A prototype phasing camera for the

Srikrishna Kanneganti*a, Brian A. McLeod*a, Mark P. Ordwaya, John B. Roll, Jr.a, Stephen A. Shectmanb, Antonin H. Bouchezb, Johanan Codonac, Roger Enga, Thomas M. Gaurona, Timothy J. Nortona, Phil Streechona, David Weavera aHarvard-Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA 02138; bGiant Magellan Telescope Corporation, P.O. Box 90933, Pasadena, CA, 91109-0933; cSteward Observatory, , Tucson Arizona, 85719

ABSTRACT

Achieving the diffraction limit with the adaptive system of the 25m Giant Magellan Telescope will require that the 7 pairs of mirror segments be phased. Phasing the GMT is made difficult because of the 30-40cm gaps between the primary mirror segments. These large gaps result in atmospheric induced phase errors making optical phasing difficult at visible wavelengths. The large gaps between the borosilicate mirror segments also make an edge sensing system prone to thermally induced instability. We describe an optical method that uses twelve 1.5-m square subapertures that span the segment boundaries. The light from each subaperture is mapped onto a MEMS mirror segment and then a lenslet array which are used to stabilize the atmospherically induced image motion. Centroids for stabilization are measured at 700nm. The piston error is measured from the fringes visible in each of the 12 stabilized images at 2.2 microns. By dispersing the fringes we can resolve 2! phase ambiguities. We are constructing a prototype camera to be deployed at the 6.5m Magellan Clay telescope.

Keywords: segmented telescopes, piston measurement,

1. INTRODUCTION

The Giant Magellan Telescope (GMT) will be comprised of seven 8.4-m diameter primary mirror (M1) segments paired with seven adaptive secondary mirrors (M2)1. To reach the diffraction limit of the 25m telescope, each of the seven segment pairs must have exactly the same spacing. Existing segmented telescopes such as the Keck Telescope have relied on edge sensors located in the gaps between the segments to provide the necessary control. The Keck edge sensors are stable enough that they are calibrated interferometrically on sky only at month-long intervals2,3. The GMT segment phasing is more difficult for several reasons. The 30-40cm segment gaps combined with the fact that the mirrors are made of borosilicate glass (non-zero thermal expansion) leads to the concern that a metrology system mounted to the edges of the segments will not be stable enough on timescales longer than a few minutes to maintain the necessary ~30nm differential piston required to meet the GMT adaptive optics error budget4,5. While the metrology systems being designed6 may ultimately have the necessary long-term stability, the baseline plan is to continuously monitor stars at intervals of 60 seconds to maintain the calibration of the high-bandwidth metrology systems7. When the AO system is operating in natural guide star mode, the piston measurements will be derived from the on-axis star as sensed by a pyramid sensor8. However, when operating in guide star mode there are no available stars close enough to the optical axis to maintain the required phase coherence between segments. This led us to an approach where we would measure the differential piston across the segment gaps using stars several arcminutes off-axis. We form Young’s double slit images from the twelve 1.5m square subapertures that span segment boundaries. By limiting the measurement to a small subaperture, correcting the tip-tilt of the wavefront in each subaperture, and measuring the fringes in the K-band, enough fringe contrast is preserved to make the necessary piston measurement9 (see Figure 1). *Contact information: [email protected] ; [email protected]

We thus set out to build a prototype of such a system and test it on sky to show that we could make piston measurements of the required accuracy using stars of typically available brightness. As the work progressed, a further complication was realized. If the M1 segments tilt with respect to each other (as they will naturally do from flexure due to gravity), and the M2 segments are tilted to compensate, an un-sensed piston variation with field angle will arise. The M1 and M2 metrology systems do not quite have the sensitivity and stability to sense the tilts independently, and no optical test is possible. Thus we were forced to conclude that high-accuracy piston measurements far off-axis would not result in a phased telescope on-axis. This is a problem that is fundamental to telescopes where both M1 and M2 are segmented. The current plan then is to make the piston measurement close to the optical axis using a pyramid sensor in the science instrument with a fainter star10. To be sensitive to piston, the star must corrected with an open-loop deformable mirror using a wavefront derived from the laser tomographic reconstruction. Because this on-instrument pyramid sensor is insensitive to 2! phase ambiguities we will maintain the off-axis piston sensor using it in a dispersed mode, but now requiring lower sensitivity. We have carried the prototype effort forward, adding a dispersion mode to it. Tip-tilt correction of the subapertures is still required to maintain fringe contrast. Thus, in section 2 we describe the instrument. In section 3.1 we describe analysis and lab results for the non-dispersed mode. The parts for the dispersion mode arrived recently and we report initial lab results in section 3.2. As of this writing, the instrument is in transport to Chile where we plan to test it on the Magellan Clay Telescope.

