for RNA transcription and protein translation in individual cells, opening up the possibility of a future in which cell properties and variability are understood in terms of, say, differential equations.
Willman, Wu, and I discussed the Virgo Cluster project, in which Wu is connecting SDSS observations to the traditional Binggeli et al Virgo Catalog and thereby critiquing both the SDSS and the Virgo Catalog. It is a pretty complex data project, because the Catalog is large enough that it is hard to do things by hand, but small enough that we don't want to permit any errors, even on troublesome entries.
In an interesting (and highly interactive) talk, van Acoleyen (Durham) showed that certain modifications to gravity (or, in Dvali's terminology, certain scalar-field interactions with the graviton) can plausibly lead to a MOND-like gravitational force law on scales the size of galaxies and acceleration on scales the size of the Universe, with Newtonian gravity on all scales where measurements are precise and unassailable. Though it is in some absolute sense unlikely to be correct (indeed, it has some fine-tunings), and though it might not reproduce the CMB and growth of structure (the calculations had not been done), it is a reminder that the "dark matter" might one day be obviated by a new conception of gravity. Might.