Strong Gravitational Lensing and Dark Energy Complementarity
Extracted source text ยท arXiv astro-ph/0401433
%\documentclass[prd, twocolumn, superscriptaddress, nofootinbib]{revtex4}
\documentclass[prd, twocolumn, nofootinbib]{revtex4}
%\documentclass[preprint]{aastex}
\usepackage{epsfig}
%\topmargin0.0cm
\newcommand{\beq}{\begin{equation}}
\newcommand{\eeq}{\end{equation}}
\newcommand{\beqa}{\begin{eqnarray}}
\newcommand{\eeqa}{\end{eqnarray}}
\newcommand{\om}{\Omega_m}
\begin{document}
\title{Strong Gravitational Lensing and Dark Energy Complementarity}
\author{Eric V.~Linder}
\affiliation{Physics Division, Lawrence Berkeley National Laboratory,
Berkeley, CA 94720}
\begin{abstract}
In the search for the nature of dark energy most cosmological probes
measure simple functions of the expansion rate. While powerful, these
all involve roughly the same dependence on the dark energy equation of
state parameters, with anticorrelation between its present value $w_0$
and time variation $w_a$. Quantities that have instead positive
correlation and so a sensitivity direction largely orthogonal to, e.g.,
distance probes offer the hope of achieving tight constraints
through complementarity. Such quantities are found in strong
gravitational lensing observations of image separations and time delays.
While degeneracy between cosmological parameters prevents full
complementarity, strong lensing measurements to 1\% accuracy can
improve equation of state characterization by 15-50\%. Next generation
surveys should provide data on roughly $10^5$ lens systems, though
systematic errors will remain challenging.
\end{abstract}
\maketitle
\section{Introduction} \label{sec.intro}
Dark energy poses a fundamental challenge to our understanding of the
universe. The acceleration of the cosmic expansion discovered through
the Type Ia supernovae distance-redshift relation can be interpreted
in terms of a new component of the energy density possessing a
substantially negative pressure. Further observations from the cosmic
microwave background radiation indicate the spatial geometry of
the universe is flat; in combination with the supernova measurements
this implies that the unknown ``dark energy'' comprises roughly 70\% of
the total density, in concordance with large scale structure data
indicating that matter contributes approximately 30\% of the critical density.
Such a weight of dark energy leads to its dominance of the expansion, causing
acceleration, restricting the growth of large scale structure, and holding the
key to the fate of the universe. But apart from its rough magnitude, its
nature is almost unknown -- whether it arises from the physics of the
high energy vacuum, a scalar field, extra dimensional or ``beyond Einstein''
gravitational effects, etc. One way to gain clues to the underlying
physical mechanism is to characterize the behavior of dark energy in terms
of its equation of state ratio as a function of
redshift (essentially a time parameter), $w(z)$. This is conventionally
interpreted in terms of the ratio of the dark energy pressure to energy
density, but can also be used as an effective parameterization to treat
generalizations of the cosmology framework of the Friedmann equations
of expansion \cite{linjen} or
simply in terms of the expansion rate and acceleration itself \cite{lingrav}.
To obtain the clearest focus on the class of physics responsible for the
accelerating universe, we seek the tightest constraints on the equation
of state. This must come not merely from high statistical precision of
a probe, but from robust control of systematic uncertainties. The
greatest accuracy and confidence
in the measurements will come from independent crosschecks and complementarity
between methods of probing the cosmology. Many studies have considered such
complementarity between probes (e.g.\ \cite{fhlt,linap,seo,linbo,wl,cluster}),
with a
promising future for next generation surveys carrying out such measurements.
However, all the methods of supernova distances, cosmic microwave background
power spectrum, weak gravitational lensing, cluster counts, and
baryon oscillations
possess a similar fundamental dependence on the equation of state through the
Hubble parameter, or expansion rate.
This paper investigates whether a truly complementary probe exists that has
nearly orthogonal dependence to the previous ones in the plane of the
equation of state parameters of value today, $w_0$, vs.\ time variation,
$w_a$. Such a probe, if practical, would offer a valuable contribution
to uncovering fundamental physics
and deserve further consideration among next generation experiments. Section
\ref{sec.anti} considers the characteristics such a method would possess and
identifies two promising candidates related to strong gravitational
lensing. In \S\ref{sec.fisher} we analyze the
sensitivity of these probes to the cosmological parameters and the constraints
and complementarity they offer. In the conclusion we summarize the
prospects for strong lensing as a cosmological probe and discuss some issues
regarding surveys and systematic uncertainties.
