NICMOS and VLA Observations of the Gravitatonally Lensed Ultraluminous BAL Quasar APM~08279+5255: Detection of a Third Image
Extracted source text ยท arXiv astro-ph/9908052
%% AASTeX v5.0 LaTeX 2e macros.
%\documentclass{aastex}
\documentstyle[12pt,aaspp4,psfig]{article}
%\documentstyle[12pt,aaspp4,psfig]{article}
%\usepackage{emulateapj}
%\usepackage{apjfonts}
%\usepackage{psfig}
%
%% preprint produces a one-column, single-spaced document:
% \documentclass[preprint]{aastex}
%% preprint2 produces a double-column, single-spaced document:
% \documentclass[preprint2]{aastex}
\def\ltsima{$\; \buildrel < \over \sim \;$}
\def\simlt{\lower.5ex\hbox{\ltsima}}
\def\gtsima{$\; \buildrel > \over \sim \;$}
\def\simgt{\lower.5ex\hbox{\gtsima}}
%
% MATH FUNCTIONS:
\def\erf{\mathop{\rm erf}\nolimits} %error function
\def\sech{ \mathop{\rm sech}\nolimits} %hyperbolic sec
\def\csch{ \mathop{\rm csch}\nolimits} %hyperbolic csc
\def\arcsinh{\mathop{\rm arcsinh}\nolimits} %arc hyperbolic sin
\def\arccosh{\mathop{\rm arccosh}\nolimits} %arc hyperbolic cos
\def\arctanh{\mathop{\rm arctanh}\nolimits} %arc hyperbolic tan
\def\arccoth{\mathop{\rm arccoth}\nolimits} %arc hyperbolic cot
\def\arcsech{\mathop{\rm arcsech}\nolimits} %arc hyperbolic sec
\def\arccsch{\mathop{\rm arccsch}\nolimits} %arc hyperbolic csc
\def\arccot{\mathop{\rm arccot}\nolimits} %arc cot
\def\arcsec{\mathop{\rm arcsec}\nolimits} %arc sec
\def\arccsc{\mathop{\rm arccsc}\nolimits} %arc csc
\def\ylm{\mathop{\rm Y}_l^m\nolimits} %spherical harmonic
\def\ylmp{\mathop{\rm Y}_{l'}^{m'}\nolimits} %spherical harmonic primed
\def\real{\Re e} %real part
\def\imag{\Im m} %imaginary part
% UNITS:
\def\km{{\rm\,km}}
\def\kms{{\rm\,km\,s^{-1}}}
\def\parsec{{\rm\,parsec}}
\def\kpc{{\rm\,kpc}}
\def\mpc{{\rm\,Mpc}}
\def\msun{{\rm\,M_\odot}}
\def\lsun{{\rm\,L_\odot}}
\def\rsun{{\rm\,R_\odot}}
\def\pc{{\rm\,pc}}
\def\cm{{\rm\,cm}}
\def\yr{{\rm\,yr}}
\def\au{{\rm\,AU}}
\def\g{{\rm\,g}}
\def\om{\Omega_0}
\def \ca {{\it ca.\/}}
\def \magnitude {$^{\rm m}$}
\def\kr{${\cal K}_r$}
\def\kz{${\cal K}_z$}
\def\kzz{${\cal K}_z(z)$}
\def\mss{{\rm M}_\odot \rm pc^{-2}}
\def\msss{{\rm M}_\odot \rm pc^{-3}}
% MISCELLANEOUS:
% angles in degrees
\def\deg{^\circ}
%\degg produces degree symbol so that 3\sec5 produces 3.`5 with the degree
%symbol and the period aligned.
\def\degg{\hbox{$\null^\circ$\hskip-3pt .}}
%\sec produces arcsec symbol so that 3\sec5 produces 3."5 with the second
%symbol and the period aligned.
\def\sec{\hbox{"\hskip-3pt .}}
\def\half{{\scriptstyle{1\over2}}}
%\s produces tilde in mathmode or horizontal mode.
\def\s{\ifmmode \widetilde \else \~\fi}
\def\={\overline}
\def\scre{{\cal E}}
\def\spose#1{\hbox to 0pt{#1\hss}}
\def\larrow{\leftarrow}
\def\rarrow{\rightarrow}
\def\llangle{\langle\langle}
\def\rrangle{\rangle\rangle}
\def\etal{{\it et al.\ }}
\def\cf{{\it cf.\ }}
\def\eg{{e.g.\ }}
\def\ie{{i.e.\ }}
%\lta and \gta produce > and < signs with twiddle underneath
\def\lta{\mathrel{\spose{\lower 3pt\hbox{$\mathchar"218$}}
\raise 2.0pt\hbox{$\mathchar"13C$}}}
\def\gta{\mathrel{\spose{\lower 3pt\hbox{$\mathchar"218$}}
\raise 2.0pt\hbox{$\mathchar"13E$}}}
%\Dt and \dt put Newton's notation dots above upper and lower case chars
\def\Dt{\spose{\raise 1.5ex\hbox{\hskip3pt$\mathchar"201$}}} % upper case
\def\dt{\spose{\raise 1.0ex\hbox{\hskip2pt$\mathchar"201$}}} % lower case
\def\del{\nabla}
\def\delv{\bb\nabla}
% TSS ADDITIONS:
\def\cte{C^{te}}
\def\sch{Schwarzschild\ }
\def\schs{Schwarzschild's\ }
\def\=={\equiv}
\def\jla{J_{\lambda}}
\def\jmu{J_{\mu}}
\def\jnu{J_{\nu}}
\def\pomega{\varpi}
\def\sigla{\sigma_{\lambda}}
\def\sigmu{\sigma_{\mu}}
\def\signu{\sigma_{\nu}}
\def\dotsfill{\leaders\hbox to 1em{\hss.\hss}\hfill}
\def\min{{\sl min}}
\def\zmax{{\sl zmax}}
\def\xmax{{\sl xmax}}
\def\sun{\odot}
\def\earth{\oplus}
\def\jy{{\rm\,Jy}}
\def\mJy{{\rm\,mJy}}
\def\GHz{{\rm\,GHz}}
\def\mum{{\rm \mu m}}
\def\com{center of mass}
\def\IGI{IGI-I}
\def\IGItwo{IGI-II}
\def\Gyr{{\rm\,Gyr}}
\def\kmsd{{\rm\,km/s/degree}}
\def\msunpc{{\rm\,M_\odot/pc^3}}
\def\Ger#1{\noindent{\bf[$\diamondsuit$ #1]}}
\def\Rod#1{\noindent{\bf[$\heartsuit$ #1]}}
\def\Ed#1{\noindent{\bf[$\clubsuit$ #1]}}
\def\Mike#1{\noindent{\bf[$\spadesuit$ #1]}}
\slugcomment{Submitted to the Astronomical Journal}
\newcommand{\ffffff}[1]{\mbox{$#1$}}
\newcommand{\scnd}{\mbox{\ffffff{''}\hskip-0.3em .}}
\newcommand{\scmd}{\mbox{\ffffff{''}}}
\newcommand{\minu}{\mbox{\ffffff{'}\hskip-0.3em .}}
\newcommand{\apm}{APM~08279+5255}
\newcommand{\ly}{damped Ly$_\alpha$}
\newcommand{\lya}{Ly$_\alpha$}
\newcommand{\mg}{\ion{Mg}{2}}
\newcommand{\four}{\rm 450\mu m}
\newcommand{\eight}{\rm 850\mu m}
\newcommand{\thirteen}{\rm 1350\mu m}
\newcommand{\K}{${\rm ^oK}$}
\lefthead{Ibata et al.}
\righthead{\apm}
\begin{document}
\title{NICMOS and VLA Observations of the Gravitatonally Lensed
Ultraluminous BAL Quasar \apm:\\
Detection of a Third Image}
\author{
Rodrigo A. Ibata\altaffilmark{1},
Geraint F. Lewis\altaffilmark{2},
Michael J. Irwin\altaffilmark{3},
Joseph Leh\'{a}r\altaffilmark{4},
Edward J. Totten\altaffilmark{5}}
\altaffiltext{1}
{European Southern Observatory, Garching bei M\"unchen, Germany \\
Electronic mail: {\tt ribata@eso.org}}
\altaffiltext{2}{
Fellow of the Pacific Institute for Mathematical Sciences 1998-1999, \\
Dept. of Physics and Astronomy, University of Victoria, Victoria, B.C., Canada
\& \\
Astronomy Dept., University of Washington, Seattle WA, U.S.A.
