It was a low-research day today. But I did get in a short and valuable discussion of CMB foregrounds with Boris Leistedt (NYU). The approach I want to pursue is to make a latent-variable model, which posits a set of scalar fields, and nonlinear functions that convert them into (high resolution) maps, that are compared to the data through the relevant beams. I think this will (almost provably) beat current approaches. I also had some conversations with Bedell about optimization. We are trying to fit for stellar spectra and radial velocities, and (as usual) we are finding that out-of-the-box optimizers don't work well!
Today I had the great privilege of spending the day with the group of Karin Öberg (CfA) at Harvard and also the NG Next team. Öberg's group is doing so many great things, related to astronomical observations of proto-planetary disks and also real lab experiments on ices and solid-state chemistry relevant to interstellar and accretion-disk physical conditions. Here are some highlights:
Ellen Price (CfA) showed a consistent chemical model in which they evolve the molecular contents of gas as it orbits in the evolving accretion disk. This is based on a NLTE chemical model built by Ilse Cleeves (CfA). She can see big changes as gas crosses the snow line (or various snow lines for different species). Edith Fayolle (CfA) showed absolutely incredible ALMA observations they have (with Cleeves and also Ryan Loomis, CfA) of proto-planetary disks around young stars. In these observations, there is so much, I could fill a whole separate set of blog posts: They see various kinds of organics that weren't expected to be formed in abiotic conditions. They also can image the disk in two spatial dimensions and the radial velocity dimension in thousands of chemical species. This is unprecedented detail on a disk, and also unprecedented information about molecules in these conditions. We discussed ways we could simultaneously model all of this and make very sensitive measurements of what is going on at various laces in the disk. As part of this discussion, Öberg and I discussed the problem that there isn't a good out-of-the-box imaging pipeline for ALMA, in part because different users with different targets have very different priors and goals.
But then we switched to lab stuff! Mahesh Rajappan (CfA) described the Öberg-lab experimental setups, in which they can deposit ices, including multilayer things, and then radiate them or heat them, to measure solid-state chemistry processes directly. Jennifer Bergner (CfA) is doing lab experiments to find and measure configurational rate constants for chemical processes in ices. These rates relate to the processes by which molecules find one another and reorient to permit solid-state reactions to take place. She was working in particular on O+CH4 to CH3OH. One theme of the day's conversations is that the organic chemistry of proto-planetary disks is seriously complex and contains everything that is needed for life (we think).
At the end of the day we discussed research synergies. I think the biggest is in building consistent models of the thousands of molecules, in the dynamical disk. One incredible idea is that a forming gas-giant planet should be hot (gravitational or accretion energy); this could affect the local chemistry in the disk: We could see the thermal signature of a forming planet in molecular species! That's a great goal for the near future. Öberg's group (and especially Cleeves and Loomis) have the data in hand, or coming when ALMA gets to their targets.
Today began with a meeting about GALEX, where Steven Mohammed (Columbia) showed that there is great metallicity information in the overlap of GALEX and Gaia, and we discovered that something must be seriously wrong with the astrometry in our re-calibration of the data.
Andy Casey (Cambridge) organized a phone meeting in which a bunch of us discussed possible scientific exploitation of the data in the ESO HARPS archive, which contains thousands of stars, each of which has tens to thousands of epochs, each of which is signal-to-noise of hundred-ish, and resolution of 100,000. Incredibly huge amounts of data. Huge. Casey asked each of us to describe low-hanging fruit, and take on short-term tasks. One thing we might do is re-factor the archive into something more directly useful to investigators.
Sjoert Van Velzen (JHU) gave the astrophysics seminar about tidal disruption events. He has a great set of results, starting from search and discovery, going through theory and models, and continuing on to multi-wavelength follow-up. The most intriguing result is that the TDEs are amazingly over-represented in post-starburst (E+A) galaxies (which I used to work on). It is hard to imagine any origin for TDEs that would so strongly concentrate them into these environments. It makes me wonder whether the things they are seeing aren't TDEs at all?
After the seminar, Boris Leistedt (NYU) posted to the arXiv our new paper on photometric redshifts. The idea is that we use what we know about Doppler Shift and bandpasses and calibration of photometry, but let the galaxy SEDs themselves be inferred, latent variables. This combines the best properties of machine-learning methods (that is, flexibility, non-parametrics) with the best properties of template-based methods (that is, regularization to physically realizable models, a generative model, and interpretability). It seems to work very well!
It's job season and my head is only just above water! Adam Riess (JHU) gave a nice colloquium at NYU today about the distance scale, and the comparison between the distance ladder and the cosmic microwave background.
