Active Optics and Wavefront Sensing at the Upgraded 6.5-meter MMT

Transcription

1 Active Optics and Wavefront Sensing at the Upgraded 6.5-meter MMT T. E. Pickering a,s.c.west b,&d.g.fabricant c a MMT Observatory, 933 N. Cherry Ave., Tucson, AZ 85721, USA; b Steward Observatory, 933 N. Cherry Ave., Tucson, AZ 85721, USA; c Harvard-Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138, USA ABSTRACT Shack-Hartmann wavefront sensors have been commissioned and are now in routine use at both of the optical foci (f/9 and f/5) of the converted MMT. Both units are of moderate resolution with 14x14 square apertures across the pupil for f/5 and 13x13 hexagonal apertures for f/9. They share a common software interface that fits a set of 19 Zernike polynomials to the wavefront errors. Zernike focus and coma are corrected by moving the secondary mirror, third order spherical by a combination of secondary motion and primary bending, and the rest by primary bending alone. In this paper we will describe the two wavefront sensor systems and how they have performed thus far. Keywords: Shack-Hartmann, Wavefront Sensing, Active Optics, Large Telescopes 1. INTRODUCTION Modern large-aperture telescopes require active primary mirror support systems to maintain the mirror s figure in the face of changing loads due to gravity and to wind forcing. Borosilicate mirrors, such as the 6.5-meter MMT primary, can be distorted by temperature gradients. Tight collimation tolerances are imposed by the very fast MMT f/1.25 primary mirror. Care must be taken to measure and correct the wavefront if the excellent images allowed by the site are to be preserved. Shack-Hartmann wavefront sensors are a common and powerful tool for measuring wavefront error in optical systems. 1 They form the basis for many adaptive optics systems (including the adaptive f/15 secondary system for the converted MMT 2 ) and have proven to be very effective for measuring large telescope optics (e.g. Magellan 3 ). In addition to the adaptive f/15 system, the MMT has two other operational foci: an f/9 system designed for compatibility with preconversion MMT instruments and a wide-field f/5 system capable of providing highlycorrected images over fields 0.5 diameter in imaging mode and 1.0 diameter in spectroscopic mode. Shack- Hartmann wavefront sensors for both of these foci have been commissioned and put into routine use over the past year. Both of these systems use probe mirrors that are positioned on the telescope optical axis between scientific observations. The design and implementation of the f/9 and f/5 systems are discussed in 2 and 3, respectively. The software used to perform the wavefront analysis is described in 4. Results from the two systems and their overall performance to-date are discussed in 5. Author Addresses: tpickering@mmto.org, swest@as.arizona.edu, dgf@cfa.harvard.edu

2 Shack Hartmann Optical Path CCD 1:1 Viewer Optical Path Steering Mirror Motors Figure 1. The as-built f/9 Shack-Hartmann device prior to installation in the top box. Two selectable light paths provide the wavefront sensor optics or a 1:1 image relay. The system was built almost entirely from off-the-shelf parts. 2. F/9 WAVEFRONT SENSOR Much of the early testing and tuning of the converted MMT s optics was performed using an interferometric Hartmann device designed by S. C. West, 4 first at prime focus and later at the f/9 Cassegrain focus. Although it has superb sensitivity to small aberrations, large wavefront errors require very fine spatial sampling to avoid nλ redundancy. This dense spatial sampling reduces the efficiency of this system. A medium resolution Shack- Hartmann wavefront sensor is more suitable for nightly operations, and it was decided to build a compact unit that could be installed within the existing f/9 top box. The limited field-of-view of the f/9 top box made a continuously operating off-axis design impractical, so the wavefront sensor is used on-axis. This requires science observations to be interrupted, though spectroscopic observers can still use the top box comparison sources during wavefront sensor observations. Figure 1 shows a labeled picture of the completed instrument before its installation in the f/9 top box. The whole assembly is mounted on a two-axis motorized slide with one axis used to adjust focus and the other to select between wavefront sensor and direct imaging optics. The wavefront sensor optics consist of a 45 mm focal length 13x13 hexagonal geometry lenslet array, an 80 mm focal length collimator lens, and an RG610 blue cutoff filter. The camera is an Apogee KX-260 CCD (512x512 with 20 micron pixels and 14-bit readout). The effective wavelength of the filter plus CCD is 7800 Å and in practice a 9th magnitude K star can saturate the camera within 30 seconds in good seeing. Example reference and data images taken with the system are shown in Figure 2.

