Strong Gravitational Lenses in the Era of Wide-Field Surveys

Furlanetto et al selected 50-odd clusters in SDSS stripe 82 by richness, in two redshift bins, and then re-imaged in g, r and i at SOAR. Image quality is comparable to median LSST, but depth is only 22-23mag. Interesting sample for algorithm validation, but they are also interested in arc statistics

Intersting way to estimate the mass of SMM galaxies

Mertens, Yoho & Starkman estimate the effect of our own motion on the convergence and shear and shear we observe, concluding that the kappa introduced is at the 0.001 level while gamma is unaffected. Given a sufficiently large all-sky ensemble, I guess this effect might be detectable as a modulation of the time delay distance signal - but its pretty small.

Hojjati, Kim and Linder (LSST DESC TDC Good Team Number 1!) try out their Gaussian Processes time delay estimation, with promising initial results - albeit on just one blinded test dataset. interestingly, they find some benefit to including microlensing in the fit, even when none is input to the mock data - which suggests that measuring microlensing parameters might potentially be tougher than just estimating time delays.

We actually ran several blinded datasets, but just illustrated the results by one case. Chris Kochanek was pretty keen that we should use a gamma=1 exponential kernel, i.e. exp{-|r/L|}, which you also adopt. We're running that to see if it performs even better. Any updates on when the Time Delay Challenge 0 will commence?

Shortly! We have the data, we're just getting it on the web for you. BTW, Eric is referring to our Flotsam code - an independent implementation of GPs for time delays, with different assumptions (and bugs :-) https://github.com/eggplantbren/Flotsam

Collett et al are attempting to use all the information in the galaxy catalog for the field around a strong lens (well, the stellar masses and photometric redshifts, at least) in order to reconstruct, using a halo model, the external convergence at the lens (like Wong, Keeton et al have been doing). In fact, their code (Pangloss) returns the PDF Pr(kappa|data), which can then be propagated into a time delay distance measurement. The model is pretty simple right now, and so everything has to be calibrated to N-body simulations (not dissimilarly from how the Suyu/COSMOGRAIL "H0LiCoW" group have been treating this problem in B1608 and RXJ1131). The goal for the future is to break free of this calibration by increasing the flexibility and physical realism of the mass model.

Jackson reviews this sub-field. Could be a good starting point for new investigators!

Eulaers et al present two new time delay measurements, including one (in SDSS J1206+4332) to 3% accuracy! HS2209 is a new delay, now measured to +/- 25%. They adopt a "netflix" approach to the measurement, combining the results from four different techniques. Interested to see how they weight them.

MacLeod, Ramsey, Agol & Kochanek use a high resolution 11.2 micron image of MG0414 taken with Michelle on Gemini North to investigate a lens model that includes an additional (third) mass component near to the merging pair of images. The radio and mid IR positions and flux ratios constrain the substructure's Einstein mass to be 10^(7.3+/0.2)Msun. They do not extrapolate to its halo mass, or consider the possibility that it is not at the lens redshift (0.96), but it's a pretty solid detection. They point out how useful JWST will be for this kind of work in future.

Would be nice to make similar forecast for Euclid (~40 epochs in the Deep Caps)

Li, Hjorth & Richard quantify the unlikelihood of seeing a multiply-imaged supernova in A1689, and other clusters.

Short et al (2012) revisit lens statistics: "The complete Herschel-ATLAS data set should be sufficient to yield competitive constraints on \Omega_\Lambda." By complete set they mean 1000 lenses, and the constraints they derive are +/-0.01 in Omega_Lambda. How optimistic are their strong assumptions about what is known about the sources, lenses, selection function and the rest of cosmology? They do conclude that the sample "will be most powerful in constraining uncertainty in astrophysical processes," but the cosmology opportunity is clearly worth keeping an eye on. It'd be good to see the systematics parameterised and marginalised over.

This is a very interesting paper indeed - Karpenka et al are trying to detect the weak lensing effect on SNe Ia, using a halo model for the line of sight structure. The interesting part is that they infer the parameters of the halo model simultaneously with the cosmological parameters, with the individual convergences playing a fleeting role before being quickly marginalised out. This is the Right Thing To Do - very computationally intensive, but feasible, as they show!

Although, to qualify that comment about computational intensity: they do not include scatter in their assumed scaling relation between halo mass and galaxy luminosity. To do that properly needs a whole slew of additional parameters, one for the true mass of every galaxy in sight...

It is similar to what I wrote a paper about in 2001. It is good to see this idea is being pursued. It has always seemed to me that you would do better with galaxy-galaxy lensing because of the much larger numbers.

It's an EEL! In Auger et al (2011) we searched for and found a sample of early type - early type lens systems like this one, and found that their sources were compact too - but our sample was limited by the SDSS spectra to lower lens and source redshifts. We did find that you need to treat the lensing and PSF convolution quite carefully to make accurate measurements of size and flux though, especially given the strong degeneracy between effective radius and Sersic index. High resolution IR imaging with a stable, well characterised PSF would indeed help a lot.

In this paper we estimate the number of gravitational arcs detectable in a wide-field survey such as that which will be operated by the Euclid space mission, assuming a LCDM cosmology. http://cgiocoli.wordpress.com/research-interests/moka/

Formed in summer 2012, the LSST DESC sets out its plans for the next few years, preparing to try and measure the accelerating expansion of the Universe very accurately indeed. Most of the work between now and 2020 will be on trying to predict, understand and mitigate against systematic errors in this measurement - so analysis pipelines need defining, implementing (in pieces) and testing on simulations. The strong lensing case is primarily time delay distances from ~1000 lensed quasars and SNe, but multiple source plane systems could play a role as well.

The state of the art in time delay estimation: Tewes at el test three different methods for interpolating noisy, sparse, microlensing-infected, gap-ridden optical lensed image lightcurve data, and illustrate the benefits of a) long light curves and b) accurate (not under-estimated) uncertainties. Will these methods be good enough for accurate cosmography with 100-1000 lenses though?

Furlanetto extend the usual elliptically-symmetric Sersic model for source galaxies to make the major axis a segment of a circle and the ellipse into an "ArcEllipse". They then explore ways of fitting this to pixelated, blurry data, and compromise on a method that underestimates flux but captures (quickly) the L/W ratio found by other methods. Since they base their measurements on SExtractor segmented images, this parameterisation could be adopted by Gavazzi et al's RingFinder code (if something equivalent is not already implemented). Interestingly, the number of parameters needed to specify an ArcEllipse is the same as that needed for a Sersic lensed by an SIS+Shear lens, though the predicted arc surface brightness is different.

Still reading the paper but... I remember that I talked to E. Bertin about how efficient SExtractor is to segment arc-like images. He was very negative about this. I'd like to hear your opinion on that!

In my experience SExtractor can do a very nice job at segmenting arcs. The tricky part is finding a SExtractor setup that works for arcs of all shapes, sizes and surface brightnesses. You can certainly bias your sample by choosing one setup over another. This is just a variant of the usual astronomical dynamic range problem of course - all methods have to solve this.