More work yesterday on central stellar densities (which are effectively measures of bulge masses); I got everything to run on my full sample of 140,000 galaxies. I also got a reminder from Moustakas that I need to review the "secular evolution" literature, which claims that bulges can grow from disks without external drivers such as mergers, tidal impulses, or starbursts. I am skeptical!
I wrote code to measure central stellar densities on SDSS galaxies and tested it on a sample of 1000. I discussed the project briefly with Moustakas, who was reluctant to agree that the lack of red galaxies with low central stellar densities indicates that blue galaxies never quiescently stop forming stars. But I don't see any way around it. I guess the sense in which Moustakas and I agree is that my observation puts a constraint on the gas contents (or sources of gas) of blue galaxies.
I started extracting SDSS data for my project on young early-type galaxies. I need the aperture photometry (which almost no-one uses), so the extraction is taking its time. While I waited, I worked on images of Local Group members for Willman. Here's one (click on it to enlarge):
I started working on getting the young early-type galaxies out of the Sloan Digital Sky Survey. My plan is to do the simplest possible thing, ie, to simply find galaxies that, if
shut off and faded for a few Gyr, would have the structural properties of typical red galaxies of the same stellar mass. No by-eye or subjective hocus-pocus.
Mierle and I had a solid pair-coding session this morning, with Skype (tm) for voice interaction and old-school unix screen used as a tool for me to see what Mierle was typing in his office in Toronto. This is a standard operating procedure for us now, and is very efficient. Today was no exception, we found and fixed some bugs and think-os in astrometry tweak for astrometry.net. The conversion of all of our code from IDL (tm) to C is an ambitious project, to say the least!
Later in the day I made a set of fake data on which tweak is required to succeed. From these, Mierle will construct a set of functional tests that tweak is required to pass after any change in the code.
Spent quite a bit of the afternoon on the spectral extraction code mentioned previously. I hardened the algorithm a bit by making it fit to the detected spectral positions, and then re-centroid on the spectral images and re-fit, etc. I also formatted all my output to be consistent with our existing quick-look pipeline (although better, I hope).
I spent the last few days traveling; hence the lack of posts. The postable work includes figuring out the most robust methodology for associating nod-and-shuffle spectra in the PRIMUS 2d images with entries in the file used by the milling machine to cut the masks, and simultaneously figuring out which holes in the mask are primary and which are secondary, etc.
I also figured out what might be a fast and robust way to find and fit the satellite trails in calibrated astronomical images. Time to register "satellitetrail.net"! Along the same lines, I also worked on the fitting and subtraction of what in SDSS data are known as "filter reflections"—glare we get because the filter edges get exposed to the light of stars outside but near the field.
In group meeting, Wu gave a nice summary of Julianne Dalcanton's result that low-mass disk galaxies don't appear to have dust lanes. Dalcanton et al suggest that this has to do with gravitational instability in a rotating disk; the large disks are unstable to gravitational collapse of dust and gas, while the small disks aren't. This hypothesis links the presence of dust lanes with metallicity evolution in a non-trivial way, because you get higher metallicity when the gas from early generations of stars collapses efficiently to make new stars. Wu and I are interested in whether this relates to the trends of PAH emission with luminosity (since PAH emission comes from illuminated, dusty regions).
After Wu, Renbin Yan (Berkeley) showed beautiful results on post-starburst galaxies, LINERs, and Seyferts in DEEP2 and SDSS. He finds that a lot of red galaxies contain LINERs, and many post-starburst galaxies contain LINERs. He has many punchlines, but among them are that certain emission-line ratio selections can reliably select red galaxies (because some kinds of weak AGN—or whatever they are—appear only in early-type galaxies), and that there aren't enough post-starburst galaxies in DEEP2 and SDSS to explain the evolution in the red sequence since a redshift of unity. This is great, because other paths to the red sequence are hard to find (given that red galaxies do not look like faded versions of blue galaxies).
I had a great day today, because I spent virtually the whole day coding!
I worked on making my code to find the spectra in a PRIMUS CCD image more robust and more accurate (in its model of the nod and shuffle behavior). It seems to work on all of our data. What remains is for me to reformat the output to what is expected downstream in our two-d spectral reduction pipeline.
I spent a lot of the day babysitting the re-build of our RC3 page. Then, when it all completed, Blanton and I realized that I had made a significant mistake. I fixed the code, deleted everything, and re-started!
Now that Blanton has made a full estimate of the sky level for every image taken in the Sloan Digital Sky Survey, it has become time to re-make our images of very bright galaxies. I worked on that today. I provide an example below.
It has also become time to provide the world with images of low-redshift luminous red galaxies, where the default SDSS pipelines underestimate flux. This has been an embarrassing issue for SDSS, without a lot of sensible discussion. Any volunteers for re-measuring the fluxes, sizes, and other parameters of a few 104 galaxy images?
I conversed with Matt Kleban (NYU) for a while today about modifications to gravity that could remove the need for dark matter or for a cosmological constant. I mentioned my heresy about acceleration—that it is not demonstrated at high significance with SNe alone; it requires the combination of SNe and LSS or CMB constraints in the context of a physical cosmogonic model (ie, the many-sigma demonstration of acceleration is indirect). This encourages me to investigate a(t) in other models, but unfortunately almost all
theoretically allowed modifications to gravity create strongly coupled, long-range gravitational interactions. There is a conflict, in other words, between models that can be computed (and therefore constrained) and models that are theoretically possible/allowed. This is not a good situation, because the non-zero dark energy can only be said to be experimentally demonstrated if it beats some plausible competitors.
In the PRIMUS project, we are taking spectra of thousands of galaxies simultaneously. Because we know the properties of the instrument and of our slitmask, we know pretty well where those spectra are on the CCD images we read out of the IMACS instrument. However, we do not know exactly; there can be offsets, scale differences, rotations, and shear. I spent the day writing automated, robust code to find the spectra in the CCD images, given a slitmask file and a set of IMACS CCD readout images. The low dispersion of our spectra makes everything more difficult—except this, which is in fact made easier by low (and non-linear) dispersion.
Masjedi gave a nice talk at group meeting about merging at redshift around 0.6 from COMBO-17 as estimated by Bell, and Kathryn Johnston (Columbia) gave our astro seminar on the merging history of the Milky Way and M31. There are lots of informative observations, but I haven't yet figured out a hard test of CDM.
Burles came to NYU for the day and he, Blanton, and I discussed data reduction issues. We are evolving to a
data modeling approach for the extremely non-trivial, low-resolution, nonlinear-dispersion spectral data we are taking; that is, we are building a two-dimensional model of each frame. What is usually thought of in astronomy as the
extracted spectrum is—in this context—just parameters in that model. When I say the words
we are, I really mean
Burles is. Beautiful stuff! Maybe someday soon our 140,000 (or so) spectra will yield redshifts?
I spent this evening reading carefully the important (though now somewhat out-of-date) paper on galaxy growth by merging and accretion in numerical simulations by Murali et al (2002). Unfortunately, the article gives its results entirely in terms of the volume average of the merger rate over all galaxy masses and types, when the per-galaxy growth rate might be a more stable (and definitely more easily testable) statistic. The paper finds that the merger rate as measured observationally is consistent with their findings, though the test would be more sensitive if they had produced and tested per-galaxy numbers. However, they find that for massive galaxies, the smooth accretion rate of intergalactic material is substantially larger than the galaxy merger rate (in mass per time units); I think this is not consistent with present-day observations (although it is hard to test definitively if the IGM material is ionized).
Despite the great improvement in both observations and simulations, there is a surprising paucity of good theoretical predictions for the merger rate since this nice paper.
I worked today on the requirements for our post-SDSS baryon acoustic feature large-scale structure project. I finally figured out that requirements should have a hierarchical form that flows from high-level science goals down to science requirements down to technical hardware and software specs.
[Sorry for the lack of posts; I have been on vacation since Monday.]
Last Monday (the 20th), I gave our informal lunch talk on the uses of strong gravitational lenses (multiply imaging or very high magnification systems) in cosmology. I am a bit suspicious that strong lensing will be crucial in the coming era of cosmology, but the talk gave me an opportunity to highlight some places where I think it might be important:
- Strong lenses might put robust limits on the radial profiles and evolution of collapsed objects, when simulations and theory become good enough to make the relevant predictions. As it is, there is an interesting lack of low-redshift gravitationally lensed arcs from massive clusters.
- High magnification lenses can be used to find very faint, very high-redshift sources. There are already some quasi-believable redshift 10 candidate galaxies out there, and some definite redshift 6 galaxies.
- Well-understood lenses lensing variable quasars can be used to make measurements of distances (times distance ratios) to pin down the world model.
All of these projects are immature now, but I got the opportunity to quip that as we transition from
precision cosmology to
accurate cosmology (Ben Weiner's joke), we might need one or more of these techniques. Lensing is a blunt tool, but it does do some things very well.