3 Figure 2. Representative reference (left) and data (right) images taken with the f/9 Shack-Hartmann wavefront sensor. The data image is a 5 second exposure taken near zenith in 0.5 seeing. 3. F/5 WAVEFRONT SENSOR Maximizing the corrected field-of-view available to the instruments was a primary design consideration for the converted MMT s f/5 focus. No space was available in the first generation instruments (Hectospec/Hectochelle and Megacam) for continuous, off-axis wavefront sensing. The f/5 wavefront sensor is deployed between science observations like the f/9 wavefront sensor. Because they do not use the entire available field of view, future f/5 instruments will include a provision for continuous wavefront sensing. The f/5 wavefront sensor system consists of a commercial Shack-Hartmann wavefront sensor and a separate CCD camera with a filter wheel that can be used for time-critical imaging while a spectroscopic instrument is mounted at the f/5 focus. The wavefront sensor we use is a slightly modified Puntino unit from Spot Optics. We modified it by adding a negative field lens to image the system pupil on the lenslet array, and we replaced the original 0.3 mm pitch lenslet array with a coarser 0.6 mm lenslet array (14x14 samples across the pupil). The collimator and lenslet array focal lengths are both 40 mm. The wavefront sensor camera is a SBIG ST9XE (512x512 with 20 µm pixels). A beamsplitter sends half of the light to the wavefront sensor and the other half to an acquisition camera, a Pixelink PL-A641. The direct imaging CCD camera is an Apogee AP8p that contains a thinned 1024x1024 chip with 24 µm pixels (about 0.14 /pixel). The wavefront sensor package mounts inside the instrument rotator between the wide-field corrector and the instruments. It has four axes of motion: 1. A long-travel linear slide that carries the optics along a line intersecting the optical axis. 2. A select stage that puts either the Puntino or direct imaging camera in the beam. 3. A focus stage mounted on the select stage for focusing both the Puntino and direct CCD. 4. A tilt stage that carries the first fold mirror and allows the instruments to be aligned with the chief ray off-axis. Figure 3 is a drawing of the wavefront sensor and the instrument rotator area, and Figure 4 is a drawing of the system with the cover removed. These four axes are controlled by a Delta Tau PMAC motion controller. There is an additional axis of motion within the Puntino for selecting between sky or a reference light source. The motors, electronics, and cameras are all controlled by a Windows XP computer embedded within the wavefront sensor electronics box. Short exposures of 9.5 magnitude stars can saturate the f/5 wavefront sensor s 15-bit readout when the seeing is good. Representative reference and data images are shown in Figure 5.

4 Figure 3. Cutaway schematic of the f/5 wavefront sensor system as it is installed. Figure 4. Drawing of the f/5 wavefront sensor package with its cover removed. The main stage assembly is deployed to the on-axis position.

5 Figure 5. Representative reference (left) and data (right) images taken with the f/5 wavefront sensor. The data image is a 5 second exposure taken near zenith in 0.5 seeing. 4. WAVEFRONT ANALYSIS AND SOFTWARE 4.1. Acquiring and Reducing Wavefront Sensor Data The software used to control the wavefront sensors and analyze the acquired data is a modular mixture of IRAF and custom C code glued together with perl, TCL, and ruby scripts. Desktop menu entries are provided on the operator s workstations to bring up f/9 or f/5 versions of the interactive interface. All of the CCD images are acquired and saved in FITS format. These data, along with logs of wavefront sensor activity and output from the analysis tasks, are archived at the end of every night. Reference images are acquired and analyzed each time a wavefront sensor is mounted on the telescope. The f/5 system has proven to be very stable over the course of several installations. The differences between reference images taken during different runs are well within the measurement errors of the system due to centroiding accuracy. The f/9 system has shown some run-to-run variations in its alignment, but once mounted both systems are quite stable and show no measurable effects of flexure as a function of the telescope s zenith angle. The centroiding of the spots is performed using a scripted interface to IRAF s DAOPHOT package. The data are initially passed through DAOFIND to locate and identify spots. The results of DAOFIND are passed to PSF, which determines a position-dependent model of the point-spread function for each image. This model and the initial spot locations are passed to ALLSTAR, which uses the model to recentroid all of the spots. Along the way, spots that are saturated, corrupted by bad pixels, or too strongly distorted are flagged and rejected. This multi-pass method is more computationally expensive than computing raw centroids or fitting a simple analytic model to the spots, but it has proven to be accurate and robust over a wide range of conditions. An average spot width is calculated from the point-spread function model and used to derive a measure of the seeing Measuring and Correcting the Wavefront After the spot centers are measured, the pupil size and center are determined and the spot centers converted to relative pupil coordinates. If multiple exposures are taken, the spot centers from each image are averaged together. Spot displacements are then calculated relative to a selected reference set, usually the one determined at the beginning of a run. Wavefront slopes are calculated from the spot displacements and modeled with a 19-term Zernike polynomial fit. The resulting fit is displayed in the form of a pupil phase map, an image point-spread function map, and a bar graph showing the RMS wavefront error contributions of each Zernike component. An example is shown in Figure 6 of wavefront analysis done in good conditions after two iterations of corrections.

