Liquid mirrors: a progress report E.F. Borra, R. Content, M.J. Drinkwater, L. Girard, L.—M. Tremblay and S. Szapiel1 Centre d'Optique, Photonique et Laser, Département de Physique, Université Lava! B. Gibson and P. Hickson Department of Geophysics and Astronomy, University of British Columbia C. Morbey Dominion Astrophysical Observatory, Herzberg Institute of Astrophysics 1 Now at Institut National d'Optique 1. ABSTRACT Having realized that liquid mirrors are useful in science (e.g. astronomy, atmospheric sciences, optical testing, etc...), we have undertaken work to determine whether they are technologically feasible. We have built a testing tower equipped with a scatterplate interferometer interfaced with a CCD for data acquisition and a microcomputer for data analysis. With this equipment, we have tested a 1 .5-meter diameter f/2 liquid mirror, showing that it is diffraction limited; interferometric measurements give Strehl ratios of order 0.8. We discuss work that we are carrying out to improve the basic technology. We describe briefly a 2.7-rn diameter liquid mirror and astronomical observatory presently under construction. 2. INTRODUCTION It is rather straightforward to show that, in a rotating fluid, adding the vectors of the centrifugal and gravitational accelerations gives a surface that has the shape of a parabola. Using mercury one gets therefore a reflecting parabola that could be used as the primary mirror of a telescope. The focal length of the mirror L is related to the acceleration of gravity g and the angular velocity of the turntable coby L=g/(2w2). (1) For large mirrors of practical interest the periods of rotation are 5 to 20 seconds and the linear velocities of the rims range between 5 and 20 km/h. Liquid mirrors have many interesting properties that make them useful to several fields of Science. For example, they can be used to make very large and inexpensive aspherics useful in Astronomy, as the primary mirror of a zenith telescope, as LIDAR receiver or as large collimators for optical applications. To give an idea of this cost advantage, we quote the costs of mirrors that we have built: a 1.5-rn diameter liquid mirror costs 20,000 dollars, and a 2.7-rn liquid mirror costs30,000$. In Astronomy, they promise a major revolution for deep surveys and, in particular, for cosmological studies as the technology should allow us to build very large reflectors at low costs. Theoutstandinglimitation of liquid mirrors, that they cannot be tilted, is not a serious handicap in Cosmology. Note that what is actually important is not whether it is possible or not to tilt a telescope but, rather, how wide a field is accessible from a single telescope; in this respect, the work of Richardson andMorbey SPIE Vol. 1236 Advanced Technology Optical Telescopes IV(1990) / 653 Downloaded from SPIE Digital Library on 20 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms 1 relevant. They have shown that using stressed mirrors as correctors would allow us to observe in small regions over a 1 5 degree wide field. Their design is somewhat cumbersome but it should be possible to make more practical designs (G. Lemaiire, private communication). Atmospheric scientists have also expressed great interest for these inexpensive large mirrors for LIDAR applications. Liquid mirrors have, or promise, interesting properties for many optical applications: very high surface quality, low-scattering, very high numerical apertures, variable focus that can be controlled with a very high precision. 3.OPTICAL SHOP TESTS Early simple optical tests of a 1-m prototype liquid mirror2 indicated that the basic concept was sound. We therefore decided to put the necessary effort to build better testing facilities beginning in the fall of 1986. We built a sturdy 7-meter tall testing tower capable of testing mirrors having diameters as large as 2.5meters. Fig. 1 Exploded view illustrating the basic mirror setup SYNCHRONOUS MOTOR Fig. 1 gives an exploded view that illustrates the basic mirror setup. The mirror and bearing rest on a three-point mount that is used to align the axis of rotation parallel to the lines of force of the gravitational field of the Earth. This alignment is easily made to within a couple of arcseconds with a spirit level. For optimal tuning of the mirror in the laboratory, we improve it with a Ronchi test or scatterplate interferometer. The mirror rotates on a commercial air-lubricated bearing having radial and axial errors of 1 micron and a coning error of 0.2 arcseconds. We use airbearings because they are convenient for small systems. Larger systems will probably use oil-lubricated bearings. The turntable is driven by a synchronous motor which is coupled to it via pulleys and a thin mylar belt. The synchronous motor is controlled by a custom built variable-frequency AC supply stabilized with a 654 / SPIE Vol. 1236 Advanced Technology Optical Telescopes lV(1990) BELT CONTAINER