An update on SCUBA-2: the new wide-field imaging camera for the JCMT

SCUBA-2 is a new, wide-field submm camera planned to replace SCUBA(-1) on the JCMT in 2005/6. The science goals and baseline design were described in an article in the Sept 2000 newsletter. Considerable progress has been made since then, and this article aims to provide an update on some of the major developments that have taken place in the past year. Although the SCUBA-2 project remains only partly funded, a very valuable contribution of 4M UKP was recently awarded by the UK Office of Science and Technology towards the total project costs (estimated to be 8.9M UKP).

The science drivers for SCUBA-2 are well-established, but for anyone not familiar with the instrument, and as a way of an introduction, it's worth re-emphasising the main science goals:

- Deep imaging. This is very time consuming with SCUBA - relying on co-adding lots of frames of data over periods of many hours (especially difficult at 450 microns, where it is rare to have extended periods of good, stable weather). SCUBA-2 aims to reach the extragalactic confusion limit at 850 microns in around 1-2 hrs, instead of ~50 hrs at the present time.

- Maximise the survey potential. Even though SCUBA has been a big step forward in terms of mapping large areas of sky, only an area about the size of a full moon has been mapped to any great depth (near the confusion limit). SCUBA-2 aims to map large areas of sky at least several hundred times faster than SCUBA to the same S/N.

- Improve image fidelity and map dynamic range. SCUBA requires a minimum of 128 secs to produce a fully-sampled map at both 450 and 850 microns. In addition, because SCUBA can only record an AC signal (i.e. chopped signal) we are constantly subtracting two images of the sky. SCUBA-2 will aim to improve data quality by instantaneously sampling the sky, and operate in a mode that avoids the necessity to sky chop.

- Imaging at two colours simultaneously. This has been successful with SCUBA - for example, it allows meaningful studies of dust properties, as well as utilizing periods of good weather to exploit the higher angular resolution available at shorter wavelengths. SCUBA-2 aims to continue in this mode.

- Act as a "pathfinder" for submm interferometers. By coming on-line in 2005/06 SCUBA-2 should have at least a few years of observations before ALMA begins full operation. Wide-field surveys will be crucial to fully-exploit the capabilities of the new generation interferometers.

The instrument challenge is therefore to take these science drivers and incorporate new developments in detector technology to design the first "Submm CCD Camera"! To achieve the science goals requires that the instrument have:

- Per-pixel sensitivities to be dominated by the sky background photon noise (fundamental limitation). This requires improvements of a factor of three over the current SCUBA bolometers.

- The maximum (undistorted) field-of-view allowed by the telescope. This turns out to be 64 sq-arcminutes (c.f. to only 4.3 sq-arcmins for SCUBA) - a factor of 16 times larger field.

- Fully-sampled imaged planes and DC-coupled electronics (no sky chopping) to improve image fidelity and map dynamic range. This requires 25,600 and 6,400 pixels at 450 and 850 microns, and ultra stable electronics at low (DC) frequencies.

- A dichroic beamsplitter to split the short-wave and long-wave channels onto two separate arrays. Thus, simultaneous observing will be available with two colours.

Conventional bolometer technology (such as that used in SCUBA) is not practical for the pixel count required for SCUBA-2. The SCUBA-2 arrays will utilise superconducting Transition Edge Sensors (TES), with the signals read-out using time-division SQUID multiplexers. The array development is being led by the National Institute of Standards and Technology (NIST) in Boulder, and the Scottish Microelectronics Centre (SMC) in Edinburgh. Earlier this year research agreements were signed with both NIST and SMC to produce prototype and science-grade arrays for SCUBA-2. An "Array Technology Meeting" was subsequently held in Boulder which led to the baseline design for a single-pixel and initial concepts for an array structure.

Several crucial decisions were made at the Boulder meeting. Figure 1 shows a schematic drawing of the novel SCUBA-2 pixel design, together with a concept for the array geometry. The design consists of an upper detector chip and a lower multiplexer (MUX) chip, which are held together with indium bump bonds. The detector chip consists of two silicon wafers diffusion bonded together, with the top wafer having micro-machined square wells to provide support for the array. The upper surface of the bottom wafer forms the radiation absorber, and is ion-implanted to give a surface impedance match to free space. The backshort for the absorber is formed by the TES device itself, which covers the entire underside of the pixel, and is a one-quarter wavelength distant from the absorber. This integral backshort design greatly simplifies the construction of the array. Electro-magnetic modeling of the pixel and array geometry shows that the absorption efficiency is likely to be at least 90%.

FIGURE 1: (a) Single pixel schematic diagram (b) Concept of the array architecture design

The hybrid detector design means that each pixel has a SQUID amplifier and coupling transformer directly underneath on the MUX chip. The MUX design is critical in that it has to minimise signal crosstalk due to magnetic coupling between adjacent pixels. The maximum area available for the MUX is therefore limited by the size of the pixel (~1mm and 0.5mm at 850 and 450 microns). At 450 microns there is simply not enough space for all the coupling inductors in the low-crosstalk design. Rather than undergo an extensive re-development programme for a separate 450 micron MUX, it was decided to make the 450 micron pixel size the same as the 850, with the consequence that the 450 array would now have 6400 pixels and be instantaneously undersampled by a factor of 2 (pixels spaced by F-lambda in the focal plane).

