Advancing next generation adaptive optics in astronomy: from the lab to the sky

(1)

From the Lab to the Sky

by

Paolo Turri

B.Sc., University of Padova, 2007 M.Sc., University of Trieste, 2012

A Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree of

DOCTOR OF PHILOSOPHY

in the Department of Physics and Astronomy

c

Paolo Turri, 2017 University of Victoria

(2)

Advancing Next Generation Adaptive Optics in Astronomy:

From the Lab to the Sky

by Paolo Turri B.Sc., University of Padova, 2007 M.Sc., University of Trieste, 2012 Supervisory Committee Dr. D. R. Andersen, Co-Supervisor (Department of Physics and Astronomy)

Dr. K. Venn, Co-Supervisor

(Department of Physics and Astronomy)

Dr. A. W. McConnachie, Departmental Member (Department of Physics and Astronomy)

Dr. Colin Bradley, Member

(3)

Supervisory Committee

Dr. D. R. Andersen, Co-Supervisor (Department of Physics and Astronomy)

Dr. K. Venn, Co-Supervisor

(Department of Physics and Astronomy)

Dr. A. W. McConnachie, Departmental Member (Department of Physics and Astronomy)

Dr. Colin Bradley, Member

(Department of Mechanical Engineering)

ABSTRACT

High resolution imaging of wide fields has been a prerogative of space telescopes for decades. Multi-conjugate adaptive optics (MCAO) is a key technology for the future of ground-based astronomy, especially as we approach the era of ELTs, where the large apertures will provide diffraction limits that will significantly surpass even the James Webb Space Telescope.

NFIRAOS will be the first light MCAO system for the Thirty Meter Telescope and to support its development I have worked on HeNOS, its test bench integrated in Victoria at NRC Herzberg. I have aligned the optics, tested the electronic hardware, calibrated the subsystems (cameras, deformable mirrors, light sources, etc.) and characterized the system parameters.

Development and support for future MCAO instruments also involves data analy-sis, a critical process in delivering the expected performance of any scientific instru-ment. To develop a strategy for optimal stellar photometry with MCAO, I have observed the Galactic globular cluster NGC 1851 with GeMS, the MCAO system on

(4)

the 8-meter Gemini South telescope. From near-infrared images of this target in two bands, I have found the optimal parameters to employ in the profile-fitting photome-try and calibration. As testimony to the precision of the results, I have obtained the deepest near-infrared photometry of a crowded field from the ground and used it to determine the age of the cluster with a method recently proposed that exploits the bend in the lower main sequence. The precise color-magnitude diagram also allows us to clearly observe the double subgiant branch for the first time from the ground, caused by the multiple stellar populations in the cluster.

As the only facility MCAO system, GeMS is an important instrument that serves to illuminate the challenges of obtaining accurate photometry using such a system. By coupling the knowledge acquired from an instrument already on-sky with experiments in the lab on a prototype of a future system, I have addressed new challenges in photometry and astrometry, like the promising technique of point spread function reconstruction. This thesis informs the development of appropriate data processing techniques and observing strategies to ensure the ELTs deliver their full scientific promise over extended fields of view.

(5)

List of Tables

2.1 Main characteristics of the ALPAO DMs in HeNOS. . . 17 2.2 Main properties of the Point Grey cameras in HeNOS. . . 17 2.3 Comparison of NFIRAOS and HeNOS design parameters. . . 19 2.4 Distances of the optics in the conjugated space of the deformable

mir-rors. Beamsplitters 1 and 2 are respectively in front of DM0 and DM11. The surface of an optical element is in parentheses. . . 34 2.5 Measurements of r0 at 500 nm using the WFS and the science camera,

compared to the design values. . . 35 3.1 Observation log of NGC 1851 with GeMS/GSAOI. . . 67

(9)

List of Figures

1.1 Median (solid line), first and third quartiles (dashed lines) C_n2 profile at the Observatorio Astron´omico Nacional de San Pedro M´artir. The solid horizontal line is the altitude of the observatory. The ground and jet stream layers are clearly visible. Image from Avila (2012). . . 5 1.2 Schematic operation of a Shack-Hartmann WFS with a flat (on top)

and distorted (on bottom) wavefront. Image from Rousset (1999). . . 8 1.3 The dramatic evolution in a decade of the classical AO PSF has been

driven by many factors, including increased density of actuators on the DM and faster frame rates. . . 9 1.4 NFIRAOS mechanical and optical design. The optical paths are in red. 12 2.1 The HeNOS bench at NRC Herzberg. In the top panel, the “science”

light rays are in blue, the LGS SHWFS path is in red and the PWFS path is in green. The photo in the lower panel has been taken with-out arranging the bench, to show the typical difference between ideal planning and real implementation of an experiment. . . 21 2.2 Grid of natural stars on the HeNOS science camera full field. . . 22 2.3 LGS generator on HeNOS. On the right is the plate with holes holding

in a square configuration the four optical fibers. A series of lenses delivers the light to the telescope. . . 23 2.4 HeNOS DM0 on the left with a beamsplitter cube in front of it. On

the right is the optical flat that substitutes PS1 during the calibrations. 24 2.5 Two of the three phase screens used on HeNOS, complete with rotary

stages and motors. On the left is one of the two Lexitek, on the right is the UCSC phase screen. . . 25

(10)

2.6 Out of focus LGS asterism on the HeNOS science camera. In the original image are visible also four vignetted ghosts. In the thresholded image the white circumferences are fitted to the circles of confusion and the four white crosses are their centers. . . 30 2.7 Stacked first diffraction ring of the natural stars on the science camera

using a 3 mm iris. . . 32 2.8 Science camera images of a natural star without (on the left panels) and

with (on the right) the three phase screens in the optical path. The diffraction-limited image before using PS0 is different from the one before PS1 and PS2 because of a change in the quasi-static aberrations. 37 2.9 Phase maps of ALPAO DM277 measured with the interferometer. . . 39 2.10 Creeping of DM277 measured by an interferometer as low-order optical

aberrations. . . 41 2.11 Focal plane sharpening iterations with different sets of Zernike modes.