Today was the usual research-packed day at Flatiron. In the stars group meeting, Megan Bedell (UChicago) told us about her multi-epoch survey of Solar twins. Because they are twins, they have similar logg and Teff values, so she can get very precise differential abundances. Her goal is to understand the relationships between abundances and planets; she gave us mechanisms in which the stellar abundances could affect planet formation, mechanisms in which planet formation could affect stellar surface abundances, and common causes that could affect both. She has measured 20-ish detailed abundances at high precision in 88 stars with (because: multi-epoch) SNR 2000-ish!
Doug Finkbeiner (Harvard) and Stephen Portillo (Harvard) told us about probabilistic catalogs; a project they are doing that builds on work Brewer, Foreman-Mackey, and I did a few years ago. They find (like us) that a probabilistic catalog—a sampling of the posterior in catalog space—can find fainter sources reliably than any standard point-estimate catalog, even one built using crowded-field software. They use HST to deliver ground truth. They aren't going fully hierarchical; we discussed that in the meeting, and the relative merits of probabilistic catalogs and delivering an API to the likelihood function (my new baby).
Neven Caplar (ETHZ) went off-topic in the meeting to describe some results on the time-variability of AGN. Sensibly, he wants to use time-domain data to test accretion disk models. He is working with PTF data, which he had to recalibrate in a self-calibration (he even shouted out our uber-calibration of SDSS). He is computing structure functions (which look random-walk-like) and also doing inference in the context of CARMA models. He pointed out that there must be a long-term damping term in the covariance kernel, but no-one can see it, even with years of data. That's interesting; AGN really are like random walkers on very long timescales.
In the cosmology group meeting, Phil Bull (JPL) worked us through a probabilistic graphical model that replaces simple halo occupation models with something that is a bit more connected to what we think is going on with galaxy evolution. Importantly, it permits him to do large-scale structure experiments with multiple overlapping tracers from different surveys. Much of the discussion was about whether it is better to have a more sophisticated model that is more realistic, or whether it is better to have a simpler model that is more tractable. This is an important question in every data analysis and my answer is very different in different contexts.
Between these two meetings, Bedell and I worked out the simplest representation for our Avast model of stellar spectra and Bedell went off to implement it. She crushed it! She has code that can optimize a smooth model given a set of noisily measured different epochs, accounting for differences in throughput and radial velocity. Not everything is working—we need to diagnose the optimizer we are using (yes, optimization is always the hardest part of any of my projects)—but Bedell did in one afternoon more than I have got done in the last three months! Now we are in a state to make bound-saturating radial-velocity measurements and look for covariant spectral variations in an agnostic way. I couldn't have been more excited at the end of the day.
Today was a low-research day. Megan Bedell (Chicago) arrived in NYC for the start of a two-day visit. We discussed our plans; we want to actually accomplish something this week, related to our project to find stellar spectral variations that co-vary with stellar surface motions (to improve radial-velocity measurement precision).
I discussed with Lauren Anderson (Flatiron) our project to use photometry and parallax to transfer spectroscopic labels to stars without spectroscopy (and, first, to de-noise the spectroscopic labels). This got me confused about how to explain the project to spectroscopists and non-spectroscopists alike: We have a way to use Gaia parallaxes to put logg values onto stars, but making no use whatsoever of stellar structure or evolution models, nor even scalings. Not even in the training set of labels! Indeed, I think we have a way to measure stellar masses with no use of physical models of stellar structure. I called Hans-Walter Rix (MPIA) to discuss further.
At lunch time there was an excellent brown-bag talk on light scalar dark matter by Ken Van Tilburg (NYU). He made beautiful, simple arguments about computing the properties of light scalar dark matter, and also very simple arguments about limiting the mass scale. When the dark matter gets very light, it becomes like a field of radio waves, but with a strange dispersion relation (because the particle rest mass isn't zero). This leads to highly observable effects. Huge interesting regions of parameter space are unexplored, experimentally, but there are prospects for both astrophysical and laboratory tests. There is an interesting regime at the massive end, where occupation numbers get small and the dark matter could even show macroscopic wave–particle duality effects. Overall it was highly educational, and a perfect example of the interdisciplinarity of the CCPP.
