You are hereThe Multiple Mirror Telescope
The Multiple Mirror Telescope
Telescopes for The 1980s. pp. 63-128
Copyright © 1981 by Annual Reviews Inc. All rights reserved
THE MULTIPLE MIRROR TELESCOPE
Jacques M. Beckers and Bobby L. Ulich
Multiple Mirror Telescope Observatory,1 Tucson, Arizona 85721
Robert R. Shannon
Optical Sciences Center, Tucson, Arizona 85721
Nathaniel P. Carleton, John C. Geary, and David W. Latham
Smithsonian Astrophysical Observatory, Cambridge, Massachusetts 02 138
J. Roger P. Angel, William F. Hoflmann, Frank J. Low, Ray J. Weymann, and Neville J. Woolf
Steward Observatory, Tucson, Arizona 85721
On May 9, 1979, the Multiple Mirror Telescope, or MMT, was dedicated. The construction and tuning of the MMT is now complete except for the automatic coalignment of the six telescopes which remains to be fully implemented. The six telescopes can be aligned very well manually by the observer so that a major program of astronomical observing has been going on at the MMT since the dedication date.
The MMT was conceived in the late 1960s when scientists at the Smithsonian Astrophysical Observatory and at the University of Arizona decided it would be possible to use six existing 1.8-meter “egg crate” mirrors to construct six parallel Cassegrain telescopes whose images were to be combined into one, by so-called beamcombining optics. In this way they could effectively create a telescope with a combined collecting area equivalent to that of a 4.5-meter-diameter single mirror, thus making it, in that sense, the third largest optical telescope in the world. The need for such a large collecting area motivated the final decision to construct the MMT. The telescope is designed to be optimal for both optical and infrared astronomy of faint and low-contrast objects. Its large edge-to-edge distance in the primary mirror assembly (6.9 m) creates, in addition, the opportunity to do very high resolution imaging, both at inlrared and optical wavelengths using, for example, Michelson and speckle interferometry.
The MMT is located in Arizona on the top of Mount Hopkins ( I l O O 53‘ 04”.4 longitude, 31 O 41’ 19”.6 latitude) at an elevation of 2600 meters, a site selected because of its general location in the astronomically active Tucson area, its good seeing and photometric properties, its reasonably high altitude making it a dry and relatively dark site, and its large number of clear nights. Figure 1 gives a general overview of the Mount Hopkins
site. The telescope is located on the relatively small, flat top of the rather sharply peaked mountain. This places it effectively high above the general surroundings. In addition to the design of the building and telescope itself, the location is, we believe, the source of the very good image quality obtained with the MMT. With our limited observing experience to date the average image quality is about one arcsecond (full width at half maximum of seeing disk) with images of 0.5 arcsecond being recorded rather frequently.
Figure 2 shows a closeup of the telescope itself. It is supported by an altitude-over-azimuth mount, which, under computer control, can track objects at both sidereal and nonsidereal rates. The telescope axes use ballbearing supports rather than the conventional pressurized hydrostaticbearing support. They are driven by a torque motor-pinion gear arrangement rather than by the traditional worm gear. The alt-azimuth mount cradles the optical support structure (OSS) which contains the six 1.8-m Cassegrain telescopes, and the beamcombining optics, as well as the 76- cm Cassegrain guide/alignment telescope, and which was designed to provide the required stiffness at a minimum cost in weight. The telescope and its mount is located within a building that rotates with the telescope azimuth. This building, in addition to housing the telescopes, contains the telescope control room, the data collection and analysis room, workshops, a conference room, and offices.
The optical paths in the MMT are shown in Figure 3. Each Cassegrain telescope has a f/2.72 1.8-m-diameter primary mirror. The secondary mirror changes the beam to a f/31.6 focal ratio. The beam is then transferred via a flat tertiary mirror and beamcombiner to the MMT’s quasi- Cassegrain focus located on the central axis of the OSS. The image scale there equals 3.6 arcseconds/mm. The steepness of the facets of the beamcombiner determines the “final f-ratio” of the MMT, which is defined by the envelope cone of the six converging beams. The angle of this cone and the size of the unvignetted field of the MMT are related. For the existing f/9 beamcombiner, this field equals 52 arcseconds. Vignetting sets in slowly, however, so that the MMT has a “useful field” (< 50% vignetting) of 4 arcmin. A “faster” beamcombiner will increase the field of view at the cost of a lower filling factor of the incident light cone and of a somewhat larger image deterioration due to the larger image plane inclinations. Because of the polarization effects caused by the tilted tertiary and beamcombiner mirrors, the MMT as it is now is not suitable for polarimetry, nor can it be used well for interferometry using other than opposite mirror pairs.
The MMT presents, in many ways, a major departure from traditional telescope technology. This departure, risky as it may have originally appeared, has resulted in a relatively low-cost facility and a new technology that will undoubtedly change the ways telescopes of the future will be built. The most significant departures from conventional optical telescope technology are (a) the use of tight-weight “egg crate” mirrors, which resulted in substantial weight savings of the telescope, (b) the use of an alt-azimuth mount, which simplifies the gravitational effects on the structure, thus again resulting in a greater economy. Ah-azimuth mounts have, of course, been used elsewhere, as in the USSR 6-meter telescope and in radio telescopes. The MMT mount represents, however, a major advance in precision and speed of tracking, (c) the use of a ball-bearing support rather than hydrostatic bearings resulting in cost savings and less maintenance, ( d ) the use of spur gear drives rather than worm gears, and (e) the use of multiple coaligned light collectors rather than a single monolithic mirror. The MMT has proved that all of these new concepts are viable, although many engineering refinements are still to be made, as is the case in all new telescopes. Observing runs began o’n the MMT in 1979 and already the telescope has proved itself to be a very good astronomical facility, producing major scientific returns for its sponsoring institutions.
The MMT was constructed jointly by the Smithsonian Institution and the University of Arizona on the basis of a Memorandum of Understanding signed on December 23, 1971. Manpower and funding were provided by both institutions to complete the telescope. The NSF provided grants for specific tasks related to the construction of the telescope and its auxiliary instrumentation. After the dedication of the MMT in May 1979, the two sponsoring institutions, on July 1, 1979, entered into a new Memorandum of Understanding for the operation of the MMT, creating the Multiple Mirror Telescope Observatory (MMTO). The MMTO is charged with the responsibility of operating and maintaining the MMT facility, as well as with the completion of a number of remaining engineering tasks. It has an engineering, technical, operations, and administrative staff of about twenty, and an annual budget of about one million dollars which is divided equally between the two sponsoring institutions with the University of Arizona deriving part of its contribution from an NSF grant. The major unfinished engineering task relates to the automatic coalignment of the six telescopes. Presently the telescopes can be successfully coaligned manually to sub-arcsecond accuracy. The automatic laser coalignment system can maintain this to 1 arcsecond rms. The latter is, however, not good enough, especially because of the exceptionally good seeing often encountered on Mt. Hopkins, which requires maintaining the coalignment to a fraction of the 0.5-arcsecond seeing disk. Continuous updating of the coalignment on star images themselves, as well as improved laser coalignment, are both being pursued.
This and other engineering efforts proceed at the same time that a substantial astronomical observing program is in progress. Up to 75% of the nights available in a month have been used for astronomical observing. The MMT is of sufficient complexity to require a staff of telescope operators to run the telescope and to assist the observing scientist in the acquisition of data. It is particularly suited for deep sky spectroscopy and imaging, for high resolution spectroscopy, for infrared observations, for precision photometry of faint objects, and for optical and infrared interferometry. Section 7 of this paper, written by R. Weymann, elaborates on the science done and to be done on the MMT. Other sections (with individual authors indicated) will deal with the history of the MMT [Sections 2 (Beckers), 3 (Woolf), and 4 (Weymann)], with the design and performance of the MMT [Section 5: 5.1-5.3 (Carleton), 5.4 (Shannon, Beckers). 5.5 (Hoffmann). 5.6 and 5.7 (Low). 5.8 (Ulich), 5.9 (Carlton. Beckers)], with MMT instrumentation [Secgion 6: 6.1 (Beckers), 6.2 (Angel, Latham), 6.3 (Low), 6.4 (Geary)], and with concepts for future multiple objective telescopes [Section 8 (Woolf)] .
2. EARLY MULTIPLE OBJECTIVE TELESCOPES
The history of astronomical telescopes is to a large extent a history of the manufacture of ever larger and larger telescope objectives. The size at which lenses and mirrors could be manufactured has been the major factor in determining the size of astronomical telescopes in the past. Because astronomers have always wanted telescopes with larger and larger collecting area to probe weaker and weaker objects, the idea of combining the light from a number of smaller astronomical objectives into one, with the goal of creating an image with the brightness of an effectively larger telescope, must have occurred to many in the past. The astronomical literature contains a number of references to concepts and implementations of such multiple objective telescopes. Some of these have already been reviewed by Ingalls (1951), Jacchia (1978), and Green (1979).
Table 1 summarizes the literature describing devices that were actually constructed and tested. Some of these consist of arrays of single mirror segments, which combine to form effectively a single objective in the sense that a single image is formed (Type A). Others, like the MMT, form initially separate images, which are combined by means of relay optics or electronics into one final image (Type B). Table 1 excludes multiple aperture devices aimed at interferometric studies only. By far the greatest early effort was that by Guido Horn-d’Arturo who, as the director of the Bologna Observatory in Italy, pursued the construction of the so-called “tessellated mirror” from 1932 to 1953 with major interruptions caused by the Second WorJd War. His was a zenith-pointing, fixed telescope 1.8 meters in diameter with limited celestial tracking provided by the motion of the photographic plate. It was Horn-d’Arturo’s concept to place a number of these zenith-pointing telescopes, each covering 80 arcminutes in latitude, 150 km apart in geographical latitude to study a larger area of the sky. He also viewed his tessellated (or segmented) mirror as a prototype for much larger telescopes 13 and 18 meters in diameter. Although these larger mirrors remained stationary, pointing at the zenith, he proposed increasing sky coverage by using complex pointing prime focal plane optics (Horn-d’Arturo 1955).
Table 2 summarizes the literature on concepts for multiple objective telescopic devices. Although recent designs for multiple objective telescopes have received much attention, it is interesting to examine some of the pre-MMT literature. The paper by Synge (1930) is especially fascinating because it appears to be the earliest description of an MMT-like device (Type B). His description of ideas for the optics and the support arrangement for the optics in fact resembles the MMT so closely that the
MMT, although conceived independently, could be viewed as the realization of the Synge concept. The paper also points out the weight (and hence cost) advantage of such a telescope (“the weight of such a telescope would tend to increase as the square of the equivalent aperture, while that of an ordinary telescope would increase as the cube”). The Fastie (1961, 1967) articles are also of interest because they describe an optical configuration actually now used to couple the MMT to the MMT spectrograph (see discussion in Section 6.2). By aligning the star images of the multiple mirror device along the spectrograph slit, and by aligning the principal axes of all incident light cones by means of an array of small refracting prisms, the large aperture (4.5-m) telescope can be coupled to a spectrograph which has the higher spectroscopic resolution of a spectrograph designed for a 1.8-meter telescope, without loss of light-gathering power.
Most of the design references listed in Table 2 are very recent. Multiple objective structures are considered in almost all concepts for future large telescopes, not only because the size of mirrors that can be constructed is limited, but also because the weight and cost of multiple aperture telescopes is lower. We may expect the length of Tables 1 and 2 to grow rapidly in the future.
3. EARLY HISTORY OF THE MMT PROJECT
After the pioneer work of Synge and Horn-d’Arturo on the two types of multiple objective telescope, there was a lull while technology developed to the point that precision telescopes of this kind became practicable. It was, indeed, an attempt to develop a large telescope of Type A by P. Connes that led to the reinvention of the MMT.
In the late 1960s, Frank Low at the Lunar and Planetary Laboratory (LPL) of the University of Arizona had developed techniques for the detection of faint objects at lO-Fm wavelengths using a 1.5-meter telescope. He believed then that he was successful in part because he used a moderate-size aperture, and he wondered how to increase sensitivity still further. While visiting the Connes telescope in 1967, he decided that, although there would be thermal problems associated with the approach, the general process of adding mirrors was appropriate, and one could operate a number of 1.5-meter telescopes with a cQmmon focus. Then, in 1968 when he had returned to the US, he discussed this possibility with Aden Meinel at the Optical Sciences Center of the University of Arizona. Meinel was concerned about increasing the light grasp and resolution of surveillance telescopes in space. After considering various ways of combining light, Meinel calculated the optical details of a Type B system. He also developed the methods of obtaining a phased field for such a telescope using prisms (Meinel 1970).