\\
Electronic mail: {\tt gfl@uvastro.phys.uvic.ca} \\
Electronic mail: {\tt gfl@astro.washington.edu}}
\altaffiltext{3}
{Institute of Astronomy, Madingley Road, Cambridge, CB3 0HA, U.K. \\
Electronic mail: {\tt mike@ast.cam.ac.uk}}
\altaffiltext{4}
{Havard-Smithsonian Center for Astrophysics, Cambridge, MA, U.S.A. \\
Electronic mail: {\tt jlehar@cfa.harvard.edu}}
\altaffiltext{5}
{Department of Physics, Keele University, Keele,
Staffordshire, UK \\
Electronic mail: {\tt ejt@astro.keele.ac.uk}}
\begin{abstract}
We present a suite of observations of the recently identified
ultraluminous BAL quasar \apm, taken both in the infra-red with the
NICMOS high resolution camera on board the Hubble Space Telescope, and
at 3.5cm with the Very Large Array. With an inferred luminosity of
${\rm \sim5\times10^{15}\lsun}$, \apm\ is apparently the most luminous
system known. Extant ground-based images show that \apm\ is not
point-like, but is instead separated into two components, indicative
of gravitational lensing. The much higher resolution images presented
here also reveal two point sources, A and B, of almost equal
brightness ($f_{\rm B}/f_{\rm A}=0.782 \pm 0.010$), separated by
$0\scnd378 \pm 0\scnd001$, as well as a third, previously unknown,
fainter image, C, seen between the brighter images. While the nature
of C is not fully determined, several lines of evidence point to it
being a third gravitationally lensed image of the quasar, rather than
being the lensing galaxy. Simple models which recover the relative
image configuration and brightnesses are presented. While proving to
be substantially amplified, \apm\ possesses an intrinsic bolometric
luminosity of $\sim10^{14}\rightarrow10^{15}\lsun$ and remains amongst
the most luminous objects known.
\end{abstract}
\keywords{gravitational lensing -- infrared: galaxies -- quasars:
individual(\apm)}
\section{Introduction}\label{introduction}
Discovered serendipitously in a survey of carbon stars within the
Galactic halo, the bright ${\rm (m_r = 15.2)}$ z=3.87 broad absorption
line quasar \apm\ was found to be positionally coincident with a
source in the IRAS Faint Source Catalog (\cite{ir98}). The bolometric
luminosity, inferred from these optical and far-IR fluxes, exceeds
${\rm 5\times10^{15}\lsun ~(\Omega_o=1, H_o=50~km~s^{-1}~Mpc^{-1},
{\rm assumed throughout})}$, making \apm\ the most luminous object
currently known. Submillimeter photometry is consistent with thermal
emission from a massive ${\rm (\sim 3\times10^8\msun)}$, warm dust
component (\cite{le98}), with \apm\ displaying an overall spectral
energy distribution which is similar to other ultraluminous systems at
high redshift (e.g. H1413+117~\cite{ba95}). Recent observations with
the 30m IRAM telescope have also detected a massive quantity of
molecular gas; such an environment is ripe for star formation, and
this process may indeed be responsible for a substantial fraction of
\apm's phenomenal luminosity (\cite{do98}). From their CO
observations, Downes et al.\ (1998) determined a redshift of 3.911 for
\apm. This differs by $\sim2500\kms$ from that determined from the
high ionization broad emission lines (\cite{ir98}). Since it is
essentially impossible to measure accurately the systemic redshift of
a complex BAL QSO like \apm\ from optical spectra, we have chosen to
adopt the value of z=3.911 throughout this work.
Gravitational lensing can distort our view of the distant universe,
enhancing the flux of galaxies and AGN, giving them the appearance of
extraordinary systems. This has been the case for several
ultraluminous systems discovered in recent years [e.g. H1413+117
(\cite{ma88,kn98}), IRAS FSC 10214+4724 (\cite{ro91,br95,ei96}) and
the proto-galaxy candidate MS1512-cB58 (\cite{ye96,wi96,se98})]. Given
the extreme inferred luminosity of \apm, the possibility that
gravitational lensing is influencing the observed properties must also
be considered.
Analysis of the point spread function (PSF) derived from the first
ground-based observations, obtained with the Jacobus Kapteyn Telescope
(JKT), revealed that \apm\ possesses a non-stellar profile. Rather,
its profile was found to be better represented by a pair of point-like
sources, separated by $\sim0\scnd4$ (\cite{ir98}). This was
interpreted as indicative of multiple imaging of \apm\ by a massive
system along the line-of-sight, probably one or both of the strong
\ion{Mg}{2} absorbers seen at z=1.18 and z=1.81. The two components
were found to possess very similar brightnesses, within 10\%; with
this, the magnification of the optical continuum was estimated to be
$\sim40$. Follow-up observations with the adaptive optics bonnette at
the Canada-France-Hawaii telescope confirmed the existence of multiple
components in \apm, with a separation of $0\scnd35\pm0\scnd02$
(\cite{ld98}). The relative brightness of the two images was found to
be $1.21\pm0.25$, consistent with the JKT observations. The
submillimeter-infrared flux, which arises in a larger emission region,
is subject to less enhancement [for IRAS FSC 10214+4724 the optical
flux is thought to be magnified a factor of $2\rightarrow3$ times that
of the dominant infrared emission (\cite{ei96})].