The top panels show the median Strehl ratio at each iteration with the most recent best result represented by a blue dot. The bottom panels display the Zernike coefficients of the best iteration. . . 47 2.12 HeNOS GUI window for controlling the single-conjugate mode. The

four panels on the right show the WFS camera image (top-left), the magnified wavefront slopes (top-right), the reconstructed wavefront (bottom-left) and the DM0 command (bottom-right). The buttons on the left have functions like applying a preset DM0 command, mea-suring the slopes and calculating the correction to apply. . . 50 2.13 Multi-conjugate interaction matrix of HeNOS. Each of the four LGS

wavefront sensors has 714 lenslets, for a total of 1428 rows of x and y slopes. The first 97 columns are for DM0 actuators, the following 277 for DM11. . . 51 2.14 Wavefront standard deviation during during a HeNOS single-conjugate

loop. Two different correction algorithms are compared. . . 52 2.15 Image from one of the four Shack-Hartmann wavefront sensors using

the LGS elongation with a uniform sodium profile. In the panels on the right are shown the spots of three subapertures at different distances from the centre of the pupil. . . 53

(11)

2.16 Natural stars on a region of the HeNOS science camera, after focal-plane sharpening. the image is on a logarithmic scale. A grid of ghost stars, shifted respect to the main sources, is clearly visible. . . 55 2.17 Original LGS system used during the early phases of the HeNOS bench. 57 2.18 WFS camera image of four lenslets when using the old LGS setting.

The asterism is clearly not a square. . . 58 3.1 Color image of NGC 1851 obtained with our J and Ks observations. . 64

3.2 Dithering pattern of the 12 long exposures in the Ks band. The same

pattern was used for the 17 J band long exposures, with 5 of the pointings repeated. . . 65 3.3 Example of a processed Ksexposure of 160 s of NGC 1851 with GSAOI.

The orange star symbols show the position of the five LGSs. The three TTSs are circled in red. . . 66 3.4 Example of a raw J band image with a large concentration of hot

(black) pixels. The cross-like pattern visible around them is due to inter-pixel capacitance. . . 68 3.5 Example 400× 3.300 _{region of a K}

s exposure before and after the image

processing with IRAF. . . 69 3.6 Image of a PSF star in the J band that shows significant elongation,

before and after the interpolation. The red contour in the right panel shows the half-maximum values and the blue contour shows the fitting ellipse. . . 70 3.7 Strehl ratios of the PSF stars in a Ks long exposure and their

inter-polation map using a smoothing spline. The position of the TTSs and LGSs are indicated by the red circles and the magenta squares, respectively. . . 72 3.8 Average FWHM, Strehl ratio, and ellipticity of the PSF stars in both

J and Ks exposures (blue and red, respectively). The diffraction-limit

FWHM values are shown using dash-dot lines for comparison. I have not analyzed exposure #10 in the J band due to a particularly poor MCAO correction. These exposures are numbered sequentially in the order they were observed. . . 73

(12)

3.9 Performance maps of a Ks long exposure, based on the measuremnt of

its PSF stars. The map of the position angles shows the orientation of the ellipses fitting the PSF stars’ contours at half maximum and the segments length is proportional to the ellipticity. . . 75 3.10 Maps of the ellipticities of three consecutive Ks long exposures. . . . 76

4.1 Detail of the same region of NGC 1851 on a long exposure and stacked image from GSAOI in Ks band. . . 81

4.2 One chip short exposure with GSAOI of NGC 1851 in Ksband. On the

top are shown the PSF stars used to measure the profile. The bottom panel is the same image after cleaning most of the stars to measure the two-iterations PSF. . . 83 4.3 Cleaning cuts applied to the final catalogs. On top, every star is

plot-ted with its magnitude versus χ, sharpness, and magnitude error. The dashed lines show the limits of the cuts. On the bottom is the dis-tribution on a logarithmic scale of the difference in position of stars matched between the J and Ks bands. The dashed line is the

maxi-mum difference allowed. . . 86 4.4 Image of a bright star of NGC 1851 in a Kslong exposure. The centroid

calculated by DAOPHOT is marked with a green cross. The blue square superimposed has sides of 2rc = 44 px and is the region within

which the adaptive optics correction is applied. Shown for reference is also a magenta circle with a radius of 50 pixels. . . 87 4.5 Radial profile of the same star of Figure 4.4, where each point is a pixel

from the image. The estimated background level is indicated by the green dashed line. The inset panel shows the detail of the radial profile around the control distance rc= 22 px. . . 88

4.6 Uncalibrated CMDs of Ks image using different PSF radii and the

HST visible photometry. The two blue horizontal lines include the stars around the brightness of the MSTO used to measure the width of the sequence. The lower panels show the histogram of stellar colors in this region and the red line is the best fit Gaussian. . . 90 4.7 Uncalibrated color of the MSTO as a function of the PSF radius in a

J and Ks image. The width of the sequence is represented by the error

(13)

4.8 A bright star in a Ks long exposure compared with its PSF models

adopting different PSF radii. The color scale is linear with a cutoff at 10% of the peak. . . 91 4.9 A region of a J band long exposure where the stars have been

sub-tracted after the profile fitting. The central bright star has left a clear halo since its PSF model used is smaller than the real PSF. The white spots along the edge of the profile are spurious detections that are automatically fitted using the PSF model. . . 92 4.10 Six PSF stars in a Ks and their profile fitting residuals using different