Low reported on the MMT concept at a meeting in Pasadena on July 28, 1969. Six months later, he also reported on it to the infrared panel of the National Academy of Sciences studying the future of US astronomy in the 1970s. During a part of this period, Robert Noyes of the Smithsonian Astrophysical Observatory (SAO) was visiting the Lunar and Planetary Laboratory.
When the Smithsonian Astrophysical Observatory set up an observatory on Mt. Hopkins in the mid-l960s, it left the summit area of the mountain clear to permit the construction of a giant telescope. After the design phase and much of the construction of the 1.5-m telescope was complete, attention was turned to the problem of making a giant telescope for the summit at a moderate cost. Fred Whipple gathered together a team consisting of Nat Carleton, Larry Mertz, and Tom Hoffman. They explored a number of schemes, each with some disadvantages. The most favored of these options was an Aricebo-like bowl made up of mosaic elements (Mertz 1970). In a phone call to Whipple, Gerard Kuiper at the Lunar and Planetary Laboratory mentioned the developing MMT concept. Meinel informed Whipple that he could obtain some surplus lightweight mirror blanks from the US Air Force, and so a number of explorations to develop a joint project were initiated.
The Smithsonian team embarked on a detailed study of the MMT concept to search for hidden flaws. Ray Weymann of Steward Observatory persuaded the National Academy of Sciences Optical Panel, of which he was a member, to support this effort. And the first mechanical studies were made of how to support the optical components. These revealed that the original mechanical concept of the telescope, while possibly appropriate for space, had too much flexure, and, indeed, that there was a problem in keeping the foci together. Weymann and Mike Reed then came up with the laser alignment scheme.
The entire joint project, however, nearly foundered over the choice of site, with some favoring Mt. Hopkins, and others favoring Mt. Lemmon. Mt. Lemmon, being higher, might be a better infrared site, while Mt. Hopkins, having a lower sky brightness, was a better site for visible wavelengths. Because of Meinel’s support Mt. Hopkins was chosen, though even after a road was started up it, the exact location of the future telescope was uncertain.
4. R. J. WEYMANN: PERSONAL RECOLLECTIONS
In the following personal recollections of some of the events and decisions concerning the construction of the MMT during the years that I was closely associated with the project, I cannot vouch for precise detail or exact chronology-many of the documents no longer exist and many of the comments were mercifully unrecorded. Nevertheless, I believe in essence the account is accurate.
For some time, the group at the Lunar and Planetary Laboratory (mainly G. P. Kuiper and F. J. Low, but also including H. L. Johnson) had discussed with A. B. Meinel, then at the Optical Sciences Center of the University of Arizona, the construction of a large MMT-type telescope to be located on Mount Lemmon. The availability of several lightweight 1.8-m blanks made this a realistic possibility. At that time the LPL group conceived of the instrument primarily as an infrared telescope. Kuiper was especially interested in its use for IR spectroscopy with the Michelson spectrometers then being developed.
My first “official” introduction to the program came after a conference on comets in April 1970 at which time Frank Low, Gerard Kuiper, and Fred Whipple described to me their concept of the project. Whipple, as Director of Smithsonian Astrophysical Observatory, had himself, along with several colleagues at SAO, been interested in approaches to a lowcost large aperture telescope and responded favorably to an invitation to join in a collaborative effort between SAO and LPL. As Director of the Steward Observatory, my reaction was that such a device offered equal promise as a telescope for optical spectroscopy and for small-field optical astronomy in general. Moreover, if the University of Arizona were serious about creating one of the leading Astronomy Departments in the country, then participation in the MMT project offered an opportunity not likely to occur again. Fortunately, the University Administration also took this point of view, and Steward Observatory soon became involved in the project.
At that time the National Academy of Sciences Survey Committee, chaired by J. L. Greenstein, was being convened, and the Optical Panel, of which I was a member, was drawing up a list of priorities. At an early meeting in Pasadena a southern copy of the Hale 5-m telescope was designated the top priority optical project. This left me disturbed. For one thing (let us be frank) it was a case of the rich getting richer, since it was envisioned that the instrument would be administered by the Pasadena group. This would squelch healthy competitron, which is often felt to be highly desirable for progress in science, especially by those without ready access to first class facilities. More fundamentally, however, it seemed a less-than-visionary project to propose for the coming decade. Bruce Gregory, acting as executive secretary for the committee, put it succinctly: “How will a duplicate of an instrument, designed and conceived in the 193Os, fare against the plans being advocated by the radio astronomers on the scale of the VLA?” With this in mind, the Panel was persuaded to hear a presentation by Aden Meinel on the MMT concept (and also one by Dick Miller on optical interferometry). The final upshot of the deliberations was that the Optical Panel and full Committee report included recommendations for construction of an MMT prototype and studies of its possible application to a large (10 m-15 m) instrument.
Meinel made an effective and enthusiastic presentation, but, unfortunately, left the clear impression in the minds of NSF representatives present that Defense Department funding was all but secured and that little help would be required from NSF, an impression that plagued us for several years. Meanwhile, several key decisions concerning the telescope were being made and the division of responsibilities roughed out. One of the early decisions was to fabricate f/2.7 primaries rather than a much slower system. Proponents of the f/2.7 design argued (correctly, I believe) that the advantage of a very compact almost “square” structure outweighed the disadvantages of more difficult figuring and the unavoidable requirement to slump the existing blanks, and the risks that process entailed.
One of the most controversial issues involved the selection of the site: Kuiper and Low of LPL strongly advocated the Mt. Lemmon site: it was developed and accessible, making construction cheaper, and was slightly higher than Mt. Hopkins, which had some slight advantage for infrared observations. SAO, of course, preferred Mt. Hopkins, because they already had a facility there. I, too, preferred Mt. Hopkins, partly because I was concerned about the effect of sky brightness on optical astronomy, but also because the realities of funding seemed to require major support through SAO and this seemed unlikely to materialize if Mt. Lemmon was chosen as the site, a view that Meinel also shared. The decision for Mt. Hopkins was, thus, a very disappointing one for the LPL group and caused their enthusiasm for the project to markedly diminish for several years.
The division of responsibility for the telescope design and construction was never a major issue: SAO personnel, especially Nat Carleton and Tom Hoffman, had considerable interest and expertise in the area of the building and the mount, whereas the University of Arizona, because of the presence of the Optical Sciences Center, took responsibility for the optics. Mike Reed, a former graduate student at Arizona with a background in electronics, was recruited by Meinel to lead the work on the alignment system. The only grey area, initially, was responsibility for the optical support structure or “tube,” but funding and engineering realities both dictated that this should fall within SAO’s domain. A formal Memorandum of Agreement, committing both institutions to a major effort in building the telescope and setting out the basic guidelines involving equal access to the telescope, was signed on December 23, 197 1. Arizona lacked a project leader to oversee the University of Arizona work on the project and, after considering several possibilities, W. F. Hoffmann, then at NASA, was selected. Hoffmann accepted the position and joined the project in early 1973. He and Nat Carleton assumed the day-to-day task of overseeing and coordinating the University of ArizonaSAO effort.
One further important design decision was made subsequent to this: in considering various structural constraints and opportunities afforded by the compact structure and limited “swing” of the tube associated with its alt-azimuth mount (an equatorial mount was never really seriously considered), Tom Hoffman and Nat Carleton suggested bringing the building structure up very close to the telescope, abandoning totally the concept of the traditional dome. They proposed enclosing the telescope with a small box-like “barn,” entirely rotating from ground level up. This idea was at first fiercely resisted, partly because of concern over restricting access to the telescope but also for reasons that now seem fairly irrelevant. While it was realized that the very open structure would minimize inside-outside air temperature differences, there was also concern that the box-like structure, coupled with the telescope proximity to the ground, might give rise to bad seeing. Happily, one of the pleasant surprises to date at the MMT has been the frequency of excellent seeing.
Funding for the telescope proved a continuing problem. Despite the Greenstein Report, there was very little sympathy at NSF for funding Arizona’s efforts on the MMT. Partly this arose from the delicate question of multi-agency funding, but partly it arose from some longstanding misunderstandings, difficulties, and personality conflicts between SAO/ SI and NSF. A “catch-22” situation developed: it was said to be necessary to first demonstrate with a small prototype the soundness of the concept, but then this approach was criticized on the grounds that what was really required was a comprehensive proposal for a full-scale MMT! In 1971 S,40 presented its plans to the House and Senate Interior Subcommittee and, with the support of the Arizona delegation, obtained initial funding in fiscal year (FY) 1972 for their portion of the project.
Arizona then turned to the State Legislature for support. At that time a proposal to request enabling legislation for Pima County to enact a lightpollution ordinance was also being considered and I had had conversations with Senator Douglas Holsclaw about this legislation. Late one evening he called me very apologetically to say that the legislation had failed to clear the Senate Majority caucus, and he added that, as a general maxim, special legislation whose full implication and background wasn’t clearly understood usually failed, even if it didn’t cost anything. This was alarming, since the implications of the MMT appropriation might not have been all that clear either and it certainly did cost something. In response to my question, Senator Holsclaw admitted that the MMT appropriation was indeed not doing very well either, and agreed that additional information might be helpful. I collected all the material I had and took it to the post office at 1 a.m.! The MMT appropriation did pass (along with the light-pollution legislation).
This event marked the first time that I felt the MMT would become a reality, or at least the University of Arizona’s participation in it, though funding continued to be, and remains, very tight. Optical Sciences had several cost overruns, and Reed had to cut many corners. In this connection, 1 would like to record the contribution that Greg Sanger made to the project. He was given day-to-day responsibility for the Optical Sciences operation while he was still a graduate student working on a dissertation; his responsibilities were later taken over by R. Parker. Despite all of the grumbling about the progress of the optics and the budget problems, the fact is that the MMT optics turned out to be very good, to the credit of both of them, as well as Dr. R. Shannon who had overall responsibility for the project at OSC.
A major criticism of the entire project was that it should not have been begun until funding was fully identified. Realistically, given the particular set of circumstances surrounding the MMT, to wait until full funding was located would have been tantamount to abandoning the project. A second criticism was that the project lacked a single leader with clear authority to make decisions, but, instead, developed a large and complicated administration. The platitude was (and is) that you “can’t build a telescope by committee.” In fact, the collaborative nature of the program, the geographical separation of the two groups, and the way the responsibilities were divided made the administrative machinery that evolved more or less inevitable, and certainly necessary.
This is not to say that errors of judgment and design philosophy were not made. In retrospect, I think we failed to draw a clear enough distinction between the secondary actuator system and the laser stabilization scheme for maintaining automatic alignment. Possibly too little attention was given to using stars for alignment, partly because too much weight was given to being able to offset guide during full moon, which, given the small field and the state of TV guiding at that time, seemed to require an internal alignment system. Some decisions such as this might have benefited from extensive outside peer review-in fact, such a review by several eminent astronomers was held on November 20, 1975, but earlier and more extensive reviews would perhaps have been useful.
In any event, the “committee criticism” has always seemed to me somewhat beside the point. In a project as complex as the MMT, the essential ingredient is a group of dedicated and competent individuals who can have their ideas carefully evaluated and their progress coordinated by a larger group whose members enjoy a certain amount of mutual trust and goodwill. The real danger of the committee structure, as I see it, is that it can become a battleground for debates revolving, not around the merits of the technical issues at hand, but around the self-esteem of groups and individuals within them.
I do not know what concepts embodied in the MMT will eventually be incorporated into the large ground-based telescope envisioned for the future in the Greenstein Report and subsequently in the Field Report, but I am convinced that the construction of the MMT has been a significant and worthwhile effort that will benefit not only the SAO and Arizona users of the instrument but the whole astronomical community as well.
