Presented at the end of the 2019 Santa Barbara Gaia Sprint at the Kavli Institute for Theoretical Physics
2019 SB Gaia Sprint
Wrap-up — KITP — 2019-03-29
Hans-Walter Rix (MPIA, Heidelberg)
Learned many things, among them
● the small-scale structure of phase-space -- abundance space
is strong and detectable -- Kamdar, Coronado
● the phase-space spiral twists, as it should -- Johnston
● wide-binary twins are real (& probably very useful) -- El-Badry
● GDR2 as a time-domain survey can and will be ``calibrated’’ -- Casey
● Gaia preselection of close, eclipsing binaries works perfectly -- Schwab
Homework: help Anthony Brown find X MEuro to “do the selection function”
Ellianna Schwab Abrahams (UC Berkeley, CTDI)
What Worked (Time Domain Gaia): What Didn’t Work (yet!):
● Measurements of flux RMS detect
variability! Pulled lightcurves out of
TESS for candidate variables confirming
close eclipsing binaries, RR Lyrae and
CVs. (With Mowlavi, Rix and Brown)
● Built a DSD neural net to predict RVs.
RVs correlate enough with other features
to predict within largest uncertainty (With
Jamal, Hogg and Casey)
● Need to add Bayesian
CV in both
TESS + Gaia
Thanks Rix, Brown, Casey,
Gaensicke & Gaia DPAC!
Alice Quillen (Rochester)
I tried using current velocities and backwards orbits of moving groups,
listed by Gagne+18 and with ages less than 60 Myr years, to see if the stars
in them formed near spiral arms and find out where were the arms?
Young moving groups seem
associated with the “Local spur”.
This arm may have wound up -- or
two different armlets are required.
Now 60 Myr ago
-Ringing might have also been caused by a dwarf
plunging through the center of the Galaxy.
-VVV measuring proper motions in the bar’s X
(resonant model for vertical heating?)
Galactocentric angle degrees
Joel C. Zinn
Ohio State University
1. Responded to referee report on
comparison of Gaia
distances/radii to asteroseismic
2. Learned about many cool things
in discussions with people, e.g.:
a. Halo is not relaxed (Quillen)
b. White dwarf merger rates are
consistent with Q branch
number density… no Kozai?!
c. Halo stars have unexpected
abundance patterns (Matsuno)
Semyeong Oh (Cambridge)
Linear velocity field of 40 m/s/pc
may be detectable with
astrometry + DR2 (statistical) noise
+ 30% RV (1 km/s uncertainty)
for the Hyades cluster.
● important systematic bias for proper
motions of bright sources
● T. Brandt’s Hipparcos-Gaia
calibration removes this and
reduces the precision by 5-10 for
Max Planck Institute for Astronomy
● blue and red patterns correspond to density fluctuations in
the disk — spiral arms?
● quadrupolar velocity structure is seen in multiple models of
the Galactic bar.
● vertical & radial velocity dispersions increase where we
expect the bar to influence the stellar dynamics.
(Flatiron Institute, CCA)
Compared the velocity tensor elements of
Eilers’ & Hogg’s APOGEE/Gaia data to the
velocity tensor elements of simulated bars
I did this
….what we should
expect to see in the
various terms of the
We see the bar in APOGEE/Gaia data and it looks
similar to simulated bars
[Will insert plots]
Comparison to Eilers’ plots
with Eilers, Hogg, Laporte, D’Onghia, Rix
and advise from Price-Whelan, Gerhard, Widrow
David W Hogg
(NYU) (MPIA) (Flatiron)
The radial-velocity variations we see (with APOGEE & Gaia) in the Milky Way disk
are going to provide evidence for density variations.
(Eilers, Pearson, Laporte)
Related to this, I formulated a method to use disk simulations to build an
empirical-theoretical model of disk patches.
If we are seeing the bar at the Galactic Center, it is at the wrong angle!
(Gerhard, Widrow, Pearson, Laporte, Rix)
We need to sharpen up a proposal to build a set of Gaia selection functions.
