====== Challenge 1: Moving groups ====== 
===== Objectives =====
  * Test our ability to identify known young moving groups in velocity space as a function of distance and RVS precision.
  * Investigate Gaia-like data, to quantify the expected errors and reveal new methods for identifying groups with accurate distance information

===== YOUNG MOVING GROUPS =====

There are around a dozen young (<100 Myr) sparse moving groups known within the
solar neighbourhood (d<150 pc). They are the remnants of recent low-density star
formation events, some of which are associated with the Sco-Cen OB association (Beta
Pic, TW Hydrae, Eps Cha), the Pleiades (AB Dor) and IC 2391 (Argus). 

We know that these groups are reasonably coherent in velocity space (see below).
Most of the scatter in this plot is due to uncertain distances, usually found from
proper motions and the mean group space motion (derived from a handful of early-type
members in each group observed by Hipparcos). While several of the groups have
distinct velocities (Argus, Octans, AB Dor), many are found within a “good box”
(Zuckerman & Song 2004) around the solar peculiar motion.

{{:tests:uv_groups.png|}}

Unlike open clusters, association members are very sparse, with their nearest
stellar neighbours highly unlikely to be group members.

{{:tests:groupc_xy.png|}}
{{:tests:groupc_xz.png|}}




With the high quality Gaia data to first order it is possible to visually identify
similar over-densities in phase space. To test the effectiveness of RVS velocities,
we offset several known moving in velocity and increasing distances from the Sun,
trying to recover the samples in UVWXYZ space. Even at 300 pc it is trivial to find
over-densities in distance-velocity space (e,g, near U=0 km/s,V=0 km/s).

{{:tests:groupc_uv_.png|}}

However, it’s necessary to consider full 3D velocities, as the over-densities are
not always apparent in some projections:


{{:tests:groupc_uw_.png|}}
{{:tests:groupc_uwmemb_.png|}}

This could obviously be automated with well-known clustering methods. 

*** To do: test on the 500 pc Group D sample ***

The weak point is the RVS velocities, which have typical errors of at least several
km/s at these distances and spectral types:

{{:tests:groupc_rverr.png|}}
{{:tests:groupc_rv.png|}}

This has the effect of smearing out the UVW over densities, reducing their contrast
over the background field stars. 



Young stars are typically several magnitudes above the main sequence and well
separated from older field stars, as seen in the G-band HR diagram. With accurate
distances for millions of stars, we can use this property to increase the contrast
between the groups and the field stars.

*** GOG does not appear to be respecting the absolute magnitudes  *** Need to
investigate this properly.



Also combine with other ancillary catalogues, e.g. GALEX, 2MASS, WISE


===== NEXT STEPS =====

  * What about log(g)? Is there sufficient information in the BP/RP spectra to select
low-gravity objects?

  * (Re-)defining empirical samples of young stars in Gaia photometric/astrometric system

Most of the scatter in the CMD and phase space plots above is due to the uncertain
kinematic distances of members without Hipparcos parallaxes. How these look with
essentially zero distance error remains to be seen. 

With the mean space motions of known groups nailed down, it will be trivial to
identify new faint group members from their Gaia proper motions and distances along,
with confirmatory RVs and youth indicators from ground-based spectroscopy.

===== FAKE GROUP GENERATOR =====

Probably very instructive to quickly code up a package to sample “perfect” IMFs and
isochrones and create fake groups with known velocity dispersions, spatial
structures (spherical, filaments, sub-structure) and binary properties. These can be
convolved with realistic Gaia errors and embedded in GUMS samples to test our
methods for identifying groups.