5. DESIGN AND PERFORMANCE OF THE MMT
5.1 Optical Support Structure
The basic optical layout of the MMT, based upon the six 1.8-meter primary mirrors, was the driving concept in the design of its supporting structures, except for a few important early thoughts that flowed the other way. One of these was the general realization that the six-shooter layout should make the support designers’ job easier, since it provided the opportunity to locate major structural members in between the six barrels. Another was the decision, stemming from Meinel’s thinking on conventional telescopes, that the primary focal lengths should be 2.7 times their diameters, rather than some larger value that might have permitted grinding and polishing the blanks in their original flat configuration. The argument was that the savings from a more compact telescope structure, and afortiori those from a smaller dome structure, must outweigh the additional costs of optical figuring. This decision was made without a detailed cost comparison, as was also the decision to use an alt-azimuth mount. This latter decision was based not so much on cost comparison (the savings are obviously substantial) as it was a decision that the problems or field rotation would not be so onerous with the alt-azimuth mount as to require the extra cost of an equatorial mount. After these early considerations, the structural designers essentially accepted the problem of filling in support around more than 100 optical components of the main telescopes and of the guide-alignment system, as well as 2000-kg instruments plus associated guider optics and the like. Hoffman (1978, 1980) describes the engineering concepts included in the optical support structure, mount, and building.
For simplicity, we wished to confine our active-optics system to correction of the tilt and focus of the secondary mirror in each optical train. It would obviously be preferable to correct the translation and tilt of all primaries and secondaries so that their axes always coincided with an ideal hexagonal figure. The simpler scheme should be adequate, however, so long as two conditions are met: 1. the misalignments of the six individual Cassegrain telescopes (due to gravitational and thermal flexure) must be small enough to maintain image aberrations below a tolerable limit, and 2. the directions of the incoming beams, as seen at the focal plane, must not vary so much that their interaction with the optical systems of instruments would suffer errors such as vignetting.
The relation of these conditions to specific structural deflections is a complicated vector problem, but we may summarize it crudely. The growth of the images due to misalignment will be less than 0.2 arcsec if the relative displacement of primary and secondary mirrors is held below about 0.7 mm, and the angle between the primary and secondary axes is held below about 2.5 arcmin. These displacements and tilts can cause image motions in the final focal plane of up to 25 and 300 arcsec, respectively. The directions of the beams will not shift by more than 1% of their angular diameters if the primary mirrors are, in addition, held to about 2 mm of their nominal positions.
These criteria controlled the configuration and stiffness of the support structure. They also imposed the important consideration that the structural members should not differ too much in thermal response time, so that expected temperature changes of a .few degrees Celsius per hour would not warp the structure.
Another important constraint on the structure was its interaction with two servomechanical systems-those of the main-axis drives, and those of the secondary-mirror controls. For these systems to operate stably at given bandwidths, the resonant frequencies of structural modes with which the servos might interact must be kept well above the highest desired frequency of response. For the main drives, stiffness against wind buffeting and other disturbing torques was calculated to require resonances above 5 Hz or so for the main structural modes. For the secondaries, the possibility of correcting for the residuals of these major disturbances required resonances of the secondary supports above 50 Hz. The resonant frequencies of a structure are, in fact, closely related to its deflections under load, and so it turned out to be a pleasant coincidence that these two different constraints could be satisfied about equally well by a given structural design.
It was decided fairly early to use a truss structure (i.e. one with long, thin members that act principally in simple compression and extension). A structure of extended cylinders, plates, etc., was considered less suited for picking up the many localized loads of mirror cells, instruments, and the like. The materials considered were steel, aluminum, and fibre composites, with steel being chosen as the most cost-effective.
Starting from this point, the detailed design of the optical support structure demanded considerable ingenuity (particularly with an added constraint of using only standard-size steel members wherever possible), but it was duly accomplished with even some room to spare. One potentially useful feature of the design was to locate the attachment points of the primary cells so that they were not on the main load-bearing paths to the elevation-drive axles. This makes it possible to adjust the cross sections of specific mirror-hanger members so as to correct any deflections that might exceed tolerances, and to do so with relatively little effect upon other critical deflections. We have not had occasion, however, to use this feature, because the predicted performance of the structure has been so well realized in practice. We have not measured many of the critical deflections directly, but have let well enough alone, since the optical performance of the telescope is up to specifications.
5.2 The Telescope Mount
The optical support structure, weighing 45 metric tons with all of its cargo, is in turn supported by a massive steel yoke. The alt-azimuth geometry permits it to rotate in elevation on two simple preloaded spherical roller-bearing assemblies. The yoke is built of internally braced box sections of 25-mm plate with central reinforcement by a 75-mm wall cylinder, supported on an angular-contact thrust ball-bearing race, which is 2.5 m in diameter and contains 130 50-mm balls. This bearing was much less costly than a hydrostatic bearing system, and preliminary tests of the yoke showed it to rotate very smoothly. Under a full load of 120 tons it operates with an average coefficient of friction of 3 x lo-', and variations are only a small fraction of the average. The yoke and bearing have a compliance of 0.1 arcsecond per 3000Nm moment about axes perpendicular to the rotation axis, to resist temporary imbalance and wind forces. The torsional stiffness about the azimuth axis is sufficient for the lowest fundamental frequency oscillation to be about 5 Hz (but see below).
Both axes of the telescope are driven by essentially identical systems employing dc torque motors and straight spur gears of moderate quality. Each axis has two motor-and-gearbox combinations that oppose one another during tracking to eliminate backlash, and the motor speeds are governed by tachometer feedback in a stable servo rate loop. The precision of the system for tracking is derived from on-shaft 24-bit encoders of the Inductosyn type. These encoders have an absolute accuracy of 1 arcsec and locally have the full 24-bit precision (0.08 arcsec). The operation of the telescope is ordinarily completely controlled by the Nova 800 minicomputer, which is dedicated to this purpose. It converts object COordinates to the current epoch, calculates corrections for refraction and aberration of light (and for telescope flexure and misalignment), and commands appropriate position sequences for slewing and tracking in altazimuth coordinates (see Section 5.8).
The analytical predictions of the behavior of the yoke and drives have been well borne out, in general. One omission in those predictions was discovered, however, in the factory assembly and testing. The mount has a mode of oscillation corresponding to a rocking on its azimuth bearing, since the balls of this bearing are confined in a race whose cross section has a radius somewhat larger than that of the balls themselves. This motion was considered in the general calculation of resistance to external forces, but its interaction with the drive system was not properly included. What generates this interaction is the placement of the two azimuth drive gearboxes close to one another in one side of the yoke (chiefly for convenience). Thus, when they exert a torque upon the mount, they also exert a net force as well, and this force can excite the rocking mode. Placement of the drives on opposite ends of a diameter would have prevented this interaction, but would have provided a constant tilting of the mount due to the torque bias that prevents backlash (though this latter effect can be compensated in the pointing-correction matrix). At present this rocking mode contributes to the low resonance frequency of the mount (2.2 Hz) and thus limits the stiffness in the azimuth drives.
5.3 The MMT Building
We expected from the start that the operation of the MMT and its associated instrumentation would be remotely controlled, for the most part, with the observer viewing the focal plane via a low-light-level television system. At the same time, we recognized that the MMT was itself a complex and novel piece of equipment, and that it would generate a large variety of novel accessory instruments. Therefore, in planning the housing for the MMT, we strove to provide both easy access to the telescope itself, and a large amount of laboratory space immediately adjacent to it, for the electronics used during observations and for the calibration and checkout of optical instrumentation.
The alt-azimuth geometry offers unique advantages over an equatorial mount for the design of the telescope housing, since the rotation of the housing follows the telescope in one coordinate. Therefore, instead of a shell dome on top of a building whose working space is necessarily below the level of the telescope, we designed an enclosure in which the telescope is completely embedded, and which is slaved to it in azimuth rotation.
There were important spatial, structural, mechanical, and thermal considerations in the design of the building to house the MMT. (Figure 4 shows a schematic of the building.) Spatially, we wished first of all to provide an observing floor at the level of the top of the yoke, and then to work up and down from that level to provide additional space. We needed to be very efficient in the use of space, both to make access convenient and
to make the best use of a very restricted area on the mountaintop site. Structually, we needed to provide support for the working space and for a shutter mechanism to expose the telescope. Mechanically, we wished to ensure smooth rotation of the building, and accurate following of the telescope. The building should permit observing in winds up to 72 km/hr, and should survive winds up to 225 km/hr. Thermally, we had to prevent heat from the enclosed spaces or from the sun-warmed exterior from creating temperature gradients in the optical paths.
The resulting building, shown in Figures 1, 2, and 4, is unusual by astronomical standards. It is rectangular (width 19.5 meters, depth 13.4 meters, height 16.8 meters) because it is less expensive to enclose space in such a shape, and the penalty in terms of wind torques and buffeting is calculated to be small. The lower part functions as a deep structural platform extending from the observing floor to well under the first floor, with principal truss planes that box in the telescope yoke. On this platform rest two three-story towers containing rooms and serving as the stowing location for bi-parting shutter leaves. Figures 5 and 6 give cross sections of the buiIding. The entire building, weighing 450 tons, rests on four massive conical steel wheels, 13 cm wide and 91 cm in diameter, which roll on a flat steel track located about 1 m below ground level on a foundation independent of the telescope pier. The wheels are supported by an articulated linkage that allows the full width of the wheel always to be in contact with the track, thus distributing the load evenly, and they are guided by four wheels pushing radially outward against another track. Two of the support wheels are driven by 15-hp dc torque motors in a servo configuration very like that of the telescope itself, with the position information conveyed to the drive by a transducer connected between the yoke and the building.
The telescope is thermally isolated from the outside and from the heated rooms on either side and below by foam sandwich panels. The working spaces are also heavily insulated from the outside. Since all the surroundings conspire to produce temperatures at least somewhat in excess of the nighttime average, however, the telescope and its environs are designed to be actively cooled by means of refrigerated coils in the floor (which is a double slab, with foam insulation between upper and lower parts) and by conventional air-conditioning units that chill and stir the air, maintaining the telescope uniformly at about the expected nighttime temperature. In addition, the telescope floor and telescope yoke are insulated to reduce the temperature differences in the telescope chamber. All warm exhaust air from the building is dumped into the basement, which is carefully sealed by a skirt from the rotating building that dips into a liquid-filled moat at the top of the foundation. The air is then drawn
through a tunnel and exhausted about 50 m from the building in the prevailing downwind direction.
Electric power to the installation is furnished by two transformers, one to supply machinery and heavy loads, and one for instrument power. The latter is supplied via a motor-generator set that serves as a low pass filter, which removes rapid fluctuations in the power and which stabilizes the power over the long term. Grounding (always a problem on a dry mountaintop) has been carefully considered both for lightning and for instrument grounds. Copper is buried all around, and particularly in the testboring holes that were made in the rock. There is a consistent scheme of branching grounds to avoid ground loops.
5.4 The Optics
The optics consists of six main telescope systems (Figure 3), each of which is a classical Cassegrain with a 1.8-m-diameter parabolic primary with focal ratio f / 2 . 7 , and a hyperbolic secondary producing a final f/3 1.6 for each of the individual telescopes. The focal location is adjusted by longitudinal motion of the secondaries to obtain a focus at various focal locations ranging from the quasi-Cassegrain focus formed by the f/9 beamcombiner 787 mm behind the primary vertex (5 focus A) to the zero spherical aberration focus 360 mm behind the primary vertex (= focus B). Calculations indicate that the spherical aberration change due over this focusing range is acceptable (< 0.2 arcsec).
In the center of the system there is a 73-cm diameter guide telescope used for offset guiding. This system is a Ritchey-Chrktien type of design with a mild aspheric plate for correction of astigmatism near the focal surface. The focal surface is spherical and the guide star sensors, for examining the wide field of view, track on a spherical surface of radius 56.4 cm. Parameters for both the main and guide telescopes are listed in Table 3.
The guide telescope was manufactured using traditional techniques. Since the primary varied only by 1.45% from a parabola, the primary was, in fact, figured using an autocollimation test from a flat to leave a given residual in the interferometer pattern measured from the focus. The secondary received its final figuring in a full autocollimation test of the assembled guide scope. The guide scope primary is a rather thick mirror blank 73 cm in diameter by 15 cm thick and is supported in a strap and multiple mechanical back support mount.
