Accuracy of Determination of Host Galaxy Morphologies in the GOODS Fields.
Fields
E.T. Chatzichristou (Yale U.), C.J. Conselice (Caltech), C.M. Urry (Yale U.), A.M. Koekemoer (STScI), N.A. Grogin (Johns Hopkins U.)
We present realistic simulations of galaxies in the GOODS fields, especially AGN host galaxies, to determine the accuracy with which host galaxy properties can be measured through multi-color HST ACS images. In particular, we study the effect of dilution when
galaxies are observed at increasing distance. Spatial resolution might bias the detection of emissin from the central source (AGN) or its surrounding star formation regions, and the accuracy with which we can measure the host morphological and structural parameters.
In extreme cases, the AGN contribution may completely dominate the total emission coming from a galaxy or, on the other hand, a faint/partly obscured AGN can be completely dominated by the host disk emission. We simulate host galaxies of various spectral types
and luminosities, to investigate the effects of limited spatial resolution on the observed AGN/host galaxy dominance and the accuracy with which we can recover the AGN host characteristics, with various fitting routines of these synthetic images.
Accuracy
Galaxy Simulations
Two-Dimensional Galaxy Modeling
We use nearby galaxies, selected to be the best analogs to higher
redshift galaxies whose structural parameters we wish to determine. We
then use these as models to simulate them as they would appear at various
redshifts, in ACS images (see also Giavalisco et al. 1996, AJ 112,369) .
This is done by applying a scaling factor (ratio of sizes between original
and redshifted galaxy) that takes into account the image scales, the
distance to and angular diameter of the galaxy. Then, the surface
brightness of the redshifted image is reduced by the factor (1+z)4 and the
flux scaled by a factor that depends on the telescope (original and HST)
apertures, the filters used, the exposure times, etc. Finally we add the
background and readnoise appropriate for the ACS, and the final
simulated images are convolved with the ACS point spread function.
Two-dimensional surface photometry fits the bulge/disk/point source
components of a galaxy. We can assess how well these parameters are retrieved
by examining residual maps of fits. We used the fitting algorithm GALFIT (Peng
et al. 2002, AJ 124,266) that can fit a galaxy simultaneously with an arbitrary
number of components and with a variety of models, including the generalized
Sersic profile (that incorporates the deVaucouleurs and exponential disk cases),
the Gaussian and Moffat functions and the Nuker law. There is a large degree of
freedom on whether to fit or keep fixed a number of variables: the isophotal
diskiness/boxiness, position angle and ellipticity, the absolute or relative
brightness of the various components and their centers, etc.This detailed
modeling allows for the accurate fitting of even complex cases of galaxies with
bars, spiral structure and dust lanes.
Ideally, one would like to do this for a sample of galaxies with a range
of Hubble types and a range of nucleus/host galaxy dominance. Here we
show a sample of three nearby galaxies that host AGNs of different
spectral types (and thus different orientation of the central source towards
the observer, according to the standard orientation model). All three hosts
have a strong disk component with spiral structures, dust lanes, and bars
or rings, thus they represent the most difficult cases to model in particular
at high redshifts. The nuclear types (Seyfert 1, 1.5 and 2) are chosen so as
to represent a range of relative dominance between the AGN and the host
galaxy. The galaxies are simulated to how they would appear in deep
GOODS-ACS z-band images at z~0,0.5,1 ignoring any morphological
k-corrections.
Results – A Case by Case Study
NGC 1566 The host galaxy dominates the light in this Seyfert 1 galaxy. Its
morphology is complex (pronounced dust & spiral structures). At z=0 the
nucleus is several times fainter than any of the other two components that
form the host (both flatter than a deVaucouleurs). At higher z the gaussian
(used to fit the compact nucleus) “robbed” some of the light from the disk
component suppressing it (the Sersic exponent becomes lower). At those z’s
the gaussian FWHM is only a fraction of a pixel, limiting the fitting accuracy
of this component.
NGC 7213 A similar set of components fits this Sy 1.5 galaxy. At z=0 the
nucleus comprises ~30% of the total light, the steeper “bulge” component
dominating the host light. At higher z’s there is a trade off between the bulge
and nuclear components: the gaussian FWHM becomes progressively larger
while the bulge becomes progressively flatter to compensate and ultimately
robbs light from the compact source. In fact at z=1 the gaussian FWHM=1.2
pix and the bulge itself is small (2 pix), making it hard toi disentangle
accurately the 2 components. The third fainter disk component remains
unchanged.
NGC 0262 The nuclear component in this Sy 2 galaxy is too faint to be
accurately fitted by our two-dimensional method, even at z=0 and is
probably included in the bulge (deVaucouleur) component that comprises
~97% of the light at all redshifts. At z=0 the rest of the light is equally
distributed into an exponential bar and a flatter disk component. The latter
has a small scale length and thus at higher redshifts the two components are
blended into one, while the total B/D ratio is preserved. At z=0.5 the bar is
still perceptible but not anymore at z=1.
NGC 1566
For the modeling of our galaxies we have consistently used at least two
components, without imposing any a priori knowledge for any of the parameters
(even for the simulated high redshift galaxies for which we knew accurately the
parameter values after having modeled them at z~0). In all cases we used the
most generalized law, the Sersic profile, and after fitting and removing the host
galaxy model we let the algorithm determine whether a nucleus is present and
subsequently fit it with a Gaussian function. The method described here is more
accurate than the one-dimensional fitting techniques or the two-dimensional fits
with fixed (deVaucouleurs and exponential) components, that are often
employed. However it is also a longer and individualized process (e.g. it is
impractical to do this with a large number of galaxies) and the interpretation of
the results is also more complex.
Right: For all galaxies we show
the original images (left), the
individual components fitted by
Galfit (middle) and the residual
images (right), in positive gray
scale. For clarity, we show the
individual components only for
the galaxies at z~0 (exact
redshifts are listed in Table 1)
shown in the upper panel. For
the galaxies simulated at z=0.5
(middle panel) and z=1.0 (lower
panel), the simulated images are
shown left and the residual
images right. The detailed
parameters for all the fitted
components are listed in Table 1.
NGC 7213
Left: Surface brightness plots
showing the 2-dimensional
decomposition of each
galaxydata into individual
components. Symbols are the
observed data, dotted lines
indicate the individual
components and the full line is
the final model (sum of all
components). Units on x-axis are
pixels (0.05”) and on y-axis
mag/arcsec2 (scale arbitrary).
NGC 0262
Conclusions
The aim of this work is to assess under what conditions (e.g. galaxy
brightness, redshift and S/N) the parameters of a galaxy can be accurately
retrieved. This should give us some confidence in the usefulness of the
“real” morphological measurements performed for a large number of
galaxies in the GOODS fields by two-dimensional surface brightness
profile fitting techniques (see several posters in this Session) .
We find that the degeneracy between various components can be very
important at higher redshifts where one lacks resolution. As far as the
accuracy of nuclear source extraction is concerned, one is not able to
separate the bulge and nuclear components uniquely in the highly
redshifted images (and even at low redshifts in some type 2 nuclei), unless
the nucleus dominates the host light. On the other hand, the B/D ratio can
vary between nearby and high-z galaxies that are marginally resolved, this
depending on the prominence of the disk component and the existence or
not of a central source (there is a “trade-off” between the bulge and nuclear
components at high z, that alter the B/D ratios). Finally, the detailed
modeling of the host galaxy light at z=0 into components such as bars, can
be lost at higher z merging into one disk component.