While the above observations revealed that \apm\ is a good candidate
for gravitational lensing, the scale of the PSF in the images is
comparable to the image separation, and no detection of the lensing
galaxy was made. The uncertainty of the relative image/lens
configuration leads to uncertainty in the lens model and hence the
inferred lensing magnification, without which the intrinsic properties
of \apm\ cannot be determined. In an effort to confirm the lensing
hypothesis and identify the lensing lensing galaxy, \apm\ was observed
with NICMOS on the Hubble Space Telescope and with the VLA, and the
results of these observations are presented here. In
Section~\ref{observations} the details of the observations are
presented, while a lens model for \apm\ is presented in
Section~\ref{lens}. The conclusions of this study are presented in
Section~\ref{conclusions}.
\section{Observations}\label{observations}
\subsection{NICMOS}
On the 11th of October 1998, we observed \apm\ with the
NICMOS\footnote{ Based on observations with the NASA/ESA Hubble Space
Telescope, obtained at the Space Telescope Science Institute, which is
operated by the Association of Universities for Research in Astronomy,
Inc. under NASA contract No. NAS5-26555.} infra-red camera on board
the Hubble Space Telescope (HST). The highest spatial resolution
NICMOS camera NIC1 was used to obtain 12 exposures of 14~sec each in
the J-band (F110W), and 12 exposures of 40~sec each in the H-band
(F160W). A further 12 exposures of 10~sec each were secured in the
K-band (F205W) with the NIC2 camera (as a K-band filter is not
available on NIC1). Each set of exposures were taken with 1'' offset
pointings in a spiral dither pattern. The main difference, for the
present purposes, between the NIC1 and NIC2 cameras is the pixel
scale: the pixel scale is $0\scnd043$ on NIC1 and $0\scnd075$ on NIC2.
The orientation of the instrument was such that the position angle of
the CCD y-axis on all exposures with the NIC1 camera was $66.84^\circ$
E of N, and all exposures with the NIC2 camera have the CCD y-axis
$66.03^\circ$ E of N.
The raw frames were pre-reduced with the package CALNICA, using the
appropriate calibration files from the STScI archive. This processing
includes bias subtraction, flat fielding, dark correction, and a
photometric calibration for converting detector counts to physical
flux units.
A visual inspection of the preprocessed frames (see the top row of
Figure~\ref{fig1}) revealed two bright images separated by
$0\scnd35$--$0\scnd4$, as previously inferred from ground-based
images. The brightest image, which we hereafter refer to as component
``A'', is located towards the NE; the slightly fainter component ``B''
is on the SW end of the system. Clearly visible on the preprocessed
F110W and F160W frames is a third, previously undetected, image. This
component (which we shall label ``C''), is substantially fainter than
components A and B, and is located between those two bright images.
The gravitational lensing models which are compared to the data in
\S\ref{lens} below require an accurate description of the observations
in terms of relative fluxes and positions of the components of this
system. To this end we have analyzed the data in two alternate ways,
measuring magnitudes and positions on individual preprocessed frames,
and on a combined image stack in each passband.
Since the three components are at best $\sim 3$ pixels separated from
each other, it is necessary to perform a careful point-spread function
fitting analysis, a requirement of which is the determination of a
good PSF model for each camera and passband. The NICMOS PSF depends on
many factors. The factors that arise from spacecraft settings
include: the chosen camera and filter, the focus setting, and the
position of the source on the detector. The PSF is wavelength
dependent, so the source spectrum also affects its PSF. It was for
this last reason that we adopted PSFs simulated with the TINYTIM
algorithm (which models the effects of all of the above constraints),
instead of picking stars from the STScI archive of NICMOS observations
to construct a PSF. We assumed that \apm\ has a flat spectrum in $\nu
F_\nu$ over the region from $0.8 \mu m \rightarrow 2.35 \mu m$, as
suggested by the spectral energy distribution displayed in Lewis
\etal\ (1998), their Figure~2. This choice is supported by the
photometric results of the present study, listed in Table~1 below.
In the first pass of data-reductions, we constructed a TINYTIM PSF
appropriate for the camera, filter, focus position and assumed source
spectrum, and used the approximate location of component A in each
frame as the PSF position. This PSF was then fit to the data on each
data frame using the ALLSTAR PSF fitting program (Stetson 1987) to
measure the magnitudes and positions of the three components. These
position measurements are accurate to typically better than
$0.1$~pixels (judging from the RMS scatter in
$|\vec{x_A}-\vec{x_B}|$). This information enables us to use TINYTIM
to construct better PSFs, taking into account the actual sub-pixel
location of each component on each frame. Refined magnitudes and
positions were subsequently obtained by re-running the ALLSTAR
program; the mean and RMS values of these measurements are listed in
Table~1.
A combined stacked frame was constructed, using the positions of
components A and B on each frame to define the frame registration.
The individual preprocessed data frames were resampled onto a finer
grid at a scale of $0\scnd025/{\rm pixel}$, using the STSDAS
``DRIZZLE'' algorithm, and then medianed to give a combined frame in
each of the F110W, F160W and F205W passbands; these high resolution
frames are reproduced in Figure~\ref{fig1}, and are shown as a
color-composite map in Figure~2.
In the middle-row panels of Figure~\ref{fig1} we have chosen image
brightness cuts that emphasize a peculiarity of our dataset: the first Airy
ring is not uniform, but instead appears brighter on the higher row number
side of the object centers. This feature of the PSF is not predicted by the
TINYTIM software\footnote{The anonymous referee brought to our attention
that other NICMOS data (e.g. the F160W NIC2 exposures of RXJ0911.4+0551)
also display a non-uniform Airy ring.}.
