The case for a bolometric millimetre
camera at the IRAM 30-m telescope
F.-X. Désert 
and A. Benoit 
Laboratoire d'Astrophysique, Observatoire de Grenoble BP 53,
414 rue de la piscine, F-38041 Grenoble Cedex 9 France,
e-mail: Francois-Xavier.Desert@obs.ujf-grenoble.fr
CRTBT-CNRS, 25 ave. des Martyrs, BP 166, 38042 Grenoble Cedex 9
France,
e-mail: benoit@polycnrs-gre.fr
Received Jan. 13th, 1999; accepted Jan. 21st, 1999
We describe here the important astrophysical results that could be
obtained by using large format (say 
) bolometric
detectors at 1 and 2 mm with the IRAM 30m telescope: having a
confusion-limited 1 mm extragalactic survey containing a large fraction
of high redshift objects, mapping star formation regions in our
Galaxy at 1 mm, and mapping the Sunyaev-Zeldovich effect at 2 mm in tens of
high-redshift clusters. We also show a first optical implementation and
the key points of this project.
Franceschini et al. (1994, see Burigana et al. 1997), Guiderdoni et al. (1998), Blain et al. (1998a, 1998b) have given number count estimates at the sub-mJy level for a wavelength of 1.3mm, for various galaxy evolution models. It seems that an episode of high rate of star formation is required at high redshift to explain both
m on the
JCMT (Hughes et al. 1998, Eales et al. 1998).
above a flux of 1 mJy at 1.2 mm. This corresponds to one
source per 30 diffraction beams. So a deep survey with the IRAM 30m
should aim at confusion-limited maps with a noise per beam of
0.2 mJy. With an estimated sensitivity of 50 mJy.s 
(see
below), this means that the camera field must be observed for 13
hours, to reach that level. In 100 hours of integration, the number
of detected sources with flux above 1 mJy (at the 
level) can
be expected to be about 70. That would be a major breakthrough in
order to study the statistics of this population, even allowing for a
factor 2 uncertainty in these numbers.
If SCUBA is already finding this population, why should we try to do this in the atmospheric window at 1.2 mm? The answer lies in the now famous positive K-correction that happens for high redshift objects. If SCUBA has a rather strong redshift selection around 3, one can expect a deep 1.2 mm survey to be biased towards redshift 5 objects. Hence, we would probe the evolution of the Universe at large redshifts, for which we know next to nothing. The large collecting area and high angular resolution of the IRAM 30m telescope would give us a substantial advantage in the search of primeval galaxies. At these wavelengths, the galactic cirrus contamination is much less than in the submillimetre domain, because high redshift objects look colder than the high latitude cirrus clouds.
Millimetre interferometers cannot achieve this mapping speed because their field of view is much smaller. Competition with the future LSA/MMA for surveys has to be carefully studied in this research area.
Surveys at 2.1 mm could be quite important as well; see a first BIMA
attempt by Wilner & Wright (1997). The confusion limit would be reached at
0.5 mJy (0.4 galaxies per arcmin ) in probably less time than at
1.2 mm ( in a few hours). But this is very much dependent
on the assumptions about the very high redshift Universe (z between 4
and 10).
Blank sky surveys should be done in areas where many complementary data have been accumulated. Obviously the HDF, CFRS and deep radio survey fields are prime targets. Mapping fields around clusters seems also a very powerful technique to observe the high redshift Universe, as done with SCUBA by Smail et al. (1997).
The gain in mapping speed will provide much more information on the cold clouds at the origin of the star formation in our Galaxy and nearby ones but also it will allow to probe the evolution of the circumstellar material around single and multiple young stars.
This is particularly true for some crucial subjects which are today strongly limited by the sensitivity of current bolometer arrays. Among them, one can present here a few major topics:
) of the
proto-stars (Class 0 and Class I objects). These mappings are fundamental
because they allow the determination of the total amount of mass
surrounding protostars, contrary
to mm interferometers which resolve out the extended envelope surrounding
these objecs (Gueth et al., 1997).
At a wavelength of 2.1 mm, it seems that the measurement of the
Sunyaev-Zeldovich effect is the least affected by radio sources and
dusty galaxies (see the review by Birkinshaw 1998). Mapping the SZ
effect with a comptonisation parameter y sensitivity better than
per diffraction beam (20 arcsec FWHM)
in the core of clusters would be possible in only ten hours. This
would be a factor 10-100 increase in mapping speed as compared to
current bolometer experiments like SuZie and Diabolo.