PSF variabilities. Most of the residuals visible in the cubic version are actually faint companion stars. . . 94 4.11 As Figure 4.6, but for different degrees of spatial variability in the PSF

model. . . 95 4.12 As Figure 4.6, but for J band images that adopt different outer radii

for the sky annulus. . . 97 4.13 As Figure 4.6, but for different iterations of PSF cleaning. . . 98 4.14 Example of the difference in using single- and multi-frame photometry

on the edge of a Ks image. . . 99

4.15 Difference of the PSF stars’ positions determined by single- and multi-frame profile fitting on a Ks image. The arrows’ lengths are magnified

by 2000. The bottom panel is the histogram of the differences in posi-tion between the two methods. . . 99 4.16 As Figure 4.6, using single- and multi-frame profile fitting. This

pho-tometry is deeper than the uncalibrated CMDs in the previous sec-tions because the input catalog includes the faint stars detected in the stacked images. In this particular exposure, the multi-frame profile fitting performs marginally better than the single-frame. . . 100 4.17 Ks photometric zeropoints of the long exposures without and with the

correction for the seeing in the reference catalog. . . 102 4.18 Magnitude errors of the PSF stars estimated by DAOPHOT for one of

the Ks long exposures. In the first case, the difference between

instru-mental and calibrated magnitude shows a large spread. When I plot the instrumental minus calibrated magnitude minus the δ correction for seeing, the spread is greatly reduced. . . 103

(14)

4.19 Detail of a region of NGC 1851 as seen in Ks band by NEWFIRM on

CTIO for the reference catalog and by GSAOI on Gemini South. The red circles are centered on the stars used for the photometric calibration and have a diameter of 1.200. . . 104 4.20 Histogram of the distance of every star in the reference catalogue to

their closest neighbor. The peak at 1.200 is indicated by the dashed line. 105 4.21 Relation between the correction for the effect of nearby neighbors δ

and the difference between the instrumental magnitudes k in the GeMS data and the corresponding reference magnitudes Kr in the seeing

lim-ited catalog. The best fit linear relationship with a slope of 1 is shown by the dashed line. . . 106 4.22 Median photometric zeropoints in Ksband of the four chips, compared

to the contribution log₁₀g to the instrumental magnitudes from their gain. The dashed line has a slope of 2.5. . . 107 4.23 Ks CMDs calibrated using the reference catalog, without and with

the correction for the seeing and with only one zeropoint per chip, independent on the exposure. . . 108 4.24 Difference between instrumental and reference magnitudes—corrected

for zeropoint and nearby neighbors—as a function of color. The Pear-son correlation coefficients show that there are no significant trends with color for any of the chips, indicating that the color coefficients of the chips are close to zero. . . 109 5.1 Calibrated and cleaned CMDs of NGC 1851 with NIR GeMS/GSAOI

photometry, in combination with HST ACS optical data. The optical-NIR CMDs have inset panels showing a zoom-in around the subgiant branch. The three plots also have a version with isochrones overlaid. . 111 5.2 Position of the stars plotted in the J − Ks CMD of Figure 5.1. . . 114

5.3 Double population in the subgiant branch of NGC 1851 detected with GeMS photometry. In the top panel is the section of the V − Ks CMD

with the SGB selection box and its median line. The bottom histogram is the distribution of offsets from the SGB median line. Overlaid is the best-fit double Gaussian. . . 116

(15)

5.4 On top: smoothed fiducial line of the V − Ks CMD with the position

of the MSTO and MSK in red. On bottom: curvature values of the raw (red) and smoothed (green) fiducial lines. The green circular dot shows the position along the fiducial line of the main sequence knee. The green square shows the point of maximum curvature on the fiducial

line. . . 118

5.5 Standard error values along the raw fiducial line in the V − Ks CMD. 119 A.1 Exposure #1 in J band. . . 127

A.2 Exposure #2 in J band. . . 128

A.17 Exposure #1 in Ks band. . . 143

(16)

(17)

ACKNOWLEDGEMENTS I would like to thank:

David Andersen and Alan McConnachie, for teaching me how to be an as-tronomer and for supporting me during these five years.

the whole Adaptive Optics group at NRC Herzberg, for all the stimulating discussions.

Peter Stetson, for keeping his office door open to my unending questions. Timothy Davidge, for sharing with me his experience on observing.

(18)

PROVERB

Dalle stelle alle stalle

(From the stars to the stables)

(19)

Introduction: Diffraction-Limited

Imaging from the Ground

Since the invention of the telescope in the 17th century, astronomers have wished for telescopes with ever-larger apertures in order to achieve new discoveries. The main reason is that a large collecting area A can capture more photons from a target in the same amount of time than with a smaller instrument, thus increasing the signal-to-noise ratio (S/N), allowing the detection of fainter signals. Another advantage in increasing the aperture diameter D is the reduction in size of the smallest resolvable element in diffraction-limited conditions θ = 1.22 λ/D, the angle of the first zero of the Airy function (Airy, 1835). Not only do more details become visible, but a diffraction-limited poit spread function (PSF) also concentrates the signal of a point source on a smaller region of the focal plane. Combining this with the increase in aperture, the background-limited exposure time is proportional to the forth power of the aperture diameter.