Today was an all-day meeting regarding this new project (as yet un-named) to measure the baryon acoustic feature at multiple redshifts to determine the expansion history of the Universe and the physical properties of the dark energy, using the SDSS Telescope after SDSS-II operations complete. A large group from Princeton, LBNL, MIT, Harvard, CITA, CWRU, and NYU (plus Arizona, FNAL, and Portsmouth by phone) converged here at NYU to discuss the science, the technical requirements, funding (of course), and data reduction and management. It was wonderfully productive, but I was talked into shepherding the
Productive and lively discussion at coffee today (Ben Weiner visiting from NOAO, Zabludoff, Moustakas, Blanton, myself) about good policies for the journals and large surveys to adopt in regards the public release of data and code.
My vision (with which no-one agreed) is that the journals should require all papers be released with a tarball that contains all the data and code, such that a reader can unpack the tarball, type make, and produce all of the analyses and figures for that paper! Of course there are implementation issues, but if we want to
stay clamped to the manifold of repeatable science (Sam Roweis) we have no other options. I could write about this for hours, but I have an NSF proposal due. I hope to return to this later, because in the course of the discussion I discovered many new reasons to support my view.
We are very far from my vision right now, and it is hard to get there incrementally.
It is great to have Zabludoff and Zaritsky here! We spoke briefly about the suggestion that galaxies in clusters are strongly affected by the cluster itself. If this happens it is pretty subtle: (1) The difference between the galaxy populations at high density and low appears not to be that disks have been removed at high density, but that bulges have been augmented. (2) Galaxies in clusters do not look like the centers of galaxies not in clusters; ie, they are not stripped field galaxies. You could say
duh! to this, but there are still many papers appearing that treat severe stripping as a competetive hypothesis.
Some of my unbloggable (consult rules at right) activity of the last year has led to the arrival of Ann Zabludoff (Arizona) and Dennis Zaritsky (Arizona) at NYU for the next nine months. Today they both gave informal lunch talks, presenting some of their research on galaxy evolution. Ann noted her result that the differences between the disk-dominated and bulge-dominated populations has to be explained as a difference in bulges, not disks. Dennis showed that there are clear relationships between the orbit and star-formation history of the Large and Small Magellanic Clouds.
I started a huge task (built on Blanton's "gmosaic_make" code that combines GALEX AIS images automatically) to create near-ultraviolet images of galaxies in the SDSS main sample whose optical colors suggest star-formation within the last Gyr. A quick inspection suggests that the K+A galaxies (those that are deficient in emission-line luminosity) also have lower ultraviolet, but I am not close to showing that post-starburst galaxies can be identified ab initio with GALEX imaging. That's the hope.
Various members of the CCPP gave talks to visiting NSF Theoretical Physics Program Director Fred Cooper on a one-day site visit. I learned about some of Gia Dvali's new ideas about testing Planck-scale physics with tabletop experiments: He has noticed that if there are Planck-scale black holes created at early times that subsequently evaporate, their evaporation products may have topological
charge that could in principle be detected at low energies with some kind of futuristic Aharanov–Bohm-like experiment. Crazy! Also, Andrei Gruzinov gave a nice explanation of Kolmogorov turbulence.
Wu and I discussed error propagation for her Spitzer photometry project, and Quintero and I figured out a great idea for finding post-starburst galaxies at high redshift.
I have to apologize to my loyal reader, because he or she has been waiting for a post for days! But I spent the end of last week preparing for and attending a NASA Extragalactic Database (NED) Users Committee meeting, in which we advised NED on their near-term and long-term priorities. NED is a pretty unique service, because its data ingestion involves a great deal of human scientific judgement. This gives it a crucial role in diversifying the current initiatives for massive data analysis.
Lisa Randall (Harvard) gave a nice talk pointing out (and calculating in detail) the fact that some of the generic large extra dimensions theories (in particular Randall-Sundrum 1) have a first-order phase transition at the weak scale and that such theories in general produce gravitational radiation in the LISA bandpass, at LISA-sensitivity levels. So we might discover extra dimensions or other non-trivial properties of the dark sector with LISA. Not bad!
Had a great visit to UC Davis, where I gave a talk to the cosmology group (which is pretty damned large). Spent the rest of the time talking with Chris Fassnacht and Matt Auger about galaxy evolution vs. strong gravitational lensing (Auger says do the former, not the latter, because we will understand the scale factor and galaxy mass distributions better by other means within the next decade or so). Also argued with Lloyd Knox and company about my brave new world fantasy in which we have constrained realization cosmological simulations that actually reproduce all observations, exactly not statistically. They all told me I am crazy; okay it is hard, but isn't science supposed to be hard?
Iggy Sawicki (Chicago) gave a very lively talk today about quantitative empirical tests of the leading cosmology-inspired modification of gravity: DGP (named after Dvali, Gabadadze, and Porrati, all here in the CCPP at NYU). The talk was lively because it is basically impossible to do exact calculations of the evolution of fluctuations (it is even hard to do the scale factor!) in this model; the arguments are all about the applicability of various possible approximations. In general, GR calculations are hard; modifications to GR are harder. We have a dangerous possible future in which no substantial competitors to GR ever get worked out. We might end up settling on GR just because it is the only exactly computed model!
After the talk, Blanton and I had a long discussion of the statistics involved in
ruling out models. A trivial point, but one that is often misunderstood, is that it is very different to ask the question
Is this model consistent with the data? than it is to ask the question
Is model A a significantly better fit to the data than model B? Sawicki showed that DGP is a significantly less good fit to the data than GR (ie, Lambda-CDM), but he did not show that either model is consistent with the data. I actually don't think that either is consistent with the data, in the technical sense (chi-squared on the order of the number of degrees of freedom). But of course this just reminds me that we don't make important decisions about the fundamental physics of the Universe on the basis of calculations in gaussian statistics!
Okay, I admit that discussion of the anthropic principle does not count as
research (see rules at right). But really, the people who use the anthropic principle are confused! I agree that you can't observe a universe that cannot contain observers. No duh! But a universe that contains observers arbitrarily different from us is observable. Recall: The anthropic principle is invoked to reduce the range of possible predictions for the properties of the Universe from the vast range allowed by, eg, string theory. You can't use observations of our Universe (such as that it contains galaxies and carbon-based life) to trim down the possible fundamental predictions, because then they aren't predictions. You must only use the fact that the Universe is observed, which requires only observers, not carbon-based observers, not observers in galaxies, not observers around stars, etc. Interestingly, many people who have worked on this problem for many years, including Susskind and Vilenkin (who spoke today at NYU) have got this wrong, and they call the requirement that there be galaxies the
The fact that the Universe contains galaxies is an observation; the requirement that it is in principle observable is properly the anthropic principle. Of course we don't know how to compute observability, given a stated theory (indeed, we cannot even predict the proton from the standard model, let alone observers!). Until observers are predictable, the anthropic principle is not useable for science, in practice.
Jim Pizagno (Stony Brook) gave a great and very lively group-meeting talk about his work on the Tully–Fisher relation—the narrow power-law relationship between galaxy stellar mass or luminosity and disk rotation velocity—and the scatter about the relation. He did the very striaghtforward thing of attempting to measure rotation curves for a complete sample of galaxies, not selected by subjective morphology. After subtracting out estimates of observational uncertainties in quadrature, he gets a scatter of about 0.4~mag, which is small by any theoretical comparison (and the theoretically predicted scatters don't include scatters in dust content). His talk brought up a lot of side issues too, relating to the metallicities and star-formation histories of the galaxy disks.
After his talk, Pizagno and I discussed his project to make stellar-mass images of SDSS galaxies and re-analyze his spectra in the context of these two-dimensional maps. Unfortunately, the ball is in my court!
Much of the debate at the merger conference in Baltimore last week was about the evolution of the red, old, early-type galaxies; what we usually call (stupidly) the
red sequence. Today Masjedi, Blanton, and I discussed Masjedi's job applications and dissertation, which all relate to the growth of galaxies on the red sequence. Most studies of the luminosity function of red galaxies suggest an increase of between tens of percent and a factor of two in the total stellar mass on the red sequence since a redshift of unity. As I have mentioned here many times, Masjedi can see this process in action among the SDSS LRGs and can quantitatively limit the mass accretion rate at the most massive end.
Unfortunately, it is at the most massive end that there is the most disagreement about the evolution of red galaxies; Masjedi points out that this can be partially resolved with photometrically selected red galaxies below the spectroscopic limit in the SDSS. He is also working on extending his work on mass accretion to lower-luminosity red galaxies, where the community is in agreement that there has been substantial evolution.
Mierle and I walked through the linear tweaking of the WCS of a small portion of a huge NOAO KPNO 4-m Mosaic camera image. My IDL implementation works just fine; now we are trying to get Mierle's much better C implementation to work. The migration from IDL to C makes the code more portable, and independent of closed-source products.