%
Note that it cannot be due to some structure of
the source, as it is coincident with the Airy ring in each
passband. The deviation away from the model PSF is significant --- a
normal TINYTIM PSF model that fits the core of component A in the
F110W filter underestimates the flux in the first Airy ring by
approximately 10\% of the total flux (i.e. the first Airy ring is
approximately 50\% brighter than expected). We implemented a simple
fix of this problem by altering the model PSFs: we applied a linear
ramp to the first Airy ring, while leaving the central region of the
PSFs ($0\scnd1$~pixels in F110W, $0\scnd15$~pixels in F160W, and
$0\scnd2$~pixels in F205W) unaltered. The form of the linear ramp was
chosen to be $PSF'(x,y) = S (y-y_c) PSF(x,y)$, where $S$ is the slope
of the ramp, and $y_c$ is the PSF center. As can be seen from the
middle-row panels Figure~\ref{fig1}, the region above (towards high
row number) component A is free of contamination from other
components, as is the region below component B; it was to the data in
these two regions that we fit the multiplicative slope $S$, which gave
$S=200\%/{\rm arcsec}$ in F110W, $S=120\%/{\rm arcsec}$ in F110W, and
$S=66\%/{\rm arcsec}$ in F205W.
\begin{deluxetable}{lcccc}
\footnotesize
\tablenum{1}
%
\tablecaption{Photometry and relative positions of the components of
\apm.}\label{table1}
%
\tablewidth{0pt}
\tablehead{
\colhead{quantity} & \colhead{Image} & \colhead{F110W} &
\colhead{F160W} & \colhead{F205W}}
\startdata
Magnitude &A&$13.450 (13.450\pm 0.017)$&$13.091 (13.105\pm 0.011)$&$12.317 (12.240\pm 0.027)$\\
&B&$13.738 (13.755\pm 0.016)$&$13.357 (13.372\pm 0.011)$&$12.603 (12.509\pm 0.011)$\\
&C&$15.368 (15.240\pm 0.026)$&$14.920 (14.743\pm 0.024)$&$14.255 (14.028\pm 0.125)$\\
Color (${\rm M_{F110W}-M}$) &A&$ 0. $&$ 0.359 (0.345\pm 0.020)$&$ 1.133 (1.210\pm 0.032)$\\
&B&$ 0. $&$ 0.381 (0.383\pm 0.019)$&$ 1.135 (1.246\pm 0.019)$\\
&C&$ 0. $&$ 0.448 (0.497\pm 0.035)$&$ 1.113 (1.212\pm 0.128)$\\
Flux (mJy) &A&$ 2.577 (2.579 \pm 0.040)$&$ 3.825 (3.775 \pm 0.033)$&$ 4.035 (4.334 \pm 0.108)$\\
&B&$ 1.977 (1.947 \pm 0.026)$&$ 2.994 (2.952 \pm 0.029)$&$ 3.101 (3.381 \pm 0.032)$\\
&C&$ 0.441 (0.496 \pm 0.012)$&$ 0.710 (0.836 \pm 0.018)$&$ 0.677 (0.840 \pm 0.101)$\\
$\nu F_\nu (10^{-15} {\rm W/m^2})$&A&$ 6.847 (6.850 \pm 0.106)$&$ 7.140 (7.046 \pm 0.062)$&$ 5.844 (6.277 \pm 0.156)$\\
&B&$ 5.252 (5.172 \pm 0.069)$&$ 5.589 (5.511 \pm 0.053)$&$ 4.491 (4.897 \pm 0.047)$\\
&C&$ 1.170 (1.317 \pm 0.031)$&$ 1.325 (1.560 \pm 0.034)$&$ 0.981 (1.217 \pm 0.146)$\\
Relative Flux $F/F_A$ &A&$ 1. $&$ 1. $&$ 1. $\\
&B&$ 0.767 (0.755 \pm 0.015)$&$ 0.783 (0.782 \pm 0.010)$&$ 0.768 (0.780 \pm 0.021)$\\
&C&$ 0.171 (0.192 \pm 0.005)$&$ 0.186 (0.221 \pm 0.005)$&$ 0.168 (0.194 \pm 0.024)$\\
\tableline \\
$x_A-x_B$ &B&$ 0\scnd220 (0\scnd220\pm 0\scnd001)$&$ 0\scnd220 (0\scnd220\pm 0\scnd001)$&$ 0\scnd214 (0\scnd214\pm 0\scnd001)$\\
$y_A-y_B$ &B&$ 0\scnd307 (0\scnd308\pm 0\scnd001)$&$ 0\scnd309 (0\scnd309\pm 0\scnd001)$&$ 0\scnd308 (0\scnd309\pm 0\scnd002)$\\
$((x_A-x_B)^2+(y_A-y_B)^2)^{1/2}$
&B&$ 0\scnd378 (0\scnd378\pm 0\scnd001)$&$ 0\scnd379 (0\scnd380\pm 0\scnd001)$&$ 0\scnd375 (0\scnd375\pm 0\scnd001)$\\
$x_A-x_C$ &C&$ 0\scnd060 (0\scnd065\pm 0\scnd002)$&$ 0\scnd055 (0\scnd056\pm 0\scnd002)$&$ 0\scnd066 (0\scnd055\pm 0\scnd004)$\\
$y_A-y_C$ &C&$ 0\scnd139 (0\scnd142\pm 0\scnd002)$&$ 0\scnd130 (0\scnd132\pm 0\scnd001)$&$ 0\scnd141 (0\scnd129\pm 0\scnd008)$\\
$((x_A-x_C)^2+(y_A-y_C)^2)^{1/2}$
&C&$ 0\scnd152 (0\scnd156\pm 0\scnd002)$&$ 0\scnd141 (0\scnd143\pm 0\scnd001)$&$ 0\scnd156 (0\scnd140\pm 0\scnd007)$\\
%$X_B$ &B&$ 0\scnd378 (0\scnd378\pm 0\scnd001)$&$ 0\scnd379, (0\scnd380\pm 0\scnd001)$&$ 0\scnd375, (0\scnd375\pm 0\scnd002)$\\
%$Y_B$ &B&$ 0\scnd000 (0\scnd000\pm 0\scnd001)$&$ 0\scnd000, (0\scnd000\pm 0\scnd001)$&$ 0\scnd005, (0\scnd006\pm 0\scnd001)$\\
%$X_C$ &C&$ 0\scnd148 (0\scnd153\pm 0\scnd002)$&$ 0\scnd137, (0\scnd140\pm 0\scnd001)$&$ 0\scnd153, (0\scnd136\pm 0\scnd007)$\\
%$Y_C$ &C&$ 0\scnd032 (0\scnd030\pm 0\scnd002)$&$ 0\scnd031, (0\scnd031\pm 0\scnd002)$&$ 0\scnd029, (0\scnd030\pm 0\scnd007)$\\
$\alpha_A - \alpha_B$ &B&$ 0\scnd196 (0\scnd197\pm 0\scnd001)$&$ 0\scnd198, (0\scnd198\pm 0\scnd001)$&$ 0\scnd199, (0\scnd200\pm 0\scnd002)$\\
$\delta_A - \delta_B$ &B&$ 0\scnd323 (0\scnd323\pm 0\scnd001)$&$ 0\scnd324, (0\scnd324\pm 0\scnd001)$&$ 0\scnd318, (0\scnd318\pm 0\scnd001)$\\
$\alpha_A - \alpha_C$ &C&$ 0\scnd104 (0\scnd105\pm 0\scnd002)$&$ 0\scnd098, (0\scnd099\pm 0\scnd001)$&$ 0\scnd104, (0\scnd097\pm 0\scnd008)$\\
$\delta_A - \delta_C$ &C&$ 0\scnd110 (0\scnd116\pm 0\scnd002)$&$ 0\scnd102, (0\scnd103\pm 0\scnd002)$&$ 0\scnd116, (0\scnd101\pm 0\scnd005)$\\
\enddata
%
\tablenotetext{}{The entries not given in brackets are derived from
the median-combined image stacks, while values in brackets are the
mean and RMS scatter of measurements on individual frames. The
relative positions are listed in the system of the coordinates
$(x,y)$, which are aligned, respectively, with the CCD row and column
directions. Also listed are the positions in the equatorial
coordinates relative to object A.}
\end{deluxetable}
Using these PSF models with an improved estimate of the brightness
distribution around the first Airy ring, we obtained ALLSTAR
measurements of the magnitudes and positions of the three components
A, B and C on the median-combined image stacks in each color. These
measurements are listed in Table~1.