This might be crucial
for the follow-up of XMM observations (made with a beam of 15 arcsec)
of clusters of galaxies. The other Cosmic Microwave Background (CMB)
anisotropies at small scales that are and will be detected by other
experiments (the Ryle Telescope and the VLA) could receive an
independent confirmation-validation at these clean wavelengths.
Sensitivity would be the same as quoted above ( 
units).
Millimetre interferometers cannot achieve the sensitivity quoted above
for extended sources because of large antennas and the lack of short
spacings.
So far, the mapping speed improvements came by adding single elements
together. The empirical limit seems to be reached at typically 100
elements. It is limited by the workload (the patience of technicians
and engineers: ask SCUBA people for example) of putting things
together and by the homogeneity of the array. In general, the worst
pixels are pulling down the overall sensitivity of the instrument.
Several recent developments in bolometer technology have made
integrated arrays possible. Four projects are in various stages of
completion: BOLOCAM is an East Coast+Caltech project (Mauskopf & Bock
1998) of 150 integrated 300 mK silicon nitride spider-web pixels with
cones separated by one diffraction size to be put at the CSO first and
then on the future 50m (at 1 and 2 mm). SHARC (and further) is an
operational camera (made by Moseley et al. at NASA-GSFC) of a 24
pixel single line that can be stacked to others in the future, and
that works at a temperature of 300 mK and a wavelength of 450 m
at the CSO (Wang et al. 1996). In France, the CEA-LETI-Grenoble (P.
Agnese) is developping a square array as the baseline
for the SPIRE bolometer instrument onboard FIRST (200 to 500 m).
It uses the Silicon chip making process to make a fully integrated
array. Another development is with the NbSi thin layers by the
IN2P3-CSNSM-Orsay (L. Dumoulin).
So if these technologies are available in Europe, what could be the best use of them in the millimetre domain? In what configuration? We argue here that to make full use of the multiplex advantage, the cone for each pixel must be dropped (as is now planned for SPIRE). A cone optimises the f/D ratio and hence minimises the pixel size. In case of lenses, the pixel scale at the focal plane is necessarily larger. A cone also clearly defines the entrance acceptance angle, effectively defining the pupil and reducing sidelobes. So, why dropping the cones? First, the new bolometer technology allows larger pixels without loss in sensitivity (heat capacity is reduced by making thinner bolometers). Then, by using a cold pupil common to all pixels, one can still prevent most of the sidelobes and heatload on the detector.
Moreover, additional problems arise from cones that can be solved by using an appropriate filled array. The most efficient (straight instead of parabolic) cones, as in the best known examples (37 bolometer MPIfR and SCUBA), are packed at a spacing of only twice the diffraction size on the sky, thus mapping at a time, a fraction of 1/4 of the available sky. The sky map must be filled with a drizzle technique using 16 different positions to have a fair sampling. This is a likely source of noise, because the map is not fully acquired at the same time. Another matter of concern is the anomalous refraction which is known to happen at Pico Veleta and at the JCMT. Even a strong source has an apparent jitter in front of a detector, giving a so-called source noise. Calibrations and photometrical measurements are thus more difficult. When reaching the confusion limit, anomalous refraction may be a strong limitation. Therefore, it seems that a cleaner and more efficient solution is to have a filled array of pixels which not only covers the largest possible area of the sky, but also samples correctly this area (i.e. at half the diffraction-size per pixel). This is the current basic design of most infrared cameras. The SHARC experiment is already designed this way.
So far the available arrays are modulated with a wobbling secondary. A total power readout technique could alleviate the use of a wobbler. This is already in use by small bolometer arrays: SuZie, NOBA, and Diabolo at POM2. In the case of a large array, the most promising observing technique is to fix the telescope in local coordinates ahead of the target and let the sky drift with the diurnal motion. Local effects and flat field can thus be disantengled from the real sources.