Today was day 4 of Galactic Archaeology and Stellar Physics. The day ended with my summary talk, which was unfair, scoldy, and mean and which was shouted (by me) over slides available here. Those slides will be incomprehensible without the things I said alongside them, but they will give you a sense of what themes I assembled (in real time) from the talks. A few non-representative highlights from today:
Binney kicked off the day with a discussion of analytic modeling of galaxies. His talk contained many valuable insights. For example, he showed that very simple distribution functions (in action space) can nonetheless create very non-trivial distributions in configuration space. He gave his usual—but excellent—argument for working in action–angle space: It is a consistent, continuous, conjugate coordinate system in which inference is possible. He also showed that high-quality modeling of the nearby Solar neighborhood can make good predictions for the position—velocity relationshiops for a larger patch of the Galaxy; that is, good modeling makes for highly predictive theories. At the end of his talk, he was asked about a radical dark-matter model, and he answered in terms of researcher utility, which was music to my ears.
Wegg showed an extremely good model of the mass function and microlensing optical depth towards the bulge, and constrained the dark-matter fraction. His results rule out a strong cusp in the dark matter, which is consistent with other things that were said at the meeting.
Grillmair and Carlberg talked about cold stellar streams. Grillmair showed marginal evidence for many more streams, consistent with a steep mass function in such objects. He name-checked our work on chaos and stream fanning. Carlberg shocked me by saying that the stellar streams are shorter than expected in theory. I just straight- up disagree with that, but maybe he is right when you take the full complement of streams together.
Côté brought us back to the subject of nucleosynthesis. He showed that there are many competing nucleosynthetic models that can produce the same data, but if you look on the inside, they imply very different things about the latent (physical) parameters. He breaks some of these by looking to LIGO and the rate of neutron-star mergers, which are probably involved in r-process. I loved the connections he drew between stellar chemical abundances, nuclear physics, and gravitational wave astronomy.
Day 3 of Galactic Archaeology and Stellar Physics opened with talks by Lind and Ness about traditional and new ways of measuring stellar parameters and chemical abundances. Both of them were effectively very critical of the traditional method, where there are large inconsistencies in atomic assumptions between giants and dwarfs, and there are many nuisances that affect the data as strongly as the chemical abundances in question. Lind also compared 1D and 3D models, and LTE and NLTE models and made some general statements about each quadrant. She implicitly suggested that limb darkening (or the spectral version of that) and time-domain spectroscopy might both be filled with information, because some of the 3D effects show up most strongly in the variations of the spectrum with position and time.
These introductions were followed by a set of talks that assess various aspects of the feasibility of chemical tagging—finding pairs or groups of widely separated stars that were born together in the same molecular cloud or association. This subject is dear to my heart! Blanco-Cuaresmo clearly articulated the two questions of chemical tagging, which have also come up here in this forum a few times. My phrasing of these questions would be the following: Two stars that were born together: How different can they be? And: Two stars that were born apart: How similar can they be? He then proceeded to do stuff with PCA and k-means that I didn't love; I don't think vanilla machine learning will solve this problem. However, he did (inadvertently, it seems) show great evidence that chemical tagging is conceivable. Similarly Kos did things with t-SNE that I didn't love, but which also show great evidence for an optimistic view! Carrera showed that open clusters have amazingly uniform chemical abundances. Ting showed argued that we might have to take a more probabilistic approach to chemical tagging than the original hopes. He called this the “pessimistic regime” of chemical tagging; no reason for pessimism there, but I get why he called it that.
In related news (and related to things Price-Whelan and I talk about), Mike Ireland showed an example of running the clock back on Gaia TGAS (plus spectroscopic RVs) to find the ages of disrupting stellar associations. He finds that you get more accurate ages if you take a probabilistic approach, which is music to my ears.
The afternoon was dominated by the Galactic Bulge, which appears to have sub-components formed by monolithic collapse and by long-term evolution out of the disk. The X-shape is primary evidence that the latter is the dominant process, though controversies continue. Unfortunately only one speaker showed the absolutely gorgeous Ness & Lang image, which I have the honor to have elicited with a tweet (tm) a year or so ago.
The day ended with Andy Casey talking about anomalies in the Solar abundances that run systematically with condensation temperature. Tantalizing to think it might have to do with the fact that the Sun hosts rocky exoplanets! These anomalies exist all over, however (or so it seems). They probably have something to do with dust depletion and dust accretion, which can spatially separate the high condensation-temperature elements from the low condensation-temperature elements. The talk was a reminder of how hard it is going to be to get a straightforward interpretation of anything in the high dimensional chemical-abundance space.
Today was day 2 of Galactic Archaeology and Stellar Physics. Again, a great day; here only a few highlights:
Else Starkenburg gave a review on first stars. There is no true first star known—nothing with primordial abundances—but there is one at or near −7 (that is, 7 orders of magnitude below Solar. This star, like many extremely metal-poor stars, is low in iron but very high in carbon relative to iron. That is a mystery, with many conceivable solutions. Starkenburg spoke about binary-star (mass transfer) ideas. The talk left me wondering: Do we know what a primordial-abundance star would look like? Afterwards, Schlaufman argued to me that we do, at least pretty accurately.