I continued working on the automatic detection of satellite trails in astronomical imaging, which is formally trivial, but quite non-trivial in the face of real data issues. In the
Hough transform image below, the white trapezoidal dot in the upper left is the signal of a satellite trail.
This is a clear detection, but the transform involves doing of order one million medians of a thousand values each. That's not fast.
Masjedi gave a great talk on his close-pair merger rate estimates and the growth of luminous red galaxies at the conference today. He is one of the few participants in the meeting who committed to very specific numbers on the merger or accretion rate, and, though there are caveats, it is a beautiful peice of work.
There was a amusing, lively, and chaotic summary session to which I contributed. I think there is a picture emerging—for the growth of red galaxies—but very few components of the picture have hard numbers on them.
Another great day at the meeting on galaxy–galaxy mergers at STScI. Among the many contributions was a nice discussion of the role of dissipation in shaping merger remnants by Rothberg (STScI), who cited some nice old papers by Carlberg, Gunn, and Kormendy that I have not previously seen; they discuss the issue I go on about: you can't make (high density) ellipticals from the mergers of (low density) spirals. Along the same lines, Novak (UCSC) showed that realistic simulations of disk–disk mergers tend to produce oblate galaxies in prolate halos, with a perpendicular alignment of the symmetry axes. This is testable right now by weak lensing.
McIntosh (Amherst) can show that the merger probability is
relatively insensitive to environment (where
here the mass of the hosting dark-matter halo). This relates strongly
to results in preparation here at NYU by Blanton and Ignarra. He also
quipped that we have
A growing but dead red sequence and a
star-forming but static blue sequence. But we all knew that
McIntosh is a poet.
van Dokkum (Yale) showed his evidence for dry merging from
very low surface-brightness tidal features in red galaxies. He gave a
wonderful presentation of the material, and drew out the
tension with Masjedi's work, and Brown's (NOAO). He nicely set
up Masjedi for his talk tomorrow! There is a lot to do with the
inter-comparison of Bell, Brown, van Dokkum, and Masjedi, because
they all have slightly different things to say about the merger
Just like yesterday, there were good talks too numerous for me to mention, but one highlight was Puech's (Paris) result on the near-infrared Tully–Fisher relation at redshift 0.7: When he trims his sample to galaxies with very clean two-dimensional velocity maps, he gets exactly the local relation. So it isn't just the luminosity function of blue galaxies that is static!
Today was the first day of the Galaxy Mergers: From the Local Universe to the Red Sequence meeting at STScI. The many very good talks today included one by Schiavon (Virginia) in which he did very detailed analyses of the mean galaxy spectra we produced for red galaxies in this paper. He finds non-trivial age and metallicity variations with galaxy mass and environment, as we expected (but were unable to do ourselves, since models are complicated).
Bell (MPIA) argued that any galaxy evolution scenario that cannot be ruled out observationally at the order-of-magnitude level is, by definition, plausible. This position is ascetic in its conservatism, but I love it.
Fall (STScI) showed evidence that the formation of star clusters is very regular (statistically) and depends primarily on the local (two-dimensional, don't ask me why) gas density. He noted a conspiracy that the mass function happens to be very similar to the luminosity function; this is not generic for sources like these because their luminosities change very quickly with time, as I wrote once in this paper. I love a conference that gives me opportunities like this to toot my own horn!
Mierle and I learned today that all of the papers describing the WCS conventions are pretty difficult! None of them give the complete formulae for converting pixel location to sky coordinates or the reverse. They describe parts of the transformations in words. None of them give fully worked-out examples against which code could be checked.
Furthermore, the WCS conventions themselves are ill-defined, in that there are mutiple different parameters/keywords that mean the same thing, and potentially conflict. So it is a mess. We decided to work internally in astrometry.net with a simplified, well-defined, minimal WCS parameterization, and push all this ugliness to the original
writing of WCS at the beginning and the end. We started a polemical and pedagogical paper on the subject.
On Monday, I rewarded myself for weeks of work on that NSF proposal with a day of work on my digital galaxy atlas project. I worked on two things, both pretty technical:
I worked on the problem of producing, for each galaxy image, a noise map that gives approximately the correct pixel-to-pixel noise or uncertainty in the intensity. This is relatively trivial, in that we have a noise model (even an implemented noise model), but it requires some decisions (a) where I have smoothed the data either explicitly or implicitly, and (b) where overlapping fields contribute to the image.
I also worked on automatic identification, fitting, and subtraction of all satellite trails in the images. For this project I am using a Hough transform, in which the space of x and y (pixels in the original image) gets transformed into the space of angle and offset for linear trails (each of these is a pixel in the Hough transform image). I think I need to do some kind of median filtering because in the dumbest possible implemetation (recall that I always start with this!) pairs and triples of stars dominate the signal.
Eric Peng (DAO, Victoria) gave a great group meeting talk about globular clusters, which separate into metal-poor and metal-rich populations. He has suggestive evidence that the metal-poor globular clusters (which probably formed at very early times) may trace the dark matter better than galaxies or stars! This would be wonderful if true, and the hypothesis makes many testable predictions.
Patrick Huggins (NYU) gave the astro seminar, about planetary nebulae. He showed that they show strong evidence for (time variable) jets, and he discussed the mysteries associated with that. He noted that a lot could be explained if all of the AGB stars that go through the PN phase are in binaries, but this seems very hard to explain in terms of the binary fraction.
Jim Pizagno (Stony Brook) talked to me about his mass images of galaxies (that is, transforming optical images into mass images using relationships between color and stellar mass-to-light ratios). He is close to a very robust system that will be of very wide applicability.
For two days now I have done nothing but NSF proposing. I think it may be done, but by the rules that is not research! Today I went to a nice talk by Eric Dufresne (Yale) about electrostatics and entropy in ion solutions. Evidently there is a lot that isn't understood; what I am hoping is that this area of
soft condensed matter produces some very general principles, the way that, for example, thermodynamics did at the turn of the century.
[I missed a few days on travel.]
On Thursday, Burles, Bolton, and I had a data analysis pow-wow at MIT. We are facing some crazy data for the PRIMUS project and they are not yet tamed. Burles feels that a lot of the problem is fitting the sky through the slits, with non-zero instrument resolution, and then tracing and extracting the object spectra from the two-dimensional spectral images. I hope so, but it is going to take some real work to do it right.
Today, Moustakas gave a nice overview at lunch (on the blackboard) of the fundamental observables in galaxy evolution: luminosities, colors, and emission line strengths. All of these are just measurements of the spectral energy distribution with different levels of crudeness, but they are differently sensitive to age, star-formation history, dust, and chemical composition, all of which are themselves inter-related. To figure out how galaxies form, and how stars form within galaxies, a consistent picture must be drawn for all these observables, that is also consistent with what we know (and we know a lot) about the dark sector.
Interspersed between NSF proposal paragraphs, I followed up some failure fields in our blind astrometry demo project (in which we throw away all the astrometry information for every SDSS field, randomize their order, and solve them all again, totally blind). Almost all of the failures can be attributed to holes in the USNO-B1.0 catalog (not our problem), I believe, but for those few failures that can't, some are bad SDSS data where, eg, the telescope jumped or fell out of focus. Even these are very rare: Our success rate is more than 99.8 percent.
Today was all teaching and all grant proposal writing, so it would be against the rules for me to point out that I submitted this pedagogical note on air resistance to the arXiv.
Mierle and I agreed on a short-term plan for getting our astrometric WCS tweak system on the web, and making sure that our code produces files that our
customers can use.
Today was all computers all the time. Piet Hut (IAS) came in for the day to have lunch and discuss ambitious coding problems, including the development of enormously complicated code, and the realization of incredibly complex constraints in constrained realizations. Hut is working on some ideas for collaborative coding which, if they pan out, could transform open-source and multi-developer projects. He is also doing some very ambitious simulations that combine stellar evolution with stellar dynamics in dense environments (such as the centers of globular clusters).
I also worked on the backups for the astrometry.net code repository. We have multi-site, off-site backups. If you don't, you are making a big mistake (trust me, I know from direct experience).
I started to put together the image test-bed for the astrometry.net pre-alpha testing. I will be contacting the alpha users (no doubt my most loyal readers) soon for example images.
At coffee we discussed a recent paper on cluster gas fractions which claims that x-ray observations of clusters were in (weak) disagreement with WMAP results on cosmological parameters. The odd thing is, they find that the discrepancy is a strong function of the radius at which you do the measurement, and that the discrepancy gets small at large radius; perhaps there is no discrepancy at all when extrapolated to the cosmic mean density?
Mierle and I worked through some small remaining issues in Blanton's simplexy code, which is a very simple peak-finding and measuring code, suitable for finding stars (or other sources) in astronomical imaging, with very few free parameters. The hope is that we can hard-code simplexy so it works on almost all possible input images without much modification. Checking that it does is the next order of business; so far we have only tested it on a single (real) image.
