1. Introduction
A number of considerations make large interferometric arrays operating at millimeter and submillimeter wavelengths particularly suitable for the study of the young universe. First, such arrays can be constructed with large collecting areas, giving them high sensitivity for the detection of faint emission from distant objects. Second, interferometers provide the high angular resolution required for the study of distant objects. Third, these arrays yield images of the sky, with the positions of imaged objects determined to high accuracy. Finally, the arrays are capable of spectral line imaging, yielding the kinematics of objects observed. This powerful performance combination has prompted three such arrays to be proposed: The U.S. Millimeter Array (MMA), the European Large Southern Array (LSA), and the Japanese Large Millimeter/Submillimeter Array (LMSA). This paper presents the MMA---its performance specifications together with high-z science examples. Discussions are being held between the European Southern Observatory (ESO) and the National Radio Astronomy Observatory (NRAO) on the feasibility of merging the LSA and the MMA. Furthermore, discussions are being held with the Nobeyama Radio Observatory on the possibility of joint operation with the LMSA. If these discussions prove successful, the sensitivities quoted for the MMA could be increased by factors of three to five.

2. Millimeter Array
The MMA consists of forty antennas of 8-meter diameter operating efficiently (surface accuracy is 25 µm) in all the atmospheric bands from 40 to 850 GHz (7 mm to 350 µm). The combined collecting area is 2000 m² . The antennas have a pointing accuracy of 1 arcsecond rms in a median wind speed of 6 m/s. The drive system can change antenna position by 1.5 degrees in one second with an accuracy of 3 arcseconds rms, a fast-switching capability imposed by the need to rapidly and repeatedly calibrate the phase of the array. The primary beam is 40 arcseconds at 230 GHz (1.3 mm); the pointing accuracy is one-tenth the primary beam at the highest frequencies. The reflecting panels are machined aluminum mounted on a carbon fiber epoxy backup structure. The mounts are of steel and the antennas are transportable by a rubber-tired vehicle.

The forty antennas can be distributed in four configurations with maximum baselines ranging from 70 m in the most compact array to 3 km in the most extended. The largest (3 km) configuration gives an angular resolution of 0.1 arcsecond at 230 GHz, other configurations and frequencies scaling accordingly. The receivers required to cover all the atmospheric bands will be housed in a cryogenic dewar at the Cassegrain focus and will operate at the follow- ing frequencies: 30-48, 67-95, 94-133, 132-183, 182-236, 235-307, 305-400, 397- 517, 602-720, 787-950 GHz. The lower bands will be transistor amplifiers (InP HFETs) and the rest will incorporate SIS junction mixers. All receivers will be dual polarization. The performance goal is twice the quantum limit, that is, the receiver noise temperature is to be 2hv/k = 21 K at 230 GHz, scaling accordingly for other frequencies. Each polarization is to have a bandwidth of 8 GHz.

A forty-antenna array has 780 independent baselines. A conventional lag correlator is being considered with 1024 lags (spectral channels) for 2 GHz of bandwidth. The correlator output will consist of 12.8 Mbytes of data per integration. Integrations as short as 32 mS are being contemplated; at that dump rate, observations of a few hours will yield Tbyte data sets.

Operation is to be remote, with no observer presence at the array site required. This means that the software must be end-to-end integrated, from proposal to observing program to first-order image and final image. Data is to be archived. AIPS++ will be the data analysis package. More detailed information on the MMA can be accessed at: http://www.mma.nrao.edu.

3. MMA Site
The array requires a site with outstanding atmospheric transparency and stability. The NRAO has conducted a careful and thorough search for such a site, beginning with studies of two sites in the continental United States, then a study of Mauna Kea (at an elevation of 3720 m where an array with 3 km baselines could be placed), and finally in the desert of northern Chile. A superb site has been located in Chile, one that rivals the South Pole in its atmospheric transparency and Mauna Kea. The Chile site also offers improved phase stability compared with Mauna Kea.

Complete site testing data can be accessed at http://www.mma.nrao.edu.

4. MMA Performance - Continuum and Spectral Line
With the receiver noise and bandwidth specifications, and considering the median conditions at the Chile site, the MMA point source sensitivity for continuum radiation is ~0.1 mJy (1 ) in one minute of integration, for all bands except the submillimeter, where it increases to 0.5 mJy at 650 GHz and higher for 850 GHz. For spectral line observing, one need only scale the bandwidth of 8 GHz to the spectral resolution desired. At 25 km s^-1 resolution at 230 GHz, which might be typical of an extragalactic observation, the sensitivity is ~3 mJy (1 ) for a one minute integration in the millimeter bands, and three times that at 650 GHz. For 1 km s^-1 resolution, more appropriate for a galactic molecular cloud observation, the numbers are five times larger. These are high sensitivities and present the opportunity to do remarkable science.

5. High Redshift Science with the MMA
The following examples are only intended to be illustrative of the power of the MMA. The actual science done with the MMA will surely include programs in the area of these examples, but there will be much more that utilizes the power and flexibility of the MMA in ways that cannot now be imagined.

The first example is the observation of continuum emission from dust in infrared luminous objects at high-z. Such objects have been intensively studied since their discovery as a class by IRAS. And these studies have been carried to high redshifts with existing facilities. The power of a large array like the MMA for the detection of such objects is unmatched by any facility, existing or proposed, on the ground or in space.