Figure 1. Phasing subapertures superposed on GMT pupil (left), and resulting images formed from the marked subaperture with varying amounts of piston between the segments (right).

2. INSTRUMENT DESCRIPTION The instrument will be located on the East Nasmyth platform of the Magellan Clay telescope. The optics are all mounted on an optical bench, which in turn is mounted on a cart with casters. Between the optical bench and the cart are a set of ball transfer plates which allow fine adjustment of the location of the instrument relative to the telescope. The instrument is not bolted onto the Nasmyth rotator, but instead operates in a gravity invariant mode. An overall view of the optics is shown in Figure 3, with a close up view in Figure 4. In the sections below we describe the optical components, following the light from where it enters the instrument, through warm reimaging optics and then into the cryostat where it is split by wavelength and reimaged onto CCD and IR detectors.

Figure 2. Photo of the phasing camera prototype.

Figure 3. Overall view of instrument optics.

2.1 Focal plane assembly Light source/pinhole For lab testing we insert into the beam at the telescope focus location an incandescent light source shining through a Melles-Griot 10 "m pinhole. The pinhole assembly is on a manually actuated stage to move it in or out of the beam. On the same stage as the pinhole is a mirror with a 0.7mm diameter hole in the center. This 2 arcsec diameter hole will pass light from the telescope into the instrument. The outer parts of the field will be directed to an acquisition camera. Because the mirror feeding the acquisition camera is at the focal plane, the optical quality does not have to be very good. Thus, we use aluminum reflector panel, in which it is easy to drill the on-axis hole. The acquisition camera is a StellaCam3 video camera connected to an Axis M7001 Ethernet-based video encoder. The camera is capable of operating at video rates or delivering still images. Out-of-the-box it comes with a streaming video web interface. It can also take single exposures with controllable exposure time.

Figure 4. MEMS relay and cryostat optics

2.2 Warm optics The first set of reimaging optics is warm and converts the f/11 beam to f/27. This produces an image scale nearly identical to what the GMT will produce. It also provides a convenient collimated beam in which to introduce a turbulence phase screen and phase step components.

We test the system in closed loop using a Kolmogorov phase screen on a rotary stage, all manufactured by Lexitek. The pupil scale projected at the phase screen is 1.3mm/meter. We have two annuli with equvalent r0(500nm)=0.15m and 0.10cm. Our lab tests were conducted with the plate rotating to give an equivalent windspeed of 17m/s, which is somewhat worse than the median speed integrated over all altitudes. Because the experiment is being carried out on a monolithic telescope, we must artificially provide a phase difference between halves of the phasing aperture. A phase discontinuity can be introduced by inserting at a pupil plane two pieces of glass with identical thickness adjacent to each other in the beam. With a 3mm thick window and a microstepping motor, phase differences from 10nm to >25 microns can be introduced. Microstepping non-linearity is eliminated with a high resolution encoder that is attached to the motor shaft. To characterize how accurately we can measure a phase difference of zero, we can translate the phase adjuster completely out of the beam.

Figure 5. Phase adjuster assembly with fixed and rotating glass windows (blue). The entire assembly is mounted on a slide for removal from the beam. 2.3 MEMS Assembly

The MEMS relay (see Figure 6) consists of an Aluminum coated 100mm radius spherical mirror in double pass fed by two 5mm right angle prisms. This Offner-style relay forms a pupil on the MEMS mirror. The prisms are epoxied to an Aluminum annulus, with features machined into the aluminum to locate the prisms. A cylinder registers off the annulus and the mirror surface is spring loaded against the top of the cylinder. Underneath, the MEMS mirror board is attached to x, y, z translation, and x,y tilt stages.

Figure 6. MEMS relay assembly

MEMS mirror The IRIS-AO PT-111 MEMS mirror has a maximum tilt of +/-5mrad, which corresponds to 0.55 arcsec. This should be adequate most of the time to compensate for seeing induced image motion. If there are momentary larger excursions, it will result in a temporary loss of lock in the tilt correction and will reduce the fringe contrast. Figure 7 shows the pupil mask overlayed on the outline of the PT-111.

Figure 7. Outline of PT-111 MEMS segments, with the phasing apertures overlayed. Apertures 1 and 2 are dispersed and undispersed, respectively, have 50cm projected gaps and can have differential piston added with the phase adjuster. Apertures 3 and 4 have no differential piston and have 70 and 90 cm gaps.