We now summarize the observational results.
\begin{itemize}
\item
The reduced chi-squared value of the PSF fits to the three components
in the median-combined stacked frame is $\chi^2 < 2$, indicating that
to good approximation, the three components are point sources. The
residuals of these PSF fits are displayed in the bottom-row panels of
Figure~\ref{fig1}.
\item
The colors of components A and B are identical to within the
uncertainties. If we accept that the measurements from the stacked
frames are more reliable than the mean value of the measurements from
individual frames, then the colors of component C are also consistent
with those of A and B. The simplest hypothesis is therefore that the
components A, B and C have identical colors.
\item
Given that the colors of the components are identical, we can average
their relative brightnesses over the three passbands. We find then
that the relative brightness of components A and B is $0.773\pm 0.007$
($0.772\pm 0.012$), where the value in brackets is the mean of the
measurements on unstacked frames, and the uncertainties are the RMS
scatter in the three measurements. The relative brightness of
components A and C is $0.175\pm 0.008$, ($0.202\pm 0.013$). Thus, we
find good consistency between the measurement methods, which indicates
that the relative brightnesses of components A, B and C are well
constrained by the data.
\item
The fluxes of the components in $\nu F_\nu$ are approximately constant
as a function of wavelength (slightly brighter in H).
\item
Averaging over the position measurements in each passband, the
distance from A to B is $0.377\pm 0.002$ ($0.379\pm 0.001$), and the
distance from A to C is $0.150\pm 0.006$ ($0.146\pm 0.007$). The
location of component C is not on the straight line connecting A and
B. This can be seen by measuring the position of component C in a new
coordinate system $(X,Y)$, where $X$ points from A to B, and $Y$ is
orthogonal to $X$ (and points in a north-westerly direction). In
these coordinates, $X_C=0.146\pm 0.007$ ($X_C=0.143\pm 0.007$) and
$Y_C=0.031\pm 0.001$ ($Y_C=0.030\pm 0.0005$). While these RMS
uncertainties in the $Y_C$ positions likely underestimate the true
uncertainty, it is clear that to a high confidence level, the three
components of \apm\ are not co-linear.
\end{itemize}
\subsection{VLA}
We obtained 3.6~cm radio observations of APM~08279+5255, using the
hybrid BnA~configuration of the NRAO\footnote{The National Radio
Astronomy Observatory is a Facility of the National Science
Foundation, operated under cooperative agreement by Associated
Universities, Inc.}
%
Very Large Array (VLA), on 18$^{\rm th}$ June, 1998. We integrated on
the target for about 2.5~hours, divided into half-hour scans,
interleaved with a nearby phase calibrator (B0820+560). The
interferometer data were calibrated and mapped using standard
AIPS\footnote{ AIPS (Astronomical Image Processing System) is
distributed by NRAO.}
%
procedures, and the flux densities were calibrated to 3C\,147 (Baars
et al. 1977). The target proved to be very faint in the radio, and
no significant polarized emission was detected. Thus only a single
iteration of CLEAN mapping was performed, without self-calibration.
The resulting radio map (see Figure~3) shows a very faint source, with
a peak flux density of only 0.26\,mJy/beam, where the off-source map
RMS was 0.013\,mJy/beam, close to the expected thermal noise. The
peak of the radio emission is at $\alpha$=08:27:58.00,
$\delta$=+52:55:26.9 (B1950), displaced by $\sim0\farcs6$ from the HST
A-image position, probably due to a combination of HST and VLA
astrometric errors. The total VLA flux of the source is
$0.45\pm0.03$\,mJy, integrated over a $\sim1\arcsec$ square aperture.
There is also a marginally detected source (peak SNR$\sim5$), roughly
$3\arcsec$ to the southwest, with a total flux density of
$0.09\pm0.03$\,mJy. To best display the radio structure, given the
elliptical natural beam, we used a $0\farcs4$ FWHM circular restoring
beam.
The radio source is clearly resolved along the A--B image axis. We
fitted two Gaussian components to a $1\arcsec$ square region enclosing
the radio source, with a relative offset fixed to the HST A--B offset,
and widths matched to the $0\farcs4$ convolving beam. The two
components had an A/B flux density ratio of 0.6, consistent with the
HST image brightness ratio of 0.7, but with a noticeable residual
between them. Adding a third component at the C-image offset
decreased the fit $\chi^2$ by $\sim30\%$, and yielded B/A and C/A flux
ratios of 0.9 and 0.5 respectively. However, we could not
definitively confirm the C-image: the fit $\chi^2$ remained less than
twice the best value provided that the C image contained less than
twice the A image flux density.
\section{Gravitational Lensing}\label{lens}
\subsection{Image Configuration}
Using ground-based data, both Irwin et al. (1998) and Ledoux et al.
(1998) determined that \apm\ comprised a pair of point-like images
separated by $\sim0\scnd4$. With the limited information available,
namely the image separation and relative brightnesses, both modeled
the lensing configuration as a singular isothermal sphere and this
suggested that the quasar source in \apm\ is magnified by a factor of
20--40. Taking this into account, intrinsically \apm\ still ranks
amongst the brightest systems known.
The data presented here greatly enhances our view of \apm, clearly
resolving the system into a pair of bright point-like images either
side of a fainter third image. But before the degree of gravitational
lensing in this system can be fully explored, the nature of the third
image must be investigated. Two possibilities present themselves;
either it is the lensing galaxy which is responsible for splitting the
bright image pair, or it represents a true `third' image of the quasar
source.