The Extremely Large Telescopes (ELT) scheduled for construction in the next decade—the European Extremely Large Telescope (E-ELT), the Thirty Meter Tele-scope (TMT), and the Giant Magellan TeleTele-scope (GMT)—have collecting areas that are an order of magnitude larger than the present largest optical telescopes. Ob-serving with these telescopes will reach unprecedented depths and spatial resolutions on account of their enormous apertures, opening new possibilities of research on ev-ery astronomical scale, from the smallest bodies in our Solar System to the largest structures in the early Universe. Some of the main science cases for ELTs (for a comprehensive review of TMT science cases, see Skidmore et al. (2015)) include:

(20)

• The smallest bodies in the Solar System can provide indispensable insights into its formation and early evolution. The large collecting area and high spatial resolution of ELTs will provide a more accurate census of asteroids, Kuiper belt objects and comets. Together with their orbital parameters, astronomers can backtrack their dynamical history. Spectroscopy will be extended to the most distant reaches of the Solar System to measure the composition of these bodies. • The improved diffraction limit will allow repeated observations over spans of decades of features on planets and moons, without requiring missions to travel across the Solar System. Astronomers will gain information on phenomena like the wind patterns in the atmospheres of the gas giants, the hydrological cycle on Titan, and the the volcanic and cryovolcanic activity of several moons of Jupiter and Saturn.

• The Kepler space telescope survey has indicated that planetary systems are common around stars in this part of the Galaxy. Most of these systems do not resemble ours. The high sensitivity and resolution of ELTs, combined with coronography and extreme adaptive optics systems, can produce images of ex-oplanets with shorter orbital periods and lower luminosity than before, and deliver a more complete picture of how planetary systems form and if planetary systems like our own are commonplace.

• The search for Earth-like exoplanets will be an important objective of ELTs. The current 10m class facilities are beginning to characterize the atmospheres of the closest hot-Jupiter planets but a larger aperture is necessary to image smaller rocky planets in the habitable range of temperatures. The next step would then be the spectroscopic measurement of biosignatures in their atmo-spheres. This goal is only possible with the extremely high S/N and resolving power that 30m and 40m telescopes will achieve.

• ELTs will be optimized for near-infrared wavelengths, where dust is more trans-parent than in the optical. They will therefore have the ability to probe inside the heavily occulted star-forming regions, where collapsing fragments of giant molecular clouds give birth to stars. Understanding how stars are formed and how the characteristics of the progenitor clouds influence the process will shed light on the initial mass function, a critical piece in understanding the stellar

(21)

populations of galaxies. The large aperture of ELTs will be critical to observe M-dwarfs at the faint end of the initial mass function.

• In the last couple of decades we have learned that Galactic globular clusters differ significantly from the paradigm of a simple stellar population, coeval with homogeneous abundances. Multiple studies with space and ground telescopes indicate that these systems were several times more massive at the time of their formation than they are now. To have a better picture of their history and, by extension, the history of the Milky Way, we can take advantage of observations of globular clusters in the Local Group. With the high angular resolution of Extremely Large Telescopes, individual stars can be separated enough for accurate photometry, astrometry, and spectroscopy, providing a more complete and accurate view of their stellar populations.

• Galaxies are known to host super-massive black holes in their centers. It is still unknown how they form but it is clear that it must be linked to the formation of galaxies. Measurements with ELTs of the stellar dynamics in the cores of galaxies will tell whether there is a correlation between the black hole mass and galactic characteristics like mass, radius, morphology, and environment.

• The current limit in studying galaxy formation at high-redshift is the photon starved regime at which 10m optical telescopes currently operate. A large pri-mary mirror and a suite of instruments optimized for the near-infrared will allow astronomers to detect proto-galaxies, measure their number, their lumi-nosity function, and accretion history. With long observations, we will be able to infer spectroscopically the chemical composition of these early galaxies, as astronomers search for a primordial metal-free stellar formation.

• We still do not have a good understanding of what constitutes the dominant part of the Universe: dark matter and dark energy. The astrometric anomalies of strong gravitational lensing produced by massive galactic halos and clusters will constrain the nature of the dark particles. Observations of high-redshift supernovae and their rate of acceleration will help us to discriminate between different dark energy models.

(22)

1.1 Atmospheric Turbulence

The ability of an ELT to produce sharp images is limited by the Earth’s atmosphere above the aperture in the direction of the target. As with every fluid, the atmo-sphere can flow in two regimes: laminar (layers moving orderly) or turbulent (chaotic motion). The transition between the two is determined by the viscosity of the fluid (Reynolds, 1883). Our atmosphere behaves turbulently, causing a continuous mixing of the air at scales smaller than the aperture size of professional optical telescopes.

The refractive index of air n is dependent on the temperature and pressure of the atmosphere (Smith & Weintraub, 1953), meaning that the wavefront entering the telescope is distorted by variations of the refractive index integrated along the altitude across the aperture. The point spread function of an imaging system is the squared modulus of the Fourier transform of the pupil function. An aberrated wavefront produces a PSF with a lower Strehl ratio (SR) and a larger full width at half maximum (FWHM) than a flat wavefront. The width of the atmospheric FWHM is called “seeing” and it is dependent on the statistics of the turbulence, not on the telescope diameter (for telescopes larger than ∼ 1 m). As a result, ground-based observation gain only a factor D2 _{instead of D}4 _{in background-limited exposure time.}

The statistical structure of the turbulence has been studied by Kolmogorov (1941) and it produces the refractive index power spectrum

Φn= 9.69 · 10−3Cn2κ

−11/3_, _(1.1)

where κ = 2π/l is the angular wavenumber of an eddy of scale size l and C2 n is

the refractive index structure constant that describes the turbulence intensity. von K´arm´an (1948) presented a slightly different turbulence power spectrum

Φn= 9.69 · 10−3Cn2 κ 2_{+ κ}2

L0

−11/6

e(−κ2/κ2l0). _(1.2)

introducing the L0 outer and l0 inner scales of turbulence that define the range of

frequencies outside of which the turbulence is negligible.