2.4 Cryo optical components We have chosen to operate all the optics at liquid nitrogen temperature. To eliminate the thermal background radiation it is necessary to cool the optics significantly. Because we already have the requirement that the detector be cooled to 77K, the simplest course of action is to cool everything to 77K. The requirement that the optics operate cold, and that they work at 2.4 microns prohibits the use of cemented doublets. We find that a design that alternates stock CaF2 field and relay lenses gives excellent image quality. In the visible channel we use other less expensive glass types. The infrared cryostat, including optics mounts and the detector mount, was supplied by Universal Cryogenics, in Tucson, Arizona. A CaF2/quartz lens pair forms a pupil on the lenslet array, the Okotech APH-Q-P1000-R16.86 fused silica array with a 1mm hexagonal pitch and 36mm focal length. The lenslets are matched 1 to 1 with the MEMS mirror facets. The phasing aperture mask is laser machined from 0.008 inch thick anodized foil, and is mounted in contact with the lenslet array. A dichroic mirror splits the light so that the visible beam (#<1µm) is reflected and exits the cryostat to the tip-tilt camera. The IR beam (#>1.5µm) is transmitted. There is a small lateral translation of the transmitted beam due to the thickness of the dichroic substrate. The beam is then reimaged to a scale of 0.18"/pixel by a pair of CaF2 lenses: a field lens followed by a single relay lens. Just in front of the detector is a filter wheel that is populated with H and K band filters, and a dark position.

The detector is an engineering grade Teledyne HAWAII-2RG, on loan from the University of Arizona, controlled with the SIDECAR ASIC Cryogenic Development kit. The SIDECAR is connected to Teledyne’s data acquisition “JADE” card via a ribbon cable. The ribbon cable passes through an epoxy potted panel in the wall of the cryostat. The software interface to the SIDECAR is made using a custom Tcl/Tk interface that communicates with the Teledyne supplied software on two levels. Setup of the device operating modes is controlled through the high level IDL HxRG socket server. The frames are read out using the lower level Teledyne-supplied “HAL” server using the Windows COM protocol, which eliminates a significant amount of overhead and allows real time display of 100x100 pixel subframes frames read at 20Hz. The visible reimager follows the same strategy as the IR channel: a field lens followed by single lens relay. The field lens is located inside the IR cryostat. Then the beam passes outside through a window through a warm relay lens before going into the tip-tilt camera, a Princeton Instruments ProEM512B electron multiplication CCD camera. The pupil can be viewed on the visible light camera by replacing the final relay lens with a longer focal length lens. These two lenses are mounted on a manually operated slide that is located in ambient air between the IR and visible cryostats. The pupil viewer was used to align the optical components, and will be used at the telescope to ensure that the instrument is aligned with the telescope pupil.

3. ANALYSIS AND RESULTS 3.1 Without dispersion To determine the differential piston !! between two subapertures we extract the long-exposure K-band fringe image from the larger parent image and Fourier transform it. The resulting transform, F(I), the mutual coherence function (MCF), has a region that corresponds to the fringe spacing. A weighting function corresponding to that region, !!" is then multiplied by F(I). The weighting function is a section of the optical transfer function, which may be computed as !! ! !!" ! ! ! !! ! !! , where !!!! are the sections of the pupil corresponding to the 2 adjoining segments. The values in the complex plane are then summed, and the phase of the resulting complex number is the final result, i.e. !! ! !"#! !!!"!!!!!!. Figure 8 shows the analysis schematically and Figure 9 shows our lab results overlayed on results of a simulation9. The lab results are slightly worse than what was predicted by the simulation, possibly because the sampling in the as-built optics is almost critically sampled whereas the simulation had sampling nearly twice as fine. Nonetheless, the results show that the piston can be recovered to better than 50nm on a typically available star.

Figure 8. Analysis of non-dispersed images. The raw image contains multiple spots (upper right), one of which is extracted (lower right). The transform is taken and the phase is extracted (upper left).

Figure 9. Simulation results (gray scale) of differential piston versus visible and infrared star brightness. Selected contour intervals are shown in blue. Lab results are overplotted in orange, showing slightly worse perforamance than predicted, but still better than 50nm RMS for typically available stars.