Several lines of argument point towards the latter possibility. First,
morphologically image C is point-like and it possesses identical colors to
the brighter components. The spectra of \apm\ with both the 2.5m Isaac
Newton Telescope (\cite{ir98}) and the 10m Keck I Telescope
(\cite{sara99a,sara99}) reveal the presence for two \ion{Mg}{2} absorption
systems at z=1.18 and z=1.81. If we adopt the lower of these redshifts as
the potential redshift of the lensing galaxy, then its distance modulus is
$44.6$ ($H_0=50 {\rm km s^{-1} Mpc^{-1}}$, $q_0 = 0.5$). Thus if image C is
the lens, its absolute magnitude would be have to be prodigious, ${\rm M_J
\sim -32}$ (note that the K-correction is approximately zero, as the SED of
component C is flat in $\nu f_\nu$). This is inconsistent with the expected
Faber-Jackson luminosity of the lens given its small $\sim 130 \kms$
velocity dispersion (Irwin \etal\ 1998), predicted by a simple isothermal
sphere lensing model. Finally, examination of the Keck I HIRES spectra
(\cite{sara99a,sara99}), which have a resolution of $\sim6\kms$
(0.04\AA/pixel), reveals a number of absorption systems both along the
line-of-sight to the quasar, the \lya\ forest, and associated with the
quasar source, the broad absorption lines. In large spectral intervals,
through regions of significant optical depth, \lya\ absorption is saturated,
and the spectrum is effectively black with a signal that is consistent with
zero. Since flux from the (foreground) lensing galaxy must fill in these
troughs, their darkness can be used to place an upper limit to the
brightness of the lens. Indeed, between 5500\AA\ to 5900\AA, the darkest
200~pixel-wide region is centered at 5769.2\AA, and the $3\sigma$ upper
limit to the mean flux in this 200~pixel region is 600 times fainter than
the mean flux over the interval 5500\AA\ to 5900\AA\ (we have adopted the
\cite{sara99a} error spectrum as a reasonable estimate of all sources of
noise in the spectrum, including the uncertainty arising from scattered
light in the HIRES spectrograph). This implies that the -- as yet
undetected -- lensing galaxy must be at least $\sim 7$~magnitudes fainter
than the quasar, that is, ${\rm V \simgt 22}$. Thus, the hypothesis that
image C is the lensing galaxy leads to an unrealistically red color: ${\rm
V-K \simgt 8}$. These facts provide compelling evidence that C is a third
image of the high redshift quasar source. This, however, can not be
conclusively demonstrated without further observations of \apm, and in the
following sections, which deal with gravitational lens modeling, both the
third image and lensing galaxy possibilities as a source for C will be
considered.
\subsection{Modeling}
Given \apm's apparent position as the most luminous object currently
known, it is important to determine how much the action of
gravitational lensing is enhancing our view of this distant source.
The excellent resolution of the NICMOS images reveals a third
point-like source between the brighter quasar image, and while
available evidence points to this being a third image of the quasar
source, the possibility that this emission arises in the lensing
galaxy is not ruled out. In terms of modeling the gravitational
lensing in \apm\ both possibilities are considered.
\subsubsection{C as a Third Quasar Image}\label{gfl_model}
If image C represents a third image of the quasar source, then extant
observations have failed to detect the lensing galaxy and hence its
characteristics and position relative to the quasar images remains
unknown to us. However, due to the fact that the image positions are
not co-linear, we know that this lens must possess some degree of
asymmetry, either due to some intrinsic ellipticity or shearing from a
nearby companion. Similarly, to produce the relatively bright third
image, a finite core radius is required~(\cite{wa93}). We would wish,
therefore, to employ a finite core, and an elliptical mass
distribution in modeling the gravitational lensing in this system.
Any such model, however, requires a minimum of 7 parameters (two
source positions, an ellipticity, core radius, mass orientation, slope
and normalization), while the data only offer 5 constraints (three
relative positions and two relative image brightnesses). To this end,
we constructed a simple model with a general search of parameter
space.
To reduce parameter space, the mass distribution is taken as being
isothermal at large radii, turning over in the inner regions to give a core.
Sampling the image characteristics for a range of source positions,
ellipticities and core radii, resulted in the model presented in
Figure~\ref{fig4}. The left-hand panel presents the position of the source
relative to the caustic distribution in the source plane. The right-hand
panel presents the image plane with the corresponding critical lines. The
small circles represent the resultant positions of the images; these are
superimposed on the NICMOS image. Both the relative image positions and
magnitudes match those outlined in Table~\ref{table1}. The lensing mass is
oriented 76$^{\rm o}$ east of north, with a small deviation from circular
symmetry $(\epsilon\sim0.01)$. The core radius of the lens is $\sim
0\scnd21$, and assuming that this galaxy is at a redshift of z=1.18, its
mass interior to the Einstein ring is ${\rm \sim 2\times10^{10}\msun}$.
While this recovers the observed image plane characteristics it is, of
course, highly degenerate, but it does demonstrate that \apm\ is consistent
with being a gravitational lensed system.
We do note, however, that more complex lens models are possible and
\apm\ may represent lensing by a `naked cusp' (\cite{wa93}).
Similarly, the location of at least one other lensing system along the
line-of-sight to \apm\ can lead to a more complex caustic network in
the source plane. With either of these, the observed image
configuration can be reproduced with quite different degrees of
magnification. The true nature of the gravitational lensing
configuration in this system will remain uncertain until the lensing
galaxy has been detected.
\subsubsection{C as the Lensing Galaxy}\label{joes_model}
The observed three image configuration is consistent with a
fundamental theorem of gravitational lensing which dictates that an
odd number of images are {\it always} formed by non-singular mass
distributions (\cite{bu81}). It is, however, at odds with
observations of most gravitationally lensed systems, as these are seen
to possess an even number of images (c.f. {\tt
http://cfa-www.harvard.edu/castles/}); this seemingly incompatibility
between gravitational lens theory and observations is usually solved
by invoking a (near-)singular core in the lensing galaxy which can
drive the magnification of one of the images to zero (\cite{na86}).
The inverse of this suggests that the lensing mass distribution must
contain a finite core radius if the image at C is to be appreciably
magnified; this is exactly the situation with the previous model
(Section~\ref{gfl_model}).