6 Figure 6. Screenshot of the main wavefront analysis display. The pupil phase map is shown in the upper left, an image point-spread function map in the upper right, RMS wavefront errors for the specific components in the lower left, and resulting primary mirror actuator correction forces in the lower right. The boxsize for the point-spread function map is 0.5. Some of the wavefront errors are corrected by adjusting the secondary. These include correcting global tilts by moving the secondary about its zero-coma point, correcting coma by moving it about its center-of-curvature, and correcting defocus by moving it along its focus axis. Results of a finite element analysis performed by BCV Progetti 5 are used to calculate the primary mirror actuator forces required to correct the other Zernike modes. Third order spherical aberration is actually corrected using a combination of primary and secondary mirror adjustment. Bending enough defocus into the primary to offset the defocus term in spherical aberration reduces the amount of force required to correct a given amount of spherical aberration by about a factor of three. The defocus added to the primary can then easily be compensated by refocusing the secondary, generally by a few tens of microns. One of the interface windows allows the operator to select which modes to correct and then to apply the corrections. 5. RESULTS AND PERFORMANCE 5.1. Before Correcting Table 1 summarizes the average beginning of night wavefront measurements for the f/9 and f/5 wavefront sensors, giving an indication of the telescope s initial condition before wavefront correction. The primary is usually not in thermal equilibrium when these data are taken. The telescope structure usually has not stabilized thermally early in the night, and this degrades secondary collimation. Significant scatter in each of the modes is therefore to be expected. The data from the f/5 wavefront sensor are taken from two runs in October and November 2003 and a third in March and April The data from the f/9 system are taken from January through mid-march plus the last half of May Although the scatter is large, there are some noticeable trends. Both setups show small and marginally significant amounts of trefoil and astigmatism. For the higher order terms, the f/5 system shows a bit of quadrafoil, while the f/9 system has some 5th order astigmatism and perhaps some third order spherical, 5th order trefoil, and 5th order coma as well. This is likely due to different average thermal conditions of the primary and secondary mirrors. The f/9 data came mostly from this past winter, while the f/5 data came mostly from

7 Table 1. Mean wavefront aberrations and their RMS values as measured at the beginning of the night before any corrections are applied. Global tilts and defocus are left out because they can vary according to telescope and instrument configuration. Note that the RMS values are not intrinsic errors but rather show the spread of obtained values that are influenced by initial secondary orientation and the thermal states of the primary and f/9 secondary. Zernike coefficients are expressed as amplitudes. F/5 F/9 Zernike Mode (nm) (nm) Astigmatism (45 ) 467 ± ± 591 Astigmatism (0 ) -494 ± ± 1089 X Coma 458 ± ± 1233 Y Coma 327 ± ± 823 3rd Order Spherical -91 ± ± 545 X Trefoil 417 ± ± 331 Y Trefoil -179 ± ± 242 5th Order Astigmatism (45 ) -44 ± ± 228 5th Order Astigmatism (0 ) 17 ± ± 267 Quadrafoil ± ± 191 Quadrafoil 2-81 ± ± 287 5th Order X Trefoil -85 ± ± 189 5th Order Y Trefoil -45 ± ± 110 5th Order X Coma 42 ± ± 215 5th Order Y Coma -57 ± ± 188 6th Order Spherical 4 ± ± 194 late fall and early spring during generally warmer weather. The f/9 data also includes wavefront distortions from a borosilicate secondary with no thermal control of its own (the f/5 secondary is Zerodur r glass). During cold weather daytime work on the telescope can warm the chamber, and there is less time to run the thermal system between the end of the day shift and the start of observing. In addition, dry winter and spring nights often have a larger dt/dt after sunset than the thermal system can follow, exacerbating temperature gradients in the primary and f/9 secondary. The daytime procedure for the thermal system was changed in early March We discovered that when the thermal system is powered down, warm air could back-flow from the chamber into the exhaust ports and introduce temperature gradients in the primary. The simple work-around is to block the exhaust ports with shower caps when the thermal system is off. This change has resulted in a marked improvement in the earlynight thermal system performance which partially accounts for the better early-night wavefront results for f/5. Data for the f/9 system from the latter part of May 2004 also seem to show improvement, though more of a baseline is needed to draw a firm conclusion After Correcting Table 2 tabulates averages of wavefront measurements taken after corrections have been completed. In general these consist of two or three corrective iterations. The averages for the lower order terms are all much smaller than at the beginning of the night, though there is still a bit of scatter. Some scatter is expected since the telescope operators do not always correct the primary if the conditions and/or observing program do not warrant it. Also, correction of modes higher in order than trefoil is disabled by default since the low order modes tend to dominate the wavefront errors, and high order modes tend to sort themselves out as the primary becomes isothermal. The higher modes require significant correction forces at the primary mirror actuators. There is still some indication of residual 5th order astigmatism, trefoil, and coma in the post-correction f/9 data. These are likely thermal artifacts and may be due to the borosilicate secondary. They are not apparent in f/5, though quadrafoil remains at about the same amplitude. The source of this quadrafoil is not yet clear and