(Brown, Fouesneau, Rix, Casey, Price-Whelan)
(University of Wisconsin - Madison)
What I did:
Compare the contributions of different sectors of
the U-V plot in different volumes to begin to
disentangle moving groups from open clusters
What I Learned:
Getting to work on astronomy research all day is
even better than I thought
Cosmological galaxies are not
in an inertial frame
What I learned this week:
Velocity of disk COM
relative to halo COM
Acceleration of COM calculated different ways
Position of disk COM
relative to halo COM
theme of the week: diffuse disequilibrium
(in science and life)
Andrew Wetzel (UC Davis)
● Reconstruct gravitational potentials of MW-like
galaxies/halos using Latte simulations (Sanderson,
○ Initial analysis (see right) is a best case scenario:
no M31, no LMC
○ Milky Way is not an inertial frame
● Use simulations to test effects of LMC on stellar halo
● Use simulations + synthetic Gaia surveys to
reconstruct 3-D dust distribution (Anderson)
● Public service notice: use cosmological Gaia mocks!
(stolen from Sanderson)
Siddharth Mishra-Sharma (NYU)
- How to perform Gaussian process regression on a sphere (with lots of help and code from Tim
Brandt and Hamish Silverwood).
- Code at https://github.com/smsharma/GPSphere (with Tim).
- Constructed model of the Gaia DR2 quasar proper motion field with GP regression.
- Can use to characterize measurement systematics?
- But not yet to look for substructure!
- Used this method to construct un(??)biased power spectrum estimator.
Alexandra Borukhovetskaya(University of Wisconsin - Madison)
undergraduate student working with Elena D’Onghia
What I came to do:
Learn Arcturus ages, metallicities, and
What I did:
1. Found ages and metallicities for known
members of Arcturus.
2. Ran algorithm (SNN) with adjusted
parameters to find several new
3. Compared ages and metallicities of
candidate members to those of known
What I plan to do in the future:
Continue modifying and improving the
algorithm to find new members and gain insight
into the origin of Arcturus.
Harshil Kamdar (Harvard)
What I Did: finished writing up a paper on v wide (>10 pc)
co-moving pairs due to disrupting star clusters in the disk, worked w/
Coronado & Rix on interpreting small-scale structure in combined
chemodynamical space (Gaia+LAMOST).
What I Learned: sampling points to follow a correlation function is
ICC University of Barcelona
-Red Clump & Giants RVS samples corrected using cross
match to 2MASS - thwarted at the last minute by a
string encoding problem in Pandas.
-GPflow & GPyTorch running in 2D with heteroskedastic
noise (i.e. uncertainties - surprisingly non-trivial).
-GPR on e.g. vRvR data (left) - needs work.
-Derivatives of latent function f(x) with GPflow
machinery (1D example above) - still quite rough, red
line should match green line.
Anthony Brown (Leiden Observatory)
What I did
● Struck gold!
○ Python notebook with example on how to create
a 6D ‘Gold’ sample from Gaia DR2
● Put my RUWE code on github
● Answered tons of questions on Gaia DR2
select source_id, ra, ra_error, dec, dec_error, parallax, parallax_error,
parallax_over_error, pmra,pmra_error, pmdec, pmdec_error, ra_dec_corr,
ra_parallax_corr, ra_pmra_corr, ra_pmdec_corr, dec_parallax_corr,
dec_pmra_corr, dec_pmdec_corr, parallax_pmra_corr, parallax_pmdec_corr,
pmra_pmdec_corr, radial_velocity, radial_velocity_error,phot_g_mean_mag,
phot_bp_mean_mag, phot_rp_mean_mag, bp_rp, g_rp, bp_g,
2.5/log(10)*phot_g_mean_flux_over_error as phot_g_mean_mag_error,
2.5/log(10)*phot_bp_mean_flux_over_error as phot_bp_mean_mag_error,
2.5/log(10)*phot_rp_mean_flux_over_error as phot_rp_mean_mag_error,
sqrt(astrometric_chi2_al/(astrometric_n_good_obs_al-5)) as uwe
and radial_velocity is not null
and rv_nb_transits > 3
Do some radial velocity cleaning...
What I learned
● Phase spiral is not ‘just moving groups’
● Some of use do want to work on the selection function
○ Look for funding
● So many creative ways to use the Gaia data!