Unless this is the first true bona-fide three image lensed system,
however, we must also consider the case that C represents the lensing
galaxy and the `true' third image has been demagnified by an almost
singular core. With this, the lensing potential of C was modeled as
an elliptical, isothermal potential (c.f.~\cite{ko89}). The lensing
configuration offers 5 constraints: two positions, relative to the
lens, for each image, and the relative image brightnesses. The model
has 3 free parameters, the ellipticity, orientation and normalization
of the mass distribution. Using C as the position of the lens is not
used as a constraint, rather it is also considered another pair of
free parameters. With 5 free parameters and 5 constraints a search for
an exact model fit can be made.
The resulting minimized model lensing configuration for this model is
presented in Figure~\ref{fig5}; here the grey represents the position
of the source relative to the lensing caustics, while the observed
images and critical lines are solid. The best position for the lens
is $0\scnd0989$ West and $0\scnd1026$ South of image A, consistent
with the position of image C (Table~\ref{table1}). The mass has a 1-D
velocity dispersion of 126$\kms$ and is oriented at 104$^\circ$ east
of north, an offset of $28^\circ$ from the model presented in
Section~\ref{gfl_model}, although the ellipticity of the mass profile
is substantial, with $\epsilon=0.2$; the differences in the models are
understandable given the different interpretations of the nature of C.
The source has an impact parameter, relative to the lens, of
$0\scnd01$ (SSW). Again, the location of the source is similar to
that seen in Section~\ref{gfl_model}. This analysis illustrates, from
a gravitational lensing point of view, that C could represent the
lensing galaxy.
But what of the nature of C? As discussed previously, if it is the
lensing system then its brightness indicates that it is not a `normal'
galactic system. One possibility is that C is itself a luminous
quasar, explaining both its apparent brightness and point-like
appearance. Such quasar-quasar lensing has been addressed
(\cite{wa97}), although only high resolution imaging and spectroscopy
will uncover C's true nature.
\subsection{The Intrinsic Luminosity of \apm}\label{lumo}
Using the models described above, the degree to which the quasar source is
magnified can be determined. Considering the first model of
Section~\ref{gfl_model}, which assumes image C is a third image of the
quasar, the total magnification of the quasar source is $\sim90$.
Similarly, the model described in Section~\ref{joes_model}, in which image C
represents the lensing galaxy responsible for splitting the quasar light
into the A and B images, the corresponding magnification of the quasar
continuum source is $\sim7.5$. We expect the magnification of the far-IR
continuum source to be a factor of 2--3 times lower than the magnification
of the quasar continuum source, due to the larger size of the emission
region. Taking these into account, the intrinsic luminosity of \apm\ is
$\sim10^{14}\rightarrow{10^{15}\lsun}$ and it retains its place amongst the
most luminous systems currently known.
\section{Conclusions}\label{conclusions}
We have presented new observations of the ultraluminous BAL quasar
\apm, using both NICMOS on the Hubble Space Telescope and the VLA at
3.5cm. These clearly demonstrate the composite nature of the system,
separating \apm\ into a pair of images, A and B, of comparable
brightness. The NICMOS images also reveal the presence of a third
point-source, C, between the brighter two. The VLA image shows
structure corresponding to the background quasar being a faint radio
source and we conclude that intrinsically \apm\ is a radio quiet
quasar with ordinary radio properties. Although the radio source is
resolved along A--B, we can not determine from these radio data
whether it comprises the A+B images or the A+B+C images.
There are two possible interpretations of the source of C: either this
central image is a detection of the foreground lensing galaxy, or it
represents a third image of the high redshift quasar. Several
arguments indicate that this latter proposition is a more likely
description of the available data. A lensing model in which the three
observed images are true images of the quasar source was presented.
This recovers both the image configuration and and relative
brightnesses with a total lensing magnification of $\sim90$.
An alternative model in which image C is the lensing galaxy, was also
explored. The position of C was found to be coincident with the
position of the lensing galaxy predicted from a simple lensing model,
indicating that this scenario may not be implausible. However, to
account for its apparent brightness, C must also be an intrinsically
luminous source. If this latter model is correct, the simple lensing
configuration it offers, with a corresponding time delay of 5~days,
provides an ideal example of a ``golden lens'' (Williams \& Schecter
1997) with which ${\rm H_o}$ can be determined from photometric
monitoring.
Considering the magnification of the above models, the intrinsic
luminosity of \apm\ is $\sim10^{14}\rightarrow10^{15}\lsun$, placing
it amongst the most luminous systems currently known, and its apparent
brightness makes it ideal to study aspects of the high redshift
Universe (\cite{hi99,sara99,sara99a}).
Recent near and mid-IR KECK images were brought to our attention just
prior to submitting this article (Egami et al. 1999). Taken in
excellent seeing, these also clearly reveal the presence of the third
image in the same location and with a similar brightness to that found
in the the NICMOS images. Given its apparent brightness, they also
conclude that this represents a third image of the quasar, rather than
the lensing galaxy. They also construct a very similar lensing model
to the one presented in this paper, although ours possesses a slightly
higher total magnification (90 compared to their 71). Overall, our
conclusions are in excellent agreement with theirs.
\acknowledgements We thank the anonymous referee for constructive
comments, and Dr. E. Egami for pointing out an error in the
orientation of one of our lensing models.
\newpage
\begin{thebibliography}{}
%
\bibitem[Barvainis et al.\ 1995]{ba95}
Barvainis, R., Antonucci, R., Hurt, T., Coleman, P. \& Reuter, H.-P.,
1995, \apj\, 451, 9
%
\bibitem[Broadhurst \& Leh\'{a}r 1995]{br95}
Broadhurst, T. \& Leh\'{a}r, J.,
1995, \apj\, 450, 41
%
\bibitem[Burke 1981]{bu81}
Burke, W.L.,
1981, \apj\, 244, L1
%
\bibitem[Downes et al.\ 1998]{do98}
Downes, D., Neri, R., Wikland, T., Wilner, D.J. \& Shaver, P.,
1998, \apj\, 513, L1
%
\bibitem[Egami et al.\ 1999]{eg99}
Egami, E., Neugebauer, G., Soifer, B. T., Matthews, K., Ressler, M. \&
Becklin, E. E.,
1999, \apj\, Submitted
%
\bibitem[Eisenhardt et al.\ 1996]{ei96}
Eisenhardt, P. R., Armus, L., Hogg, D. W., Soifer, B. T.,
Neugebauer, G. \& Werner, M. W.,
1996, \apj\, 461, 72
%
\bibitem[Ellison et al. 1999a]{sara99a}
Ellison, S. L., Lewis, G. F., Pettini, M., Sargent, W. L. W.,
Chaffee, F. H., Foltz, C. B., Rauch, M. \& Irwin, M. J.
1999a, \pasp, {\it In Press}
%
\bibitem[Ellison et al. 1999]{sara99}
Ellison, S. L., Lewis, G. F., Pettini, M.,
Chaffee, F. H. \& Irwin, M. J.