PRESSURE REGULATOR TO AIR SUPPLY THREE-POINT MOUNT Downloaded from SPIE Digital Library on 20 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms crystal oscillator. We can control the rotational velocity of the table and thus the focal length of the mirror by changing the frequency of the power supply. Our tests show that focus is stable. We have made containers with a variety of construction techniques, from simple flat plywood sheets to light weight composite material containers. Final figuring of the containers is done with spincasting. The turntable is spun at the appropriate angular velocity while we pour liquid epoxy resin in the container. The epoxy surface takes rapidly the shape of a parabola and is allowed to harden while the table is spinning. Although spincasting seems an easy thing to do, in practice, it requires experimenting and knowledge of the type of resins available. For example, we find that epoxy resins give better results than polyester. Spincasting gives a parabolic surface that is good to about 0. 1-mm, which is adequate for our purpose. To save weight and cost, we try to work with a layer of mercury as thin as possible; it can be thought as a liquid high reflectivity coating; we have developed techniques that allow us to work with layers ofmercury as thin as 1-mm. The largest mirror that we have tested has a diameter of 1 .5 meters and f/2 numerical aperture. The testing setup and the results have been described in 3 ,where a block diagram of the testing setup is given, but we give below a short summary, along with unpublished information. The mirror is illuminated either with a He-Ne laser, or a continuum source and interference filter, followed by a spatial filter assembly. The data are obtained at the center of curvature of this mirror, we therefore must use custom-built null lenses to correct the large spherical aberration present at the center of curvature of a parabolic mirror. Note that, because we image through null lenses, we test the entire surface of the mirror. This is a major improvement over the tests in 2 where the 1-m mirror had to be diaphragmed to f/9 (diameter = 0.1 8 meters) to reduce spherical aberration. The null lenses used reimage the mirror at f/12 (f/24 at the center of curvature). We have various test equipment, such as interferometers or knife-edge and Ronchi test assemblies. For data acquisition, we use a 5 12X480 CCD detector connected to an 8-bit framegrabber interfaced with a PC/AT clone computer. Image analysis is carried out with commercial as well as our own software. Fig.2. The Airylike pattern of a 1.5-m liquid mirror observed with a CCD camera. To emphasize the faint rings, the center of the PSF was greatly overexposed. Radial cutsthrough the image show that the rings are very weak (2.5% of peak intensity for the first brightring). The first dark ring is, as expected, at a radius of 0.11 arcseconds. SPIE Vol. 1236 Advanced Technology Optical Telescopes IV (1990) / 655 Downloaded from SPIE Digital Library on 20 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms Figure 2 shows a magnified image of the point-like object, created by the laser and spatial filter, observed with the CCD camera and using a 1/30 second exposure (see 3 for another image of the PSF) ; we can clearly see the Airy-like pattern of the 1 .50-m mirror. To emphasize the faint rings, the center of the PSF was greatly overexposed and the intensity scale was altered and is highly nonlinear. Radial cuts through the image show that the rings are actually very weak ( 2.5% of peak intensity for the first bright ring). Measurements find that the rings of the Airy pattern are at the predicted location, the first dark ring is measured at a radius of 0. 1 1 arcseconds from the center of the PSF. We videotaped hours of data and find that the Airy pattern is always visible, although the intensity and symmetry of the rings varies a little, possibly from seeing in the testing tower. Figure 2 gives a spectacular illustration of the imaging quality of the mirror. To the best of our knowledge, this is the first time that the Airy pattern of a large mirror has been directly observed. Ronchi, knife-edge and interferometric tests show that the focus of the mirror is stable. At the center of curvature of the mirror, the PSF is given by the square of the amplitude of the Fourier transform of the pupil function. As a consequence, the central part of the Airy pattern as seen in Figure 2 is produced mostly by low spatial frequency components, while high spatial frequency ripples cause extended wings and scattered light far from the center of the PSF. We do indeed see low-amplitude high frequency ripples caused by vibrations originating in the building (see discussion below). Some preliminary evaluation of scattered light is discussed in 3 by comparing the azimuthally averaged PSF of the mirror to those of conventional telescopes. We found a small