Cecilia Mateu (UdelaR)
What I Did+Learned:
- Provided astropy reference
frames for streams in galstreams
library (+Adrian Price-Whelan)
- Made galstreams Python 3
- Many RR Lyrae in Pal5's tails
- Many streams are (very) difficult
to confirm, but many (some?)
do show up in PS1 photometry
(w. Nora Shipp, A. P-W, Sarah
Pearson, Giuliano Iorio)
- Learned many cool astropy tools
- Gaia RR Lyrae Gold!
Sukanya Chakrabarti (RIT)
Worked on : what caused the disturbances in the outer HI disk of the Milky Way ?
Sampling Gaia PMs
(Shanghai Astronomical Observatory)
● Got Member Probability of each star around
Pal 5 from Photometry
Sprint & Eugene Vasiliev (IoA, Cambridge)
Thx for the advice from Nora, Adrian, David, and Giuliano !
Motivation and Future Plans:
1. Get tail members with 6D → Disruption Process
2. Hunting for UFD
You will see magic happens after combining probabilities from
photometry & astrometry in Eugene’s slide!
1. How to execute a project efficiently
2. I need another PhD in Astronomy
● Happened to find Phase-Space
Slices of Gaia Sausage using
Sprint & Tadafumi Matsuno (PhD student,
NAOJ) and Kohei Hattori (U of Michigan)
You will see what we can infer from the slices
in Kohei’s slide!
Eugene Vasiliev (Institute of Astronomy, Cambridge)
Search for tidal tails around globular clusters with Gaia astrometry and DECam photometry
With Zhen Yuan and valuable advice from Adrian, Ana, Sarah P., Jason H., Nora, Giuliano
Pal 5 NGC 288
Learned Gaussian ProcessesⓇ, Matched FilteringTM.
Developed the membership selection pipeline combining
Gaussian Mixture for parallax+PM and Matched Filtering for CMD
Re-discovered tails around Pal 5, NGC 288.
Not found anything around NGC 5897, NGC 7089, …
Plan to continue the search for other cluster tails...
Kohei Hattori (U of Michigan)
phase-space slice of the Gaia Sausage
With Tadafumi Matsuno (PhD student,
NAOJ) and Zhen Yuan (SHAO).
StarGO package (by Zhen Yuan) is
powerful in finding substructure.
Red: Solar neighbor / Black: total sample
Elena D’Onghia (work with Boquan Chen)
We use SNN algorithm (Chen et al. 2018) to cluster in 5 dimension to
rebuild Orion. Two clumps are newly discovered.
This approach opens us the possibility to study how star formation occurs
This is an example of the superiority of SNN on DBSCAN.
(1) Two clusters one dense one less dense
(2)-(3) DBSCAN recoves one group
(4) SNN recovers both groups
Lorenzo Rimoldini (University of Geneva, CH)
Classification of variable objects in all Gaia DR2
(without time series):
> better than you might expect
>> it can be useful before DR3 (SDSS-V, TESS, ...)
>>> help in gathering rare variable classes
>>>> millions of new variable candidates soon!
Learned about streams, complementary data sets,
interesting color-color plots combining optical, IR
and radio, features of AGBs in the radio band,
some details about Gaia as well :)
Please email me by the end of April if you want
to make sure your class of interest is included!
Giuliano Iorio (Institute of Astronomy, Cambridge)
● Pal 5: 2 RRab, 4 RRc
● Stream: 16 RRab, 6 RRc
Membership likelihood for RRLs in Pal 5 stream
Thanks to Cecilia, Adrian, Nora, Eugene, Elianna
for advice and useful discussions
What I did:
Learned: Galstreams, what is a GP, a
lot of interesting way to use Gaia Data.
What I found:
A lof of RRLs!!
To do: Improve background model,
analyse other streams, Pal5 stream
No other variables
Productive discussions on phase space spirals with Kathryn,
Chervin … radially plunging satellites with Heidi, and GP blah blah
blah with Lauren and Hamish, to be continued in Gaia “brisk walk”
during and after the conference
Gauthier, Dubinski, LMW 2006
LMW, Chequers et al 2014
Spekkens & Sellwood 2007
Infer surface density
from moments of the
velocity field --- here,
a linear model on
pixels of simulation.
Probably makes more
sense to go after the
FR , Fphi...learn rather
than assume the
Jeans/CBE (cf. Alice
Density model residual
for helpful advice and questions
Sample of 11475
complete and ~89%
RGB selection is still
in progress, but will
and purity (~67%)
I learned: (1) It works!
