ABSTRACT

We recommend a uniform mm/submm array of 64 \Theta 12 m antennas and a collecting area of 7000 m 2 . This fits the budget goal of 400 M$. Relative to same-area arrays with smaller antennas, this array gives a lower number of receivers to build, maintain, and upgrade, a smaller correlator, more room, better pointing, better calibration, better self-cal thresholds, and more flexibility for the future.

Contents

1 INTRODUCTION: THE LSA GOAL OF 10,000 m ² COLLECTING AREA

2 THE CHARGE TO THE COMMITTEE

3 OUR RECOMMENDATION

4 ARGUMENTS FOR THE UNIFORM ARRAY

5 ARGUMENTS FOR LARGE ANTENNAS RATHER THAN SMALL ONES

6 CONCLUSION

Our original concept for a Large Southern Array was a synthesis telescope that would cover all the millimeter windows up through the 350 GHz band, with a total collecting area of 10,000 m² . Our motivation was the need for very large collecting area to complement powerful optical telescopes. There are exciting discoveries being made by, and awaiting the new generation of optical telescopes (HST, Keck, the ESO VLT, Gemini, Subaru, Carnegie, and others). In addition to these large new telescopes, the years in which the LSA/MMA will do research will be the post-Hubble years, the period of the Next Generation Space Telescope. The NGST will be able to detect early-universe galaxies at z > 8. We should build an instrument that can detect CO, C I, C + , O ++ , N ++ , and dust in such objects. This requires an increase in collecting area comparable to that of the VLA, of the order of 10,000 m ².

The second motivation was the need for very large collecting area to use the angular resolution: The goal of both the MMA and LSA projects has been to reach an angular resolution of 0.1" at 1.3 mm. With a collecting area of ~ 1000 m², the IRAM interferometer currently makes spectral-line maps at 1.3 mm with a beam of 0.6" and a noise of 1 K in 0.25 km s^-1 channels (needed for star formation studies), and 0.1 K in 25 km s^-1 channels (needed for extragalactic studies). If we reduce the beam size by a factor of 6, we must increase the sensitivity 36 times to keep the same noise levels, in degrees Kelvin. For spectral lines, we cannot get any sensitivity increase from larger-bandwidth receivers. We hope to gain a factor of 4 from some improvement in detectors and with excellent weather at the new southern site. However, if we want to really use the increased resolution for molecular spectroscopy, then we should increase the collecting area by about a factor of 10 --- the sources we look at are just too weak. The number of sources we can study does not go linearly with sensitivity. There are thresholds below which we cannot detect whole classes of sources. Protostellar disks are a particular example.

2. THE CHARGE TO THE COMMITTEE:

The MMA/LSA meeting in Charlottesville in June 1997 set up this committee to make a recommendation on the choice between:

1. --- A mixed array of 8 m and 15 m antennas, or

2. --- A uniform array of 12 m antennas.

Our committee recommends a uniform array of 12 m antennas. We think that within the budget goal in the June 1997 meeting of 400 M$, it is possible to build a mm/submm array with a collecting area of 7000 m² . This can be done with 64 x 12 m antennas. We are somewhat flexible in the exact diameter and number of antennas, provided the same total collecting area is reached. Because the studies by the European engineering groups (ESO, IRAM, and M.A.N.) are just getting started, however, we propose that the engineers be given a chance to design and build an excellent 12 m dish. We think the 64 x 12 m array is more practical than an array of smaller dishes because it yields essentially the same science with a much smaller correlator, and a lower number of receivers to build, maintain, and upgrade. Otherwise, one will unnecessarily build a lot of extra equipment without any gains in any basic astronomical observing parameter --- sensitivity, spectral resolution, positional accuracy, or angular resolution. We think it is too expensive to build a same-area array with antennas of 8 meters.