excess of scattered light due in part to the large number of auxiliary surfaces used for testing surfaces and in part to a spiral-like structure that rotates with the mirror. This spiral structure has a very small amplitude ( 30.01 MB/hr SPIE Vol. 1236 Advanced Technology Optical Telescopes lV(1990) / 659 Downloaded from SPIE Digital Library on 20 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms To minimize the presence of concentric waves in the Hg set up by vibrations, we designed a support structure whose resonant frequency minimizes them. Any waves that are set up should be damped as quickly as possible, preferably before traveling one wavelength. Following 10 concerning ourselves with the low frequency regime, where the gravity wave term dominates the capillary term in the surface wave equation (i.e.X >> 1 .2 cm), one finds that the characteristic damping time,t ,for these waves is given by: ,r= g 2p/32irv 26 A4 [secJ (2) where a --487 dynes/cm is the Hg surface tension, g--98 1cm/s2 is the gravitational acceleration, p = 13.57g/cm3 is the Hg density, and T =0.016 g/crn/s is the Hg viscosity. In other words, equation (1) is telling us that if one is within the low frequency regime (A>> 1 .2 cm), wave damping times increase rapidly. To avoid this regime, one must impose the restriction V > 8.3Hz ( i.e. A > 1.2cm) as a lower limit to mirror vibrations. The most important manifestation of this effect is the fundamental resonant frequency of the tripod support structure. By judicial selection of tripod material, cross-section, and length, a resonant frequency of '.'6OHz was established, safely distancing ourselves from the region of long surface wave damping times. An effect which must be taken into account is that due to the expansion or contraction of each tripod leg due to thermal variations. The expansion of a leg, y, under a temperature change, \T, is given by: 7= 1 Cexp AT [cm] (3) where 1 560cm is the leg length and cexp 2.25 1O cm/cmPC is the coefficient of linear expansion of the aluminum alloy. For a temperature change A T=1OC, typical for a night's observing, the leg will change its length by y -'1.3mm. This corresponds to a focal point shift of ''1.5mm, or equivalently, a defocussing of --6&csec. for every temperature change of 1C. A temperature-sensitive auto-focussing mechanism is incorporated into the LMT prime focus detector structure to compensate for such fluctuating thermal conditions. The primary observing program for this 2.7m LMT is a year-long narrowband photometric galaxy redshift survey utilizing a set of fifty 100 A filters. With a signal-to-noise ratio of ten, there is the opportunity to obtain photometric data capable of determining redshifts for approximately 1500 galaxies for which V< 18.3, over the course of the year 11 Similar spectrophotometric multicolor galaxy surveys have been discussed recently 12,13,14 Table II lists some of the expected performance estimates related to this instrument's observing program. 660 / SPIE Vol. 1236 Advanced Technology Optical Telescopes lV(1990) Downloaded from SPIE Digital Library on 20 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms TABLE II Narrowband Photometry Performance Estimates Parameter Value Number of Filters 50 Filter Bandwidth 100 A Filter Transmission 50% Typical Galaxy Diameter Sarcsec. Nightly Integrated Limiting Galaxy V (SNR=1O) 18.3 Number of Galaxies in Survey Area with V<18.3 1500 5. CONTRIBUTIONS TO THE TECHNOLOGY Liquid mirrors are a new technology and we are exploring many basic concepts to decrease costs, increase performance and reliability. Among the many things that we have explored : the use of mylar as a cover to protect from the wind, the use of monomolecular layers to decrease the effect of the air on the surface, discovery of a simple technique that allows us to work with very thin layers of mercury (and therefore decrease weight and costs) and demonstration of the Intermediate Damping Layer technique (IDL). We have shown that, with the IDL technique it is possible to improve dramatically the quality of inexpensive but poor quality mechanical bearings 15, The cost of the present prototypes is dominated by the bearing and, to a lesser extent, the container and mercury itself. The costs of these components increase with the weight of mercury. In particular, most of the load on the container and the bearing comes from mercury. We could not use mercury layers thinner than 4 millimeters in the past for, otherwise, surface tensions broke the surface. We have now developed a very simple technique that produces thin layers on any substrate. We have produced layers as thin as 0.5 mm on rotating 50 cm liquid mirrors. Ronchi tests show that surface quality is independent of thickness and that stabilization times are actually shorter for the thinner layers. Layers thinner than 1 mm can be produced but necessitate a better surface finish for the container so