Although this sounds like a step backwards, the scientific consequences are not that great. The large-area mapping speed is reduced by a factor of 2, a micro-step pattern will be needed to produce a fully-sampled map (only a 4-point jiggle is needed), and the beam size on the sky for each pixel is about 14% broader. Given the alternatives (a complex and lengthy re-development MUX programme) these losses were considered to be acceptable to minimise further risk in the array development. In fact, there are also some advantages such as the pixels at both wavelengths having the same physical size (making fabrication techniques inherently simpler), and fewer channels of detectors (making the electronics simpler and cheaper).

The optical design (see below) and the necessity to have the pixels at least a wavelength in size means that each focal plane will be approximately 90mm in diameter. Silicon micro-machining at the SMC is confined to industry standard 3-inch wafers. Hence, a focal plane layout has been adopted that consists of 4 separate sub-arrays at each wavelength. Each sub-array will have 40 x 40 pixels, and will be butted together as shown in Figure 2.

FIGURE 2: Array focal plane layout showing the 4 sub-arrays butted together.

Since the Array Technology Meeting in February, work has concentrated on the design and manufacture of single TES pixels and linear multiplexers, as well as developing the techniques needed to construct large-format arrays. The single-pixel design, as mentioned above, is largely complete and NIST are now starting to produce test single-pixels with the desired thermal and electrical properties. Crucially, this means that the pixels must have the correct noise equivalent power and time response to meet the science goals. Some problems with the Mo and Cu guns in the plasma deposition system have prevented the production of test TES pixels until now. These have been resolved and the programme can now re-commence. This will start with the fabrication of some x-ray pixels to show that earlier designs can be reproduced.

The MUX manufacture is already ahead of schedule with a 1 x 32 prototype already undergoing cold dip testing. The initial results, stepping through the addressing very slowly look very encouraging. A 4 K dip probe is being developed to allow testing at the full MUX addressing rates. Testing should commence at the end of September. Figure 3 shows the overall system read-out scheme, highlighting some of the components that will be incorporated.

FIGURE 3: SCUBA-2 signal readout scheme showing photographs of some of the components.

A considerable amount of work has also been carried out at SMC on developing and testing array manufacture techniques. Implanted wafers have already been made and have demonstrated excellent absorption efficiency in tests at Cardiff University. The SMC is also investigating the best way to manufacture the silicon backshort and sourcing vendors for the bump-bonding process.

In parallel with the array development, work is proceeding at a healthy rate on the instrument design at the ATC. The optics design is a crucial part of this, as it has a significant bearing on the size of the focal plane, the size and shape of the cryostat needed, and the location of the instrument on the telescope.

There are complex trade-offs in the design involving minimising the field distortion and maximising the Strehl ratio over an 8 arcminute field, as well as ensuring that the focal plane is a manageable size. Optimisation of the design using Zeemax allows for the final optics to be f/2.7, accommodating individual pixels sizes of ~1mm and a array diameter of some 90mm. However, at the tertiary mirror of the telescope, the re-imaged field is some 600mm in diameter, and so SCUBA-2 requires a complex set of optics using large mirrors with complex shapes to re-image the field onto the detector array with maximum efficiency and minimum distortion.

Another important aspect of the optical design is minimising any stray light from reaching the arrays that can potentially degrade sensitivity. This is particularly important for SCUBA-2 since the detectors have no feedhorns to define their field-of-view (they have an approximate PI steradian f-o-v). In the SCUBA-2 design the field-of-view is controlled by a cold stop some 300mm in front of the array, and it is therefore vital that a high quality pupil image be attainable at this point in the optics. Other complexities involve how the images will "sheer" as the telescope moves in elevation, and flexure issues concerned with how the mirrors are mounted.

Taking all these issues into account, we are currently considering locating SCUBA-2 on the antenna floor - underneath the Nasymth platform, but still attached to the telescope A-frame. An initial concept for this is shown in Figure 4. The main advantages of this are that it becomes much easier to control the quality of the cold-stop pupil image (the current design shows this to be the excellent), puts the instrument in an area that has more space around for electronics, and potentially allows much easier access for future add-on instruments (such as a polarimeter or an FT/grating/F-P spectrometer). Optimisation of this design is currently underway.

FIGURE 4: (a) Possible location for SCUBA-2 on the telescope.

Work is also ongoing in defining just how we will observe with SCUBA-2 on the telescope. Carrying out this work up-front is crucial, as it will likely impact on the opto-mechanical design. The two basic observing modes will be STARE-MAP - it's simplest form being a "point-and-shoot" mode for sources comparable in size or smaller than the field-of-view, and SCAN-MAP - rastering the telescope across an extended source that is larger then the f-o-v. It is likely that each of these basic modes will have several configurations depending on the source morphology and how deep an integration time is required. One of the fundamental differences between SCUBA-2 and SCUBA is how the (dominant) background emission is determined and subtracted. Instead of sky chopping, which increases the confusion limit and limits the visible source-size scales, it will be up to the astronomer to decide just where the sky background is.