4. ARGUMENTS FOR THE UNIFORM ARRAY

The strongest scientific reasons for a large mm and submm array lie in the mainstream mm/submm science: molecular spectroscopy and dust continuum observations of galactic and extragalactic sources. There are other reasons from solar system research and the study of extended emission from other thermal and nonthermal sources. These are covered in the MMA scientific documents and previous LSA documents. We think the mainstream science that can be done with the mixed and uniform arrays is basically the same. We prefer the uniform 12 m array rather than the mixed 8 m + 15 m array for the following practical reasons:
Manufacture, Construction, and Assembly: Standardized manufacture for all antennas; standardized assembly for all antennas; simpler stations for all antennas; standardized receiver boxes and receiver cabin layout for all antennas; half the number of spare parts that would be needed for a mixed array.
Effects on Interferometer Phase: All antennas have their axis crossing point at the same height above the ground (i.e., same phase reference center); all antennas have the same degree of thermal expansion --- no differential phase effects for different types of antennas; all antennas have the same sensitivity to wind.
Operations, Maintenance, Interchangeability: One type of transporter only, for moving antennas; all antennas have the same maintenance procedures; every antenna interchangeable with every other one, on every station; all antenna parts (panels, subreflector, etc.) interchangeable from antenna to antenna.
The 5000-m Altitude Site: Construction and daily operations at high altitude will pose problems enough. We should simplify and standardize as much as possible all assembly and long-term maintenance tasks. Cost: Within a budget goal of 400 M$, it appears possible to build a mm/submm array of 7000 m² (Fig. 1 and Tables 2 and 3). With this area, realistic mixed and uniform arrays end up with about the same total number of antennas, and have roughly equal total costs. Because there is no dramatic saving in the amount of equipment or the cost, we think the possible astronomical advantage of mixed arrays, namely the more elegant solution of the zero-spacing problem, is not worth the extra complication in construction and operation.

5. ARGUMENTS FOR LARGE ANTENNAS RATHER THAN SMALL ONES:

They will work. We are convinced that well-engineered 12 m antennas can work up to the 800 GHz window, have a global surface accuracy of 25 µm, and can have the required mechanical pointing accuracy (1/10th of a beam at 860 GHz = 0.7" ). Up to now, we have not heard any good reasons to the contrary. We are confident because submm telescopes of this size already exist --- the 10.4 m CSO, the 15 m JCMT, and the 10 m HHT --- and because the preliminary study of the LSA/MMA Antenna Study Committee (report of 1997 November) shows the feasibility of new-generation antennas which will be even better than the existing submm telescopes. We think the correct design strategy for this array is to build the largest possible antennas that will meet the specs at sub-mm wavelengths. We think this could be done with 15 m antennas, but their beam sizes are slightly too small for interferometer arrays in the submm bands, and there would be no margin for error in pointing or tracking. We think the largest possible dish size for practical interferometer operation in the submm bands is about 12 m.

Lower Number of Receivers; Smaller Correlator. While it is clear that the desired collecting area could be achieved with larger numbers of little dishes, there are very important advantages in making the antennas as large as possible. The most important advantage is that they yield lower numbers of receivers, IF processors, local oscillators, cryogenerators, and everything else that needs to be built, maintained, and upgraded (Fig. 2). Because the number of baselines varies with the square of the number of antennas (Fig. 3), a great advantage of the larger dishes is the reduction in the size of the correlator. In fact, the possible points in favor of small dishes (see Table 4) all seem rather minor in comparison with the very large advantages in hardware and software reduction that can be gained with bigger dishes. This is very important for manpower and operations costs, and the feasibility and cost of upgrades. In practice, over the next 20 -- 30 years, there will be a need to replace mm and sub-mm receivers and correlators with next-generation versions to profit from progress in new detectors, superconductors, semiconductors, faster electronics, etc. A crucial consideration for a large interferometer ar ray such as this one is therefore to keep the total number of antennas as small as possible. Given the same initial investment money for antennas, frontends, and backends, one should invest a larger proportion of it in good antennas, because they are permanent, and a lower proportion in receivers and correlator, because they will be replaced.

More Room. Every time something must be added or upgraded in the receiver cabin --- a new receiver, a cryogenerator, a compressor, a calibration table, a chopper, another monitor of some process or other --- you need more room. This is a real operational advantage of big dishes. In the initial design ideas presented so far, the IRAM design has a larger cabin (4.5 m) than the other designs, but this size could be extended to the other large-antenna designs as well. The need to make the receiver cabin a reasonable size is a disadvantage for small dishes, like 8 m dishes, in that it prevents them from being packed closely together in compact configurations. That is, larger dishes might actually allow shorter spacings, relative to their diameter, than smaller dishes.