\section{Complementarity in the Equation of State Plane} \label{sec.anti}
Astronomical observations involving distances and volumes all follow from the
metric in a simple, kinematic way \cite{weinberg}. For convenience,
the dependence
can be written in terms of the conformal distance interval
\beq
d\eta=dt/a(t)=-dz/H(z),
\eeq
where $a(t)=(1+z)^{-1}$ is the scale factor as a function of proper time,
equivalently parametrized in terms of redshift $z$, and $H=\dot a/a$ is
the Hubble parameter. So in observing some source at redshift $z$ the
cosmological information is carried by the quantity $H$.
Varieties of distance observations -- angular diameter, luminosity,
proper motion -- all contain the same information. One exception is
the parallax distance (cf.\ \cite{fpoc}, \S3.2) which involves not only
$H$ but the spatial curvature $k$ as well; however this is not practical
for cosmology and CMB measurements strongly indicate a flat universe.
Another subtlety involves exotic models that break the thermodynamic, or
reciprocity, relation between the various distances
\cite{thermo,tolman,bassett} by violation of Liouville's theorem, but
this arises from photon properties and not cosmology.
Volume elements, and hence numbers of sources, are built up out of
distances and so similarly involve the Hubble parameter. From the
Friedmann equation the relation between the expansion rate and the
dark energy equation of state is
\beq
H^2/H_0^2=\om(1+z)^3+\Omega_{DE} e^{3\int d\ln(1+z)\,[1+w(z)]}, \label{eq.hw}
\eeq
where $H_0$ is the present value of the Hubble parameter (the Hubble
constant), and for a flat universe $\Omega_{DE}=1-\om$. As discussed
in \cite{linjen} this can be written more generally as defining
an effective equation of state
\beq
w(z)=-1+\frac{1}{3}\frac{d\ln\delta H^2/H_0^2}{d\ln(1+z)},
\eeq
where $\delta H^2$ encodes our ignorance of the right hand side of
eq.\ \ref{eq.hw} after the first, matter density term.
Making $w(z)$ more positive (holding $H_0$ and $\om$ fixed) increases
the expansion rate in the past, and so decreases the acceleration (since
the rate today is fixed). Distances will be less, sources appear
brighter, etc. If we try to characterize the nature of the dark energy
in the simplest, nontrivial way, through the value of its equation of
state today and its increasing or decreasing positivity into the past,
we see that cosmological measurements, depending only on $H$, don't
really care in a gross sense {\it how} $w(z)$ was more positive --
either increasing the present value $w_0$ or increasing the time
variation $w_a\sim dw/d\ln(1+z)$ does the trick.
Thus, for all such probes the quantities $w_0$, $w_a$ will be
anticorrelated; an increase in one can be (at least partially) offset
by a decrease in the other. If we plot constraints from astronomical
data in the $w_0-w_a$ plane (marginalizing over any other parameters),
the probability contours will show a degeneracy direction
tilted counterclockwise from the vertical. This is well known and
illustrated for a number of different probes in \cite{hutcoo}; note
also that a ``bare'' $H$ of course has the same dependence \cite{linbo}.
However, this has the fundamental implication that complementarity
between methods can only be partial in this physically key parameter
plane; different data sets cannot be orthogonal. This restricts the
ability to constrain the equation of state and understand the nature
of dark energy. What would be valuable is construction of a cosmological
probe whose sensitivity lies clockwise of vertical, into the unexplored
half-plane of positive correlation between $w_0$ and $w_a$. Note that
this argument does not depend on the specifics of how $w_0$, $w_a$ are
defined, but for concreteness we adopt the parametrization
\beq
w(z)=w_0+w_a(1-a)=w_0+w_a z/(1+z),
\eeq
successful in fitting smoothly varying dark energy models
\cite{linprl,lin10217}. A characteristic time variation of the equation
of state is given by $w'\equiv dw/d\ln(1+z)|_{z=1}=w_a/2$.
