We have discovered what appears to be a giant sub-mm arc, which is coincident with, but much more extended than, a previously known optical arc in the lensing cluster MS0451-03. The figure shows 850 micron SCUBA contours overlaid on an HST image, with the lensed galaxy clearly visible. Without higher resolution sub-mm imaging it is impossible to determine how many separate objects (or images of objects) are contributing to the SCUBA emission. However, the extended ridge of emission lies directly along a portion of the Einstein ring for this cluster, which gives a striking visual demonstration of the reality of gravitationally lensed amplification for sub-mm sources. There can be no doubt that the source of the sub-mm emission lies considerably beyond the z~0.5 of the cluster.

This image also demonstrates that not all cosmological observations with SCUBA are at minimal signal-to-noise levels; this single map may in fact contain more total flux than any other "blank field" which has been observed with the JCMT. (See full size image for contour levels.) The SCUBA contours are derived from a combination of jiggle-map data taken with different chop throws and directions, and come from a map made using direct matrix inversion of the difference data. This approach was necessary because of the strong extended emission in the map, which otherwise was partially chopped out. This is the first time such a technique has been applied to SCUBA data.

For those interested in the data analysis, it's worth giving a little more detail (and others should skip this paragraph!). The basic issue is how to deconvolve the off-beams when making maps with SCUBA. The standard method (for scan-maps at least) is to approximate a series of chops as samples in Fourier space. A better approach is to iteratively solve for the underlying map which, when chopped, gives the actual data. And the the direct method is to invert the matrix of differences in order to obtain the unchopped map. None of these approaches work well unless there are enough sets of different chops in the first place to properly constrain the map (or alternatively so that the matrix is not singular). For jiggle-maps an iterative method will of course work, but in fact the total number of pixels is small enough that there's no need, since direct inversion can be performed. There are no more than around 2000 pixels in a single jiggle image, and a 2000x2000 matrix is small enough to directly invert on even a modest workstation. There are some complications to do with how one deals with the noise, with the DC level and with pixels which are off the main map, but these can all be dealt with, and are discussed in a forthcoming paper on our SCUBA observations of MS0451-03.

The lensing amplification through the cluster allows for detailed study of intrinsically fainter sub-mm sources than are possible in the field, and make follow-up observations easier in other wavebands. Existing Chandra X-ray data for this cluster, together with new near-IR imaging are in the process of being analysed. This may help towards determining the nature of these particular sub-mm sources, as well as understanding the importance of lensing for SCUBA galaxies in general.