This was followed by a bunch of other low-metallicity star talks. DaCosta, Venn, and Schlaufman all spoke about searches for extremely metal-poor stars using clever photometric techniques. This ties in with my comment yesterday that we might be able to do a lot of Galactic Archaeology science with photometric surveys (possibly backed up by spectroscopy for calibration or training). It also bodes extremely well for Gaia Bp–Rp narrow-band photometry, which will be laden with stellar information.
Stello and Huber gave talks about asteroseismology. In Stello's review, I learned that the brightness variations are temperature variations (or really temperature–size variations), not pure size variations. This surprised me, and then was immediately obvious: The atmosphere reacts adiabatically to fast changes. He also very clearly connected the mode properties to the stellar properties, and explained the important point that dwarfs, subgiants, and red giants have different physics connecting their seismic modes to their masses, ages, and bolometric luminosities. Huber compared existing asteroseismology to Gaia data, showing that there is a consistent story, but also showing that for almost all asteroseismic targets, the asteroseismology will provide more precise distances than even end-of-mission parallaxes.
There were a set of nucleosynthesis and supernovae yield talks. My personal highlight here was a talk by Hampel about neutron capture physics. She starts with the observation that between s-process and r-process there is a whole range of neutron densities, and at different densities, you get different abundance yields. She then used real stellar data to measure the neutron density for an intermediate neutron density between s and r, calling it i. This talk stood out among the nucleosynthesis talks for its containing (like Stello's talk on asteroseismology), clearly explained fundamental physics.
Today was the first day of the meeting Galactic Archaeology and Stellar Physics in honor of Ken Freeman (MSSSO). As per usual when I am at a meeting, this blog can't convey the full day of talks, so I will just put here very personal highlights.
Freeman opened the conference, giving his overview of what he wants to know about the Galaxy. He is excited about the revolution happening now in which we might have 6-d phase space and 30-ish chemical abundances for stars all over the Galaxy. He brought up two themes that would be very important in today's talks, the bimodality in the alpha/Fe distribution (and its connection to different disk components), and radial migration. On the former, he uses the bimodality to separate the thin and thick disks; he is so confident that he literally calls a chemically separated component the “thick disk”. On the latter, he showed some results I hadn't seen on velocities as a function of metallicity that he argued make the radial migration clear. I have to figure that out! Relevant to things we worked on in the Gaia Sprint, he asked whether the disk components are different heights because of heating or a big event. I think we now know that at small heights it is heating. But the alpha-rich component might be thick because of an event.
Hekker discussed the SAGE project to get a uniform catalog of masses and ages for stars out of non-uniform inputs. She referenced The Cannon but is taking an opposite tack: She is trying to homogenize the data by making all the data constrain the same physical model.
Ruiz-Dern discussed red-clump stars, and in particular building a data-driven model of the relationships between spectroscopic parameters and photometric colors. She showed very good evidence that we could do a lot of the science we do with spectroscopy with photometry instead! That was not her goal, but it got me thinking in a totally new way about my project with Lauren Anderson.
Bovy discussed his results of dissecting the Galaxy into narrow chemical-abundance slices. Where Freeman had used the differing amplitudes in the alpha/Fe bimodality as a function of position to show how different different parts of the Galaxy are, Bovy used the same data to show how similar different parts are! That's a great property of a good scientific result: It can be interpreted either way! He discussed in detail what aspects of his Galaxy decomposition results are consistent and inconsistent with ideas from radial migration.
Talks by Duong and Chiappini again used chemistry to investigate the thin and thick disks, and Chiappini explicitly warned the audience that we will get different results if we split the Galaxy on chemical or structural lines. This also mirrored comments by Bovy.
Toyouchi looked at explaining the alpha/Fe bimodality with an event in the Milky Way's past. This got me thinking about the question: How can we tell whether the bimodality is a fundamental property of the chemical enrichment of molecular clouds or whether it is just the result of some very specific event in the Milky Way's particular past?
Feuillet showed amazing age-abundance relationships for the 19-ish elements that APOGEE observes. It is a goldmine of empirical results. She finds a few highly problematic elements. Like us, she finds that alpha/Fe is strongly correlated with age, at all alpha/Fe values and at all ages.
Buder talked about the GALAH survey and what has been learned and improved with the Gaia DR1 TGAS release. He announced that GALAH is using The Cannon as part of its data analysis pipeline. He said (and I believe him) that they are using it to speed up the code. I like that; it's good for my brand!