that, at some point, there is a diminishing cost advantage to thin layers. The tests of the 1.5-meter mirror were carried out with a mercury layer 2.3 mm thick. The present technique that we use to establish thin layers of mercury is however somewhat inconvenient. We can establish directly 2.5-mm thick layers but thinner layers must be pumped down from thicker ones. It would be desirable to find techniques that establish directly thin layers. We thus have explored two basic ways of doing this: decreasing surface tensions and making mercury wet the container. For example, we have carried out experimental exploratory work on the electrocapillary effect on mercury. Out of this we have identified a promising new technique to directly establish thin layers. We also have looked at the effect of contaminants and monomolecular layers. It is difficult to find substances that mercury wets, with the notable exception of metals that form amalgams with it. Unfortunately the amalgams rapidly form a surface layer of oxide that tarnishes it beyond usefulness. We have tried electrocoating an epoxy surface with mercury, with mixed results. We also have studied the effects of roughening and etching on the wettability of a surface, concluding that, contrary SPIE Vol. 1236 Advanced Technology Optical Telescopes lV(1990) / 661 Downloaded from SPIE Digital Library on 20 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms to water, smoothness helps the adherence of mercury. We have identified some other possible ways to make plastics wettable but have not yet explored them. Liquid surfaces are sensitive to turbulent winds. Sheltering from outside winds is not a major concern but, because the container rotates, this causes a relative wind which is likely to generate waves at some speed. A solution to the wind problem is to use a thin mylar sheet to shelter the liquid from the wind. Very thin mylar sheets are available from many companies. The quality of mylar over small regions can be quite good. Interferometric tests that we have made on a 8-micron thick 5-cm diameter mylar sample show quite acceptable quality ( P-V = ?J6). We have examined larger samples, that covered the entire 1.5-m mirror, with a Ronchi test; we saw large scale variations that we have not measured quantitatively. However, we did image a resolution chart, resolving 0.5 arcsec. Thinner mylar (down to 2 micron-thickness) is available and may have very good quality. We are investigating this. We have measured the transmission curves of mylar samples, it is transparent from 3,000 A to 6 microns, after which it is mutilated by deep absorption troughs. Monomolecular layers are very effective at decreasing the effect of the wind on water and dampenmg vibrations in general. We have carried out preliminary investigations of monomolecular layers on mercury. We have made an extensive library search on the properties of monolayers in general and mercury in particular. We have established monolayers mercury and have identified problems involved with this; for example, the mercury surface must be very clean. 6. ACKNOWLEDGMENTS This research has been suported by the Natural Sciences and Engineering Research Council of Canada and the National Research Council of Canada. 7. REFERENCES 1. Richardson, E. H. and Morbey, C.L. 1987, "Instrumentationfor Ground-Based Optical Astronomy Present and Future ", ed. L.B. Robinson (New York:Springer-Verlag) 2. Borra, E.F., Beauchemin, M. , Arsenault, R., and Lalande, R. 1985, Pub. Astron Soc. Pac. 97, 454. 3. Borra, E.F., Content, R. Drinkwater, M.J. and Szapiel, S. 1989, Astrophys. J. (Letters) 346, L41. 4. Wetherell, W.B. 1980, "Applied Optics and Optical Engineering ",Vol VIII, eds. Shannon, R.R., and Wyant, J.C., Academic Press, New York, p. 171 and p. 252. 5. Borra, E.F, Content, R., Poirier, S. and Tremblay, L.M. 1988, Pub. Astron Soc. Pac. 100, 1015. 6. Content, R., Borra, E.F, Drinkwater, M.J., Poirier, S., Poisson, E., Beauchemin, M., Boily, E., Gauthier, A., and Tremblay, L.M. 1989, Astron.J. 97, 917. 7. Borra, E.F. 1982, J. Roy. Astron. Soc. Canada 76, 245. 8. Gibson, B.K. and Hickson, P. 1990, " The Evolution ofthe Universe ofGalaxies: The Edwin Hubble Centennial Symposium\", ed. Kron, R.G. (Salt Lake City: Brigham Young University Press). 9. MacKay, C.D. 1982, SPIE 331, 146. 10. Landau, L.D. and Lifshitz, E.M. 1959," Fluid Mechanics ", (New York: Pergamon Press). 1 1. Tyson, J.A. and Jarvis, J.F. 1979, Astrophys. J. (Letters), 230, L153. 12. Loh, E.D. and Spillar, E.J. 1986, Astrophys. J. ,303, 154. 13. Couch, W.J., Ellis, R.S., Godwin, J., and Carter, D. 1983, Monthly Not. Roy. Astr. Soc., 205, 1287. 14. Koo, D.C. 1985, Astron. J. , 90, 418. 15. Borra, E.F, Content, R. and Boily, E. 1988, Pub. Astron Soc. Pac. 100, 1399. 662 / SPIE Vol. 1236 Advanced Technology Optical Telescopes IV(1990) Downloaded from SPIE Digital Library on 20 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms