VLA Monitoring of 1608+656

S.T. Myers (U. Pennsylvania)
C.D. Fassnacht (Caltech)
... and the CLASS act ...


Last update: 27 June 1997

Introduction

The quadruple-image lens system CLASS 1608+656 (Myers etal. 1995) consists of 4 images of the core of a background radio galaxy (Snellen etal. 1995) at a redshift of z = 1.394 (Fassnacht etal. 1996) lensed by a galaxy at z = 0.63. The radio emission from the lensed source remains point-like in MERLIN and VLBA images, though there must be a jet at some level given the double-lobed structure observed on large scales. Observations of 1608+656 made during the CLASS-2 sessions in 1995 demonstrated that the source was variable, and VLA monitoring of this system was undertaken during the 1996-1997 A and B configurations. This work has been carried out by Chris Fassnacht as part of his Caltech thesis, and I am reporting his results here as he is now at the Keck telescope observing further CLASS lenses and candidates.

Is 1608+656 a golden lens? A number of talks at this meeting will address the criteria for membership in this exclusive club, and I will use my own set of desirable quantities below. It is apparent that 1608+656 is a darn good lens, and time will tell whether this system will be elevated to those lofty heights of lensdom that mere ordinary lenses can only aspire to. However, in this imperfect Universe governed by Niven's Postulate --- ``The perversity of the Universe tends toward a maximum.'' --- this may be about as good as it gets.

Does 1608 (as it is known to us all, though it was christened 16156554 in its 8-character VLA observing schedule) have a simple geometry --- does a simple single-lens model suffice, and is it isolated? Modelers have been trying 2-galaxy lens models, and some of this work will be reported elsewhere in this workshop. There are rumors also that none of the models matches the positions of the components well, though it is not clear what a poor chi-squared means when astrometric accuracy is on the order of 5 mas or better. As for its environment, there is no sign of companions, but we have not yet extensively mapped the region around the lens. Time will tell on these counts. Certainly the 1608 system provides many constraints on the mass distribution, given the four images and the optical ring seen in HST images (see Neal Jackson's presentation). As yet, VLBA observations have shown only unresolved components of the images, though it should be possible to do better. The radio flux densities of 1608 A - D certainly show variability at the 10% level, as will be shown below, and there are features in the light curve that yield time delays.

Observations and Analysis

Given the image separations (2.1 arcsec across the quad), it is possible to monitor 1608+656 with the VLA at 8.5 GHz in the A and B configurations. This was carried out this last year from 10 Oct 1996 - 22 Apr 1996. We were able to sample at intervals of 3 to 5 days during this period. Our observations lasted about 1 hour, with an observing cycle consisting of

This cycle was repeated to fill up the allocated time slot. The two comparison sources were chosen to be nearby steep-spectrum (and thus non-variable) sources.

The first phase of amplitude and phase calibration was carried out in AIPS, using the standard procedure. The comparison source data and 1608 data were output as UV-FITS files, for further processing in difmap. The following procedure was adopted in our first go-round at this, and all methods and results should be considered as preliminary.

The comparison sources were mapped and master models made in difmap. For each individual observing day (``epoch''), the data was read in along with the master model, a single phase-only selfcal performed to align the phase center (we have also experimented with a global amplitude gscale with fixed gain). A series of the master models with different overall amplitude scalings were then read in, and goodness-of-fits to the data determined by the modelfit procedure (no iterations were made). These chi-squared values were output, and a parabolic fit made to find the minimum chi-squared amplitude (and uncertainty). These best scaling values were recorded to be used as normalizations for 1608.

For 1608+656 itself, data and master model were also read into difmap, and a phase-only selfcal performed. Modelfitting was then carried out, allowing the 4 point component fluxes to vary while keeping the positions fixed. The best model was then output and recorded.

VLA Light Curves

The comparison source fluxes were normalized to the mean over the observations. In each epoch, the average of the normalized comparison source amplitudes was calculated to form a scaling factor for the epoch. This was used as a ``flat-field'' curve. The four component model-fit fluxes for 1608+656 were divided by the flat-field value to get normalized component fluxes.

So far in this data-set, we have only analysed the 3 bright components A - C. Component B varies first, followed by A, then closely by C. The weakest component D should be last, and considerably later.

Preliminary results show clear variations on the order of 10% in the fluxes of the 3 bright components. There are also strong features in the curves which can yield time delays. The total span of the observations was just over 200 days, and visual inspection of the curves gives a delay of about 30 days. However, there was only one (or maybe two) good features over this period, and further observations are warranted to refine and confirm this time delay.

The best match is between the B and C components. The normalized curves were registered and a chi-squared formed. A ``chi-by-eye'' analysis gives a delay of about 34 days between B and C. The curve for the brightest component A is somewhat different. Our visual fit gives a delay of around 29 days. It must be stressed that these numbers are very preliminary, and we are currently working on a full analysis of the data. This information is being provided as a courtesy to the workshop participants, and should not be quoted, referenced, or used in publication without the authors' consent.

We still have a number of things to work on in the area of error estimation and control. An exhaustive chi-squared analysis (or other statistical analysis) is still under way. The odd behavior of the A component is suspect, and we are checking the calibration and analysis pipeline. Monte-Carlo testing should be done also.

Conclusions

We have a good case for measurable time delays in this system. There is clear variability, and indications of identifiable features in the VLA light curves which yield a shift of about 30 days. There is some discrepancy between the A component and the other two brighter components in our data. However, the A component is highly magnified (about twice that of the B and C components) and thus slight differences in its behavior might be expected. This will have to be verified in careful analysis. Though from a modeling standpoint this system has some flaws, it is clearly a superb lens from observing point of view. We should in principle be able to pick up the fainter D component, giving 3 delays.

In October we will be proposing for the next VLA monitoring sessions in A and B array. There are a number of things that we should consider for improving over the first run. We should probably observer longer, to improve the sensitivity. Do we want to observe more often, given the 5 day or so delay between A and C? Do we want to add more comparison sources to improve the flat-field curve? What is our limiting uncertainty? Is our reduction pipeline optimal?

One of our problems is the gap between B and the next A configuration of over 1 year. We have proposed VLBA monitoring to bridge the gap, but did not receive the time. It may be possible to use the VLA in C and D for rough total-flux monitoring. Merlin is yet another possibility. This is something we should address for future observations.

Comments on CLASS in general

CLASS has proved to be a remarkably successful program. As of this date, in the two CLASS-1 and CLASS-2 phases, we have observed around 7800 target field, and identified 5 new lenses (and there are still a number of candidate doubles to be checked). (When combined with JVAS, over 10000 targets have been surveyed yielding 11 lenses!) The reduction pipeline is effective, and automatic mapping and modelfitting has given us a powerful database. I am working on the CLASS survey paper, and will make the CLASS data available with publication. We hope to get time in the next A configuration for around 3000 more targets for CLASS-3 to finish the uniform sample. As well, we will return to monitoring 1608+656, as well as other CLASS lenses.


References


CLASS WWW Sites:


The Authors