Blanton, Moustakas, and I discussed the things you can learn from galaxy emission lines, in particular star-formation rates and the electron densities and temperatures in ionized regions in galaxies. Moustakas believes that he can rule out a suggestion in the literature that the electron densities in star-formation regions are a strong function of redshift.
Mierle and I checked out the results of Blanton's super-simple simplexy code that detects and measures the positions of sources in astronomical images. It seems to work pretty well. Now Mierle is packaging Blanton's code into something that reads a (possibly multi-HDU) FITS file, runs simplexy, and writes the output to a (possibly multi-HDU) FITS binary table.
Moustakas gave group meeting, about the evolution in the mass–metallicity relation. Last year he came and gave a job talk on the same subject, and one very interesting thing has changed: If he factors in reasonable luminosity evolution for galaxies, his evolution in the luminosity–metallicity relation is severely diminished. His results are now (nearly) consistent with no chemical evolution whatsoever for massive galaxies. On the other hand, he does find that at high redshifts there is a (rare) population of massive but low-metallicity galaxies. Maybe these are proto-galaxies?
Mierle assigned me the job of building the testbed for the astrometry.net tweak and online systems. I have started, and soon I will be contacting the alpha users for sample images.
Keir Mierle (Toronto) just showed up for two months of hard work on astrometry.net. We spent a long time going through the gory details of tweak, which is the process by which we take a pretty-good astrometric WCS solution for an image and make it precise (and, we hope, accurate) and encode it in the image header in a standards-compliant way. None of this is very difficult; all of the difficulty comes in dealing with the heterogeneity of input images and desired output WCS formats.
Mierle's main project, if we can stabilize tweak quickly, will be to work on the general blind
camera calibration problem, in which a large number of images from a single camera are used to figure out the optical properties, and variations of those properties with focus, temperature, gravitational loading, etc., just by looking at the images themselves. This is the problem astronomers working on deep imaging call
astrometry, and frankly we are not competitive if we can't do this very well.
John Moustakas arrived to start his postdoc this week! It was great to spend some time chatting about mergers, galaxy spectra, and the Spitzer Space Telescope. He told us that he and Tremonti (Arizona) have evidence that high-redshift star-burst/AGN galaxies have very high velocity outflows, presumably driven by star formation or black-hole accretion. It is going to be a fun time having Moustakas in the group.
The school year is just starting, so I didn't get much research done, but I did return the proofs to ApJ on this article.
Yesterday I made a figure showing how we parameterize quadrangles of stars for the astrometry.net blind system. The parameterization is clever, because it is rotation- and scale-free, but if stars are uniformly (ie, Poisson) distributed on the sky, the parameters uniformly fill a compact, finite, simply shaped volume in four-dimensional parameter space. This is not a requirement for a good system, but it makes a few things much easier.
[for yesterday] Zolotov, Willman, and I specified Zolotov's first project with stellar velocities in the Galaxy halo in models and in data. I then read papers in preparation by Maller (on inclination effects in galaxies), by Sheldon (on using the weak lensing by clusters of background sources to infer the masses of the clusters statistically, and the cluster-mass cross-correlation), and by Quintero (on environmental relationships).
Yesterday, Sheldon and I discussed the recent press about the definition of the word
planet, under discussion at the IAU. We agreed that this press is not good press, because it implies that astronomers are scholasticists, interested more in the definitions of words than the origins of our Solar System. Not all publicity is good publicity!
On the research front, Blanton and I discussed a problem I have been thinking about a lot lately, which came up in a talk by Dan Maoz while I was at MPIA this summer: How do you figure out the distributions of properties of things, separated by type, when your type assignments are imperfect? Maoz faces this for imperfectly typed supernovae, but this comes up frequently (for example when galaxies are placed in redshift bins using noisy photometric redshifts). Unfortunately, the correct methodology is not written down anywhere convenient to the typical astronomer.
In between working on the first paper for astrometry.net there was a lot of discussion here at NYU about the offset between the mass center and the baryon center of the subclump in the
bullet cluster, which got a lot of press because it says some things about dark matter. This is a very hard observation for modified gravity theories to reproduce (at least the ones that want to replace dark matter with a modification to GR). More interesting to me is that the sub-clump is moving incredibly fast (hence
bullet), although the theorists say this is not incredibly unlikely.
During a week in the wilderness, I worked on the paper describing our success with blind astrometry with SDSS data. Thanks to the heroic work of Lang and Mierle, our success rate is over 99.8 percent and we have no false positives after having solved (blind) over 300,000 SDSS r-band fields.
Yesterday Bhardwaj and I measured the average current star formation rates in her samples of elliptical galaxies, using the H-alpha line luminosity, as a function of total galaxy stellar luminosity (stellar mass). Her next job is to compare these estimates to the star-formation rate estimates from GALEX UV imaging, and from the suite of models that explain the total galaxy spectral energy distributions.
Zolotov and I figured out the first-year physics problem of estimating the angular momentum vector for a galaxy disk in a numerical simulation of a forming galaxy, and transformed all the particle positions and velocities into the simulation equivalent of galactic coordinates.
Masjedi, Blanton, and I discussed the current state of redshifts for the PRIMUS prism spectroscopy project. Blanton is taking over for Masjedi while Masjedi gets some papers written. We concluded, as we always do, that calibration is everything.
Zolotov and I worked on preparing simulation outputs so that we can "observe" the stars in simulated galaxies much as the SDSS observes stars in the Milky Way, in preparation for some comparisons.
It was a sad day yesterday, as it was the last day for Quintero at NYU; we will miss him!
Quintero led group meeting; he spoke about a new project we have conceived to find the progenitors of bulge-dominated, red, dead galaxies. We think we will be able to use their statistics to test or constrain the emerging picture that AGN feedback must be involved. We can also refine the limit from Quintero et al (2004) on the birth rate of new bulges.
Yesterday, Willman, Zolotov, and I found a specific focus for our project on stellar velocities in the SDSS.
I also discussed and read a very nice new paper by Marla Geha (OCIW) et al on the gas in dwarf galaxies, to appear on arXiv soon. They find that when you add gas to stars, there is a very straightforward baryonic Tully-Fisher relation, implying that there is very little in the way of mass-dependent expulsion or stripping of baryonic material from low-mass galaxies. They also find a very strong dependence of the gas fraction distribution on small-scale environment; I think it is the strongest environment dependence known for galaxies.
On the last day of the Oort meeting, the four groups (which were working to answer four questions posed by Blandford at the beginning) made their presentations. The two groups that did publishable calculations were the Hubble Constant group (my group) and the substructure lensing group. Our group concluded that there is a strict 99-percent upper limit on the Hubble Constant from the known lens systems of 78 km s-1 Mpc-1, found by making the maximal assumption that mass follows light. This limit is much stronger than that from the HST Key Project or any other direct method other than perhaps the baryon acoustic feature in the LRGs.
The substructure group took a skeptical view and showed that there is no absolutely clear evidence in multiple imaging systems for truly dark substructure in dark-matter halos, in the sense that all lenses with magnification anomalies can be explained either by giving mass to luminous substructure or else line-of-sight effects like extinction or scattering.
Today Chris Kochanek (OSU) and Paul Schechter (MIT) faced off over microlensing and substructure, but unfortunately, they resolved their differences over breakfast, so no sparks flew. Schechter gave a beautiful, pedagogical description of the effects of micro- and milli-lensing (ie, the perturbations to smooth lens models by substructure and clumps). Kochanek showed that he can fit the detailed microlensing light curves of real gravitational lenses by brute force: He actually creates billions of totally random microlensing histories, and finds the few that fit each lens; from these he infers the masses of the stars, the velocities of the stars, and the sizes of the quasar emission regions.
In the afternoon, the Hubble Constant group worked late getting ready it's presentation for tomorrow: We have a much better upper limit on the Hubble Constant than the HST Key Project
On the third day of the Oort meeting here in Leiden, Gary Bernstein (Penn) told us how to analyze future data sets for weak lensing. It was nice to see some of the nitty-gritty, and in fact Bernstein and collaborators are working on some ideas that I love: For very high-accuracy data analysis, especially when instrumental effects are large and signal-to-noise is low, the best approach is to explicitly model the data (rather than
measure things in it); in this approach your
measurements are just parameters of your reduced-chi-squared-unity model of the data.
Bhuvnesh Jain (Penn) followed Bernstein with a discussion of the limiting systematics for future weak lensing surveys. He identified redshift distribution as a significant issue; this might be an independent motivation for the PRIMUS project. Interestingly, Jain was very confident that the atmospheric distortions to the point-spread function will not be a problem (I didn't agree).
Wayne Hu (Chicago) gave his talk today about
precision cosmology. Blandford forced Hu to slow way down when he was discussing the specific assumptions that go into the determination of the distance to last scattering. Hu's discussion was absolutely great; he argued that there are very few assumptions involved, except maybe adiabaticity (which is tested, but not at the level required). He spent some additional time encouraging us to determine the Hubble Constant. He espoused some heresies including that he considered a small amount of spatial curvature much more plausible than a non-trivial equation of state for the dark energy (I agree); these are degenerate for some experiments.