The primaries for the major telescope are of lightweight construction, of egg-crate fused silica form. Since these mirrors were originally intended for a different application, all six mirrors had to be returned to Corning Glass Works to have the front radius slumped to an approximation of the final radius to be placed on the mirrors to obtain the f / 2 . 7 parabolas. Optical work on these primaries was carried out on two large polishing machines in the Optical Shop of the Optical Sciences Center. The optical operations involved first generating and polishing the back surface of the mirror convex, edging the front and back plates, and then generating. grinding. and polishing the front surfaces of the mirror to spherical surfaces whose radius matched the final edge radius of the desired parabolas. Interferometric tests at the spherical stage confirmed the absence of astigmatism in the mirrors. The mirrors were tested on top of an air bladder support of somewhat similar form to that to be used in the final telescope. Support of the mirrors during fabrication and testing used an air bladder in the test setup, and two layers of ordinary rug cushioning under the mirror on the fabrication table. The mirrors were moved from the table to the test setup. Matching of the radii of the six mirrors to within 3 mm (+ 0.03%) was accomplished using an invar steel tape measurement while doing a knife edge test at the center of curvature on the shop floor. Parabolization was carried out by polishing in the center of the aspheric, leaving the edge of the mirrors essentially untouched. The edge of the mirrors thus served as a continuing radius reference during the fabrication process. In addition, not working the edge portions avoided the usual edge-turn-down problems that occur in a mirror, which is desirable since the edges of the mirrors are to be used in the optical alignment system for the telescope. Initial roughing-in of the parabola required knife edge and wire testing on the shop floor. Final testing involved measurement with a null lens and an interferometric setup in the test tower.
The initial goal for the mirrors was to put 90% of the energy within one-half arcsecond. Most of the mirrors closely approached this (see Table 4). Figure 7 shows a composite of the energy distribution curves for the mirrors. As can be seen, there is some dispersion in quality of the mirrors, with mirror six being the worst of the lot, due partially to some residual astigmatism in the surface. Interferometer photographs of the six mirrors taken through the null arrangement are shown in Figure 8. The basic parabolic figure of all the six mirrors is attested to by the straight fringes. The major distinguishing feature of these mirrors turns out to be a number of fine circular zones that were left in the fabrication process. The geometrical energy distribution curves shown in Figure 7 were obtained after computer reduction of the interferograms on the mirrors. Full diffraction calculations of the primary mirrors show the FWHM (full width at half maximum) of the point spread function to average only 0.14 arcsec. Table 4 contains information on the rms surface error and energy
distribution on the six mirrors, both on the test setup and when tested in the final flotation cells at various elevation angles. A full description of the MMT optical design and performance is given by Ruda & Turner (1980).
The MMT construction represents the first time that lightweight structured mirrors of this sort have been used in a variable altitude, gravity environment. The primary mirrors are each located by three axial hardpoints with most of their weight floated on an air bag whose pressure varies with altitude. Seven percent of the total mirror load is carried on the three axial hardpoints. The radial location of the mirror is defined by counterweights through an 11:l ratio lever system acting on a pair of chains that surround the lower half of the front and rear plates of the mirror. These are adjusted so that the mirror hangs free of astigmatism when one is viewing at the horizontal position. Only a few pounds force is generated against the radial locating points. These radial locating points are referenced from an invar ring cemented to the back of the mirror near its center.
There are actually twelve secondary and twelve tertiary mirrors for the telescope, six of each for visual use, and six for infrared use. The infrared secondary mirrors are smaller (and thinner) than the optical ones to eliminate the radiation coming from the edges of fhe primary mirrors and to provide a low inertia for the mechanically oscillating infrared secondaries. In fabrication, all twelve secondary mirrors were tested against a concave hyperbolic test plate, which itself had been tested in a null lens interferometer system. Therefore, there is no variation in radius between the secondaries.
The optics were tested while in place on the telescope using star images. Figure 9 shows a Foucault knife edge test of one of the mirrors as well as an interferogram. The circular zones left over after the fabrication process are very evident also in the out-of-focus images for all six telescopes. They are, in fact, the dominant feature in all of these tests, other deformations being of lesser importance. The knife edge test shows, of course, the first derivative (slopes) of the surface irregularities in the direction normal to the knife edge, the out-of-focus image shows the curvature of these irregularities (second derivative). This curvature shows up remarkably well for relatively slow telescope systems like the MMT (f/3 1.6). A closer evaluation of the six telescope knife edge images shows only a very small amount of coma in some telescopes and, except for two telescopes, no astigmatism. Adjustments in the mirror support remain to be made to eliminate this residual astigmatism and coma. The stellar image quality is very good and mostly determined by seeing (Section 5.9). Image sizes with a FWHM of less than 0.5 arcseconds had been reported and sub-arcsecond-diameter images are common. Even so, this image size is mostly determined by atmospheric seeing both inside and outside the telescope chamber and not by the optics. Future improvements in the chamber thermal conditioning will undoubtedly improve the image quality further. Whether then the telescope optics will become the limiting factor remains to be seen.
The average diameter for the non-vignetted field equals 35 arcseconds as compared to 52 arcseconds as the theoretical optimum (Beckers 1980).
Future adjustment of the optics will increase this field of view. The vignetting of the telescopes sets in very slowly, however, so that even for a 4-arcminute-diameter field the vignetting is less than 50%.
5.5 Telescope Coalignment
REQUIREMENTS In order to produce and maintain an accurately superposed image from the six individual telescopes it is necessary to be able to (a) independently adjust each image position, (b) determine the initial superposition, and (c) continuously detect deviations of individual images from the composite image.
For the MMT, the first is accomplished by remotely controlling the tilt of each of the secondary mirrors about two axes, the second by television monitoring of the focal plane either visually or under computer control with centroid determining routines, and the third by either periodic updates of the second or with an internal artificial star system using laser beams.The desired accuracy of the coalignment depends on the overall quality of the telescope optics, tracking, and site and the intended uses of the instrument. The individual Cassegrain telescopes as determined from optical shop measurements allowing for primary, secondary, tertiary, and beamcombiner figure imperfections and for spherical aberration and miscollimation provide 90% of the light from a point source in a .7-arcseconddiameter circle and a FWHM of the point spread function of about 0.2 arcsec. Image quality (FWHM) at the site averages 1 arcsec and has been as good as 0.5 arcsec. Current spectroscopic, imaging, and interferometric uses of the MMT can take advantage of images less than 1 arcsecond, dictating superposition accuracy which does not noticeably degrade the individual telescope performance, e.g. = 0.1 arcsecond superposition error. It should be noted that this provides a more stringent requirement on the coalignment than that specified when the telescope was conceived and designed. At that time, typical seeing was expected to be 2-3 arcseconds, rather than the 1 arcsecond frequently experienced now, and the usefulness and power of the MMT for imaging and interferometry was not fully appreciated.
The range of individual image wander is determined by the rigidity of the supporting structure, which in turn is dictated by the maximum allowable image degradation due to miscollimation. The criteria given in Section 5.1 imply a maximum image motion of 300 arcseconds. The observed maximum relative image motion of the six telescopes is -t 20 arcseconds over an elevation range 17-88 degrees with typical and maximum drift rates of .05 arcseconds/minute and 0.2 arcseconds/minute during stellar tracking.
ADJUSTMENT OF IMAGE POSITIONS The actuator for the secondary mirror tilt is designed to meet the requirements for image superposition. The secondary actuator characteristics are as follows:
Minimum angular step: 0.059 arcsec motion in focal plane Range: 600 arcsec Maximum slew rate: 12 arcsec S C '
The mirror pivot consists of two 12-mm-diameter steel rods to provide both precise angular control and high axial rigidity to allow for a secondary chopper mechanism for infrared use. The tilts are achieved by two spring-loaded micrometer screws coupled to a 1 00-steps-per-revolution stepper motor by a 110-to-1 harmonic reduction gear. One step of the stepper motor provides a 0.058-pm motion of the micrometer, .23 arcseconds tilt of the secondary, and .067 arcsecond motion of an image at the focal plane.
The actuator assembly also provides for remote focus control using precision ways that assure no change in mirror tilt during focus motion. The torque motor drive provides for a smallest increment in mirror position of 1 gm. Motion of the tilt and focus motors will be under computer control with commands initiated either manually with a control paddle or automatically from a focal plane monitoring and image centroiding device. The current actuators are a modification of the original MMT actuators, which had torque motor drives for all three motions and less precise setting accuracy. Additional coalignment control for precise setting of the direction of the principal rays of the individual telescopes can be achieved by an adjustment of the tilt of the tertiary mirrors. The control system for this has been designed but not yet implemented.
INITIAL SUPERPOSITION The initial superposition of the six images requires accurate determination of the relative position of the center of each of the images and feeding this information to the secondary actuators to provide for the necessary correction. Since the images share some motion due to seeing and tracking errors, it is important that the relative position determination be simultaneous for all six images. A camera consisting of a GE TN2200 series CID camera controlled by a Z-80 microprocessor has been built for this measurement. The scale on the 128 x 128 array chip equals .16 arcseconds per pixel. In operation, an exposure of the six separated images will be taken, stored in memory, and analyzed for the centroids of each of the images to an accuracy of < .05 arcseconds, the entire operation taking about ten seconds depending on stellar magnitude. The measurements are then entered into the computer controlling the secondary actuators to provide the appropriate corrections. The CID camera, while providing excellent dimensional stability, is currently limited to a small field and relatively bright stars (mu = 5 without intensifier, mu = 10 with intensifier).
A second system for superposition of images utilizing a sensitive intensified focal plane acquisition and guiding TV has been developed (see Section 6.2.2). This provides for viewing of stars of up to mu = 18 continuously in a 4-arcminute field around the slit of a spectrograph or hole into a photometer. The optics provide a focus of the telescope pupils where a set of prisms provides a small deviation of each image and where a pupil selector can be used for examining individual images as desired. With this system, the images can be manually superposed to an accuracy of .2 arcseconds in a few seconds of time. Provisions are being made for automating this with centroid-determining software at a higher accuracy of 0.1 arcsecond. At the same time this system will serve as an autoguider for the MMT.
DETECTING DEVIATION FROM SUPERPOSITION Depending on the telescope orientation, superposition must be updated every 30 to 600 seconds (typically 120 seconds) to maintain .l -arcsecond superposition accuracy with the CID camera on a bright star, or with the focal plane guiding television on a field star either manually or automatically. The optics and computer hardware and software for the guiding television are being designed to make continuous guiding on the individual images straightforward with the capability of updating as frequently as ten times per second.
For instrumentation and conditions (such as daylight observing or no guide star in the field) for which the above approaches are not suitable, the artificial star system using laser beams will be used to provide an internal mechanism for determining telescope misalignment drift. The basic elements of this laser alignment system are shown in Figures 10 and 1 1. It is described in detail by Reed (1978) and McDonough (1980). In the center of the hexagonal MMT telescope array is a 73-cm f/17.6 guide-alignment telescope of Ritchey-ChrCtien design, designed for telescope alignment and guiding (Table 3).
For alignment purposes, this telescope serves as a collimator for a lasergenerated point source. In order to maintain the desired 0.1-arcsecond collimation accuracy a second laser system provides continuous automatic collimation of this telescope. The outgoing beam is transferred to the periphery of the six large telescopes by three 1.8-meter-long periscopes and thence into the telescope apertures by six roof-prism-90-degree-prism combinations which function as elongated cube corner reflectors. Both the periscopes and the cube corners have the property that the outgoing rays remain accurately parallel (or anti-parallel) to the incoming rays if the device is tilted, so long as it behaves as a rigid body. A prototype of the periscope has been tested and found to have a stability of better than 0.05 arcseconds.
Because a single marginal ray cannot distinguish between a change of focus and a tilt, a second beam is introduced into each aperture with a pair of pentaprisms, as shown in Figures 10 and 11. Both beams are then sensed by silicon detectors at the focal plane, the focus beam by a split detector that senses the radial direction only (the pentaprisms giving invariance only in this dimension), and the main beam by a quadrant detector that senses both radial and azimuthal motion. These signals, processed by the computer to remove noise and seeing effects, provide the desired pointing accuracy by averaging over time, and are then used to provide corrections to the tilt and focus of the secondary actuators.
Pairs of thin wedge prisms, shown in Figure 11, provide for remotecontrolled two-axis tilt adjustment of each of the beams for initial setting of the telescope focus and image position. To date, the laser system has been operated only with the alignment beams, not the focus beams, directly controlling the torque-motor-driven secondaries without deglitching and averaging. In these experiments the superposition was maintained to 1 arcsecond rms over the short term (1-2 hours), and the long-term drift reduced by an order of magnitude. Direct simultaneous monitoring of the laser beams and stellar images at the focal plane indicate that considerably better performance can be achieved by a fully operating laser system.
The methods of maintaining the image superposition have evolved from the original design since the telescope has been available to use in 1979. Considerable further evolution is likely as experience is gained with this unique instrument.