Figure \ref{fig.compl} illustrates this anticorrelation for the
distance-redshift probe (dashed ellipse). A hypothetical probe that
involves positive correlation between the equation of state parameters
(dotted ellipse) can give nearly orthogonal constraints and so the
joint probability contour (solid ellipse) between the two types of
probe tightly characterizes the equation of state. Furthermore, since
reduced data quality, e.g.\ from systematic uncertainties, often
lengthen the confidence contour along the degeneracy direction (long
dashed ellipse), an orthogonal probe immunizes against systematic
errors. However, degeneracies with other cosmological parameters
interfere with both these desiderata, as discussed in \S\ref{sec.fisher}.
\begin{figure}[!hbt]
\begin{center}
\psfig{file=fakecorr.ps,width=3.4in}
\caption{This cartoon illustrates the value of orthogonality in the
equation of state plane for tightly constraining dark energy
parameters. The short dashed ellipse represents a distance-redshift
probe, or any other that has an anticorrelation between the equation
of state parameters. Adding a positive correlation probe (dotted)
gives a joint contour (inner solid curve) with considerably smaller
uncertainties.
This also helps to reduce the impact of systematics (long dashed
curve and outer solid curve). However the situation is more complicated
when marginalizing over a larger set of parameters.
}
\label{fig.compl}
\end{center}
\end{figure}
The growth of large scale structure, or equivalently the evolution
of gravitational potentials, and hence, say, the number of galaxy
clusters of a certain mass, might seem to offer a different dependence
than distances. But the growth equation also has $H$ as its predominant
ingredient and indeed in \cite{hutcoo} we see that the growth factor
still possesses the anticorrelation. An exception might occur for
models where the dark energy (or alternative gravity theory) is
inhomogeneous itself and can act as a source to the matter density
perturbations, but this is not expected to occur for scalar field
dark energy on subhorizon scales \cite{caldwell,dave}.
If simple measures of distances or products of distances all give
anticorrelation, yet they involve differing levels of sensitivity.
This provides the hope that opposing these quantities through ratios
might break the pattern of degeneracy. Indeed this was found in
\cite{linap} on consideration of the cosmic shear test\footnote{A note
on the name: this probe involves the observed shearing of
a sphere due to the properties of spacetime itself. Despite a perfectly
isotropic space, a sphere will appear sheared because the spacetime is not
isotropic -- directions along and perpendicular to the null geodesics differ.
This shear is due to cosmic properties, unlike shears in gravitational lensing
due to anisotropies not in spacetime but in space, though the latter is
unfortunately sometimes called cosmic shear.}
originated by \cite{alpac}. Here the key quantity was
$H(z)d_A(z)\sim H(z)\int dz/ H(z)$, where $d_A$ is the angular diameter
distance. For a certain redshift range, $z\approx1.3-2.3$ (for $\om=0.3$),
the competition between the two ingredients causes a positive
correlation between $w_0$ and $w_a$. A critical pitfall however was
that systematic uncertainties could blow up the constraint contour in
a preferred direction such that the resulting contour actually exhibited
an anticorrelation (compare Figs.\ 5 and 8 of \cite{linap}). Another
indication of positive correlation was found between discrete values
of $w(z)$ in neighboring redshift bins by \cite{songknox} for
ratios of distances entering weak gravitational lensing calculations.
Here we investigate distance ratios that appear in strong gravitational
lensing measurements. Specifically, we consider $r_{ls}/r_s$, related
to the angular separation $\Delta\theta$ between multiple images of a
source, $r_{ls}/(r_lr_s)$, related to the time delay $\Delta t$ between
the images, and $r_lr_{ls}/r_s$, connected with the length scale of
caustic properties, or the cross section, of the lensing. Here $r_l$
is the comoving distance to the lens, $r_s$ to the source, and
$r_{ls}$ between the source and lens; in a flat universe $r_{ls}=r_s-r_l$.
Figure \ref{fig.distratio} illustrates the sensitivity of each of these
ratios to the cosmological parameters as a function of redshift.