(2) How to cross-match on the Gaia Archive
(3) The memory limit of my laptop
Color code: amplitude of local
corrections with respect to DR2.
Some catalogs: no patterns
Others: stripes along RA or Dec + large
Banding due to old ground-based
positions used to build proper motions.
Larger scale patterns: reference system
differences, or rotation.
Ad e t in -bi g o
as id r e r .
TODO: apply to DR2 pm correction for rotation (Lindegren’s doc); check current results vs. astropy method
for position propagation.
Discovered this week: interesting intersections between dynamics in the Galaxy and the Solar System….
Ideas for an article! ….Sprinting ! It works!
Juna Kollmeier (@ Doc right now unfortunately)
GG Catalog Progress
GG Brown Dwarfs: Faherty (Done)
GG White Dwarfs: Gaensicke (Done)
GG: CMD: Brown (Done)
GG: 6D Phase Space: Brown (Done)
GG Spiral (Is there a spiral?)
GG Red Clump (Hawkins?)
GG Low Z (Simon?)
Where to host these?
I’m Andy Casey (Monash) and this week I self-calibrated and simulated the
astrometric and radial motion jitter for single stars (w/ Rix and Price-Whelan)
the “binary fraction” simulated AEN
- p.d.f. Of a source being a
single star given
- radial velocity
semi-amplitude for binary
Aim: 3d bulge kinematics from RVS
Newberg et al. (2007)
m-M = 14 - 18.5
[Fe/H] = -2
Age = 12 Gyr
Alexa Villaume (University of California Santa Cruz)
Observed color - Synthetic color
Asked a lot of questions
about sources of
uncertainties and quality
flags in DR2 (thank you
Anthony and Morgan!)
Learned: Apples-to-apples comparisons between
synthetic and observed photometry is complicated
and many corrections are needed
Timothy Brandt, UCSB
Kareem’s wide binaries can calibrate parallax uncertainties. Formal uncertainties
are 20-30% underestimates? Only addresses spatially uncorrelated errors...
Hints of magnitude-dependent zero-point offset (couple μas/mag)?
Emily Cunningham (UCSC)
1. I worked with Sanderson and Wetzel on how we can quantify and measure the impact of an
LMC-like interaction on the velocity field of a mostly accreted stellar halo.
2. I estimated the parameters of the velocity ellipsoid over different octants in the sky using RR
Lyrae! In the Milky Way (Pan-STARRS) and the Ananke mocks for the Latte simulations.
Milky Way m12m: Latte/Ananke
w/ Cecilia Mateu: coordinate frames for stellar streams in galstreams
→ Next: we will clean up the library and provide other metadata for streams
w/ Nora Shipp: what “known” streams in the north can be confirmed?
→ Next: compute significance for candidate rediscoveries, repeat with DECaLS?
w/ Cecilia+Giuliano Iorio: RR Lyrae in the Pal 5 stream, >15 candidates
→ Next: fit stream models to new distance information, improve MSTO filter
w/ Ana Bonaca: Improving models of the GD-1 stream
1) How to use new RV measurements to improve fit of “spur”
2) Better global models of the stream (Fail!)
→ Next: Figure out what is going on with the stream track?!
Adrian Price-Whelan (Princeton)
2D density model
Viz work! Thanks to
Nora for data on
SiO maser emission
No Coincidences in Gaia
Coincidences in Gaia+2MASS (~400, %100 MIRA!)
David Mykytyn, New York University
● Managed to learn a decent kernel density for the
main sequence stars
● Very close (alas not close enough yet) to be able
to generate the “binary sequence”
● Hopefully this will happen very soon?
● Learnt KDE in PCA space, and transformed it
backwards into the real CM space
What I learned this week among other things
● Estimating Gaia’s small field of view correlations
● Wide-binary identical twins are real - good for parallax validation
● Effort to have Gaia selection function(s) [but will be painful to build]
Morgan Fouesneau (MPIA, Heidelberg)
What I did this week
● Dynamics of M31 HII regions using biased Gaia astrometry
● (incomplete) Empirical star cluster model
Actually stars! Not looking at stars - epic fail!
No coherent rotation
Consistent with random
Van der Marel et al. 2019
Jatan Buch(Brown U.)