In the afternoon we continued on the Hubble Constant. I spent time re-reading Eisenstein et al (2005), a paper on which I am proud to be a co-author, though I can't take credit for some of the best parts. The baryon acoustic feature in the LRG correlation function constrains strongly the distance to redshift 0.35, but less strongly the Hubble Constant, because uncertainties in the world model enter. Similarly, I realized, the lensing time delays only measure the Hubble Constant in the context of a specific world model; really each system constrains a combination of cosmological parameters. It's my job to figure that out tomorrow.
Yesterday (Monday) was the first day of the Oort Workshop in honor of Roger Blandford (Stanford) and about gravitational lensing. Blandford opened the meeting with five questions and has encouraged the group to split into teams to answer them by the end of the week. I joined the Hubble Constant group, where the question is
Can we use lensing to determine the Hubble Constant? At first we were all extremely negative, but slowly a consensus emerged, that each lens can, in principle, give you a very strong upper limit on the Hubble Constant. This is because the predicted time delays among images depend on the radial concentration of the mass, and the mass profile is very unlikely to be more concentrated than the stars.
Wayne Hu (Chicago) convinced us that this was pretty important, because the CMB only measures the Hubble Constant if you assume flatness and vanilla cosmological constant. A strict upper limit can rule out strange models (like NYU-born DGP), and a strict lower limit (also possible, we think) can rule out spatial curvature and scalar fields that have different redshift evolution from the cosmological constant.
The upshot is that the Hubble Constant group is working towards writing a paper on the subject, which is ambitious, but not obviously impossible.
[for yesterday:] I was pleased to find Cox et al on the construction of bulge-dominated galaxies from disk-dominated galaxies in merger simulations. They find that they cannot reproduce the observed properties of bulge-dominated galaxies without allowing for dissipational gas dynamics and rapid central star formation. I have been saying this in a hand-waving way for years, and I am very happy to see it all done quantitatively. This really does mean that we can observe the creation of new bulges by looking for starbursts and post-starburst galaxies, as we have been arguing since Quintero et al 2004.
[for Tuesday:] I worked on the possibility of measuring the 3-d velocity of Galactic substructure using blue horizontal branch stars, which all have luminosities in a narrow range (so their magnitudes tell you their distances), and which have radial velocities (for some) and proper motions (for all) in the SDSS.
Phil Hopkins (Harvard) spoke here today about systematically fitting all AGN luminosity functions at all wavelengths by convolving a bolometric luminosity function with the observed range (different at each luminosity) of spectral energy distributions and dust attenuations. His fits look great, and make many predictions. In an aside, he noted that the faint end of the AGN luminosity function at high redshift may well be set by the fading power-law of the AGN after its high-luminosity event. If so, this would be an application of the idea in this obscure paper.
Quintero and I worked on focusing his latest paper onto the processes that control the morphology–density and star-formation–density relations.
Rix and I found that we could get an intelligible proper motion distribution as a function of stellar magnitude for what are plausibly turn-off stars in the halo and in the thick disk, inspired by Girard et al. The relative velocity of a typical thick-disk star relative to the Sun increases with distance, but the proper motion involves an inverse distance (because it is an angular measure of motion), so the distribution for the thick disk is very simple.
Davé's students Finlator and Oppenheimer gave nice talks today about high-redshift star formation in numerical models of structure and galaxy formation. Finlator showed that the star-formation histories of galaxies at early times are very constrained (if the star-formation prescriptions in the models are close to correct); this will either aid strongly the fitting of high-redshift spectral energy distributions of galaxies, or else show that there are issues with their numerical simulation of cosmic star formation. Oppenheimer showed that the long-debated early enrichment of the IGM, which has been suggested because CIV (triply ionized carbon) is seen at all redshifts, is something of a coincidence; the carbon abundance is rising, but the fraction of the carbon in CIV form is falling. Though it looks like the IGM was polluted early and only once, in fact pollution has been a continuous process; there is no requirement of any
Pop III stars to enrich the IGM at early times.
Rix, Bell, and a cast of thousands argued this afternoon about how the black hole masses
know about their host bulge masses. Some predictions were made.
[Sorry for the long gap between posts; I was hiking in the Alps!]
Nice lunchtime talk with Sandy Faber (UCSC) and Frank van den Bosch (MPIA) and Davé and Bell regarding the build-up of mass on the red sequence. I pushed my position that starbursts have to be involved. Faber continued the discussion with her talk at the University; she showed that
downsizing—the evolution of the mass scale of assembly to lower masses with time—can be seen in all measures of merging and activity.
Dani Maoz (Tel Aviv) gave a great talk today about the problem that (a) galaxy clusters contain enormous amounts of mass and enormous numbers of stars and are strongly gravitationally bound, and (b) they contain a large amount of iron (a few parts per thousand by mass), and yet (c) they show very low type Ia supernova rates. Either the supernovae go off very early in the lifetime of a stellar population, or else the metals are put out there some other way. The latter explanation—since there are no other ways of expelling metals into the intergalactic medium—is not plausible. But if the type Ia supernovae are prompt, there are issues in understanding supernova rates in the field. It is an extremely rich subject area that is only getting richer as supernova projects evolve, and Maoz is doing his statistics right.
ps: Maoz does most of his observing with a 1-meter telescope!
Rix and I argued today (while walking up from the Bismarckplatz to MPIA) about the future of observational cosmology, in imaging, spectroscopy, radio, optical, and simulations. I made my pitch for constrained realizations, which Rix thought was crazy. It is, of course.
There was a one-day mini-symposium here (at MPIA, Heidelberg) on star formation in a cosmological context. There were good, long talks all day. A few randomly recalled notes:
Though Grebel (Basel) introduced her work entirely in the context of the substructure problem, I think I agreed more with Springel (MPA Garching) who, under questioning, admitted that the substructure problem may be a sub-problem of the luminosity function problem (the problem that the distribution of galaxy luminosities or stellar masses looks nothing like the CDM-predicted distribution of dark-matter halo masses). Grebel showed that the nearby dwarfs—ie, those close to their host galaxies—are not only older in stellar populations than more distant dwarfs, but also more metal-rich. Springel convinced me that cosmic rays might be an important dynamical component of the ISM and affect star formation in galaxies, particularly low-mass galaxies.
Abel (Stanford) showed amazingly high dynamic range (1014.5) adaptive-mesh simulation results. He argued that the details of ultra-high redshift supernovae were unimportant for initial
pollution of the IGM with metals; the distribution of the metals into the IGM is done by a combination of SNe, winds, and radiation (which makes low-density regions that are easy to expand into). Gallagher (Wisconsin) showed details of star formation in starbursts in the local Universe; he convinced me that Perseus A is the most incredible galaxy in the Universe. I am inspired to find SDSS analogs.
I worked on the weekend on statistical subtraction of color-magnitude diagrams and proper motion distributions in and out of globular clusters observed by SDSS. I am trying to assess the possibility of measuring proper motions of clusters of stars statistically. The challenge is to subtract the foreground correctly; since in general the foreground is populous, the results depend strongly on how you subtract it.
Lots of conversations with Rachel Somerville (MPIA) and Davé today about the idea that mergers trigger starbursts and AGN, and the AGN shuts off star formation so that the fading remnant galaxy can join the red sequence. This picture is observationally tractable, but it is difficult to robustly constrain the fraction of AGN that are merger-triggered, the fraction of mergers that trigger AGN, and all the relevant timescales. We came up with some ideas for Quintero to test when he arrives here next week.
Yesterday and today, Rix and I found what proper motions exist in the SDSS data and tried to assess their usefulness for constraining halo properties. It looks a bit dicey.
I gave a few-minute talk today that expanded into a half-hour of discussion about the merging of smaller galaxies into LRGs (Masjedi's new results). The low accretion rate we get implies that either the accretion rate is a very strong function of cosmic time, or else that LRGs formed very early. Romeel Davé (Arizona), who is also visiting, noted that early formation of LRGs is not necessarily unreasonable from a theoretical perspective, because
cold mode accretion (which can rapidly build bulges, in principle) is much more common at earlier times.
Yesterday and today, Rix and I found what proper motions exist in the SDSS data and tried to assess their usefulness for constraining halo properties. It looks a bit dicey.
I gave a few-minute talk today that expanded into a half-hour of discussion about the merging of smaller galaxies into LRGs (Masjedi's new results). The low accretion rate we get implies that either the accretion rate is a very strong function of cosmic time, or else that LRGs formed very early. Romeel Davé (Arizona), who is also visiting, noted that early formation of LRGs is not necessarily unreasonable from a theoretical perspective, because
cold mode accretion (which can rapidly build bulges, in principle) is much more common at earlier times.