5.6 Infrared Capabilities of the MMT
GENERAL CONSIDERATIONS As the result of several years of development at the University of Arizona, the general principles of design for groundbased IR telescopes were proven by modifying small instruments such as the 0.71-m and 1.55-m telescopes. Low & Rieke (1974) have described these principles in detail and they discuss the limiting performance of instruments using these designs. The MMT represents an effort to extend these concepts to the design of a new type of large telescope that retains all of the performance advantages of proven smaller telescopes without the risk and uncertainty of applying similar solutions to a much larger scaled-up design. In addition, the MMT design adds the fundamental advantage of an unfilled array in those applications where maximum angular resolving power is important. An equally important motivation, though still the subject of considerable debate and future study, was to achieve a design that would reduce significantly the cost of very large telescopes capable of meeting the basic needs of both optical and infrared astronomy. The full performance of the MMT in the infrared has not yet been measured, but certainly no other telescope of competing infrared performance has yet been built. We briefly list here the most important infrared features of the design as they bear on observations in the infrared.
The general distinguishing features of a telescope designed for the IR are: 1. minimum background noise emitted by the telescope, i.e. minimum obscuration, minimum number of uncooled mirrors, and small mirrors with easily maintained coating, 2. stability of the telescope background, i.e. extreme internal rigidity under wind loading and infrequent articulation of uncooled mirrors, 3. a means of background subtraction such as secondary mirror modulation. Smooth, sub-arcsecond tracking and arcsecond pointing are also important.
OPTICAL COSFICLRATIO~ The optical configuration, which IS that of six identical 1.8-meter f/33.8 telescopes arranged in a circle around a common axis, permits all of the features of the well-proven f/45 1.55-m IR telescope to be preserved. The IR secondaries are undersized slightly to serve as a “cold” stop against the sky. Two of the three uncooled mirrors are easily maintained and can be equipped with special IR coatings; the primaries are comparatively small and are relatively easily maintained with a fresh, clean coating. Thus, it should be possible to achieve and maintain an overall emissivity of the telescope as low as 5%. Clearly, the fourth mirror in the system, the beamcombiner, must be cooled along with the detector, the spectral filters, and the field optics and baffles.
CHOPPING SECONDARIES Because of their importance in determining IR performance, the chopping secondaries merit special attention. It has been shown that no other simple method of background subtraction performs as well as scanning the IR beam on the sky by tilting the secondary mirror. Using a servo-controlled linear motor with sensitive position transducers, it is possible to produce extremely accurate square-wave or linear scans over angles ranging from a few arcseconds to several tens of arcseconds. Because of the small size of the secondary mirrors, chopping speeds above 20 Hz are possible with low power dissipation and without image degradation from mirror bending or from overshoot. In the MMT design the chopping secondary mechanism provides for accurate rotation of the chopping axis around the optical axis so that all “twelve” IR images may be coaligned. The chopping motion is completely independent of the image coalignment systcm, which also tilts the secondary mirrors.
PERFORMANCE The chopping secondaries and the infrared photometer described in Section 6 were first used at the MMT in October, 1980. The measured performance was exactly as predicted by scaling from the 1.55-m telescope to the full 4.5-m aperture of the MMT. In other words, the MMT performs as well in the infrared as would a fully optimized conventional infrared telescope of 4.5-meters aperture.
The sensitivity of the MMT is best illustrated by the detection of the flat spectrum radio source 0332 + 078 at a magnitude of 18.40 -t 0.25 at 2.2 pm. The measurement required one hour of integration through a 9 arcsec aperture. (The corresponding la level is 20.0 mag.) This source was previously undetected at 2.2pm in two attempts with the Steward 2.25-m telescope; it is an empty field on the Palomar Atlas plates; and it was not detected in the visible on a deep plate obtained with the Kitt Peak 4-meter telescope. It is the first example of an extragalactic infrared source that is undetectable in the visible. Its measured flux level is four times fainter than any object detected previously at 2.2 pm.
Because the MMT is the largest optical/IR telescope now in operation, when measured from edge to edge (6.9 meters), it potentially has the highest resolving power at these wavelengths. This fact, coupled with the additional consideration that the MMT is the first of a possible series of large optical/IR arrays to be built in the future, suggests that the MMT will be very important as a tool for spatial interferometry both at optical wavelengths and in the IR. An important incentive to workers in this field is the newly discovered result that the “seeing” at the MMT is remarkably good compared to that at other large telescopes in the Tucson area. To achieve the full possible angular resolution of the MMT, it is necessary both to “phase” the individual telescopes with respect to each other and to eliminate differential changes of the polarization state of the incident lightbeams. So far this has been achieved both at optical and infrared wavelengths with pairs of opposite telescopes. Because of the off-axis reflections on the telescope tertiaries and on the beamcombiner, the polarization state of the light will be modified, but this modification is identical for these pairs. The pathlengths of the two telescopes can be adjusted until an interference pattern appears.
D. W. McCarthy and F. J. Low have already operated the MMT as a two-element Michelson interferometer at 5 pm and have measured both the pathlength equality of the three pairs and the stability of a single pair. Work is proceeding toward the goal of operating the MMT as a phased array at the longer IR wavelengths. The pathlength stability, as shown by McCarthy’s data in Figure 12, is inherentb adequate for phasing at 20 pm and may be adequate for 10 pm. The coherence length @‘/Ax) is a function of both wavelength and bandwidth and can therefore be adjusted within the detecting system to extend operation to shorter wavelengths
with reduced bandwidth. K. Hege, indeed, obtained optical fringes for Ak as large as 0.01 pm. Ultimately, however, it will be necessary to devise a method for continuously measuring and controlling the pathlength for each of the telescopes.
5.8 Digital Telescope Control and Data Collection Systems
Two computer systems are used at the MMT for routine observational programs. A third will shortly be added for coalignment control. The first system, called the mount computer, is primarily dedicated to driving the telescope control servos. As shown in Figure 13, it is composed of a Data General Nova 800 minicomputer with several input and output devices. Mass storage is accomplished with two floppy diskettes. External signals are fed directly into the computer where they are processed and used to update the command signals to the telescope drives. The second computer system is called the instrument computer, and it is shown in Figure 14. The same type of computer is used here (Nova 800) along with terminals similar to those used with the mount computer. The instrument computer is used to remotely control instrument functions and to record, Calibrate, and display data. A hard disk is available with 50 M byte storage,
The MMT is pointed by positioning the altitude-over-azimuth mount under control of the mount computer system. Every 100 milliseconds the absolute shaft-angle encoders are read to 0.077 arcsecond precision, and these readings are compared with the calculated source position. The position errors are used to generate correcting torque commands in a Type I1 position/velocity servo system for each axis. These servo systems are composed of four electronic loops which control torque, velocity, and position by means of negative feedback and which also provide improved acceleration performance by means of a positive source velocity feedforward loop. The torque and velocity loops are closed in hardware, but the position loop is closed in the mount computer software, which integrates the position error while tracking to produce a velocity command signal. The desired rate is converted to an analog voltage in a digital/analog converter and applied to the hardware velocity loop. When changing sources, another algorithm is automatically applied to produce the optimum motion. In this case, the commanded velocity is made proportional to the (signed) square root of the magnitude of the position error. Thus, the telescope will tend to move in a parabolic path. Software limits on maximum velocity and acceleration result in smooth ramping up of telescope velocity to its maximum slewing speed as well as ramping down to just meet the source. The algorithm used for repositioning results in optimum telescope motion to track even a rapidly moving source.
A dynamic model of the telescope servo system has been developed in the Laplace complex-frequency domain. In addition, mechanical and electrical measurements were made to determine the relevant physical characteristics of the drive system, such as moment of inertia, spring constant, viscous damping coefficient, and motor torque constant. Computer simulations were then used to derive hardware and software parameters to optimize the performance of the position control system. In particular, this model has proved to be very useful in reducing the tracking errors caused by wind torque disturbances. Frequency spectra of errors due to such disturbances have been measured by injecting an electronic signal that disturbs the servo in the same way as varying wind or friction torques, and the agreement with the model calculations is good over several decades in frequency. A detailed description of the mount control system is given by Ulich & Riley (1980).
While tracking at sidereal rates, the encoder position readout follows the commanded position to 0.07 arcseconds rms, which is the size of the least significant bit in the encoder readouts. After correction for systematic encoder errors we find that these encoder positions track the stellar positions to accuracies of better than 0.25 arcsec for periods of 10 minutes and longer. Of course, due to foreshortening, the error on the azimuth
encoder is reduced by the cosine of elevation to yield a smaller true angle on the sky. The tracking actually improves near the zenith, where a zone of avoidance with a radius of about 0.16 degrees exists because of the limited maximum velocity in azimuth. The tracking errors are larger when a significant wind exists in the telescope chamber, with an average error of about 0.5 arcsecond in a 50 km/hr gusty wind.
The true source position must be corrected for atmospheric refraction, instrumental defects, and zero-point offsets in order to determine the desired encoder readings. Optical refraction is sufficiently well understood that its correction is straightforward and can be calculated with high accuracy. The mount control software uses ambient temperature and barometric pressure inputs to calculate the refraction coefficient and to display the value currently being used. Other corrections such as gravitational flexure, tilt of the vertical (azimuth) axis from the astronomical zenith, nonorthogonality of the telescope" axes, collimation errors, and offsets are automatically applied to the calculated apparent source position. The coefficients that represent these errors are determined by performing a least-squares fit to a large number of measurements of pointing errors. Figure 15 shows the residuals left after such a fit was carried out for one night's observations with a CID camera (Radau & Ulich 1980) at the focus of the guide telescope in the MMT. The mean radial pointing error, which would have been observed if the optimum coefficients had been used, was 0.88 arcseconds, and the standard deviation of the radial pointing error about zero was 1.12 arcseconds. Over a period of weeks the pointing characteristics change slowly, and relatively frequent calibrations are necessary to maintain accuracy similar to that shown in Figure 15. In addition, several periodic encoder errors have been detected by special experiments, and when these are corrected the pointing accuracy should be improved. Thus, even though with conventional techniques the MMT has more accurate "dead-reckoning'' pointing than other telescopes, it may be improved significantly in the future.
5.9 The Site, Image Quality, and Seeing
The general selection of Mt. Hopkins as an observing site was described in Sections 3 and 4. The decision to put the MMT on the absolute summit, though having a certain inevitability, was not made without scrutiny. Seeing tests consisting of the measurement of star-trail widths from a 10-cm telescope indicated that the summit and two slightly lower knolls to the east and north had comparable (and good) seeing conditions. The summit was chosen because the prevailing fair-weather winds are from the west, and because the summit itself, though small in area, was still larger than the other knolls.
In an effort to preserve the natural mountainside cover of scrub vegetation on the southern half of the summit, we chose to tolerate a very steep (22% grade) pitch at the end of our road, so that the road could follow the ridge line off the northeast side of the peak, rather than follow an easier grade, which would have exposed a lot of road surface and cut bank to the direct heat of the sun. The parking and walk-around area in which the building sits is covered with light-colored crushed stone. This essentially places the building on a pinnacle with steep (45"-90" inclination), undisturbed slopes on all sides except the northeast. Our expectation was that these surroundings would heat fairly uniformly and have short thermal response time, giving us the best possible conditions consistent with a low placement of the telescope (about 9 m above ground level), which was forced upon us by budget limitations.
As mentioned above, a short thermal response time was in our minds at all stages of design of the telescope and building, and we have succeeded pretty well in the open optical support structure and in the lightweight building skin backed up everywhere by good insulation. The observing floor and the yoke are necessarily massive, and, therefore, are intended to
be actively cooled. The yoke arms and the floor are insulated with a layer of Styrofoam and plywood to decrease the thermal time constant of their surfaces. As yet, our main cooling system is not active, because the refrigeration machinery has not been purchased, yet the seeing at the MMT appears to be at least equal to, if not better than, that at Kitt Peak and Mt. Lemmon. One additional thermal feature that is doubtless helping us is the extreme openness of the observing chamber, which gives a very free circulation of air even without fans, and a large solid angle open to the sky for radiative cooling.