The value of C_n2 depends on the characteristics of the atmosphere and it changes with altitude (Figure 1.1). Usually, lower altitudes (within a few hundreds meters from the ground) have the strongest turbulence for a variety of factors, such as ra-diative cooling of the ground during the night, wind shear, physical obstacles to the laminar flow (mountains, trees and the telescope dome), and convection inside the

(23)

dome. The high-altitude turbulence is often concentrated in several layers. Around 10 km a turbulent layer is often observed between the troposphere and the tropopause where the temperature inversion produces the jet streams. The C_n2profile can be mea-sured both directly—with atmospheric balloon probes (Azouit & Vernin, 2005)—or indirectly from the ground with optical instruments like SCIDAR (Rocca et al., 1974; Caccia et al., 1987), generalized SCIDAR (Fuchs et al., 1994; Avila et al., 2001), DIMM (Martin, 1987; Sarazin & Roddier, 1990) with MASS (Kornilov et al., 2003; Tokovinin et al., 2003), or SLODAR (Sch¨ock & Spillar, 1998; Wilson, 2002).

Figure 1.1: Median (solid line), first and third quartiles (dashed lines) C_n2 profile at the Observatorio Astron´omico Nacional de San Pedro M´artir. The solid horizontal line is the altitude of the observatory. The ground and jet stream layers are clearly visible. Image from Avila (2012).

The power spectrum of the wavefront phase φ caused by a Kolmogorov turbulence was found by Fried (1966) to be

Φφ= 0.023 r −5/3 0 κ

−11/3

(24)

with r0 the coherence length defined r0 = 0.42 k2 Z +∞ 0 C_n2(h) dh −3/5 , (1.4)

also known as the Fried parameter, a function of the wavefront angular wavenumber k = 2π/λ. The wavefront variance is σ2

φ = 1.03(D/r0)5/3 (Noll, 1976) and

Equa-tion 1.3 it is dominated by low-order aberraEqua-tions. The isoplanatic angle

θ0 = 0.31

r0

h (1.5)

is defined as the angle seen from the telescope under which the phase varies with a root mean square of 1 rad, where

h = R+∞ 0 h 5/3_C2 n(h) dh R+∞ 0 Cn2(h) dh !3/5 (1.6) is the weighted altitude of the turbulence (Fried, 1982).

Assuming that the wavefront phase at the telescope aperture is shifted by the wind but not distorted (frozen-flow hypothesis, Taylor (1938)), the coherence time

τ0 = 0.31

r0

v (1.7)

is the time over which the phase along a line of sight varies with a root mean square is 1 rad, with the weighted wind velocity

v = R+∞ 0 v 5/3_C2 n(h) dh R+∞ 0 Cn2(h) dh !3/5 . (1.8)

The long exposure PSF is the superposition of the instantaneous PSFs created by the different realizations of turbulence during the exposure and its full width at half maximum is F WHM = 0.98 λ/r0 (Sarazin & Roddier, 1990). For large telescopes

r0 is smaller than D, so resolution in seeing-limited conditions is worse than in the

diffraction-limited case, where (F WHM = 1.03 λ/D).

It would seem that the obvious solution to provide diffraction-limited conditions to large telescopes is to put them outside of the atmosphere. The problem with space telescopes is that, since their weight scales with the cube of the diameter, the cost to launch an ELT even in low Earth orbit is prohibitive. Also, there are currently no rockets that can house such a big instrument. Moreover, if the telescope is designed

(25)

to observe in the infrared, like the James Webb Space Telescope and the ELTs, it would have to be put at a distance from the Earth that makes it unserviceable. For these reasons and others, ground-based telescopes will still play a central role in the future of astronomy and so it becomes imperative to solve the problem of atmospheric turbulence.

1.2 Adaptive Optics

To restore the potential “D4_{” benefit of operating in seeing limit on large telescopes,}

the technique of adaptive optics (AO) is used to measure and correct the atmospheric turbulence (see Davies & Kasper (2012) for a comprehensive review). In “classic” adaptive optics, aberrations are measured in the direction of a bright source above the turbulence by using a wavefront sensor (WFS). Natural guide stars (NGS) are ideal for this task but often no bright star is available in the direction of the scientific target (within θ0). The solution is to generate an artificial star called a laser guide

stars (LGS) (see Sandler (1999) for a review on laser guide stars). This is created by a laser beam launched from the proximity of the telescope in the direction of the target (Foy & Labeyrie, 1985; Thompson & Gardner, 1987). Most lasers operate at 589 nm of wavelength to resonate with the D2 line of the sodium atoms in the ∼ 10 km thick mesopause at 90 km of altitude(Pfrommer & Hickson, 2010, 2014).

Wavefront sensors divide the pupil into regions called sub-apertures and on each of them it can measure the phase directly—as in the unmodulated pyramid WFS (Ragaz-zoni, 1996)—or its gradient—as in the Shack-Hartmann (SHWFS) (Collier, 1971; Platt & Shack, 2001)—or its second derivative—as in the curvature WFS (Roddier, 1988) (see Rousset (1999) for a review on wavefront sensors). The Shack-Hartmann is the most common WFS in adaptive optics. It uses a square array of microlenses in the pupil plane to focus a portion of the wavefront onto a detector. The position of the spot on the focal plane is dependent on the gradient of the phase (Figure 1.2).

Since LGS wavefront sensors are blind to the tip-tilt aberration (Rigaut & Gen-dron, 1992), natural stars—called tip-tilt stars (TTS)—are necessary in an LGS AO system to measure this mode, although they can be fainter than normal NGSs since they have only to determine the average phase gradient of the entire pupil. Another NGS WFS needed in an LGS AO system is called the truth wavefront sensor (TWFS), it has a higher spatial resolution and it is used to calibrate the LGS focus term caused by the sodium layer finite altitude and to track its change caused by the movement

(26)

Figure 1.2: Schematic operation of a Shack-Hartmann WFS with a flat (on top) and distorted (on bottom) wavefront. Image from Rousset (1999).

of the sodium layer during the observation. It also compensates for asymmetries of the sodium profile density that can induce radially symmetric aberrations.