8 Table 2. Mean wavefront aberrations and their RMS values measured after corrections have been applied and before resuming science observations. Again, global tilts and defocus are left out and the RMS values are observed variations and not intrinsic errors. Zernike coefficients are expressed as amplitudes. F/5 F/9 Zernike Mode (nm) (nm) Astigmatism (45 ) -4 ± ± 412 Astigmatism (0 ) -19 ± ± 440 XComa -9± ± 253 YComa 9± ± 292 3rd Order Spherical -78 ± ± 133 XTrefoil 58± ± 203 YTrefoil 22± ± 257 5th Order Astigmatism (45 ) -21 ± ± 245 5th Order Astigmatism (0 ) 48 ± ± 230 Quadrafoil ± ± 178 Quadrafoil ± ± 203 5th Order X Trefoil -46 ± ± 170 5th Order Y Trefoil -46 ± 94-4 ± 108 5th Order X Coma -33 ± ± 160 5th Order Y Coma -59 ± ± 132 6th Order Spherical -18 ± ± 167 bears further investigation. That said, it is not a large effect, and the mean post-correction aberrations for both foci would result in very nearly diffraction-limited images. Figure 7 shows a representative example of how well the MMT optics can be tuned when the conditions are good. The seeing was exceptionally good ( ) at this time, so the high order modes were corrected. The enclosed energy fraction within a 0.1 radius is 82% for the Zernike fit versus 47% calculated directly from the spot motions. Part of this difference is due to centroiding uncertainty since the RMS spot deviation of corresponds to pixels. Some minor image-to-image seeing variation between the five exposures averaged together here accounts for another part. The largest spot motions near the center of the pupil are real and do repeat from image to image, however, so there are modes of distortion not being reflected in the Zernike fit. The complete f/5 optical system was specified to achieve a 80% enclosed energy diameter of 0.36, and we observe 0.32 here Open-loop Performance Without the benefit of continuous wavefront monitoring, the ability to maintain telescope performance between rounds of wavefront sensing is at least as important as the performance of the wavefront sensors themselves. The largest factors that must be dealt with are gravity and the temperature of the optical support structure (OSS) of the telescope. The effect of gravity is calibrated by taking wavefront sensor data over a range of elevation angles. At each position, the secondary is moved about its zero-coma point to place the star at the center of rotation, wavefront sensor data taken, focus and coma corrections applied, and then the secondary s coordinates logged. After data acquisition is complete, the secondary coordinates are fit with a linear combination of the sine and cosine of telescope elevation angle. Gravity should only affect three of the axes (Z and Y translation plus rotation around X), so the others are used mainly to check for problems. The results of the fits are then entered as offsets to the secondary mirror position. Elevation data are generally taken every time changes are made to the front end of the telescope. So far, the solutions for f/5 and f/9 have been fairly consistent between mountings which is reassuring. Correcting for OSS temperature changes should in principle be quite easy, and the thermal expansion of the borosilicate primary mirror is also well-determined. The difficulty has been getting an accurate and representative measure of the OSS temperature. During an engineering run in January 2004 we tested the effects of temperature

9 Seeing = 0.24" Spot FWHM = 2.85 pixels Figure 7. Screenshot of Zernike polynomial fit (top) and spot motion analysis (bottom) of wavefront performance on a very good night after wavefront corrections have been applied. The Shack-Hartmann data was taken at an elevation of 85 in seeing with a total exposure time of 20 seconds. The RMS spot deviation is