1999, \apj, 520, 456
%
\bibitem[Hines et al. 1999]{hi99}
Hines, D. C., Schmidt, G. D. \& Smith, P. S.,
1999, \apj, 514, L91
%
\bibitem[Irwin et al.\ 1998]{ir98}
Irwin, M. J., Ibata, R. A., Lewis, G. F. \& Totten, E. J.,
1998, \apj\, 505, 529
%
\bibitem[Kneib et al.\ 1998]{kn98}
Kneib, J.-P., Alloin, D., Mellier, Y., Guilloteau, S.,
Barvanis, R. \& Antonucci, R.,
1998, \aap\ 329, 827
%
\bibitem[Kochanek, Blandford, Lawrence \& Narayan 1989]{ko89}
Kochanek, C. S., Blandford, R. D., Lawrence, C. R. \& Narayan, R.,
1989,\mnras, 238, 43
%
\bibitem[Ledoux et al.\ 1998]{ld98}
Ledoux, C., Theodore, B., Petitjean, P., Bremer, M.N.,
Lewis, G.F., Ibata, R.A., Irwin, M.J. \& Totten, E.J.,
1998, \aap\, 339, L77
%
\bibitem[Lewis et al.\ 1998]{le98}
Lewis, G.F., Chapman, S.C., Ibata, R.A., Irwin, M.J. \& Totten, E.J.,
1998, \apj\, 505, L1
%
\bibitem[Magain et al.\ 1988]{ma88}
Magain, P., Surdej, J., Swings, J.-P., Borgeest, U. \& Kayser, R.,
1988, \nat\, 334, 325
%
\bibitem[Narasimha Subramanian \& Chitre 1986]{na86}
Narasimha, D., Subramanian, K. \& Chitre, S. M.,
1986, \nat, 321, 45
%
\bibitem[Rowan-Robinson et al.\ 1991]{ro91}
Rowan-Robinson, M., Broadhurst, T., Oliver, S.J., Taylor, A.N.,
Lawrence, A., McMahon, R.G., Lonsdale, C.J., Hacking, P.B. \&
Conrow, T., 1991, \nat\, 351, 719
%
%\bibitem[Schneider et al.\ 1992]{sc92}
% Schneider, P., Ehlers, J. \& Falco, E. E.,
% 1992, {\it Gravitational Lenses}, Springer-Verlag Press, Berlin
%
\bibitem[Seitz et al.\ 1998]{se98}
Seitz, S., Saglia, R. P., Bender, R., Hopp, U., Belloni, P. \& Ziegler, B.,
1998, \mnras\, 298, 945
%
\bibitem[Stetson 1987]{st87}
Stetson, P. B.,
1987, \pasp, 99, 191
%
\bibitem[Wallington \& Narayan 1993]{wa93}
Wallington, S. \& Narayan, R.,
1993, \apj\, 403, 517
%
\bibitem[Wampler 1997]{wa97}
Wampler, E. J.,
1997, \apjl, 476, L55
%
\bibitem[Williams \& Lewis 1996]{wi96}
Williams, L. L. R. \& Lewis, G. F.,
1996, \mnras\, 381, L35
%
\bibitem[Williams \& Schecter 1997]{wi97}
Williams, L. L. R. \& Schecter, P. L.,
1997, {\it Astronomy \& Geophysics}, 38, 10
%
\bibitem[Yee et al.\ 1996]{ye96}
Yee, H. K. C., Ellingson, E., Bechtold, J.,
Carlberg, R. G. \& Cuillandre, J.-C.,
1996, \aj\, 111, 1883
%
\end{thebibliography}
\newpage
\begin{figure*}
\centerline{
\psfig{figure=Ibata.fig1.ps,height=3.5in}
}
\caption[]{The NICMOS images of \apm. From left to right, the columns
show the oversampled combined images in the filters F110W, F160W and
F205W. The same $1\scnd125 \times 1\scnd125$ region is shown in each
panel. In the top-row panels, the brightness cuts have been chosen to
show the third image, which is clearly visible between the brighter
two images. The middle-row panels have brightness cuts to emphasize
the first Airy rings. The residuals of the PSF fitting (detailed in
the text) are shown in the bottom-row panels.}
\label{fig1}
\end{figure*}
\begin{figure*}
\centerline{
%\psfig{figure=Ibata.fig2.ps,height=3.5in}
}
\caption[]{NICMOS color-composite image of the \apm\ system. The
orientation and scale is the same as Figure~\ref{fig1}. Again, C can
be clearly seen between the brighter images. }
\label{fig2}
\end{figure*}
\begin{figure*}
\centerline{
\psfig{figure=Ibata.fig3.ps,bbllx=34pt,bblly=132pt,bburx=530pt,bbury=630pt,clip=yes,width=3.5in}
}
%
\caption[]{ VLA map of APM08279+5255 at 3.6cm, obtained using the
hybrid BnA configuration. The contours increase by factors of
$\sqrt{2}$ from twice the off-source RMS noise level of
0.013\,mJy/beam. The HST image components are marked with crosses,
and all position offsets are relative to A, at
$\alpha$=08:31:41.64,$\delta$=+52:45:17.5\,. The radio source is
offset by $\sim0.6\arcsec$, probably due to VLA and HST astrometric
errors. The natural and convolving beams (FWHM) are shown at
bottom-left, respectively as open and filled ellipses. }
\label{fig3}
\end{figure*}
\begin{figure*}
\centerline{
%\psfig{figure=Ibata.fig4.ps,width=5in,angle=-90}
}
\caption[]{ The left-hand panel presents source plane with the image
position relative to the caustic structure of the lens model discussed
in Section~\ref{lens}. The right-hand panel presents the
corresponding critical lines over the image plane, coupled with the
image positions, overlaid on the NICMOS images of \apm. The
orientation of both frames is identical to that of Figures~\ref{fig1}
and~\ref{fig2}.}
\label{fig4}
\end{figure*}
\begin{figure*}
\centerline{
\psfig{figure=Ibata.fig5.ps,width=5in,angle=0}
}
\caption[]{ If C is the lensing galaxy, a simple lens model can be
constructed to account for the observed image configuration; this is
illustrated above. Here, the light grey represents the caustic
structure and source position, while the solid lines represent the
observed image configuration and critical line. The total
magnification of a point-like source with this model is only 7.5,
substantially less than the 90 expected if C is actually an image of
the quasar. In this plot, North is to the top, while East is to the
left and the coordinate system is centered on image A. The source
and images are point-like but have been smoothed with a $0\scnd03$
Gaussian to represent the NICMOS resolution.}
\label{fig5}
\end{figure*}
\end{document}