The day started with a conversation with new NYU graduate student Marc Williamson about a project I would like to do in the APOGEE data, looking for time variability in stellar spectra. There should be variations there, because of convection and star spots; the question is whether we can see it, and whether we learn anything about stars from it. I am optimistic. It also connects to ideas I have about improving radial-velocity measurements.
I had lunch with Jo Bovy (Toronto), with whom we discussed what Gaia and APOGEE will jointly reveal about the Milky Way. I like to say that there is great complementarity, because APOGEE is infrared, and its red giants span much of the disk, while Gaia is optical and its sensitivity is best in the halo. However, Gaia is a very sensitive system overall, so it is a detailed question just how much or little Gaia data we will get on the APOGEE giants. Of course I ended up deciding that even if we do get lots of Gaia information about APOGEE targets, that only strengthens my view!
The day opened with a conversation with Guangtun Zhu (formerly JHU) who has been doing great things with the eBOSS and MaNGA spectra from SDSS-IV. On the former, he has made a composite (average) spectrum and can see many things that haven't been seen before in galaxies like these. He can see fluorescence from the outer ISM (or maybe IGM) and he can see the effects of other extremely weak emission and absorption lines. He can also see that the emission lines are due to outflows, but in great detail: Different lines with different relative amounts of absorption and emission have different profiles and he has a consistent story for all of these.
I ended the day by working on the text in the paper on image modeling (image differencing) by Dun Wang (NYU) and in the paper on data-driven galaxy SED models by Boris Leistedt (NYU).
Today was the usual action-packed day at the CCA. Before the group meetings, Lauren Anderson (CCA) showed me work on stellar twins in photometric (APASS + Gaia) space, and how similar they are in RAVE-on spectroscopic parameters. It looks like she might be able to put (through a machine-learning-like method) spectroscopic parameters on every star in TGAS. This would be yet another data-driven model of stars.
At stars group meeting, Tjitske Starkenburg (CCA) and Keith Hawkins (Columbia) discussed this paper about chocolates which seems to be the only paper so far about new substructures in the Gaia data. There was some discussion about how they converted parallax to distance (apparently Binney has an opinion), and how the sub-structure is found via a cross-correlation between the data and random realizations. The evidence is a bit weak. However, some of the substructures look worth following up in chemistry and in other stellar samples.
This was followed by Michael Gully-Santiago (formerly Kavli, Beijing) showing his work on figuring out the ages and masses of really young (few Myr) stars, and (more ambitiously) getting the spectra of star spots on their surfaces! His project is very related to our spectroscopic binary work: How to measure a cold spot on a hot star? His approach is to modify the (Czekala et al) Starfish code to also fit for cold patches (at the same metallicity and logg as the main surface). For his particular case (LkCa-something), he finds a preference for a spot temperature of 2700-ish K covering 80-ish percent (!) of the star. This was followed and interrupted by lots of discussion about stellar binaries, long-term evolution, longitude effects, and so on.
Between group meetings, Sandro Tacchella (ETHZ) talked about his paper on gender bias in astronomy. The paper gathers good data, and is (properly) limited in its conclusions. It ends with a relatively sophisticated causal inference that is fairly convincing that women are cited less than men for papers with otherwise similar properties. This involved building a predictive model for citations. That led to a good discussion!
In the cosmology group meeting at the end of the day, Kris Sigurdson (UBC, NYU) spoke about CHIME 21-cm observations and foreground mitigation. Foregrounds should be smooth in the frequency direction, unlike the narrow-band 21-cm emission. We discussed the differences between filtering-like (or K-L-like) techniques that deliver minimum-variance modes, and subtraction-like techniques that try to build an explicit model of the foregrounds. We vowed to continue the discussion.
My people are tired of hearing endlessly about data-driven models of stars. But today Boris Leistedt (NYU) created a new one, and I am extremely excited about it.
The idea—which is more-or-less my unattempted Gaia Sprint idea—is to build a very flexible model in color–magnitude space, and then generate noisy parallaxes. That is, a hierarchical model for the parallaxes, with a color–magnitude diagram that is learned simultaneously. Today, Leistedt had the breakthrough that this could be done in bins in color and magnitude, with a Dirichlet model. That is out-of-the-box inference; he got it working and it looks nice! This is all on the path to removing physical models from (what you might call) the Gaia distance ladder (which starts at parallaxes, and ends with some kind of distance estimate for everything that can be detected).
(The first sentence of the second paragraph of this post uses all three kinds of dashes: em, en, and hyphen. Bring it on, typography nerds!)