Eric Bell (MPIA) pointed out in conversations about merger rates today that since the LRG mass function is very
steep, there is a big difference between the merger rate considered as
the probability that a given LRG will merge in the next Gyr and
probability that a given LRG had a major merger in its past. These differ because in the latter question you are dividing by the number of LRGs, and in the former you are dividing by the number of twice-as-massive LRGs. Of
course Bell is interested in asking the latter question and we (Masjedi and I) are asking the former.
I would say that we are asking the
question and Bell is asking the
wrong one, because close pairs tell
you about the future, not the past, but then again the evolution of the Universe is continuous, as I often argue.
James Pizagno (OSU) is also here visiting Rix, and we spent some time discussing the project of turning multi-wavelength images of galaxies, which show starlight, into stellar mass maps. Pizagno has some great code, and some very believable mass maps. In some spiral galaxies, the spiral arms disappear entirely when he goes from starlight to stellar mass. We agreed to test his code on some of my bright galaxies.
I spent the day reminding myself how TROTW (the rest of the world) accesses the SDSS data—through a public SQL server setup. We at NYU, of course, keep and spin our own copy of the entire data set down to raw (almost), so I haven't had to use the public SQL stuff for more than five years. It works beautifully and it is fast, so I think this will be fun. I am in Heidelberg for the month working on Milky Way substructure with Hans-Walter Rix (MPIA) and his group.
[This post is for Friday.] All I did all day is work on the atmospheric extinction; I was trying to use five years of SDSS spectroscopic data to infer it. Unfortunately, seeing is correlated with airmass, so my results have a
grey term due to fiber losses. Not sure what do to at this point, because there is very little we trust in the atmospheric extinction literature and folklore.
Optimistically, we (Blanton, Bolton, Burles, Masjedi, myself) decided that the PRIMUS prism spectroscopy of redshift unity galaxies will start to measure good redshifts as soon as we understand the throughput well. The throughput is not trivial to measure, and it depends on airmass, photometricity, and other factors. But we spent all day working on it and are very close to having some reasonable estimates. What I don't understand is what kind of data we need to take each PRIMUS night to ensure that, after the fact, we will be able to deteremine the throughput accurately enough for good redshift determination.
I spent a good fraction of today discussing what could be done with a large sample of stellar radial velocities, heterogeneously selected (as are the stellar spectroscopic targets of the Sloan Digital Sky Survey) with Willman, Zolotov, and Hans-Walter Rix (MPIA, Heidelberg). There are lots of opportunities to discover Milky Way substructure in this sample.
We have decided, optimistically, that we will be getting redshifts out of PRIMUS with high efficiency as soon as the calibration information—wavelength solutions, throughputs, atmospheric absorptions, and point-spread functions—is all known. We worked on all those things today.
Burles, Bolton, Masjedi, and I worked on calibration and understanding of the low-resolution prisim spectroscopy we have of very faint galaxies as part of the PRIMUS project. The spectra are so low in resolution that both the point-spread function in the spectral direction and the sensitivity variation across pixels matter for the extraction of spectra and determination of redshifts.
Willman told us in group meeting about all the newly discovered Milky Way substructure, including four new ultra-faint dwarfs. In all, about seven dwarfs have been discovered in the seventh of the sky covered by SDSS imaging so far, so we expect many tens of new dwarfs once we have good imaging all-sky. Substructure problem (ie, the problem that the simulations of Milky-Way-like galaxies put many more dwarfs in the halo than we have observed): be gone! After group meeting, Willman, Zolotov, and I discussed possible projects to find yet more substructure in the near term.
Finished the revision of the post-starburst galaxy manuscript, and resubmitted it, and posted it to astro-ph. It should come out on Monday morning on astro-ph. Thanks everyone for helpful comments and work. If you want an "advance copy" just let me know.
[Added later:] Amazingly, our scientific editor at the ApJ sent us the acceptance for this paper within five hours of the re-submission. Let's hear it for the efficient and excellent ApJ!
Quintero saved me today after I posted yesterday about infall regions and then realized, late at night, that I had used the wrong data subsample. Luckily the right one—obtained for me by Quintero minutes before he had to jet for his flight to Germany—showed the same effect! Whew! The rest of the day was spent working on text.
Masjedi gave me great comments on the recently re-completed post-starburst environments paper. He noted that some of the discussion about infall regions was not quite right in tone, and in fact we do have a marginally significant detection of the infall of galaxies into clusters. I am incorporating his comments now.
Bhardwaj is very close to having a full-up star-formation history fitting code, specialized for fitting old stellar populations with small amounts of recent star formation. This will be used to investigate UV-excess and mid-IR-excess early-type galaxies, and a host of other fun things.
It appears that Zolotov does confirm the Grillmair et al tidal stream candidate—we can't say if it is a stream, but a linear feature does appear to be there.
Michael Brown (Princeton) spoke about dry merger rates and the fading of red galaxy populations from his analysis of the NOAO Deep Wide Field Survey and the Spitzer IRAC Shallow Survey. He finds that dry merger rates must be fairly low since redshift of unity (consistent with Masjedi's work), but that significant stellar mass is being added to the red sequence over the same time period. His results have implications for the small-scale correlation function (a la Masjedi) and the abundance of post-starburst galaxies (a la Quintero) and the evolution of these with cosmic time.
I spent the day debugging code, one of my most favorite activities. I helped Bhardwaj debug the code she is writing to update and improve upon the kind of mean spectral analysis that Eisenstein and I did back when we had fewer data and less good models. I helped Quintero debug his code to compare the environmental dependences of galaxies in the models and in the real Universe. I helped Zolotov debug her code to re-find, in a simple way, the known Milky Way tidal streams.
Arrgh! I spent a significant fraction of today dealing with a problem caused by the fact that few consider standards—agreed-upon conventions for formatting data, for example—important. In this case, some of Willman's images, taken at some typical observatory, had astrometric information that conformed to no existing WCS standard or even convention! If you aren't going to obey the standards, then at least obey the conventions. One of the big goals of astrometry.net is to get all image WCS onto one of the standard formats, but that means that astrometry.net is going to have to know quite a few of the non-standard, idiosyncratic ways people have chosen—independently—to put the WCS into an image. Until astrometry.net standardizes everything, there will be no
grid or interoperability among heterogeneous data sets.
I spent today on issues related to getting our
tweak code up and working on the WWW. The tweak code takes an image with reasonable astrometric WCS (good to a few arcmin), and replaces it with more accurate WCS (good to better than an arcsec), computed at user-set polynomial order, and puts that more accurate WCS into the image header in a standards-compliant way. This will not be a unique capability of astrometry.net, but it is essential to its future, unique capabilities.
Blanton showed us some of his low-luminosity galaxy results in group meeting today. I was impressed that he can show that lower luminosity galaxies have similar baryon-to-dark-matter ratios to those of higher luminosity galaxies (though they have much higher mass-to-light ratios), showing that stellar winds and supernovae do not drive out gas preferentially in smaller galaxies. He also has reasonable evidence that the mass-velocity relation (the extension of the Tully-Fisher relation) at low masses is consistent with what one would expect if there is a CDM-predicted abundance of dark halos, occupied by galaxies in a more-or-less monotonic way by luminosity.
Adi Zolotov and I spoke at some length about possibly repeating, in a hands-off and quantitative way, the discovery of a putative tidal stream by Grillmair et al. We ought to recover that stream, and maybe others, and we ought to be able to put strong limits on what else is out there.
Masjedi convinced me that since the post-starburst galaxy environments paper contains some tantalizingly marginal results, we should update its data to the most recent, huge SDSS footprint (for the first time ever working on the Sloan Digital Sky Survey, I feel numbers limited!). So I obeyed Masjedi's direct order and started work on that. In principle, it just involves swapping one file in for another, but we all know that it never works in practice (and it didn't today).
At group meeting, Masjedi showed an amazing array of results, including that: the growth of luminous red galaxies by merging and accretion is limited to a few percent per Gyr at redshift of one third; dry mergers dominate the accretion rate; the transition from two-halo to one-halo term in the galaxy correlation function is clearly visible in the LRG-normal-galaxy cross-correlation; and some fraction of galaxies near LRGs seem to have tidally triggered star formation.
I didn't do much today except advise the troops.
David Schiminovich (Columbia) and Ben Johnson (Columbia) came down for the morning and lunch ostensibly to discuss their big Spitzer/IRS program on SDSS galaxies, but, as always we spent the entire time talking about galaxy evolution (and philosophy). Interestingly, I think they are the only people in the world doing a large infrared spectroscopy program on a wide range of optically selected normal galaxies—most large Spitzer programs are aimed at IR-luminous sources like ULIRGs and AGN. They bought my lunch, which was not commensurate with the lack of help I gave them on their IRS data—thanks!
I started to take a look at an old paper by McIntosh et al that might have something to say about my ideas about using central stellar densities to put hard constraints on galaxy evolution scenarios.