An example of how the systems analysis work has an impact on the instrument design is that a cold dark shutter is being designed into the instrument to allow "dark frames" to be taken. The frequency of the dark measurements depends critically on the low-frequency (DC) stability of the detectors and electronics. Other crucial areas include how to deal with variations in detector responsivity (which is very dependent on the sky background) and also how to cope with atmospheric sky-noise. We are currently developing various simulation tools to assist in determining how best to use the instrument, given atmospheric instabilities and variations in detector response. This work is being carried out with the assistance of Rudolf Le Poole (Leiden) and Hans von Someron Greve (ASTRON) and also David Hughes and Ed Chapin (INAOE).

The cryo-mechanical design is challenging! Two 250 x 250mm focal plane structures must be integrated into a cryostat and cooled to ~60mk. This structure is thermally coupled to a dilution refrigerator and has a 1K optics box to minimise stray light. This box will also act as a magnetic field shield for the SQUID amplifiers. Initial concepts are being considered in more detail. One novel feature is the possible use of a 4K "pulse tube cooler" to enable a cryo-free dilution refrigerator to be used. This would save a great deal of time and effort (and consequently money...) during both the lab testing phase and operation on the telescope.

The timeliness of the delivery of SCUBA-2 to the JCMT is critical. For the telescope to remain competitive, in the era leading to new facilities (such as the 50-m LMT in Mexico, the Herschel satellite, and the ALMA interferometer), it is vital that SCUBA-2 be delivered by 2006. Figure 5 highlights some of the current and upcoming facilities that are/will be available for submm astronomy over the next decade. The current plans for SCUBA-2 are to deliver to the telescope before the end of 2005 with one sub-array at each wavelength (i.e. 1600 pixels at 450 and 850 microns). This will give the community as early access to the instrument as possible. Since the production and testing of science-grade sub-arrays will take some time in any case (due largely to limited production line facilities), it has been agreed that a phased delivery of instrument capabilities is the best approach.

FIGURE 5: The road to SCUBA-2 (and beyond...)

The SCUBA-2 Project Consortium is now well established. The overall management of the project, as well as the instrument design, construction, testing and commissioning, are the responsibility of the UK-ATC. NIST are responsible for the TES detectors and SQUID MUX manufacture. Working closely with them will be the SMC, who will carry out the silicon micro-machining of the detector and support structures, and with a specialist company, develop the superconducting bump-bond process for the detector and multiplexer wafers. The Astronomy Instrumentation Group at Cardiff University will test single pixels and prototype electronics, manufacture the filters and dichroic, and participate in the array test and evaluation programme. The Joint Astronomy Centre will be involved in the commissioning of the instrument, and will be largely responsible for infrastructure requirements at the telescope.

The most critical milestone that SCUBA-2 faces is proving that the proposed array technology will actually work! Hence, there is a major project milestone in October 2002, when it must be proven that a small prototype array (which could be as big as 32 x 32 elements) can be produced. This must prove issues such as detector yield and mechanical robustness, as well as demonstrating thermal, electrical and optical properties close to that designed. In summary, the other major project milestones are as follows:

SCUBA-2 represents a major departure from existing submm continuum instruments. Incorporating state-of-the-art technology will allow the realisation of the first large-format "CCD-like" camera for submm astronomy. The science applications for such an instrument are tremendously exciting and very broad-based, ranging from the study of Solar System objects to probing galaxy formation in the early Universe. SCUBA-2 will map large-areas of sky many hundreds of times faster than the current SCUBA. The improved sensitivity and imaging power will allow the JCMT to really exploit periods of excellent weather on Mauna Kea. Undertaking wide-field surveys with SCUBA-2 are vital to fully exploit the capabilities of the new generation submm interferometers. Finally, the new technology has applications beyond SCUBA-2 (optical photon counters, x-ray spectrometers, arrays in space or on even larger telescopes) and thus represents a major strategic investment on behalf of the JCMT funding agencies.

More information can be found on the SCUBA-2 Homepage: http://www.roe.ac.uk/atc/projects

The SCUBA-2 Project Team consists of:

ATC: William Duncan (Project Director), Wayne Holland (Project Scientist), Trevor Hodson (interim Project Manager), Damian Audley (Instrument Scientist), Dennis Kelly (Systems Analysis and Software Design), Tully Peacocke (Optical Design Engineer), Peter Hastings (Mechanical Design Engineer), Mike MacIntosh (Electronics Engineer), Vicki Ramsay (Project Support)

NIST: Kent Irwin (Project Manager and TES Design), Gene Hilton (MUX Design), Steve Deiker (TES Manufacture and Testing)

SMC: Anthony Walton (Head of Group), Alan Gundlach (Project Manager), William Parkes (Design Engineer: bump bonding), Camilla Dunare (Design Engineer: ion implantation and wafer bonding)

Cardiff Univ: Peter Ade (Project Manager)

Joint Astronomy Centre: Ian Robson (Director), Dean Shutt (Chief Engineer), Tomas Chylek (Mechanical Design Engineer)