pseudo-wide-area-IFU: slit-less bright stellar spectroscopy

Mapping stellar pops/kinematicss with slit-less spectroscopy?

Science goal:

There are parts of the sky that are quite 'crowded' in stars, where we would like to get some basic spectroscopic information for all (bright enough) stars: velocity, Teff, (logg), and [Fe/H] (other elements, too?).

The prime example (accessible to optical observations) in our Milky Way may be Baade's window; and parts closer to the Galactic center, for near-IR observations.

This would lead to orbit-abundance-age(?) information for vast sets of stars, as a basis for dynamics,

and for formation studies based on the structure of the abundance-orbit distribution.

Effective ways of getting such information:

The "obvious" straightforward approach may be to do vast bona-fide IFU mosaics. But this approach is likely to hit its limits (sociological, time-allocation?) at 100-200 pointings? At any rate, e.g. Baade's window is 1000's [TBC] of squarearcmin (or MUSE FOV's).

Slit-less spectroscopy as an alternative?

The basic set-up envisioned is a follows: consider a classic long-slit spectrograph, where usually a slit selects a tiny fraction of the focal plane for subsequent dispersal by a grating/grism/prism; often a pass-band filter in the optical path limits the spectral extent.

Now, envision that same set-up, but with the aperture (slit) plate removed. Let's consider the regime of a single bright star in the field: the detector then will show a simple spectrum of this star; as the star is a point-source, the slit mainly served to eliminate much of the sky background. So, one will have a spectrum of a star (limited by, say, a narrow-ish passband filter of, say 200A), but with "200A's worth of sky", not the "2A's worth of sky" (for a slit).  Note, that to good approximation, the stellar spectrum depends on 5 numbers, (x,y)_pos, flux, Teff, v_los, [Fe/H]   (where logg and flux are degenerate at a known distance).  If you wanted to get (Teff, v_los, [Fe/H]) for that star, you'd fit model spectra, given (x,y)_pos, flux. These 6 pieces of information are to be compared to photometry, which yields three pieces of information: (x,y)_pos, flux.

Now, imagine the field being full of bright stars, and slit-less spectroscopy. It will look the same as before, just like a seemingly (!) crowded mess with 100s (1000s) of star spectral streaks, many of them overlapping. But -- and this is the core conjecture of this approach -- when it comes to information content, this image is not much more crowded. Let's presume that we have photometry, which makes (x,y)_pos, flux_normalization a "given". Then we have to solve for 

p( { Teff, v_los, [Fe/H]}_(all_stars_in_field) | {(x,y)_pos, flux}__(all_stars_in_field) ).

As long as the signal detection is linear, this is just the linear superposition of the problem above.

And, if observations at different angles are obtained (2-3), then degeneracies of directly overlapping

stars could be mitigated [HWR's view: that's almost unneeded: the famous SB2 binaries show that

separation in velocity space suffices..]

The bain is possibly: how narrow to choose the passband filter., not to get killed by sky; but still have enough spectral coverage to get stellar parameters.