- Analytic modeling of disequilibria using the continuity equation with Larry and Hamish: can’t resolve all
- Accelerations for Hipparcos-DR2 x-matched stars (courtesy T. Brandt), definitely worth exploring!
- Great breakout on simulations and mock catalogs. Got a flavor of the diversity of philosophies and techniques,
going into the available codes, and, most importantly, their scope depending on the scientific problem at hand.
> Worked on:
- Responded to referee report for a paper on estimating the local DM content in the solar neighborhood; mild
evidence of ‘excess’ matter density (also verified by A. Widmark using an independent technique).
- Will better disequilibria modeling only improve precision and rule out the excess? Can only answer on a
- Baby step: started simulating effect of DM halo (final goal: NFW population) + test particle disk on a stellar
distribution using galpy potentials (thank you J. Hunt). [Bummed couldn’t make at least 1 pretty plot! :(]
- Implemented A. Brown’s RUWE cut and found no significant change in derived results!
May Gade Pedersen
(KU Leuven, Belgium)
Recreated from Bastian+2018
What I did:
Constructed Gaia CMD of cluster with known
eMSTO. Converted isochrone clouds to Gaia
colours and ‘fit’ to observations.
Principles behind cluster membership
verification. Use appropriate Gaia
passbands for bolometric
Kareem El-Badry (UC Berkeley)
Twinning: excess of equal-mass
wide binaries is real!
Made a catalog of 3000
hierarchical triples with
Lots of other ideas:
● Companions to massive
white dwarfs with Sihao
● Parallax error calibration
with Timothy Brandt
● Kozai mechanism in
triples with Juna
Thanks for great discussions
with Rix, Casey, Gaensicke,
Oh, Hogg, Fouesneau, Brown,
Still working on the math ¯\_(ツ)_/¯
Jason Hunt (Dunlap/Toronto)
Talked about: Non-axisymmetric structure, actions, galpy, phase space
spirals, actual spirals, bars
Worked on: Lots of little things, including
paper draft from last sprint, and looked at
models of different spiral structure
Simulations from Alex Pettitt
Ana Bonaca (Harvard)
Found GD-1 models with relative radial velocity
offsets between the stream and the spur
comparable to the observed offset
Lauren Anderson (Flatiron Institute):
Worked on building a 3D dust map of the Milky Way
Simulation: (Almost) reconstructed the “true” dust density in the Latte simulation,
the foundation for Ananke. We have the right amplitudes and scales !!
Data: Applied to 105 dust estimates from Gaia with good parallax estimates
Hyper parameters: ℓ ~ 60 pc and σ ~ 80 M
/pc3 (optimized from a more local sample)
Sihao Cheng Johns Hopkins University ● relation between the white dwarf Q branch
and wide binary fraction with Kareem
El-badry, supporting the idea of Q-branch
WDs being WD-WD merger
● extinction within 250 pc is still uncertain
● tried to analyse proper motions of Sagittarius
● learned a lot. Some new ideas for my next
possible project. Thanks for discussions with
Newberg, El-badry, Anderson, Casey,
Gänsicke, Zinn, Hunt, etc.
Borja Anguiano (University of Virginia)
Kathryn V Johnston (Columbia)
?? PS-Spirals - where? when? what??
with Chervin Laporte and Larry Widrow
Did: orbits in axisymmetric MW potential
Jphi (=Lz!) => groups with shared history
dvz & dvR “hit” => z-vz spirals & R-vr sloshing
???? Towards the ORIGIN ????:
“Rewind” spiral in Jphi bins
test approach with N-body sims
limitations: impulsive assumption; local view; moving groups
red: rguide =9 kpc white: rguide =7kpc
Chervin F. P. Laporte (UVic)
PS-spirals (Larry, Kathryn, Anthony locally & Martin
Cool Smith+Jason H+Jason S), finding the OLR with
DR2 (Jason H), GP blablablah (Larry+Hamish),
velocity field of MW (Eilers+Pearson+Hogg), Moving
groups Arcturus (Elena D’Onghia)
Did: checked some quick dirty solutions to my
problems, didn’t work… but looking forward to next 3
weeks :). Made this movie of the PS-spiral in
overdensity to help you hypnotise your guests back
Learned: I might have to GP blablablah my way out of
this with RC to do what I initially intended to do.