Figure 12: A schematic optical layout
Figure 12 shows a possible optical layout of the bolometric camera at 1.2 mm. It uses one warm lens (assumed here in polyethylene with a n=1.47 index of refraction) at the 30m focal plane and one cold lens at the pupil image of the secondary. This cold pupil lens closes the 1.6K box. Note that the pixel size is here 2.6 mm (i.e. larger than the operating wavelength) and that is samples half a diffraction size. Filters (not shown) have to be placed at the cold pupil or just in front of it at 4 K or higher. The camera at 2.1 mm may require bigger pixels, hence may be 16 by 16 pixels wide. The field of view
Table 4 gives a conservative estimate of the expected sensitivity.
For that, we assume a very mediocre state of the atmosphere
and the telescope: atmospheric opacity (at the measurement elevation)
and temperature of resp. 0.4 at 1.2 mm and 250 K, telescope emissivity
and main beam efficiency of resp. 0.1 and 0.25 at 1.2 mm and 0.50 at
2.1 mm. The filtering is assumed to have an overall transmission of
15 percent in a 
bandwidth. The box
enclosing the detector must be kept at 1.6K to avoid overloading the
detectors. Most of the photon noise is due to the atmosphere, and not
to the telescope. The same calculation adapted to the present
37-bolometer array and Diabolo experiments give sensitivities which
are slightly above what has been obtained on the sky. The needed
detector sensitivity can be achieved with the present technology, on
single bolometers, especially with relatively slow time constant. This
sensitivity of arrays should be coming soon. Cooling the detector to
0.1 K might be advantageous in this respect.
| Characteristics | units | 1 | 2 |
| Wavelength | mm | 1.2 | 2.1 |
| Heat load | pW/pix | 23 | 2.5 |
| Photon noise |  | 17 | 5 |
| Assumed Pix. noise | | 10 | 5 |
Point Source 1  , 1s. | mJy | 50 | 25 |
We list here several open issues that should be dealt with before designing such an instrument.
Having a truly mapping millimetre instrument would bring the same qualitative changes as we saw 15 years ago when IR cameras arrived at the telescopes and replaced single element detectors. The modern submillimetre instruments are near or at the confusion limit in extragalactic and galactic environments. Data acquired with arrays having cones may be very hard to exploit. A true camera has a potentially large multiplex gain and cleaner behaviour at the confusion limit. The IRAM 30m user community clearly has to discuss the various options before attempting to build such an instrument. We think the challenge is really worth the efforts and that the time is ripe to start a definition study. We here suggest to continue pre-design studies and then build the instrument which could be soon fitted with prototype detectors of 5 by 5 or 8 by 8 but which would also be compatible with future 32 by 32 bolometer arrays.
We thank P. Agnese, L. Dumoulin, A. Dutrey, S. Guilloteau, J.-M. Lamarre and B. Lazareff for many helpful discussions.
Blain, A.W., Ivison, R., & Smail, I. 1998, astro-ph/9710003 v2
Blain, A.W., Smail, I., Ivison, R.J, & Kneib, J.-P. 1998, astro-ph/9806062
Birkinshaw, M., Physics Reports, submitted, astro-ph/9808050
Burigana, C., Danese, L. et al., 1997, MNRAS, 287, L17
Eales, S.A., Lilly, S.J., Gear, W.K., et al., 1998, ApJL, submitted
(astro-ph/9808040)
Franceschini, A., Mazzei, P., De Zotti, G., Danese, L., 1994, ApJ, 427,
140
Gueth et al., 1997, A&A 323, 943
Guilloteau, S. & Dutrey, A. 1998, A&A 339, 467
Guiderdoni, B., Bouchet, F.R., Puget, J.-L., et al., 1998, Nature, in press, a
stro-ph/9706095
Guiderdoni, B., Hivon, E., Bouchet, F.R., & Maffei, B., 1998, MNRAS,
295, 877
Hauser, M.G., 1998, ApJ, in press (astro-ph/9806167)
Hughes, D.H., Serjeant, S., Dunlop, J. et al., Nature 394, 241
Lagache, G., et al, 1998, A&A, in press
Mauskopf, P., & Bock, J., 1998,
http://www-lmt.phast.umass.edu/ins/continuum/bolocam.html
Motte, F. et al. 1998, A&A , 336, 150
Puget, J.-L., Abergel, A. et al. 1996, A&A, 308, L5
Smail, I., Ivison, R. J., & Blain, A. W., 1997, ApJ, 490, L5
Wang N., Hunter, T. R., Benford, D. J. et al., 1996, Applied Optics,
35, 6629
Wilner, D. J., & Wright, M. C. H., 1997, ApJL, 488, L67