Good Pointing. The flux of the minimum usable source for pointing measurements, for a given accuracy in arcsec, varies inversely as D ³ N ^0.5 , where D is dish diameter and N is the number of dishes. For arrays of constant collecting area = ND ² , this means the minimum useful flux for a pointing source is inversely proprtional to D ² . Hence to reach the same precision, it is more advantageous to have fewer large dishes than to have many small antennas. The recent study by Lucas (1997, LSA report) shows that with 64 x 12 m dishes we could use pointing sources down to 100 mJy, independent of frequency (if T_sysv = 0.5 K/GHz). The source counts at 3 mm by Holdaway, Owen, & Rupen (MMA Memo 123) imply that there would be 330 such sources/steradian at 3 mm, which gives a mean separation on the sky of 1.8 deg. This is a capability worth striving for, as it will greatly improve the ability of the array to make frequent pointing, particularly important for submm observing. Put another way, the time it takes to measure a pointing offset to the accuracy required for 1/30th beamwidth operation at 300 GHz is shorter for a large-antenna array than it is for a small-antenna area of the same total area (Fig. 4).

Good Calibration; Good Self-Cal. The flux of the minimum usable source for the normal interferometer calibration varies inversely as D ² N ^0.5 , where D is dish diameter and N is the number of dishes. For arrays of constant collecting area = ND ² , this means the minimum useful flux for normal calibration is inversely proportional to dish diameter. Self-calibration requires good signal-to-noise on most baselines involving an antenna in one atmospheric coherence time (e.g, Cornwell, T.J., 1989, in Diffraction-Limited Imaging with Very Large Telescopes, ed. Alloin & Mariotti, Kluwer, 273). The flux of the minimum usable source for self-calibration also varies inversely as D ² N ^0.5 , again meaning that the minimum useful flux for self-cal, for arrays of the same total area, is inversely proportional to dish diameter (Table 4 and Fig. 5).

Anomalous Refraction is NOT a reason to avoid large dishes. Anomalous refraction is caused by an atmospheric phase gradient across an individual antenna, on the time scale of seconds. The errors are random in time and with antenna. The normal synthesis imaging averages over the size of the (u, v) cell in both space and time, greatly reducing these errors. The large redundancy in the array also averages out these errors. Hence anomalous refraction is not a problem for normal synthesis maps or mosaic maps. At the shortest observing wavelength, 350 µm, the atmospheric opacity and phase noise will permit interferometry only under good conditions. The on-site opacity and phase measurements (S. Radford, MMA Website) show lowest opacity and lowest path length fluctuations occur in the Chilean winter. The diurnal variations of r.m.s. phase reported for the site by Holdaway et al. (1996 NRO Technical Report 51) show that best observing is at night. These two conditions correspond to the 25% quartile of the total time. For this good sub-mm time, the prediction is that the anomalous refraction contributes 0.26" to the short-timescale error for a 12-m antenna pointing at elevation 50^o (Holdaway, MMA Memo 186). Thus the anomalous refraction effect is irrelevant for observing under good sub-mm conditions.

Dynamic Range in Fast Mosaics is NOT a reason to avoid large dishes. Holdaway (MMA Memo 178) recently modeled the effects of pointing errors on fast mosaic maps at 230, 345, 490, and 650 GHz, and concluded that dynamic range degraded with dish size. We think this analysis is incorrect for several reasons. Firstly, the modeling was done prior to the meetings of the LSA/MMA antenna group, and assumed pointing errors which were higher than the current estimates. Secondly, the source models were unrealistic, with a peak source flux of 60 Jy/beam at 1.3 mm, and constant brightness temperature as a function of frequency for the extended emission -- i.e., the model sources were assumed to be blackbodies, with extents up to 1 arcmin. Thirdly, the lowest contours in the modeling are set by the errors induced by the imaging, rather than the system noise. Scaling from the noise estimates (Table 1) in comparable integration times as in Holdaway's analysis shows that the system noise --- even with receiver temperatures at four times the quantum limit --- makes it impossible to detect realistic extended emission, for any galactic or extragalactic observing project to the dynamic range levels listed in Memo 178. The only exception is at 3 and 1.3 mm, where there is anyway no problem with the mosaic's fidelity. However at all shorter wavelengths, and in all longer configurations of the array, there is no such thing as high-dynamic range imaging for these projects.