We have taken the redshift to represent the lens redshift $z_l$ and
for convenience fixed the source redshift to be $z_s=2z_l$ (since that
gives roughly the greatest probability of effective lensing). Generally
the greatest
sensitivity is to the magnitude of the matter (or dark energy) density, then
to the present value of the dark energy equation of state ratio, and least to
the time variation of the equation of state.
\begin{figure}[!hbt]
\begin{center}
\psfig{file=distratiosens.ps,width=3.4in}
\caption{The sensitivity of the distance ratios to the cosmological
parameters $\theta=\{\om,w_0,w_a\}$ are encoded in the derivatives
plotted here, normalized to 1\% fractional measurement of the distance
ratio. The larger the absolute
magnitude of the derivative at a particular redshift, the more
constraining the observations there.
}
\label{fig.distratio}
\end{center}
\end{figure}
Of most interest to our
investigation are the crossings from negative to positive sensitivity,
i.e.\ at some redshift the distance ratio switches from diminishing
to growing as we change the value of a parameter $\theta=\{\om,w_0,w_a\}$
(increasing it, say).
Since these crossings occur at different redshifts if the parameter is
$w_0$ than if it is $w_a$, then the correlation between these
quantities can shift from negative to positive. We see that positive
correlation occurs for $r_{ls}/r_s$ for $z_l\approx 0.9-1.7$ and for
$r_{ls}/(r_lr_s)$
for $z_l\approx 0-0.6$; zero crossing in equation of state variables
and hence positive correlation
does not occur for $r_lr_{ls}/r_s$. (The exact values will depend on
the fiducial cosmology, here taken to be $\om=0.3$, $w_0=-1$, $w_a=0$.)
For the ratio $r_{ls}/r_s$ we expect rapid evolution in the $w_0-w_a$
degeneracy direction as the contour will rotate from horizontal at
$z=0.9$ (no sensitivity to $w_0$, so lying parallel to the $w_0$-axis)
to vertical at $z=1.7$ (no sensitivity to $w_a$). Fig.\ \ref{fig.slflower}
shows this behavior in a ``flower'' plot, with idealized precision
to better illustrate the degeneracy direction. A strong lensing survey
measuring image separations from lenses within this redshift range
should possess good internal complementarity and the ability to constrain
the dark energy equation of state. Furthermore, in this
range strong lensing will have near orthogonality with supernova distance and
weak gravitational lensing surveys and so be a valuable complementary probe.
\begin{figure}[!hbt]
\begin{center}
\psfig{file=slflower.ps,width=3.4in}
\caption{Confidence contours for an idealized experiment measuring
the ratio $r_{ls}/r_s$ rotate depending on the redshift. Degeneracy
directions follow the Fisher sensitivities from Fig.\ \ref{fig.distratio},
with zero crossings of a parameter giving contours parallel to that axis.
Note the strong positive correlation between equation of state variables
for $z_l=1.3$, and hence strong complementarity with, e.g., distance
measurements.
}
\label{fig.slflower}
\end{center}
\end{figure}
The ratio $r_{ls}/(r_lr_s)$ appears somewhat less promising. Its
sensitivity to $w_a$ is much weaker in the key redshift range and
the evolution of the correlation between $w_0$ and $w_a$ much less
dramatic. In particular the contour never tilts far from the vertical.
This means it is not as complementary either internally or with other
cosmological probes. But it does possess three interesting
characteristics: its overall sensitivity to the equation of state is
reasonably large at $z>1$, the positive correlation region is at low
redshifts where observations are easier (though there is
less volume, hence fewer lens systems), and it has an odd null
sensitivity with respect to $\om$ at $z=0.35$. This last is where its
sensitivity to $w_a$ is maximal in the positive correlation region and
eases degeneracy with the matter density and hence reduces the need for
a tight external prior on it. So time delay measurements should
not be wholly neglected as a possibly useful probe.
\section{Constraints on Cosmological Parameters} \label{sec.fisher}
So far we have only concentrated on the degeneracy direction. What about
the actual magnitude of constraints capable of being imposed on the
cosmological parameters? If we consider one parameter $\theta$ at a time,
fixing the others, we can obtain a lower limit
on the parameter estimation uncertainty. This is given by
\beq
\delta\theta\ge \left|\frac{\partial[{\rm Distance\ Ratio}]}{\partial\theta}
\right|^{-1} \delta[{\rm Distance\ Ratio}].