After the wavefront sensor has measured the phase, the real-time control system acquires the data and calculates the correction that the AO system has to apply to inject a distortion equal and opposite to the turbulent wavefront. A deformable mirror (DM) is a reflective surface that can quickly change shape due to actuators placed behind the reflective surface (see S´echaud (1999) for a review on deformable mirrors). The maximum spatial frequency (or control frequency) that a deformable mirror can correct is fc= 1/(2d), where d is the actuators pitch projected on the aperture of the

telescope. The AO system behaves as a high-pass filter for the wavefront, such that only light within the control radius

rc =

λ

2d (1.9)

of the PSF can be affected by the correction, leaving the wings of the profile mostly unchanged with respect to the seeing-limited PSF. Because the actuators are typically arranged on an orthogonal grid, the shape of the so called “dark hole” can be a square with side 2rc.

(27)

For an adaptive optics system to work at its best, it should be able to sample and correct the wavefront with a spatial resolution of r0 (Equation 1.4) and temporal

resolution of τ0 (Equation 1.7). The region of sky where the correction is optimal has

an angular size of θ0 (Equation 1.5), inside which the turbulence is correlated to the

DM correction. Because these values are ∝ λ1.2_{, adaptive optics performs better at}

near-infrared wavelengths than at optical ones, because NIR AO systems need fewer actuators, a slower loop and correcting a larger patch of sky.

(a) Binary star observed with PUEO on CFHT (Fusco et al., 1999)

(b) Multiple star system observed with LBTAO on LBT (Close et al., 2012)

Figure 1.3: The dramatic evolution in a decade of the classical AO PSF has been driven by many factors, including increased density of actuators on the DM and faster frame rates.

1.3 Multi-Conjugate Adaptive Optics

The main shortcoming of classical AO is the limited angular size of the corrected field θ0—known as the isoplanatic angle—typically of the order of a several arcseconds. A

(28)

single guide star can probe only the cylinder of turbulence in one direction. As the angular separation to a target increases, so will the difference in aberrations sensed by the guide star and the science target, thus widening and deforming the delivered PSF. This effect, known as anisoplanatism, is mitigated by multi-conjugate adaptive optics (MCAO) systems (Beckers, 1988). In MCAO, multiple guide stars—spread over the field of view—are employed to probe the atmosphere in different directions and reconstruct the atmospheric distortion in a larger volume. Multiple deformable mirrors—optically conjugated to different altitudes—apply the correction to the layers of turbulence associated to those heights. The size of the region of sky that an ideal MCAO system can correct is called the generalized isoplanatic angle and it is determined by the generalized fitting error (Rigaut et al., 2000), the inability of the system to correct properly for turbulence layers not conjugated to a DM. Even with this limitation, the generalized isoplanatic angle θn (where n is the number of DMs

employed) is typically several times larger then the isoplanatic angle θ0, providing

diffraction-limited images over a large patch of sky.

While an MCAO system can work using only NGSs, the use of LGSs is advanta-geous because it alleviates the requirements for an asterism of bright stars close to the scientific target, thus increasing significantly the sky coverage. Although, three natural tip-tilt stars are required in MCAO to counter the “tilt anisoplanatism”, the inability of an LGS MCAO system to measure and correct the “null modes”, the five geometric distortions of the image (global tip-tilt and plate scale) caused by equal and opposite focus terms on the DMs (Ellerbroek, 1994; Rigaut et al., 2000; Ellerbroek & Rigaut, 2001; Flicker & Rigaut, 2002).

Galactic globular clusters (GGC) are ideal targets for MCAO instruments. First, they are a collection of tens of thousands of stars and many of them are bright enough to be used as tip-tilt stars. They can also be used for both measuring the performance across the image and creating the PSF model. Moreover, the use of MCAO has a clear advantage over observations of GGCs that don’t use this technology. Galactic globular clusters are crowded and extended objects that on seeing-limited images have only a few of the brightest stars easily distinguishable, while the others are either too faint or too close to other stars to be detected.

MCAO has been proven to work on sky in 2007 with the short-lived demonstrator MAD on VLT (Marchetti et al., 2003, 2007). More recently, the Gemini Multi-Conjugate Adaptive Optics System (GeMS) has started operations in 2012 at the Cassegrain focus of the Gemini South Telescope on Cerro Pach´on (Chile) (Rigaut

(29)

et al., 2012; Neichel et al., 2013b, 2014; Rigaut et al., 2014). It is the only MCAO currently on sky, the first to use LGSs, and the first facility-class instrument of its kind. It uses five sodium laser beams to measure the turbulence, arranged in a 6000 square asterism—with the fifth star in the center—launched from the back of the secondary mirror (d’Orgeville et al., 2012; Neichel et al., 2013a). Five SHWFSs (one for each LGS) measure the high-order aberrations along these five directions, with three more deployable wavefront sensors that patrol a field of 20. The deformable mirrors are optically conjugated to the ground and 9 km. The instrument was orig-inally designed with a third deformable mirror conjugated to 4.5 km but, following the failure of multiple actuators (Rigaut et al., 2014), GeMS was commissioned with only two deformable mirrors.