3.2 With dispersion With the addition of a grism behind the dichoic mirror we can disperse the light along the direction of the fringes. In our implementation we use a 100 line/mm, 24° blaze angle grating replicated on a 2 mm square 29° fused quartz prism. This disperses the K-band over 15 pixels. A similar concept operating at visible wavelengths with higher dispersion was tested by Albanese et al. on the Keck telescope as a prototype for JWST11. In the final design for the GMT we will need 12 grisms, each oriented perpendicular to the tangent line between the mirrors. In this prototype we disperse a single subaperture. For a system with zero path difference the fringes will be horizontal; as the path difference increases, the fringes will tilt. Unlike the analysis used by Albanese et al., which analyzed intensity profiles along the dispersion direction, here we use a Fourier analysis of the dispersed fringe image. The analysis of the dispersed image requires extracting the fringe image, and Fourier transforming it. To avoid introducing edge artifacts, we preprocess the image by subtracting a best-fit Gaussian to the image and then apodizing the edges before computing the transform F(I). To derive the absolute path difference we derive the integer number of waves from the horizontal location of the peak in F(I). The fractional number of waves is computed just like the non-dispersed case, but with the weighting function !!" multiplied by an additional Gaussian function to reflect the narrower peak seen in the MCF. Figure 11 shows the performance with lab turbulence (r0=0.15, v=17m/s) for a bright star. Noise in the MCF image causes the peak location to be mismeasured slightly, which sometimes leads to a 2! phase jump. We will be working on improving the image processing to make this measurement more robust. The correctly measured frames have an RMS piston measurement of 90nm which is sufficient to prevent the on-axis wavefront sensor from jumping by 2!. Fringes are visible out to approximately +/- 20 "m making it possible to use this device for fringe capture.

4. CONCLUSIONS We have completed construction of a prototype phasing camera for the Giant Magellan Telescope and will shortly be testing on the sky at the 6.5m Magellan Clay Telescope. Lab tests indicate that we will be able to use such a camera to bring the GMT into phase from 20"m out, and maintain it to within 100nm which is more than sufficient for preventing the on-instrument wavefront sensor from jumping fringes.

Figure 10. Integrated images with turbulence with fringes dispersed horizontally (upper left); MCF computed without preprocessing (upper right); Image with Gaussian subtracted and then apodized (lower left); MCF of processed image (lower right).

Figure 11. Piston error measured from dispersed images with turbulence. The analysis software currently mis-measures the peak location in the MCF and thus creates a 2! error on some frames.

ACKNOWLEDGEMENTS

We gratefully acknowledge Marcia Rieke and the NASA-funded NIRCam Team at the University of Arizona for loaning us the HAWAII-2RG detector used in this experiment.

This work has been supported by the GMTO Corporation, a non-profit organization operated on behalf of an international consortium of universities and institutions: Australia Ltd, the Australian National University, the Carnegie Institution for Science, , the Korea Astronomy and Space Science Institute, the Smithsonian Institution, The University of Texas at Austin, Texas A&M University, University of Arizona and .

REFERENCES

[1] Johns, M. McCarthy, P. J., Raybould, K., Bouchez, A., Farahani, A., Filgueira, J. M, Jacoby, G. H., Shectman, S.A., Sheehan, M., “Giant Magellan Telescope: overview”, Proc SPIE, 8444, 8444-52 (2012) [2] Chanan, G., Ohara, C., & Troy, M., Active Optics, 39, 4706 (2000) [3] Chanan, G., Troy, M., Dekens, F., et al., Active Optics, 37, 140 (1998) [4] Bouchez, A. H. et al., “The Giant Magellan Telescope adaptive optics program”, Proc SPIE, 8447, 8447-54 (2012) [5] Trancho, G., Espeland, B., Bouchez, A. H, Conan, R., Hinz, P., van Dam M., “GMT AO system requirements and error budgets in the preliminary design phase”, Proc. SPIE, 8447, 8447-202 (2012) [6] Acton, D. S., Bouchez, A. H., “Phasing metrology system for the GMT” Proc. SPIE, 8447, 8447-75 (2012) [7] Bouchez, A. H., McLeod, B. A., Acton, D. S., Kanneganti, S., Kibblewhite, E. J., Shectman, S.A., van Dam, M. A., “The Giant Magellan Telescope phasing system”, Proc. SPIE, 8447, 8447-138 (2012) [8] Esposito, S., Pinna, E., Quiros-Pacheco, F., et al., “Wavefront sensor design for the GMT natural guide star AO system”, Proc SPIE, 8447, 8447-57 (2012) [9] Codona, J. “Pairwise segment phasing with the GMT”, GMT internal report (2008) [10] van Dam, M. A., Conan, R., Bouchez, A. H., Espeland, B., “Design of a truth sensor for the GMT laser tomography adaptive optics system”, Proc. SPIE, 8447, 8447-43 (2012) [11] Albanese, M.,Wirth, A.,Jankevics, A., Gonsiorowski, T., Ohara, C. , Shi, F., Troy, M., Chanan, G. and Acton, S., "Verification of the James Webb Space Telescope coarse phase sensor using the Keck Telescope", Proc. SPIE 6265, 62650Z (2006).