A new paper on astro-ph on void galaxies by Patiri et al confirms the results of Blanton et al and Quintero et al that the environment effects on galaxy formation are directly related to star-formation histories, and all the structural (morphological) relations with environment are products of the structural–star-formation relations and the star-formation–environment relations; ie, the morphology–environment relation is
secondary to more fundamental relationships. This implies that galaxy morphologies are set by local processes (though does not prove it).
Willman gave an intriguing group meeting talk about results from her last run of simulations of a Local-Group-like dark-matter concentration in a cosmological context. There is a hope of constraining aspects of early reionization with the star formation histories of the dwarfs.
Wu showed me that we do have detections of some dwarf galaxies at 8 microns in IRAC, so we ought to have detections in the IRS spectra, unless I did some kind of calculation wrong (disturbingly possible).
Masjedi told me about what he learned from his participation in the Galaxies in the Cosmic Web conference. There is rapidly growing interest in the constraints on galaxy evolution from close pairs, internal stellar densities, and continuity—all the issues we work on here.
A major theme in my talks and research for the last few years is that the Universe we see (ie, an instantaneous picture of our past light cone) is a snapshot of a continuous process in which galaxies are evolving by forming stars, merging, accreting, fading, etc. In principle, if our sample is representative, the entire continuous process ought to be reflected, statistically, in our snapshot. At the very least, the snapshot must be consistent with the continuous process. I imagine, in effect, a kind of
continuity equation analysis of the data.
In particular, I think a lot of the trivial ideas thrown into the literature about how galaxies might be evolving are strongly in conflict with this continuity consistency requirement. I worked on that a bit more directly today, by working on my small project (mentioned earlier) on what the galaxy central surface-brightness distribution (ie, the distribution of stellar densities at galaxy centers) has to say about what kinds of processes can be taking place in galaxy evolution.
One of my soap-box issues (inherited from Roweis) is that whenever you make a measurement with data, you better be optimizing a justified scalar objective function. If you aren't, then you haven't made the
best measurement you can, in any sense. Parenthetically, this is one of my objections to the HST
drizzle algorithm for making combined images; it has no such scalar.
Marshall and I spent Friday night, past midnight, working on the correct scalar objective function for the automated lens finder. It is not an easy problem; we don't have a final solution, but we learned a lot. Some desiderata: It should be close to zero when lens models don't work well on the source, and it should be close to unity when the lens model under consideration explains most of the image intensity. It should not depend strongly on whether the images have had their foreground and sky subtraction done correctly on large scales, or other details of subtraction of the foreground galaxy light.
[This post is for today.]
Phil Marshall and I sped up our lens-finding code by a factor of two today; we only have five or ten more factors of two to go!
I gave the IPAC talk, on mass build-up at redshift of 0.1 or so.
Haojing Yan (Spitzer) showed me an estimate of the comoving density in stellar mass at a redshift of six!
[This post is for yesterday.]
Phil Marshall, Lexi Moustakas, Chris Fassnacht (Davis), and I continued working on the automated lens finding code and paper. In the usual style, Marshall and I are pair coding.
I had an extremely pleasant lunch with Richard Ellis (Caltech) and his large (and productive) group. We discussed many things, including lensing and galaxy evolution and the intersection of the two.
Brent Buckalew (Spitzer) showed me nuclear and extra-nuclear mid-IR spectra of star-forming galaxies so I can see what to expect from our dwarf galaxies. Unfortunately, the low metallicity dwarfs in his sample show very little interesting emission, which bodes ill for us. Some of Buckalew's galaxies have strong metallicity gradients, but none strong enough that there are spectra with PAH emission at the nucleus, but without in the outskirts.
Phil Marshall (SLAC), Lexi Moustakas (JPL), and I spent all morning working on our nascent paper on automatically finding strong gravitational lenses in high-resolution imaging data. Our
beautiful idea is to brute-force consider every possible lens model for every object in an imaging data set, and then use the places where the sky is blank to veto lens models that predict multiple images in the wrong places. This appears to be close to working, so naturally we start writing the paper immediately.
In the afternoon I spoke with Joe Mazarella (IPAC) and collaborators about LIRGs, ULIRGs, and their environments. This is wide-open territory with a lot of likely interesting possible outcomes.
Stomach flu got most of my time today, but I did take a look at Wu's visualization of our Spitzer/IRAC 3.6 micron imaging of low-luminosity galaxies. They all seem to be well detected in our 60-s integrations. Will they be detected at 8 microns? Our Spitzer/IRS spectra suggest not.
Not being able to see any flux in any of our Spitzer mid-IR spectra of dwarf galaxies (we didn't necessarily expect to be able to see anything), I added up all of our two-d spectra to see if we get anything on average. Here's what I got:
See that whopping spectrum? No? I see something, but you have to have astronomical eyes. Next week: figuring out what limits we can place.
I had a long talk with Chuck Steidel (Caltech) today in which we ranged over many topics, but in particular we discussed the best evidence that galaxies can and do interact with their large-scale environments, by, eg, driving massive outflows with star-formation events. Chuck pointed out that there are four times as many
metal (in astronomical parlance: elements beyond helium) atoms outside galaxies as inside, so in addition to his direct evidence for outflows, there is a pretty strong indirect argument.
I gave an informal lunch talk to the galaxy-interested denizens of the Spitzer Science Center on what we know about galaxy environments from the Sloan Digital Sky Survey and related data.
I discussed galaxy morphologies and evolution in their distribution with redshift with Kartik Sheth (Spitzer). Sheth finds that the bar fraction is indeed evolving with redshift, and thinks it has to do with the mean dynamical state of disk galaxies being hotter at earlier times. I told him that I don't believe in morphology! We also discussed tests of the hypothesis that spiral arms are transients, as I mentioned in some recent post.
Jessica Krick (Spitzer) told me about intracluster stars and cD galaxies. She finds that it is very hard to decide what is a cD and what is intracluster; there may be no real distinction between these at all. She also finds it hard to measure radial gradients in their colors, because the intracluster light is so low in intensity (and she has the best flat-fielded data ever, basically). We discussed galaxy evolution and hierarchical clustering, and we figured out a possible project (based on Masjedi's projects) to see if cD galaxies are plausibly built from galaxies that fall into the clusters (the standard paradigm).
Lin Yan (Spitzer) gave a nice talk on high-redshift sources observed with Spitzer.
Today I took Lee Armus's (Spitzer) suggestions about how to look for faint sources in IRS spectra (by differencing) and turned them into a simple data reduction script. They worked nicely, in that most of the detector artifacts went away, but I still don't see our galaxies! On the other hand, we don't expect the galaxies to be very bright in the mid-infrared, so maybe our observations just put upper limits? My expected brightness and sensitivity calculations (for the proposal) involved some serious guessing.
George Helou here (Spitzer Science Center, where I will be for the next two weeks) showed me beautiful data on the mid-infrared properties of nearby galaxies, including NGC 1377, which is a luminous lenticular, but it has a tiny knot of star formation that looks like an enormously powerful, low-metallicity starburst (perhaps a bit of gas that was accreted?). The starburst knot shows enormous silicate absorption and effective extinction, but no PAH emission at all. The more normal galaxies he showed me (primarily from the Spitzer SINGS program) generally show PAH (polycyclic aromatic hydrocarbon) emission when there is star formation, with some variation depending on the contribution of AGN to the radiation field. On the other hand, the galaxies show a huge variation in the hot and warm dust underlying the PAH emission, making analysis of the PAH features non-trivial. Fortunately his team is working on some generally useful tools. Although most investigators seem to think that PAH emission is related to metallicity, this is not at all clear from the currently available data.
Lee Armus here agreed to help Ronin Wu and me understand our Spitzer spectra of low-luminosity galaxies, for which the galaxies Helou showed me will serve as
context. We might be able to shed light on the PAH relationship to metallicity and other possible things (such as radiation field, dust temperature, and geometry).
On the plane to LA, I sketched out a short paper on using the central stellar densities of galaxies to show that red galaxies are not faded versions of blue galaxies, that if blue galaxies are to fade into red galaxies, there need to be huge star-formation events, and that we can identify the predecessors of present-day red galaxies by looking for blue galaxies with huge surface brightnesses. Working title:
Blue galaxies would rather burn out than fade away.
I spent what little time I had on Friday advising students, as I am on my way to the Spitzer Science Center to be the
astronomer in residence for a couple of weeks. It should be a time to get my Spitzer programs understood, which will be a big help to me and to Ronin Wu, who is working on our Spitzer imaging and spectroscopy of low-luminosity galaxies.
Quintero and I discussed his work on duplicating his clustocentric distance measures for galaxies in a numerical simulation. He has made a group catalog analogously to Berlind's in the simulation, and is now "observing" the clustocentric distances, in order to test our hypothesis that his environmental dependences of color and radial profile will not be matched in the simulation.
Berlind, Blanton, and I discussed the relative merits of auto-correlations and cross-correlations for measuring bias and testing hypotheses. I am a huge believer in cross-correlations with dense populations. Berlind is about to test trends in cluster ages with environment.