The corrected wavefront is imaged by GSAOI, the Gemini South Adaptive Optics Imager (McGregor et al., 2004; Carrasco Damele et al., 2012). This camera has four Teledyne HAWAII-2RG detectors (Blank et al., 2012)—a HgCdTe hybrid CMOS (Janesick, 2004) sensitive to NIR—arranged in a 2×2 array, each chip with 2040×2040 active pixels. The pixel scale is 0.019600/px and the total field of view is 82.5 × 82.500, with a gap of 2.500 between the chips. The broadband filters used in the instrument (Carrasco Damele et al., 2012) were built with transmission curves close to the 2MASS photometric system (Bessell, 2005) to provide an easy photometric conversion.

The future generation of ELTs will need MCAO systems to deliver diffraction-limited images over relatively wide fields. One of these instruments will be NFIRAOS (Narrow Field InfraRed Adaptive Optics System), the first light MCAO facility planned for the Nasmyth platform of the Thirty Meter Telescope (Herriot et al., 2012; Boyer et al., 2014; Herriot et al., 2014; Boyer & Ellerbroek, 2016). It will have two deformable mirrors (Figure 1.4) conjugated to the ground (DM0) and to 11.2 km (DM11), and six LGS wavefront sensors. It will also include a pyramid wavefront sensor that serves as both a TWFS or as the sole WFS when the instrument is used in NGS-only mode (V´eran et al., 2014). NFIRAOS promises to deliver a much higher performance than GeMS as it uses more LGSs over a smaller field, aiming to achieve 60% Strehl ratio in H band over ∼ 3000. NFIRAOS is being designed and will be built in Victoria at NRC HAA. The first light science instrument that will take advantage of the NFIRAOS correction will be the InfraRed Imaging Spectograph (IRIS) (Larkin et al., 2010; Moore et al., 2014; Wright et al., 2014).

(30)

Figure 1.4: NFIRAOS mechanical and optical design. The optical paths are in red.

1.4 Thesis Overview

For this thesis I studied how the inner workings of a multi-conjugate adaptive optics system determines the quality of the scientific output. My goal is to understand the photometric performance of future MCAO systems by pursuing two different paths, astronomical instrumentation and observations. These are often treated separately but they need to be synthesized to reach the ambitious goals of diffraction limit science with Extremely Large Telescopes. The first aspect of my work was the in-tegration and testing of HeNOS (Herzberg NFIRAOS Optical Simulator), an optical bench that reproduces several features of NFIRAOS in order to prove background and calibration algorithms that will be used on TMT. In addition, I observed the Galactic globular cluster NGC 1851 using GeMS. I used this existing facility to de-termine best techniques to measure the performance, analyze the data, and produce precise photometric science, with the future goal to transport this strategy to MCAO systems on ELTs. In Chapter 2, I describe the HeNOS experiment and the tests and experiments that I conducted. In Chapter 3, I summarize the observations of NGC 1851 made with GeMS, how I reduced the image, and how I derived the instrument

(31)

performance in terms of PSF. Chapter 4 presents the test that I conducted to obtain precise and accurate photometry with MCAO observations. In Chapter 5, I use the photometry obtained in the previous chapter to generate color-magnitude diagrams that I use to detect several features peculiar to NGC 1851. In Chapter 6 I present the conclusions to my work.

(32)

Chapter 2 HeNOS: An MCAO Test Bench

2.1 Introduction

To realize the full scientific potential of the Extremely Large Telescopes, we need AO systems much more complex than the ones currently deployed. To meet tighter wavefront error budgets on larger telescopes, AO instruments have to rely on advanced methods of calibration and operation, some of which have only been demonstrated in simulation. The associated risk can however be mitigated by using scaled-down laboratory demonstrators to validate these calibration and operational methods, so that they can be corrected and refined before being implemented on a full scale ELT instrument.

To provide direct support for the development of NFIRAOS for TMT, we have set up a MCAO test bench dubbed the Herzberg NFIRAOS Optical Simulator (HeNOS) in the adaptive optics laboratory at NRC Herzberg. HeNOS is a scaled down version of NFIRAOS and it is intended to reproduce its behavior within the limits of a labora-tory. The purpose of HeNOS is to identify and help solve issues that could adversely affect the performance of NFIRAOS and that are not apparent from computer sim-ulations alone. On a typical H-band observation with NFIRAOS, the current error budget for the wavefront error is 220 nm, corresponding to SR = 0.50. An additional 80 nm of error—easy to overlook—raises the total wavefront error to 235 nm when added in quadrature, with SR = 0.45. The faint source sensitivity is proportional to SR−2, so this small error would increase exposure times by ∼ 25%. The HeNOS experiment is designed to prevent just these type of small errors from having big impacts on the science performed with TMT.

(33)

In order to compare the computer simulations with a real life system, we have to anchor MAOS (Wang & Ellerbroek, 2012)—the NFIRAOS end-to-end Monte Carlo AO simulation tool—to measurements in the lab by comparing the expected perfor-mance on sky to the bench in a variety of well controlled conditions, such as faint and poorly corrected NGSs, non-uniform sodium layer (Pfrommer & Hickson, 2014) and field-dependent non-common-path aberrations. Other goals for the bench include de-veloping calibration procedures testing the robustness of the tomographic algorithm under realistic and varying conditions, and validating the optimization methods that operate on time scales longer than ten seconds. An example is the task updating the matched filter (Gilles & Ellerbroek, 2006) with the evolving measurement of elon-gated LGSs with a SHWFS and estimating C_n2 using SLODAR (SLOpe Detection And Ranging) (Wilson, 2002). The bench will also eventually be used to test astrometric and photometric calibration techniques for NFIRAOS.

