Verification of the kinematic analysis of the NIFS datacube for M60-UCD1

Inputs:
======

m60-ucd1_combine_best9.fits -- combined exposures (by Anil)

m60-ucd1_fwhm.fits -- FWHM of the NIFS instrumental response (by Anil)

sn25plus_bin.fits  -- the binning of the map in order to be able to perform
                      the direct comparison


Template spectra:
=================

PHOENIX intermediate resolution (R=10000) synthetic stellar atmospheres
downloaded from here:
ftp://phoenix.astro.physik.uni-goettingen.de/MedResFITS/R10000FITS/

8 spectra were used as templates:
M2 giants: T_eff = 3600, logg = +1.5; 
     [Fe/H]=-0.5, 0.0 dex, [alpha/Fe]=0.0,+0.4 dex
late G giants: T_eff = 4800, logg = +2.5;
     [Fe/H]=-0.5, 0.0 dex, [alpha/Fe]=0.0,+0.4 dex

lte03600-1.50-0.5.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz
lte03600-1.50-0.0.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz
lte03600-1.50-0.5.Alpha=+0.40.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz
lte03600-1.50-0.0.Alpha=+0.40.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz
lte04800-2.50-0.5.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz
lte04800-2.50-0.0.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz
lte04800-2.50-0.5.Alpha=+0.40.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz
lte04800-2.50-0.0.Alpha=+0.40.PHOENIX-ACES-AGSS-COND-2011-HiRes.fits.gz


Fitting technique:
==================

CE04 Penalized Pixel Fitting (IDL implementation)
No additive continuum
Multiplicative continuum (mdegree=11)
moments=2 (pure Gaussian) and moments=4 (Gaussian + h3/h4)
A linear combination of 4 to 8 models is used (aka optimal template).

Wavelength range:
=================

* full NIFS range (2.084 to 2.405um)
* one-half with CO bands (2.228 to 2.405um)


Results:
========

(a) Solar vs alpha-enhanced models. When all 8 models are mixed together, in
most cases the entire template weights turn to be distributed between the
solar composition models.  In some bins, there is a very small contribution
of the 3600K-1.5-0.0-alpha+0.4 model up-to a few per cent, none of the other
three contributes anything to any best-fitting template.  No correlation
between the 3600K-1.5-0.0-alpha+0.4 model weight and the distance from the
centre is seen.  This suggests that either the alpha-enhanced PHOENIX models
are wrong, or the [alpha/Fe] value in the Strader et al.  paper is
overestimated.

When only alpha-enhanced templates are left in the template list, the
chi^2 values come out higher than those for for the Solar-composition
models but the fitting looks ok.

(b) [Fe/H]=0.0 vs [Fe/H]=-0.5. When only 4 alpha=0.0 templates are
used, almost in all cases the entire weight goes to *one* template, namely
3600-K-1.5-0.0.  Generally, none of the other three contributes
significantly to the weight. But in case of the full wavelength range,
several peripheral bins show up-to 10-12 per cent contribution of the
3600K-1.5-0.5 template suggesting the marginal metallicity gradient.

(c) Full vs truncated wavelength range (red part containing CO bands).
No major differences in the central part, measurements become unstable
beyond 0.5arcsec from the center.

(d) Moments=2 (pure Gaussian) vs Moments=4 (Gaussian + h3 + h4)
Central part of the field changes a little bit, the outer parts (beoynd
0.3-0.4 arcsec) become noisier. It is definitely worth considering using the
pure Gaussian solution at least in the outer parts where the LOSVD sampling
is quite poor.


The attached files
==================

Every directory includes:
-- Postscript maps for binned versions of radial velocity, velocity dispersion,
h3, and h4 (even when they are all 0).
-- A FITS binary table containing maps (47x47) of parameters and their
uncertainties, and a map of chi^2

I do not include the template weights as it doesn't make much sense (see the
description above)

The directories are named according to the parameters of the fitting
procedure: 
-- moments=2/4 for Gaussian and Gaussian+h3/h4
-- full or red wl range
-- alpha=0.0 (4 templates), 0.4 (4 templates), or both (8 templates)

The adopted systemic velocity is v=1320km/s (1298km/s after barycentric
correction).