If needed, the big dishes can be under-illuminated at the highest frequencies. We are confident that the pointing specification of 1/30th of a beam at 300 GHz can be met, if the engineering groups are allowed to design and build a 12 m antenna. Even if they couldn't meet this spec, however, it would always be possible to meet this criterion, if necessary, by slightly under-illuminating the dishes at the highest frequencies (which will also reduce spillover). We could easily make a 12 m dish be a 10 m dish at 800 GHz, just by increasing the taper in the 800 GHz receiver's horn-lens combination by a factor of 1.2 (Table 1). The dishes could then operate as 10 m dishes, if needed, at 800 GHz only, and operate as 12 m dishes in all the other windows, where the pointing specs can certainly be met.

More flexibility for future expansion. We propose 64 x 12 m dishes because that matches the budget goal (Fig. 1 and Tables 2 and 3). However our longer-term goal remains a collecting area of 10000 m² , which we consider essential for the high-priority astronomical studies of both galaxies at high redshifts and protostars in our Galaxy, and we hope the array can have the possibility for upgrading. The recent NRAO conference proceedings on the future of the Very Large Array underlines the difficulty of making a major upgrade of the VLA over the past 17 years. We think the LSA/MMA will have more flexibility for future upgrades if the initial array has large-diameter antennas. If we start with 12 m antennas, then each additional 9 antennas gives another 1000 m² in collecting area. Equivalent gains with smaller dishes require many more antennas, many more receivers, and a much bigger correlator change, all of which imply a larger incremental cost for the future upgrade (Fig. 6). For example, if the budget allows 7000 m² now, and if a correlator for 90 antennas is feasible, then it would be wiser to build 64 x 12 m dishes = 7000 m² now and then add another 26 antennas later, to get an area of 10000 m 2 --- than it would be to build 90 x 10 m dishes now, and have less flexibility for future expansion. Another possibility would be to add an extra ring of panels to existing antennas. Again, this type of upgrade would be less expensive if there were fewer, larger antennas to start with.

Better fallback solutions. Should the project run into serious budget problems, then the original MMA goal of 2000 m² collecting array can be met with only 18 x 12 m dishes, instead of the original MMA number of 40 antennas. This would still be the most sensitive mm/submm array in the world. If one were forced to fall back to the original MMA number of 40 antennas, 40 receiver boxes, and the original MMA project's correlator size, then 40 x 12 m would yield a collecting area of 4500 m² , i.e., 2.2 times more sensitivity than in the original MMA concept. The Zero-spacing problem can be solved by other means. The lack of information at the very center of the (u; v)-plane will be a problem with any kind of interferometer array. Even with 8 m dishes, the zero-spacing problem would be serious at 800 GHz, limiting detections to sources of size < or equal to 3" . We think many of the zero-spacing problems can be settled with multi-beam detectors on a single dish of the same size as the array antennas. Rather than building 90 or 128 small dishes just to better fill the hole in the (u; v)-plane coverage exactly at zero spacing, it would be better to have fewer and larger dishes in the array, and take care of the zero-spacing hole with a single dish equipped with multi-beam receivers, at least for the sub-mm continuum. A good solution would be to have an extra antenna from the same production line as the array antennas, with the same dish diameter. The imaging process, incorporating results from a same-diameter single dish together with the interferometer measurements is described by Cornwell, Holdaway, & Uson, 1993, A&A, 271, 697. The single dish would profit from recent progress in mm/submm multi-beam bolometers, such as SCUBA at the JCMT, and SHARC at the CSO. The 128-beam BOLOCAM proposed by Caltech, JPL, and UMass might be ideal for this purpose. Both the SCUBA and BOLOCAM arrays cover the 1 mm, 0.87 mm, 0.45 mm, and 0.35 mm bands just by changing filters.
Such a continuum multi-beam system, on the same site and scheduled by the same organization that runs the interferometer array, would let observers obtain the zero-spacing maps --- if needed --- at about the same time and in the same excellent atmospheric conditions as for the interferometer data. It would be more practical for operations and maintenance, if only the antenna with the multi-beam bolometer would equipped with a nutating subreflector, rather than every antenna in the interferometer array. This proposed solution still needs to be checked by simulations, for mm/sub-mm source models with realistic brightness temperatures and realistic system noise.