The stellar image quality at the MMT is routinely measured by means of a CID camera placed at the quasi-Cassegrain focus. One arcsecond corresponds to six pixels on this camera. Based on the limited observing of about one year, the MMT experiences typically one arcsecond seeing (i- WHM images). Occasionally, the image quality exceeds 0.5 arcsecond and seeing worse than 2 arcsecond is uncommon. Much of this seeing appears to be internal to the telescope and telescope chamber. Because of the existence of the laser coalignment system (see Section 5 . 9 , we have a way of measuring the internal seeing in the MMT by means of the recording of the varying error signals from the laser detectors. The position fluctuations of the laser images is substantial. It typically amounts to 1-2 arcseconds peak to peak, and the rate of fluctuation falls in the 0.2-4 Hz range with the slower fluctuations dominant when the chamber is closed and with the fast fluctuation dominant when the chamber is open and when it is windy. Figure 16 shows the laser beam fluctuations for a full observing night. Also shown is the maximum temperature differential in the chamber (between the yoke and the air). Both quantities are large. We suspect a relation between the two. It is possible to estimate the importance of the internal seeing with respect to the total stellar image size by making the approximation that its effect can be evaluated purely by geometrical optics effects. The schlieren causing the image deterioration are larger than the 25 mm of the telescope aperture used to image the laser beams. Visual examinations show most of them to be smaller than the telescope aperture. If so, the displacement of the laser images can directly be related to the FWHM of a point spread function. Assuming a Gaussian distribution this relation equals FWHMlaser = 2.35 X rms laser image motion. Most of the laser image motion originates in the main telescope rather than in the guide telescope. Assuming the different sources of image deterioration to add quadratically one has
Figure 17 shows a scatter diagram of measurements of FWHM2star against (rms laser motion)2. It is obvious that often much or most of the image deterioration occurs in the telescope so that further efforts to
reduce the telescope and telescope chamber thermal effects are very much needed.
6. MMT INSTRUMENTATION
Auxiliary instrumentation for use with the MMT can be located both at the quasi-Cassegrain focus and at one of the two Nasmyth foci. The optics to feed the Nasmyth foci do, however, not yet exist so that all MMT instruments used now, or planned for the near future, use the quasi- Cassegrain focus. Because of the unconventional illumination of the images in the focal plane, an f/9 beam with a filling factor of - 50% by the six f/31.6 beams (see Figure 3), special techniques are often needed tn optimize the coupling of the instruments to the telescopes. The technique used for the MMT spectrograph has been described by Fastie (1961, 1967). Table 5 lists the seventeen instruments that are or shortly will be used with the MMT. Some of these are instruments that are constructed specially for the MMT and are, therefore, optimized for this telescope (e.g Far-IR Photometer, MMT Spectrograph, IR Spatial Interferometer). These will, when finished, become MMT facility instruments. Others are instruments built for other telescopes but adapted to the MMT to conform to the MMT pupil configuration (e.g. Fast Photometer, MHO Echelle Spectrograph) whereas still others are coupled to the MMT without further optimization.
In the rest of this section we describe in more detail the instruments that are specifically being built as facility instruments for the MMT.
6.2 The MMT Spectrograph
GENERAL CONSIDERATIONS The most basic function of a modern faint object spectrograph is to record quantitative spectra of stellar or small objects at moderate dispersion, with subtraction of sky background. In the past decade, a variety of electronic detection schemes have been used for this purpose. For example, the Wampler-Robinson scanner was an early and very successful device to read electronically the output of image intensifiers. While the basic star-minus-sky mode of operation will continue to be a key function, in designing a spectrograph for the coming decade we have aimed to take full advantage of the two-dimensional formats of the new generation of electronic detectors. Observations that need area format detectors are cross-dispersed spectra of good resolution and broad cover; long-slit spectra of extended objects; and simultaneous spectra o[ multiple objects in the field of view of the telescope, obtained through the
use of an aperture plate with several holes corresponding to the positions of objects in the focal plane. In designing the spectrograph we, therefore, want to provide a versatile instrument incorporating both the basic linear and area format capabilities, while maintaining mechanical stability and simplicity of operation.
Designing for the MMT rather than a conventional telescope presents unique problems and advantages. The plate scale is large, 3.6 arcsec/mm, which means that a stigmatic spectrograph will give excellent resolution along the slit. The field of view is 2 arcminutes for no more than 10% vignetting by the beamcombiner (Beckers 1980). We have taken as a design goal that the spectrograph should be unvignetted for a slit length of 5 cm or 3 arcminutes. When stigmatic imaging is not needed, the large plate scale and unique geometry of the pupil make it practical and attractive to construct a simple image slicer for the f/9 focus. A scheme originally proposed by Fastie (1961) has been adopted and is described in the next section. In a conventional spectrograph the slit jaws or entrance apertures are reflective, and a sensitive TV system is used to identify the star field, to focus, and to center the program object on the aperture. Offset guiding may also be accomplished by field stars on the TV screen. At the MMT as it now stands, the TV will also be used to maintain the accurate coalignment of the images from the six primaries, by correction of the secondary tilts (see Section 5.5).
No single detector yet covers the full optical range efficiently. One must still reckon to use magnetic or proximity-focused image tubes in the ultraviolet and blue, where cathode quantum efficiencies of 25% are achieved. Because of the large scale of the telescope, we set as a goal that the spectra should match the largest commonly produced size of image tubes of 40 mm. We also decided that the camera should be constructed to allow the use of magnetic intensifiers, which have the best efficiency, resolution, and distortion characteristics. Readout of the image tube will ultimately be by an electronic detector with two-dimensional format. However, as a practical matter we made it our immediate goal to put the spectrograph into operation with the Smithsonian photon-counting Reticon system.
OPTICAL DESIGN Because the two best detectors we want to use, magnetic image tubes and CCDs, are complementary in their wavelength coverage and rather different in their formats, we decided early on to divide the beam below the slit with a dichroic mirror, and operate red and blue channels simultaneously. Ultraviolet and blue light is reflected to a large collimator and optics optimized for a 40-mm image tube, while the remaining light passes through to a smaller refractive collimator and smaller optics tailored for a CCD, as shown in Figure 18. In this way the complete spectrum can be observed at once.
The scale of the instrument is set by our goal of matching detector formats. We have adopted a folded Schmidt as the best camera for use
with a bulky image tube. The field of view for good images in this type of camera is no more than 10". Thus, the focal length of the camera with a 40-mm format should be no less than 230 mm. Based on considerations of using the fastest possible camera ( Q f/l) to get the maximum scale reduction, we adopted for the blue camera a focal length of 250 mm, and a beam size of 150 mm, requiring a collimator focal length of 1.35 m. The CCD detector for the red channel is small, < 18-mm diagonal, and the above dimensions are scaled down by a factor of two in the design for the cryogenic CCD camera. This camera has no folding flat, and the CCD is supported by a spider directly at the camera focus. For both cameras the image scale at the detector is 53 wm/arcsecond, large enough that spatial structure in sub-arcsecond seeing will be resolved by good area detectors.
Another key element that led to the geometry of Figure 18 was our incorporation of cross-dispersing prisms that can be optionally introduced into either the red or blue channels. To satisfy our goal of being able to switch quickly from direct imaging to spectroscopy, and to allow quick interchange of gratings, the design calls for gratings to be mounted on turrets in both channels. The red channel can thus be set up with silvered flats at both the grating and prism positions for direct imaging at 20 arcseconds/mm, alternatively with either the grating or prism giving dispersion, or with both together for a cross-dispersed format on the CCD.
Four gratings are currently available or on order for the blue channel. For low dispersion, a 300 gr/mm grating in first order gives dispersion of 130 A/mm and covers the spectrum from 0.3 to 0.6 p at a resolution of < 10 A. The highest dispersion grating that can be accommodated by the 30" included angle of the spectrograph without substantial vignetting has about 45" blaze angle. Such a grating is being ruled at Hyperfine Inc., with 250 gr/mm for use with the cross-dispersion prism to give full spectrum cover. The dispersion ranges from 9 A/mm in the UV to 13 A/mm in the blue, and resolution considerably less than 1 A will be achieved, depending on the detector being used. The red channel will also be equipped with gratings for a range of dispersion. For the cross-dispersed mode we propose to use a grating blazed at 21 O with 125 gr/mm which will give full cover from 0.43-1.1 pm on the CCD, with dispersion ranging from 50 A/mm in the blue to 100 A/mm at 1 pm. There is adequate separation between orders in both channels so that a slit of 10 arcseconds can be imaged without having the orders overlap. There is also room to accommodate the elongated images produced by separate image stackers for star and sky spectra, as now described.
A unique feature of the design is the arrangement for reformatting a normally round stellar image into a long thin image to improve spectral resolution. The method for image slicing takes advantage of the MMT's optical configuration. The secondaries are slightly displaced so as to arrange the six images stacked in a line. At the focal plane the images are formed on small segments of a field Iens. These are laterally displaced from their original position in the parent lens, so as to form a common overlapping image of the telescope pupils. A negative lens then serves to give a virtual star image that is three times smaller than the focal plane stack, brings the focal ratio of the common beam tof/9, and restores the pupil to its original position at infinity. This method was originally described by Fastie, and is explained in detail by Angel et al. (1979). We call this arrangement an image stacker, and the spectrograph will be provided with a pair (for star and sky) that can be introduced at the focal plane when desired.
A good focal plane monitoring TV camera can resolve the entire useful guiding field of the MMT, about 4 arcminutes, with arcsecond resolution. We, therefore, designed the TV to be boresighted on the telescope axis, but with interchangeable lenses to allow for high magnification of the central field when desired. The full field (70 mm) is reflected by a plane field mirror inclined at 12t" to the telescope axis. A large Cooke triplet lens (125-mm diameter) is used as a collimator, and reimaging onto the TV is accomplished by one of several camera lenses located close to the exit pupil of the collimator. The shortest focal length camera gives a reduction factor of 5.4, bringing the full field within the TV camera format of 11 X 15 mm. In order to distinguish the overlapping images formed by different primaries, it is only necessary to insert stops at the exit pupil, where the six primaries are imaged as a cluster. Alternatively, a circle of six wedge prisms will be available to separate the six images on the TV, even though they are together at the spectrograph entrance aperture. With the aid of a projected reticle image, which will be in the form of dot-like artificial star, it will be possible to maintain accurate coalignment and guiding without disturbing an observation with the spectrograph.
DETECTOR The first observations with the MMT spectrograph utilized a photon-counting Reticon detector system. The basic approach of this detector, to centroid individual photon events with a Reticon diode array at the output of a high-gain image intensifier package, was pioneered by Shectman (1976). The MMT detector includes several major improvement5 over earlier SAO versions called the Z-machine (Davis & Latham 1978).
The heart of the MMT detector is a custom dual 1024 Reticon and associated electronics for high-speed readout and signal processing. The Reticon head is modular so that it can be interchanged between image intensifier packages. The first MMT applications used a package of three Varo military tubes, all with fiber-optic windows. The first is a special Varo 8605 40-mm electrostatic diode inverter with custom thin S-20 cathode for enhanced blue response. It is fiber-optically coupled to a standard Varo 8605 selected for high green sensitivity, which in turn is coupled by a 40- to 25-mm fiber-optic reducer to a Varo 3603. This 25-mm MCP inverter with special fast phosphor is fiber-optically coupled to the Reticon with a special split faceplate made by Galileo Inc. The split faceplate effectively eliminates the 1.0-mm dead space between the two diode arrays and thus minimizes the effects of distortion by mapping the sensitive areas of the Reticon arrays nearly onto a single diameter of the intensifier output. It also makes it much easier to align the detector along the spectrum, an important practical feature on the echelle spectrograph. On this package not only was the first tube carefully selected for maximum sensitivity, but also we tried to maximize the number of photoelectrons per event at the second cathode. This is what sets the shape of the pulseheight distribution, which in turn influences the counting efficiency attainable. With too little effective gain from the first stage it is possible to pick up extra dark events from the second cathode and to get persistence from areas that have just been brightly illuminated, such as comparison lines, despite the frame subtraction in the signal processing. Both these effects were a problem with the earlier Z-machine detectors but are essentially eliminated in the MMT detector. The spectral sensitivity of this system is basically a classical S-20 cathode response with an ultraviolet cutoff just short of the H and K lines because of the fiber-optic input window. The peak absolute counting efficiency was measured to be nearly 15% at 5000 A.
One of the most encouraging aspects of the MMT detector is the improvement in the fixed patterns. Raw data from the Z-machine detector have approximately 50% modulation with most of the power in every 2-, 4-, and 8-pixel pattern. The new system has little power in these frequencies and an overall rms modulation of about 3%. The residual patterns appear to be mostly cathode nonuniformities. Under ideal conditions flatfield corrections can attain a signal to noise of at least 200. The operational limit is not quite so good, apparently because of the way slightly different areas of the cathode get illuminated due to mechanical, magnetic, and electrostatic flexure, and also because of differences in the way stellar and lamp exposures illuminate the slit.