10 by tracking a star near the north pole (to remove elevation effects) and taking continuous data with the wavefront sensor. For the first half of the night, focus tracked very close to predictions. However, when the wind shifted and picked up later on we saw significant shifts in focus of 100 µm or more but did not measure any significant corresponding changes in OSS temperature. At the time we only had two sensors on the OSS, which were located at each end of one of the main tubes, and took the average of the two readings to be the OSS temperature. Afterwards more temperature sensors were added that have helped improve performance of the temperature corrections, at least qualitatively. We have not yet been able to redo the experiment of wavefront sensing near thepoletoquantifythis. As mentioned previously, keeping the primary and f/9 secondary mirrors isothermal is another important component of maintaining good image quality. When the primary needs to be warmed or cooled to match ambient conditions, the thinner, inner part of the mirror reacts to the changes more quickly than the thicker, outer part. The amount of spherical aberration present tracks these radial temperature gradients very closely and requires more frequent wavefront sensing to compensate. High order aberrations generally track smaller scale temperature variations in the mirror (i.e. hot or cold spots), but usually are not large enough to affect image quality as much as the lower order ones. Astigmatism is another mode that is observed to change significantly between wavefront measurements and that we have not be able to correlate well. Since it corresponds to the softest bending mode of the primary mirror, it can be quite sensitive to small temperature and wind-load variations. 6. CONCLUSIONS AND FUTURE WORK Overall, our wavefront sensor systems have worked very well so far and have significantly enhanced the overall performance and productivity of the telescope. There are, however, significant improvements that still need to be made. The telescope operator interface will benefit from greater automation to help reduce the time required to make wavefront measurements and corrections. This automation will also enable continuous WFS operation for upcoming f/5 instruments that will support it. Improved open-loop performance will help reduce the amount of wavefront sensing that is required. The open-loop handling of elevation-dependent effects works well and requires an improved hexapod controller to be made fully automatic and transparent. Indications are that this is also true for the handling of temperature corrections. We are also considering open-loop corrections of the primary, since some modes track closely with measured temperature variations. The easiest of the modes to handle will likely be third order spherical since we have most, if not all, of the data we need to derive an empirical relation between it and an azimuthally-averaged dt/dr. ACKNOWLEDGMENTS The expertise and hard work of numerous people contributed to the construction and operation of these wavefront sensors. The telescope operators (Mike Alegria, Ale Milone, and John McAfee) in particular deserve thanks for their help, patience, and support in running these systems on a nightly basis. Other MMTO staff we would like to thank include Cory Knop, Brian Comisso, Ken Van Horn, and Pete Spencer for electronics support, Shawn Callahan, Ricardo Ortiz, Court Wainwright, and Steve Bauman for mechanical support, Dennis Smith for mountain and f/9 top box support, and Bill Kindred for help in aligning the f/9 wavefront sensor within the top box. The f/5 wavefront sensor was built by members of the Central Engineering group at SAO led by Ed Hertz. Mark Mueller played a major role in the mechanical design, John Roll wrote the instrument control software, and Tom Gauron led the electrical design. REFERENCES 1. L. Noethe, Active optics in modern, large telescopes, in Progress in Optics, p. 43, G. Brusa, A. Riccardi, F. P. Wildi, M. Lloyd-Hart, H. M. Martin, R. Allen, D. L. Fisher, D. L. Miller, R. Biasi, D. Gallieni, and F. Zocchi, MMT adaptive secondary: first AO closed-loop results, in Astronomical Adaptive Optics Systems and Applications. Edited by Tyson, Robert K.; Lloyd-Hart, Michael. Proceedings of the SPIE, Volume 5169, pp (2003)., pp , Dec

11 3. P. L. Schechter, G. S. Burley, C. L. Hull, M. Johns, H. M. Martin, S. Schaller, S. A. Shectman, and S. C. West, Active optics on the Baade 6.5-m (Magellan I) Telescope, in Large Ground-based Telescopes. Edited by Oschmann, Jacobus M.; Stepp, Larry M. Proceedings of the SPIE, Volume 4837, pp (2003)., pp , Feb S. C. West, Interferometric Hartmann wave-front sensing for active optics at the 6.5-m conversion of the Multiple Mirror Telescope, Applied Optics 41, pp , July H. M. Martin, S. P. Callahan, B. Cuerden, W. B. Davison, S. T. Derigne, L. R. Dettmann, G. Parodi, T. J. Trebisky, S. C. West, and J. T. Williams, Active supports and force optimization for the MMT primary mirror, in Proc. SPIE Vol. 3352, p , Advanced Technology Optical/IR Telescopes VI, Larry M. Stepp; Ed., pp , Aug