\eeq
In Fig.\ \ref{fig.distratio} we took a fractional measurement error of 1\%.
Thus, for example, the sensitivity for $r_{ls}/r_s$ with respect to the
parameter $w_0$ at $z=0.3$ is -6.5 so the best possible estimate of $w_0$
from such a single, 1\% measurement is 1/6.5=0.15. That is, the
unmarginalized uncertainty is $\sigma(w_0)= 0.15$. This will be degraded
by degeneracies with other parameters, or by worse precision in the
measurement, and improved by additional measurements, subject to some
systematics floor.
In fact, degeneracies with additional parameters wash out the simple
orthogonality of Fig.\ \ref{fig.compl}. The parameter phase space is
actually 3-dimensional (from $\om$) or higher and the joint probabilities
do not merely trace the intersection of the contours in the equation of
state plane. This weakening of apparent complementarity also means
that the magnitude of the sensitivity with redshift matters as well
as the degree of correlation, and this can alter the favored redshift
range. All these effects need to be taken into account.
Here we seek to establish what improvements the complementarity of strong
lensing data with other probes such as distance-redshift or weak lensing
data can realistically provide. Next generation precision in all
three of these come in one
package: the proposed Supernova/Acceleration Probe (SNAP: \cite{snap}). With
some $10^7$ galaxies observed in its deep survey (down to AB mag $R=30$)
and $10^8$ in its wide survey (to AB mag $R=27.5$), the canonical
estimate of one strong lens system per thousand galaxies provides an
anticipated wealth of data. Other valuable data sets will come from
LSST \cite{lsst} and LOFAR \cite{lofar}.
\subsection{Image Separation} \label{sec.sep}
If we consider an optimistic strong lensing measurement precision of 1\%
in a redshift bin of 0.1, whether
through statistical means, e.g.\ 100 lens systems of 10\% precision, or as a
systematic floor, we find that the parameter uncertainties are large if we
do not a priori fix some of the variables. Even employing the orthogonal
combination of measurements of the strong lensing separation variable
$S=r_{ls}/r_s$ at $z=0.9$ and $z=1.7$ (cf.\ Fig.\ \ref{fig.slflower}), the
estimations are $\sigma(w_0)=0.15$, $\sigma(w_a)=0.53$ even with $\Omega_m$
fixed, and 0.28, 0.93 with a prior on $\om$ of 0.03. The lack of sensitivity
of the strong lensing separation to the dark energy equation of state
and, as mentioned above, the degeneracy with the other parameter, $\om$,
prevent as
well its complementarity with other probes from being as effective
as might have been hoped from \S\ref{sec.anti}.
Figure \ref{fig.slangle} shows contours in the $w_0-w_a$ plane, now
marginalized over $\om$, for various combinations of probes. This considers
1\% measurements at $z=0.9$, 1.3, 1.7. The improvements in parameter
estimation remain modest: about 15\% in $w_0$ and $w_a$ when added to
SNAP supernova distance data, 19\% and 3\% respectively when added to the
supernova plus weak lensing data set, 25\% and 13\% when added to
supernova plus Planck CMB data, and 17\% and 6\% when added to
all three data sets. If strong lensing $S$ measurements were available from
$z=0.1-1.7$, say (recall this is the lens redshift, with the source assumed to
lie at twice this redshift), the improvement to the triple data set
increases to 30\% and 15\%.
If the measurement floor for the strong lensing was instead at 2\%, then the
constraints in conjunction with the other three probes only gain due to strong
lensing by 6\% and 2\% for $z=0.9$, 1.3, 1.7 or 17\% and 7\% for $z=0.1-1.7$.
\begin{figure}[!hbt]
\begin{center}
\psfig{file=snsl.angle.ps,width=3.4in}
\caption{Estimates of the dark energy equation of state parameters
from various combinations of cosmological probes (SN=supernovae,
WL=weak lensing, CMB=cosmic microwave background) improve modestly
with the addition of strong lensing image separation measurements.
Outer contours of each pair of line types include the labeled data
sets as will be provided by SNAP and Planck; inner contours add the
1\% strong lensing constraint.
}
\label{fig.slangle}
\end{center}
\end{figure}
\subsection{Time Delay} \label{sec.time}
The other strong lensing quantity we consider is $T=r_{ls}/(r_lr_s)$, related
to the time delay between images. This introduces an extra parameter --
the Hubble constant $H_0$, which we marginalize over. Again, measurements
of strong
lensing by itself provide only poor estimates of the cosmological parameters,
even for combinations of redshifts where the degeneracy directions are
complementary. The insensitivity to $\om$ at low redshift noted in
\S\ref{sec.anti} does not help since the overall weak dependence and
degeneracy of the equation of
state variables prevents precision constraints.
In complementarity with supernova distances, strong lensing $T$ measurements
of 1\% precision at $z=0.1-0.6$ provide improvement by 21\% and 9\% in
$w_0$ and $w_a$. Adding them to all three other data sets helps by
13\% and 9\%, or 22\% and 15\% if the strong lensing measurements
extend from $z=0.1-1.7$. Weakening the precision to 2\% adjusts this
last case to 7\% and 5\%. Note that no external prior on
$H_0$ is required; it is determined from the data to better than 1\%.
\subsection{Image Separation and Time Delay} \label{sec.septime}
Simultaneous addition of both separation and time delay information allows
further improvements. Relative to the supernova only data, these are 36\% and
41\% for 1\% strong lensing data at $z=0.1, 0.3, \dots 1.7$; added to all
three other probes they yield tighter constraints by 31\% and 17\%. With only
2\% precision the improvements in the latter case are 15\% and 7\%. We
show the parameter constraints in Fig.\ \ref{fig.snsl2} for the three
probes without strong lensing, with strong lensing for the fiducial
redshift bins, for a reduced strong lensing redshift range, and for the
weaker 2\% strong lensing precision.
\begin{figure}[!hbt]
\begin{center}
\psfig{file=snsl2.lamsug.ps,width=3.4in}
\caption{Estimates of the dark energy equation of state parameters from
supernovae, weak lensing, and CMB measurements can be further tightened
by adding observations of both strong lensing quantities (SL$^2$): image
separations and time delays. The outer, long dashed ellipse with no strong
lensing and ``$\Lambda$'' in its center corresponds to the outer solid
ellipse of Fig.\ \ref{fig.slangle}.
Strong lensing data at 1\% precision adds valuable complementarity (solid
ellipse), even over a reduced redshift range (dotted). Higher
systematic uncertainties on strong lensing (2\%: short dashed), however,
strongly diminish its usefulness. As usual (cf.\ \cite{linbo}), dark
energy models with time varying equation of state such as the SUGRA
model (ellipses centered around ``S'') show enhanced complementarity.
}
\label{fig.snsl2}
\end{center}
\end{figure}
The highest redshift bins are not that powerful but it is important to include
the $z=0.1$ bin to constrain $H_0$, which is strongly degenerate with
$w_0$ and $w_a$. Even with $H_0$ known to better than 1\% from the data,
the degeneracy still plays an important role since the strong lensing
time delay probe is so much more sensitive to $H_0$ than to the equation of
state.
As usual, taking the fiducial cosmology to be that of a cosmological
constant tends to underestimate both the sensitivity and complementarity
of the probes \cite{linbo}. For comparison Fig.\ \ref{fig.snsl2}
includes contours for a supergravity inspired dark energy model with
time varying equation of state: S(UGRA) with $w_0=-0.82$, $w_a=0.58$.
Here strong lensing offers more dramatic improvements, by 49\% in
estimating $w_0$ and 54\% in $w_a$
for 1\% precision and 29\% and 22\% for 2\% precision. Even at 2\%
precision the final uncertainty from the four probes in complementarity
achieves an impressive constraint on the time variation of
$\sigma(w')=0.05$.
\section{Conclusion} \label{sec.concl}
While strong lensing does not completely fulfill the promise of orthogonal
constraints on the dark energy equation of state that would provide vastly
improved parameter estimates, it does offer some complementarity,
especially for models with time varying equation of state.