PERFORMANCE In May and June of 1980 a large number of exposures were taken with the MMT spectrograph and its detector. A resolution of 35 pm FWHM at the center of the diode arrays was routinely achieved, with some degradation at the edges of the intensifiers. Because of the fiber-optic demagnifier, this corresponds to about 60 pm at the first cathode. Observations of spectrophotometric standards through a 1.5 arcsec slit generally gave overall system efficiencies of about 1% including the atmosphere, slit losses, spectrograph, and detector. The observed dark rates were generally less than 10 counts/cm2s-’. This is several times higher than was attained with the same system during wintertime tests at Agassiz station where the cathode temperature was probably some 10°C less. In Section 7.1 we describe some of the scientific results obtained with the MMT spectrograph.
6.3 Infrared Photometer
The first MMT infrared photometer was designed and built by George Rieke, Bill Hoffmann, and Frank Low. The instrument is shown schematically in Figure 19. The InSb detector is cooled to 4.2 K (Rieke 1980) and will be used primarily at 2.2 and 3.4 pm. Note the longf/32 cooled collimators, the cooled beamcombiner, and the dichroic beamsplitter (all at 77 K). When used with the servo-operated, secondary mirror modulation system, it is expected that background-limited performance will be achieved with very small (< 3 arcsec) apertures. In addition to selectable cooled apertures, the system includes a large cooled filter wheel and provision for a cooled circular variable filter. Thus both broad-band and narrow-band absolute photometry can be performed efficiently
Don McCarthy and Frank Low are developing a 10- and 20-pm photometer which incorporates features that allow the six optical pathlengths to be made equal, thus allowing operation as a phased array. When “seeing” is exceptionally good, this system should operate fully phased with continuous integration and extremely high signal to noise. When less than ideal conditions prevail, the data may be reduced in the less efficient “speckle” mode. In either case, the angular resolving power is expected to be diffraction-limited.
6.4 The SAO CCD Camera
The concept of a true two-dimensional detector having close to unity quantum efficiency over a broad region of the visible and near-IR spectrum and with both low noise and sharp images is, of course, attractive, both for direct imaging applications and for spectroscopy. Although no commercially available imager can claim to s?tisfy all these desired characteristics at present, recent developments in large solid-state arrays such as charge-coupled devices (CCDs) promise an approach to the ideal in the
near future. Development of a CCD camera at SAO during the past year has been motivated by the desire to use devices that are presently available for actual astronomical observations, both to see what the characteristics of this new technology are in this setting and, we hope. also to do some scientifically interesting projects in the process.
The system that has resulted consists of an upward-looking dewar container to maintain the CCD at a lower temperature (typically below 1 50 K) for dark current suppression and a self-contained electronic module to scan the imager on command, process the resulting video signal, and transmit the digitized data to the instrument computer. At present we are using the large CCD manufactured by RCA and available in both frontside- and backside-illuminated versions. This device is organized into 512 rows of 320 elements, each pixel being 30-pm square. The resulting format measures 9.6 mm x 15.4 mm, one of the largest presently available in a solid-state imager. Quantum efficiencies have been measured to peak as high as 75% and, in the case of the backside-illuminated version, may be quite flat from blue to near-IR wavelengths.
Despite the large pixel size of this particular CCD, they are still poorly matched for direct photography to the 278 pm per arcsec scale of the MMT without some sort of optical demagnification, except perhaps on occasions of truly exceptional seeing. We are presently investigating the use of simple lens systems to provide a better match of pixel size to seeing disk to prevent signal degradation due to oversampling. It may, indeed, ultimately prove desirable to design special-purpose optics to perform this function, both to produce better images over wider passbands and to minimize scattered light, as well as perhaps to modify the present beamcombiner to produce a wider field.
It is not yet clear whether the RCA CCD or any other presently available device will prove useful as a detector for spectrographic applications. Such a detector must have exceptionally low readout noise in addition to high charge-transfer efficiency at low signal levels to prevent loss of spatial resolution. Part of the development effort using the present camera system will focus on defining the nature of the problems in these respects, in hopes that the information can be used by the developers of future advanced solid-state imagers to improve performance in these critical areas.
7. ASTRONOMICAL RESEARCH WITH THE MMT
It is a difficult task to describe possible future astronomical research with an instrument before the instrument in question is fully operational, and even before the full array of instrumentation has been defined, much less constructed. Nevertheless, some of the rather unique properties of the MMT and the characteristics of the site suggest the direction that astronomical research with the MMT is likely to take. We describe, in turn, several of the instruments now operatingar soon to be operating on the MMT and the astronomical research program to be carried out with them, as well as some results already achieved in the brief period the MMT has been operating.
As described in Section 6.2, spectrographs on the MMT can operate in either a conventional mode, with all six images superposed, or aligned along the slit in the “in-line’’ mode. To date, only the conventional mode has been used except for some very preliminary tests with the Mount Hopkins Observatory echelle spectrograph, but some interesting results have already been obtained in this mode.
Following the discovery of the twin quasars (0957 + 561 A, B) by Walsh et al. (1979), Weymann et al. (1979) used the Z-machine spectrograph to set a limit of - 15 km s-’ between any differences in the absorption line redshifts in A and B, which substantially strengthened the case for the gravitational-lens hypothesis suggested by Walsh et al. In May 1980, during the first trial of the MMT spectrograph in the conventional mode, Weymann et al. (1980) discovered the triple quasar 1 1 15 + 08, the second example of a gravitational lens (Figure 20). During this first test, Liebert, Latham & Steiner (1980) also observed the magnetic white dwarf in AM Herculis during a rare faint “off” phase and were able to measure a field strength of 20 MG from the Zeeman patterns in hydrogen lines (Figure 21). The excellent seeing at the site, together with the small amount of vignetting over a large spectral range and very good image characteristics of the spectrograph make it well suited for moderateresolution spectroscopy of faint objects where breadth of coverage is important.
A unique feature of the MMT is its use in conjunction with spectrographs operating in the in-line mode. This feature, combined with the frequent occurrence of 1-arcsec and better seeing and the utilization of echelle gratings, allows unprecedented slit transmission even for quite high spectral resolution: In the 15-cm echelle spectrograph for the MMT under preliminary design, an R-2 echelle grating feeds a 75-cm camera which matches a 1-arcsec entrance slit to a 25-pm-resolution detector yielding % 5 km S C ’ FWHM and frequently allowing most of the light to pass through the slit. Typically, a conventional coudt spectrograph under average seeing yielded comparable resolution with a slit transmitting only 5% of the light (Dunham 1956). Again, considering as an example QSO absorption lines, objects at m, - 16.5 can be studied at a resolution that will, judging from the 21-cm data to date, fully resolve the line profiles in these objects and allow definitive abundance analyses.
It is startling to realize that we can now contemplate spectra at 5 km SC’ resolution of objects at m,, 2 16.5; only two decades ago, before the introduction of image tubes, such resolution was obtainable only for objects of about 9 magnitudes brighter with existing coudt spectrographs (Dunham 1956).
7.2 Direct Imaging
Astronomical research at the MMT involving direct imaging was not one of the prime research applications envisioned for the MMT. It is, therefore, of some interest that one of the most noteworthy recent results involved direct imaging with an intensified vidicon of the triple quasar, which showed distinct elongation (incipient splitting?) of the brighter of the three images (Figure 22). This image was obtained using only one of the six telescopes to obtain the best possible resolution and one-second exposures were centroided and shifted and then superposed to increase signal to noise, achieving FWHM of - 0.6 arcseconds. For maximum signal to noise, all six telescopes should be utilized, but of course it is not necessary for the six images to be coincident. To distinguish real image structure from residual telescope aberrations and seeing effects. it is. in fact, advantageous to separate the images on the detector and to coadd them later electronically. A device has been built, based on the concepts involved in the in-line spectroscopic mode, in which six identical small fields, one from each telescope, are arranged in a 3 X 2 format on the single intensified TV camera without adding of the sky background radiation. A program currently under way with this device involves a systematic search of a large number of QSOs for multiple images. Other programs using the SAO CCD camera (Section 6.4) are also aimed at high quality imaging of small areas on the sky.
7.3 Optical Speckle Interferometry
The technique of speckle interferometry allows information on image characteristics down to the diffraction limit of the telescope. (For a general review, see Labeyrie 1978.) It is essential to record the data over times such that the speckle patterns are frozen in position (i.e. a small fraction of a second). In practice, this is achieved by a high-gainintensified TV camera readout onto video tape (Hege et al. 1980).
The use of this technique with the MMT is of particular interest for two reasons. First, it appears that the seeing characteristics of the site often result in a substantial fraction of the energy residing in a small number of speckles, so that the limiting magnitude of the technique may be somewhat fainter (perhaps down to mu - 18) than initially thought. Second, experiments to date with opposite mirror pairs show that the optical support structure is sufficiently stable that speckle information from the six mirrors acting coherently is obtainable. The diffraction pattern is, thus, very similar to that from a single 6.9-m telescope and information down to scales of - .02 arcsec available. This opens up a large number of astronomical research possibilities in solar system astronomy
(e.g. properties of Pluto’s satellite and the diameters of several asteroids), stellar astronomy (studies of preplanetary discs, inner regions of circumstellar shells), and extragalactic astronomy (structure of Seyfert nuclei and searches for sub-arcsec image splitting in the brightest QSOs).
7.4 Research in the Infrared
From its inception, the MMT was designed with infrared research in mind. To this end, the secondary support structures offer a minimum amount of emissive surface area and a set of undersize secondaries for use during infrared work is provided. A variety of IR research programs are well suited to the instrument.
Considerable experience has already been obtained using opposite mirror pairs as a Michelson interferometer (see Section 5.7). The fringe amplitude is measured by rapidly scanning the pattern over a grid placed over the detector. In this way, the visibility function for a given separation has been determined for a number of late-type giants and supergiants in the 2-10-pm region. The Michelson technique has the disadvantage that only one Fourier component of the image is measured at any one time. In principle, direct imaging and speckle interferometry can be carried out in the IR just as it can in the optical. In practice, until a two-dimensional imaging IR device is available, the technique is somewhat different: the image is scanned very rapidly across a single slit placed over the detector and a (one-dimensional) power-spectrum analysis is performed; the intrinsic telescope plus atmosphere modulation transfer function is obtained from nearly simultaneous observations of a nearby point source. The applications of this technique are, as above, to resolve late-type supergiants and possible preplanetary discs. Details on the IR structure of the nucleus of NGC 1068 should be possible. An interesting proposal has been made to operate the MMT as a phased array at 20 pm, not simply with the objective of obtaining increased resolution but to obtain a 6.9-m diffractionlimited image so that very small apertures can be used to reduce the background to an absolute minimum for 20-pm photometry. A photometer with the six IR secondaries chopping in phase will shortly be placed in operation and will be equipped with an indium antimonide detector for work in the 1-3-pm region and a bolometer for work at 5-20 pm. At 2 km most QSOs are readily observable, and the brighter ones can be studied with low spectral resolution. Giant ellipticals with z > 1 ought to be measurable. A circular variable filter spectrometer will be able to observe the broad emission lines of active nuclei and the brightest QSOs. Fourier transform spectrometers operating in the 50-25-km s ' range will be used to examine spectral features in very red objects of special interest (e.g. MWC 349) and to study processes of star formation (e.g. 2-pm spectra of the stellar objects embedded in the p Ophiuchus complex).
8. FUTURE MULTIPLE OBJECTIVE TELESCOPES
The supergiant telescopes of the future, either in space or on the ground, will almost certainly be multiple objective telescopes. In space, the need to send up the telescope in moderate-size pieces will preclude a monolithic primary. Even then, the cost of such a teleyope, at present, is prohibitive.
The supergiant ground-based telescope is within reach and is extremely worthwhile. The cost per second of dark time observing with a 15-m ground telescope and a 2.5-m space telescope is similar, though the overall cost, annual cost, and amount of dark observing time are an order of magnitude less on the ground. The ground telescope will, for example, be advantageous for spectroscopic observations of any object brighter than the sky, studies of extended objects of low surface brightness, in multiple object spectroscopy (e.g. galaxies in a cluster), in spectroscopic observations between 1 and 3 pm where detectors are limited by internal noise, in high angular resolution infrared and optical (speckle) observations, and in the rapid development of new equipment. It would be useful for submillimeter astronomy at 0.3 mm. For sub-arcsecond seeing conditions it would compete successfully with the Space Telescope in optical sky backgroundlimited research. Two studies, by Faber (1980) and Strittmatter (1980), have explored many of these options in detail. Some graphs that demonstrate the multiple object advantage are in Woolf & Angel (1980).