[I ran out of time for posting yesterday, so this post covers two days.]
Eric Linder (LBL) gave a very nice lunchtime talk on Monday about dark energy, including a classification of most dark energy models (ie, physical models) into
thawing types. He showed that the next generation of experiments (eg, SNAP or JDEM) ought to be able to distinguish these types, although more will probably have to be done to resolve a particular model within the type. In the current experiments, there are large degeneracies still available for the physical properties of the dark matter.
In the afternoon on Monday, Roman Rafikov (CITA) gave a beautiful review of what is known about our own Solar System and other planetary systems, and puzzles about the formation of the planets (and in particular their solid cores). Unfortunately I had to leave before he finished!
Today I basically worked on the big, bright galaxies in SDSS, with conversations with Blanton about sky, and Quintero about
presentation, if you will.
In other news, Masjedi has a beautiful result: The mass spectrum of merging events for LRGs. He can show that the mass accretion onto luminous red galaxies is likely dominated by L-star and brighter galaxies, and that LRGs accrete at most tens of percent of their masses since redshift unity. This agrees very nicely with other, less direct measurements, so it is a wonderful and productive result.
I worked on various things relating to big, bright galaxies in the SDSS and other surveys. I also emailed the following to Hans-Walter Rix (MPIA) in preparation for my trip to Heidelberg this summer. It was partly inspired by this post by Rob Knop.
Imagine that spiral arms are created by tidal events which are then "wound up" by differential rotation. Then we can use kinematic information to "unwind" the arms and measure a "time interval" since the last tidal event. Compare this unwinding time to various possible timescales, such as those in star formation histories or in the kinematics of companions and check all simple hypotheses. Now that's an insane project; only the optimistic could embark on that one...!
David Weinberg (OSU) gave a great talk today on galaxy evolution in a cosmological context with an emphasis on disk galaxy formation. He was able to match the properties of the distribution of galaxies in disk size, stellar mass, and rotation velocity with a model with cosmological inputs and some by-hand inputs. He predicts that high surface-brightness disks have a larger mass ratio of disk mass to dark-matter halo mass than low surface-brightness disks. In general, my main concern with these kinds of analyses is that they do not allow the disk properties to be set by late gas accretion (which seems likely to me). However, his model makes testable predictions and we can test them!
I officially posted to the SDSS collaboration the Gunn Atlas of Galaxies project, outlining its main features. I will post more about this soon.
On Sunday I worked on analysis related to NYU's future in observational astrophysics projects. We will join some kind of big project, but we haven't decided which one. There are a lot of issues, involving science, personnel, and money.
I am at Brown today, giving a talk.
I gave the Princeton Gravity Group talk today, about galaxy merger rates as observed in the nearby Universe.
Afterwards, I talked to Padmanabhan and Finkbeiner about SDSS-III, PANSTARRS, PRIMUS, and other big, new survey ideas. Finkbeiner argued that we should forget all that and look for direct evidence for dark matter (by annihilation to gamma rays or any other means available). Padmanabhan was skeptical.
I spoke with Strauss and Ho about high-redshift post-starburst galaxies found in the quasar target selection sample. It looks like there are enough (hundreds out to redshift 0.8 or so!) that we can ask about evolution and environments, in comparison with Quintero's sample.
I spoke with Gunn about the nascent galaxy atlas with his name on it. We are a long way off, but Blanton and I have made huge progress in the last few weeks.
Hester and I continued our discussion of ram-pressure stripping and how the observations in the optical speak to that.
On Monday, Janice Hester (Princeton) showed us at group meeting theoretical and observational results on ram-pressure stripping of gas from galaxies as they fall into clusters or groups. At lunch, Uros Seljak (Trieste) showed us that the dark matter can't (basically) be sterile neutrinos, using the clustering of Lyman alpha forest. In the afternoon, Blanton and I had long discussions of future surveys, with the long-term goal of deciding what we should get ourselves into.
Today I gave two talks at Eastern Illinois University, including a well-attended public talk. Thanks to my hosts for a great time.
With Burles I worked on making smoothed versions of the PRIMUS spectra, with a smoothing kernel that is constant in km/s despite the fact that the spectra have very strongly non-linear dispersion.
I also worked on details of the NED SDSS mosaics; they are nearly ready for prime-time.
I worked on subroutines for comprehensive SDSS imaging of NASA Extragalactic Database sources. The idea is that we ought to be able to turn any NED object query output into the instructions for creating an imaging mosaic of SDSS data. This is all in preparation for the nascent
Gunn Atlas, discussion of which heated up this week.
I worked on galaxy envionment indicators today.
If you are measuring galaxy environments, you can choose from some options: You can choose a measure of local density that either (1) has a fixed physical length scale (radius) or else (2) is high in signal-to-noise. The reason for this dichotomy is that the former are done, basically, by counting the (necessarily small) number of galaxies in the relevant local volume, while the latter uses something like the Nth closest or a clustocentric distance or tesselation, where the scale is set by what gives good signal-to-noise (and is therefore different in different environments). We have shown that a fixed scale of 1 Mpc is about the best possible scale, but you can't measure it at high signal-to-noise.
After this, you can choose a measure that either (a) uses redshift information or else (b) works with imaging data alone. The latter allows you to go fainter into the luminosity function, and work with cheaper (and more readily available) data, but it costs you a lot in signal-to-noise, because you have (even with good photometric redshifts) a lot of projected galaxies (either foreground or background). Purely imaging indicators also have the problem that they almost never have systematics that are constant with redshift, since foreground/background and photometric redshifts all vary strongly with redshift for reasonable (read: the usual) choices that can be made.
Spent time with Burles looking through prism spectra taken as part of the PRIMUS project. Most of the spectra look nothing like what a spectroscopist is used to looking at, since the dispersion is totally non-linear, and we are (deliberately) not dividing out the system sensitivity (we are doing all fitting/modeling in the observed frame, so we multiply the models, not divide the data; there is a difference). But we found a few clear broad-lined quasars, one of which turned out to be a z=1.6 quasar with double-peaked emission lines incorrectly given a lower redshift in the DEEP2 redshift survey. Apparently PRIMUS is finding lots of redshift errors in the various large surveys, including DEEP2, VVDS, and COMBO-17. Not to worry, though, the vast majority of published redshifts are correct! We formulated a dumb project for me: Find all the SNe among the 18,000 or so extant PRIMUS spectra. I'm on it!
In the late weekend (after taxes), I worked on fundamental photometric calibration, analyzing the contribution by Stubbs and Tonry. I don't think a great deal of what they say is truly new, but it is certainly about time someone wrote it down and got the conversation rolling. It is boring as heck, but it is required if precision cosmology is to move forward. My own contributions on this subject are here (my most-cited first-author paper) and here (my least).
This afternoon, Burles, Coil (Arizona), and I discussed the current and near-future steps with PRIMUS. Coil, Blanton, and Eisenstein are working on sample selection, mechanical
collisions, and masks. Cool (Arizona) is working on extractions. Burles is working on getting the wavelength solutions right (his arc-fitting software is incredible). Masjedi is working on redshifts. What's next? Science. Note how I am not working on anything. Very clever, no?
At group meeting, Burles told us about the gravitational lenses he and Bolton (Harvard) have been finding in copious numbers (about 40?) from looking for two-redshift objects in the SDSS spectroscopy. They have one of the largest statistical collections of gravitational lenses in existence, and they have many uses for lensing, cosmology, and galaxy astrophysics.
After Burles, Coil told us about the clustering of quasars (from SDSS) with galaxies (from DEEP2) at redshift of unity. She finds that quasars are clustered like the galaxies, and have similar bias. She finds this at great signal-to-noise by using a cross-correlation (rather than auto-correlation). Indeed, as Eisenstein, I, and others have been arguing for many years, the cross-correlation is much higher in signal-to-noise than the auto-correlation function for rare populations, and you should almost never use the latter when you can use the former. Coil's result is a great advertisement for this fact, because the cross-correlation of DEEP2 with 17 (yes, 17) SDSS quasars has a higher signal-to-noise measurement of clustering at Mpc scales than the entire 2dF QSO survey auto-correlation function!
After Coil, Marla Geha (OCIW) told us about the gas fractions of dwarf galaxies, which show an enormous range, but one that is a very strong function of environment. Dwarf galaxies with very low gas fractions are almost always close to (ie, within hundreds of kpc of) more luminous galaxies. This effect has never been seen before because prior to Blanton et al (2005), there has not been a dwarf sample selected without regard to environment! The very nice thing is that Geha's results (with Blanton and Masjedi) rule out many ideas about dwarf galaxy evolution and support others—a rare thing in the world of galaxy astrophysics, filled as it is with soft predictions.
After group meeting, Mukhanov (Munich) gave a wonderful informal talk about what inflation naively and straightforwardly predicts (and what it does not). Nice!