Furthermore, the data resources required will be innate within the deep
and wide optical and near infrared surveys of SNAP; abundant strong lensing
data should also come from LOFAR at radio wavelengths and both strong and
weak lensing from LSST in the optical. We find that strong lensing
adds reasonable value
to dark energy constraints when the measurements reach the 1\% precision
level. This is in accord with other calculations (e.g.\
\cite{yamamoto,lewisibata}). Precision arises from statistics
subject to a systematics floor.
The challenge in the next generation will not be the search for sufficient
strong lensing data, it will be achieving systematics limits sufficiently
low to allow the statistical wealth and the value of strong lensing
complementarity to come into play. Currently 1\% sounds rather optimistic
for a bound to systematic uncertainties. The main contribution is
likely to come from our ignorance of the lensing mass model. For example,
within a singular isothermal sphere model 1\% measurement of the
distance ratio $S$ requires knowledge of the velocity dispersion to
0.5\% \cite{dhk}. Cross correlation of different sources with the same
lens to remove systematics, {\` a} la \cite{bernjain} for weak lensing,
would not work, both because of the rarity of strong lensing and because
sources at different distances and positions would probe different parts
of the lensing mass distribution.
One might hope to constrain the lensing mass distribution by making
use of the simultaneous weak lensing
information that SNAP or LSST would provide, but this will be of little use
according to \cite{dalal}. Basically the weak lensing measurements of
shears average over a larger range of scales (or multipoles) than the
convergence contribution from the mass to the strong lensing. It is
like trying to locate a pin with thick gloves. Nongaussian effects
also need to be incorporated in a more rigorous treatment of strong
lensing. Relaxation
of the simplification that sources lie at redshifts $z_s=2z_l$ is
unlikely to change significantly the results shown here.
The advantage of image separations and time delays is that they are
well defined and used as markers of the cosmic
geometry. One could also use the statistics of lensing systems,
e.g.\ the number of lenses out to some redshift (e.g.\
\cite{sarbu,oguri,kuhlen}),
but then one must deal with astrophysical scatter and observational
bias and incompleteness. Furthermore,
lensing counts depend on mass density growth and volume factors
that take us further from our original goal of finding orthogonal
probes in the equation of state plane.
Since systematics are the limiting factor, we can attempt to seek out
special lensing systems that ameliorate these uncertainties,
such as strongly lensed calibrated candles
(e.g.\ Type Ia supernovae \cite{holz,ogurik}), time evolving, differentially
amplified sources (e.g.\ Type II supernovae \cite{goobar,wagoner}), or
relatively clean Einstein cross or ring images. With the wealth of
future data such subsampling may be practical.
While interrelations between cosmological parameters entering into
distances and the expansion rate prevent clean orthogonality between
cosmological probes, the various methods retain sufficient complementarity
to both crosscheck each other and further tighten constraints on the
nature of dark energy. If the systematics challenge can be met, in
special systems at least, strong lensing merits further consideration
in joining the cosmological toolbox of next generation probes.
\section*{Acknowledgments}
I thank Dragan Huterer for helpful comments. This work has been supported
in part by the Director, Office of Science, Department of Energy under
grant DE-AC03-76SF00098.
\begin{thebibliography}{99}
\bibitem{linjen}
E.V.\ Linder and A.\ Jenkins, MNRAS 346, 573 (2003); astro-ph/0305286
\bibitem{lingrav}
E.V.\ Linder, in draft
\bibitem{fhlt}
J.A.\ Frieman, D.\ Huterer, E.V.\ Linder, M.S.\ Turner, Phys.\ Rev.\
D67, 083505 (2003); astro-ph/0208100
\bibitem{linap}
E.V.\ Linder, Phys.\ Rev.\ D68, 083503 (2003); astro-ph/0212301
\bibitem{seo}
H-J.\ Seo and D.J.\ Eisenstein, Ap.\ J.\ 598, 720 (2003);
astro-ph/0307460
\bibitem{linbo}
E.V.\ Linder, Phys.\ Rev.\ D68, 083504 (2003); astro-ph/0304001
\bibitem{wl}
M.\ Takada and B.\ Jain, astro-ph/0310125
\bibitem{cluster}
S.\ Majumdar and J.J.\ Mohr, astro-ph/0305341
\bibitem{weinberg}
S.\ Weinberg, {\it Gravitation and Cosmology\/} (Wiley, 1972)
\bibitem{fpoc}
E.V.\ Linder, {\it First Principles of Cosmology\/}
(Addison-Wesley, 1997)
\bibitem{tolman}
R.C.\ Tolman, Proc.\ Nat.\ Acad.\ Sci.\ 16, 511 (1930); note
the traditional reference is I.M.H.\ Etherington, Phil.\ Mag.\
15, 761 (1933) but this actually writes the relation with one
fewer power of $1+z$ due to a different distance definition.