A number of proposals for building giant multiple objective telescopes are discussed in Hewitt (1980). There currently seems to be some hope for construction of multiple objective telescopes between 10 and 25 meters for the USSR, European Southern Observatory, USA, and the state of California. Kitt Peak Observatory and the Universities of Arizona, California, and Texas are attempting a joint program of technology development to lead to a 2 15-m telescope. It has been accepted that it is inappropriate to attempt to build a monolith of this size, and arrays and bowls are not suitable for the infrared. Thus, the technology development program has among its main tasks to explore the reasons for choosing between a multiple objective telescope of Type A or B (Type A is segmented primary, Type B is multiple objective). The Kitt Peak NGT (next generation telescope) program explored both of these options at the 25-meter size as the steerable dish and the large MMT. Further developments of the Type A system have been conducted at the University of California by J. Nelson (1980) and colleagues, while a study of a large Type B system has been carried out at the University of Arizona.
The University of California study has been concerned with developing a primary mirror that retains its shape regardless of applied forces. The relative positions of 1.4-meter hexagons are sensed, and position actuators correct any errors. The group has developed sensors and actuators as well as a method of bending and polishing precise off-axis paraboloids, which has been successful for a 0.35-meter circular prototype. The University of Arizona study, in contrast, has explored the possibilities of adaptive optics, in so far as current knowledge of seeing parameters permits. Also, they have explored the use of light pipes, and the use of the separate subtelescopes for highest visual efficiency and smallest visual images by instantaneous wavefront tilt corrections. Both studies have explored som? techniques that are applicable to both Type A and Type B systems. The experience with the MMT has pointed to two major changes that should be incorporated in future instruments. First, the experience with phasing the telescope for infrared interferometry has encouraged us to expect that we can make an MMT where all the elements are held in phase, that reaches the diffraction limit of its entire aperture rather than of that aperture's component parts. Second, we have seen the separate star images being moved around by differential seeing so that they looked like a swarm of bees. We believe that we can often use fast coalignment on stellar images to make these coalesce and so obtain higher angular resolution than we would on a rigid telescope of the same area.
Type A and Type B systems have different faults and merits so that at this time the choice between them is not obvious. A Type A system starts with a lightweight primary because the segments are thin, and this saving in weight propagates through the entire telescope. To reduce cost still further, the telescope and dome can be shrunk by reducing the focal ratio. This makes alignment tolerances very tight, though this is not a problem if alignment is carried out on a star image. More important, the provision of a large, high quality, achromatic field with low light losses becomes difficult. Currently, one can attain acceptable performance at a primary ratio of fl2.0, and ingenious optical design might push as far as perhaps f / 1.6. A Type B system has a much shorter tube for the same mirror area, so the minimum dome can be smaller around a Type B system. If dome cost is proportional to (size)*', then the minimal dome around a Type A system off / 1.7 will cost twice as much as one around an 8-mirror Type B system of fl2.7. In existing observatories, domes are typically one third the project cost, so that such factors are important. The smaller size could in principle be used in a Type B system for telescope weight and cost reduction, but it is not yet clear whether the tubes of Type A and Type B systems have the same demands for rigidity and high frequency response.
The provision of the high quality wide field is more easily met in an fl2.7 primary in Type B. For visual imaging, one should probably use multiple CCDs and electronic image combining so that use of the individual foci is no disadvantage. For multiple object spectroscopy one can as easily lead light pipes from separate foci as from a single one. Thus, in wide field visual use Type B is at an advantage. However, when one comes to developing a phased field for coherent imaging, Type B suffers from a small amount of vignetting at quite modest field angles and for some purposes this may cause problems. Thus, for phased work with wide field, Type A is advantageous. Phasing a Type A system is initially very time consuming if there are many facets. However, a Type B system has the telescope structure enter into the phasing, so that although the phasing operation is fast, it must be carried out more frequently. It is possible that
laser interferometer distance sensing might eliminate these problems with both types of systems.
The biggest questions concern the primary optics and the extent to which the telescope figure must be operable by “dead reckoning” rather than by setting up on a star or continuously using an offset reference star. Dead reckoning probably needs very stable mirror materials like ULE or Cervit. The blanks are relatively expensive, and there may be both large costs and manufacturing delays for a Type B system. On the other hand, for a system with partially adaptive optics, thin pyrex mirrors or pyrex honeycomb mirrors may be perfectly acceptable, and their use might notably reduce the cost of the Type B system. Honeycomb is not appropriate for bend and polish, and unless one keeps solid pieces very thin, they might not work out well for segments in a Type A system. Further, the focal tolerances are so tight in a Type A system that Cervit may be needed for this reason alone. So far, pyrex honeycomb has only been built in small sizes, so it is not yet an established technique.
The diffraction patterns of phased Type A and B systems are quite different, with a Type A more closely resembling that of a filled circular disk. Assuming perfect optics, all telescope systems of the same mirror area should produce the same central intensity in their diffraction patterns. Thus, as the width of a dilutable mirror system increases, and the angular resolution goes up, less of the energy finds its way into the central maximum, and one expects to find that (energy inside 1st zero) x (resolving power)2 is constant from system to system. In Table 6, we show the result of calculating these quantities, and also the comparative signal-tonoise ratio for an infrared telescope with an optimum small diaphragm. It can be seen that a Type B system will differ from a Type A system in that the Type A will have a somewhat higher signal-to-noise ratio for IR observations and a somewhat lower resolution. It is not easy to put these relatively minor differences into perspective. At wavelengths shorter than 5 3 fim, the light concentration will be set by seeing and there will be no appreciable difference between the systems. The highest sensitivity IR observations will be made from space, and ground observations will mainly refine these with measurements of image position, structure, and time variability. Here, the high resolution should be advantageous. On the other hand, for IR spectroscopy the energy concentration is likely to be more important.
If it should turn out that costs distinctly favor one kind of system, or, rather unlikely, that a major problem should be found in one kind of system, then the choice may be easy. At the time of this writing, it does not seem likely that the choice will be so painless. Thus, while it is clear that the future lies with either the Type A Segmented Primary Telescope, or Type B Multiple Mirror Telescope, the crystal ball is cloudy when it is asked which should be preferred.
With the construction of the Multiple Mirror Telescope a major new and novel facility has been added to the capabilities of the world of optical nighttime astronomy. The MMT is being used not only to gather high quality astronomical data but also to test new techniques for astronomical telescope construction. It is quite remarkable that a telescope like the MMT with the many deviations from conventional optical telescope technologies has, in fact, become a very high quality, very usable facility in a very short time after it was nominally completed. The credit for that goes to the many competent and dedicated scientists, engineers, and others who participated, and who participate now, in the design and implementation of the telescope. The MMT Dedication Symposium (SAO Special Report 385) listed over 340 people who are associated with this venture. We do not attempt to give such a full list. We do, however, want to specially acknowledge the contribution to the MMT by some who have contributed in a major way, who have dedicated a major part of their professional lives toward the realization of this unique- facility, and who are not coauthors of this chapter: T. E. Hoffmann, A. Meinel, M. Reed, G. Sanger, F. Whipple, and J. T. Williams. The help of such capable administrators as J. Gregory and P. Sozanski from SAO and R. Kassander and A. Weaver from the University of Arizona was essential to accomplish the MMT. The love and dedication towards the project of these and of many other participants are the reason for the success of the Multiple Mirror Telescope. We gratefully acknowledge the help of M. Green, V. Tersey, and G. McLoughlin in the preparation of this manuscript.
Part of the material presented in this publication is based upon work supported by the National Science Foundation under Grants AST 76- 20822, AST 76-21732, and AST 79-25421.
1The Multiple Mirror Telescope Observatory is a joint venture of the Smithsonian Institution and the University of Arizona.
Angel, J. R. P., Hilliard, R. L., Weymann,
Rarr. 1. 1980. Optical and Infrared Telescopesfor
Rasov, N. G., Dimov. N. .4., Gvozdev, M. I..
1979. The MMT and the Future of
Beckers, J. M. 1980. MMTO Technical Report
Burke, J. J., Kirchoff, W. 1968. Sky and
Chaffee, F. H. 1974. Astrophys. J. 189:427-40
Carleton, N. P., Hoffmann, W. F. 1978.
Davis, M., Latham, D. W. 1978. Instrumentation
in Astronomy 111. Proc. Soc.
Dunham, T. Jr. 1956. Vistas in Astronomy 2: I223
Faber, S. 1980. A Scientific Casefor a 10
Faller, J . E. 1978. Optical Telescopes ofthe
W. Richter, R. N. Wilson, pp. 301-11
Fastie, W. G. 1961. J. Opt. Soc. Am. 51:1472
Fastie, W. G. 1967. Applied Optics 6.397-402
Fotescu. G., Heilbrunn, G. 1909. UK Patent
Grainger. J . F. 1979. The MMT and the
Green, J. R. 1979. Literature Search on
Grundmann, W. A., Richardson, E. H.
Hall. D. N. B. 1978. Optical Telescopes of
Hanbury Brown, R. 1964. Sky and Telescope.
Aug: 64-69. See also Space Physics
Hege, K., Hubbard, E. N., Strittmatter,
P. A. 1980. Applications of Digital Image
Hewitt. A. 1980. Optical and Infrared Telescopes
Hoffman, T. E. 1978. Optical Telescopes
Hoffman, T. E. 1980. The MMT and the
Horn-d'Arturo, G. 1935. Publ. Oss. Astron.
Horn-d'Arturo, G. 1950. Publ. Oss. Astron.
Horn-d'Arturo, G. 1953. J. Brit. Astron.
Horn-d'Arturo. G. 1955. Publ. Oss. Astron.
Horn-d'Arturo, G. 1956a. Publ. Oss. Astron.
Horn-d'Arturo, G. 1956b. Optik 13 Vol 6:254-58
Ingalls, A. G. I95 1. Scientific American
Jacchia, L. 1978. Sky and Telescope 55:
Labeyrie, A 1978. Ann Rev. Astron. Astrophvs.
Larson, H., Fink, U. 1975. Applied Optics
Leach, R. et ai. 1980. Submitted to Publ.
Astron. Soc. Par. Learner, R. C. M. 1978. Optical Telescopes
Liebert, J., Latham, D. W., Steiner, K. G.
LOW. F. J. 1979. The MMT and the Future
Low, F. J., Rieke, G. H. 1974. The Instrumentation
McCarthy, D. W., Low, F. J., Howell, R.
McDonough, D. F. 1980. MMTO Technical
Meinel, A. B. 1970. Applied Optics 9:2501-
Meinel, A. B. 1978 Optical Telescopes of
Meinel. A. B. 1979. The MMT and the Future
Meinel, A. B., Meinel, M . P. 1980. Optical
Mertz, L. 1970. Optical Instruments and
Mertr, L. 1980. Optical and Infiared Telescope~
Nelson, J. 1980. Optical and Infrared Telescopesfor
Oort, J. H. 1980. Astron. Astrophysics.
Owmantown. Lord. 1828. Edinburgh Journal
Radau, R. E., Ulich, B. L. 1980. MMTO
Reed, M . A. 1978. Optical Telescopes for
Richardson, E. H., Grundmann, W. A.
Rieke, G. H. 1980. Ap.p. lied Oprics. Submitted
Ruda, M.. Turner, T. J . 1980. MMTO
Shannon, R.R. 1979. The MMT and the
Shectman, S. A. 1976. Publ. Astron Sor.
Steqhenko, N V 1979 The WMT ond the
Strittmatter. P. A. 1980. Optical and Infrared
Synge, E. H. 1930. Phil. Mag. 10:353-60.
Thomas, W. 1965. US Patent No. 3507547
Ulich, B. L., Riley, J. T. 1980. MMTO
Vaisala, Y. 1949. Urania Barcelona 220:
Walsh, D., Carswell, R F., Weymann, F. J.
Weymann, R. J.. Chaffee, F. H.. Davis, M.,
Weymann, R. J., Latham, D. W., Angel,
Woolf, N . , Angel, R. J. P. 1980. Optical