Consider the IR-luminous source Arp 220. Figure 1 shows the spectrum of Arp 220 at various redshifts from z=1 to 20. As the source is moved to higher z, the peak of the spectrum is redshifted toward the frequency being observed, compensating for the increased distance of the source. If you can detect the source at z=1, the ``K-correction'' will allow you detect it at z=20, provided you are observing at millimeter wavelengths. This happy circumstance was first noted by Brown (1990) in the proposal for the MMA, and has been developed by others as well. See, for example, Franceschini et al. (1991), and Blain and Longair (1993, 1996). The MMA is certainly capable of the detection, with a sensitivity of 0.1 mJy (1 ) in a one minute integration. Do such sources exist at high redshift? Extrapolations of the number counts in the IRAS catalog suggests that they might exist in high numbers. In Figure 2, where we have adapted the results of Franceschini et al. (1991), we see that for the MMA bands at 1.3 mm and 0.8 mm there are roughly three to four sources per square arc-minute.

Figure 1. Red-shifted spectrum of Arp 220.

Figure 2. Predicted source counts of infrared-luminous galaxies for various wavelengths shown in Żm.

detectable at the 5 level in a 25 minute integration. That is roughly one source per primary beam, so these sources may well be detected in every continuum observation, independent of the observer's intentions. The second example is the observation of spectral line emission from high-z objects, both in molecular and atomic fine-structure lines. First, consider the observation of the rotational lines of carbon monoxide. Six high-z CO sources have been detected (see Table 1). This has been possible with existing facilities because several, perhaps all, of these sources are gravitationally lensed, and because of the willingness of the observers to conduct long integrations. All could be detected and imaged in minutes with the MMA. The interpretation of all of these sources would be aided by higher angular resolution provided by the MMA. Images with angular resolution measured in 0.1 arcseconds rather than 1 to 3 arcseconds will provide much tighter constraints on lensing models and, in turn, on source properties.

Table 1. High Redshift CO Emission

Source Name	Redshift	Detection Telescope and Integration	Time	Detection Reference
F10214-2475	2.29	12M	17 h	Brown & Vd. Bout (1991)
H1413+117	2.56	30M	7.5	Barvainis et al. (1994)
BR1202-0725	4.69	Pd B	16	Omont et al. (1996)
BR1335-0415	4.41	Pd B	10	Guilloteau et al. (1997)
53W002	2.39	OVRO	100	Scoville et al. (1997)

These detections imply the important goal of detecting and imaging the underlying population of unlensed sources. Images in spectral line emission of IR-luminous galaxies from z=0.1 to z=5 would be a rich source of information on the formation and evolution of galaxies. The MMA is capable of doing that, and the task would be made easy by the larger collecting area that could be provided by a merger of the MMA with the LSA and/or the LMSA. Similarly, a larger (merged) array would make it easier to image normal galaxies in molecular line emission.

Carbon monoxide is not the only molecule with high-z emission detectable by the MMA. For example, detections of CO, HCN, HNC, H₂CO, and HCO⁺ in absorption in a number of sources to redshifts of z~0.9 have been reported. See, for example, Combes and Wiklind (1997), Gerin (1997), and Menten (1996, 1997). Extensions of these observations to higher redshift and to emission lines can be contemplated with the MMA. Finally, the atomic fine-structure lines of oxygen and nitrogen are candidates for observation with the MMA. The most important of these are the CII line at 1901 GHz (158 µm) and the NII lines at 1460 GHz (205 µm) and 2459 GHz (122 µm). Initial observations with the James Clerk Maxwell Telescope and with the Infrared Satellite Observatory have been disappointing in that the high-z CO sources appear to be much weaker in CII emission than would be predicted by the observed ratio of this line to CO in the Milky Way. However, independent of ease of observation, the lines are 5  potentially very important diagnostics of the molecular clouds in these objects: star formation, cloud heating and cooling, and metallicity.

6. Costs, Schedules, and Partnerships
The estimated cost of the MMA is $200,000,000 (1997$). The project is to proceed in two stages: a development phase of three years, followed by a con- struction phase of six years. The funding for the first year of development, $9M, has been appropriated by the U.S. government and is expected to be available from the National Science Foundation (NSF) in early 1998. NSF has urged the NRAO to seek domestic and/or international partners who would join the MMA project. At the same time, as was mentioned above, the NRAO is discussing a merger of the MMA with the LSA. This merger envisions an array much larger than the MMA, of say 7000 m² collecting area but only twice the total cost. This larger array would have more than three times the sensitivity, cutting integration times by a factor of ten, compared with the MMA alone. Joint operation with the LMSA would increase the collecting area to at least 10,000 m² , with a corresponding increase in sensitivity.

References
Barvainis, R., et al. 1994, Nature, 371, 586.
Blain, A.W. , and Longair, M.S. 1993, MNRAS, 264, 209.
Blain, A.W. , and Longair, M.S. 1996, MNRAS, 279, 847.
Brown, R. 1990, The Millimeter Array, proposal to NSF.
Brown, R., and Vanden Bout, P. 1991, AJ, 102, 1956.
Combes, F., and Wiklind, T. 1997, ApJ, accepted.
Franceschini, A., et al. 1991, A&AS, 89, 285.
Gerin, M., et al. 1997, ApJ, 488, L31.
Guilloteau, S., et al. 1997, A&A, 328, L1.
Menten, K., and Reid, M. 1996, ApJ, 465, L99.
Menten, K., et al. 1997, Highly Redshifted Radio Lines, NRAO Workshop, October 9-11, Green Bank, WV.
Omont, A., et al. 1996, Nature, 382, 428.
Scoville, N.Z., et al. 1997, ApJ, 485, L21.