\bibitem{thermo}
E.V.\ Linder, Astron.\ Astrophys.\ 206, 190 (1988)
\bibitem{bassett}
B.A.\ Bassett, astro-ph/0311495
\bibitem{hutcoo}
A.\ Cooray, D.\ Huterer and D.\ Baumann, astro-ph/0304268
\bibitem{linprl}
E.V.\ Linder, Phys.\ Rev.\ Lett.\ 90, 091301 (2003); astro-ph/0208512
\bibitem{lin10217}
E.V.\ Linder, in {\it Identification of Dark Matter\/},
ed.\ N.J.C.\ Spooner and V.\ Kudryavtsev (World Scientific, 2003);
astro-ph/0210217
\bibitem{caldwell}
C.P.\ Ma, R.R.\ Caldwell, P.\ Bode, and L.\ Wang, Ap.\ J.\ 521,
L1 (1999); astro-ph/9906174
\bibitem{dave}
R.\ Dav\'e, R.R.\ Caldwell, and P.J.\ Steinhardt, Phys.\ Rev.\ D66,
023516 (2002); astro-ph/0206372
\bibitem{alpac}
C.\ Alcock and B.\ Paczy{\' n}ski, Nature 281, 358 (1979)
\bibitem{songknox}
Y-S.\ Song and L.\ Knox, astro-ph/0312175
\bibitem{snap}
SNAP: http://snap.lbl.gov
\bibitem{lsst}
LSST: http://www.noao.edu/lsst
\bibitem{lofar}
LOFAR: http://www.lofar.org
\bibitem{yamamoto}
K.\ Yamamoto, Y.\ Kadoya, T.\ Murata, and T.\ Futamase,
Prog.\ Theor.\ Phys.\ 106, 917 (2001); astro-ph/0110595
\bibitem{lewisibata}
G.F.\ Lewis and R.A.\ Ibata, MNRAS 337, 26 (2002); astro-ph/0206425
\bibitem{dhk}
A.N.\ Davis, D.\ Huterer, and L.M.\ Krauss, MNRAS 344, 1029
(2003); astro-ph/0210494
\bibitem{bernjain}
G.M.\ Bernstein and B.\ Jain, Ap.\ J.\ 600, 17 (2004);
astro-ph/0309332
\bibitem{dalal}
N.\ Dalal, D.E.\ Holz, X.\ Chen, and J.A.\ Frieman, Ap.\ J.\ 585,
L11 (2003); astro-ph/0206339
\bibitem{sarbu}
N.\ Sarbu, D.\ Rusin, and C-P.\ Ma, Ap.\ J.\ 561, L147 (2001);
astro-ph/0110093
\bibitem{oguri}
M.\ Oguri, Y.\ Suto, and E.L.\ Turner, Ap.\ J.\ 583, 584 (2003);
astro-ph/0210107
\bibitem{kuhlen}
M.\ Kuhlen, C.R.\ Keeton, and P.\ Madau, Ap.\ J.\ accepted;
astro-ph/0310013
\bibitem{holz}
D.E.\ Holz, Ap.\ J.\ 556, L71 (2001); astro-ph/0104440
\bibitem{ogurik}
M.\ Oguri and Y.\ Kawano, MNRAS 338, L25 (2003); astro-ph/0211499
\bibitem{goobar}
A.\ Goobar, E.\ M{\" o}rtsell, R.\ Amanullah, and P.\ Nugent,
A\&A 393, 25 (2002); astro-ph/0207139
\bibitem{wagoner}
E.V.\ Linder, R.V.\ Wagoner, and P.\ Schneider, Ap.\ J.\ 324,
786 (1988)
\end{thebibliography}
\end{document}