Validating these algorithms on HeNOS is important not only to minimize the possibility of crippling errors for the performance of the instrument and to push the quality of the AO correction, but also to guide the image analysis done by astronomers when they will use NFIRAOS. A primary example is the experiment conducted on HeNOS on MCAO PSF reconstruction. Typically, precise photometry and astrometry require to measure the PSF from selected stars directly on the analyzed image. If a field doesn’t include enough bright stars or, on the contrary, the field is too crowded, then it becomes difficult to produce an accurate point spread function. The solution is then to predict it from the knowledge of the optical system, an estimation of the turbulence parameters and the telemetry of the adaptive optics loop during the exposure. This technique has already been demonstrated for single-conjugate AO systems (V´eran et al., 1997) but its extension to MCAO and NFIRAOS will require additional modeling that can be put to test on HeNOS (Gilles, 2016, 2017). This series of experiments will help to provide future users of the instrument with a new method of analysis and the knowledge to use it effectively that will maximize the science output from their MCAO observations on TMT.

To properly reproduce the features of NFIRAOS relevant to reach these goals, the main subsystems of HeNOS (described in detail in Section 2.3) have to reproduce those on NFIRAOS. The atmospheric turbulence is distributed in altitude, concentrating in several layers, the strongest of which is close to the ground. Two deformable mirrors, conjugated to the ground and to a high altitude, provide corrections on a wide field of “natural” stars (point-like sources at infinite distance) imaged on a science camera.

(34)

The commands sent to the DMs are evaluated from wavefront sensors pointed to elongated laser guide stars. The HeNOS design is driven by how the architecture of the system is scaled down from a 30m class telescope to an optical bench.

I began work on HeNOS from its earliest stages when the design by P. Span`o had to be translated on to the optical bench. I laid out the optical elements and aligned them. I integrated several electronic components (cameras, deformable mirrors, laser diodes and phase screen motors) whose controllers were prepared and tested by E. McVeigh. Together with M. Rosensteiner I then calibrated the bench and I wrote the algorithms that run the AO loop. The system is controlled by a single PC using code written in MATLAB and C++.

In this chapter, I first analyze in Section 2.2 the main parameters that define the design of HeNOS and describe how they were chosen. From there, I derive in Section 2.3 the general characteristics of the bench subsystems. Following the align-ment of the optics described in Section 2.4, I have characterized several fundaalign-mental parameters of the system in Section 2.5, with the purpose of evaluating precisely the bench performance. In Section 2.6, I then describe several calibrations and other operations necessary for operating HeNOS as a complete MCAO system. Finally, in Section 2.7 I discuss several issues that I have encountered on HeNOS during these years, providing an explanation of the causes and possible solutions.

2.2 Experimental architecture

The core components of the HeNOS hardware are the two ALPAO magnetic DMs. Their characteristics presented in Table 2.1 constrain the system design. The smaller DM97 substitutes for the NFIRAOS ground-layer DM0, while DM277 is in place of the high-altitude DM11. The number of actuators across the circular aperture of DM97 is 8.7 which is obviously not close to the 61 actuators across the NFIRAOS DM0 (Herriot et al., 2014). Our first architectural decision was therefore to simulate a telescope of aperture DH = 8 m, which implies the pitch of the array of actuators

on the back of the DM is pH = DH/a0 = 0.92 m.

The bench operates with visible light sources since it is easier to align the optics and operate the detectors with visible light rather than near-infrared wavelengths. The light sources on HeNOS are AlGaInP laser diodes with a maximum optical output power of 5 mW at λH = 0.67 µm.

(35)

DM0 DM11

Model Hi-Speed DM97-15 Hi-Speed DM277-15

Number of actuators 97 277

Aperture diameter A0 = 13.5 mm A1 = 24.5 mm

Actuators across the aperture a0 = 8.7 a1 ∼ 15.8

Bandwidth 750 Hz 800 Hz

Settling time (±5%) 1.0 ms 1.0 ms

Table 2.1: Main characteristics of the ALPAO DMs in HeNOS.

for the wavefront sensor and the science camera. They are operated using a typical exposure time of 20 ms, a gain value of 1 and linear gamma.

Model Grasshopper2 GS2-GE-50S5M

Number of pixels 2448 × 2048

Pixels on the diagonal np = 3192 px

Pixel size 3.45 µm × 3.45 µm

Quantum efficiency (@0.67 µm) 76%

Full well depth 7300 e

Read noise 10.03 e

Dark current 47.10 e /s

Maximum frame rate 15 fps

Table 2.2: Main properties of the Point Grey cameras in HeNOS.

The science camera is required to have a resolution at least equal to the Nyquist limit in order to properly sample the core of a diffraction-limited PSF. The pixel scale therefore has to be smaller than s = λH/(2 · DH) = 8.64 mas/px, resulting in

a maximum diagonal field of view F OV = np · s = 27.600. A smaller field of view

F OVH = 10.900 has actually been chosen in order to reduce the size of the optics

and decrease their cost. The science camera is now oversampled, but the loss in signal-to-noise can be compensated by increasing the intensity of the sources.

Even if the bench is simulating a telescope with an aperture smaller than TMT, it can reproduce in part the NFIRAOS performance by preserving the number of isoplanatic angles θ0 across the field of view. To set the bench parameters, I note

the NFIRAOS F OVN = 20 (Boyer et al., 2014), the isoplanatic angle in H band

(36)

(V´eran et al., 2012). By requiring that F OVH θ0,H(λH) = F OVN θ0,N(λN) , (2.1)

the bench then has to have θ0,H(λH) = 0.8500. If HeNOS uses a turbulence profile

similar to the turbulence profile expected at Mauna Kea, then r0,H(λH)

θ0,H(λH)

= r0,N(λN) θ0,N(λN)

, (2.2)

and the Fried parameter value would be r0,H(λH) = 0.0683 m. This value is much

smaller than the actuators pitch pH = 0.92 m, which would lead to a severe fitting

error (Section 2.6.2). To increase r0,H(λH) without affecting θ0,H(λH), the turbulence

profile can be stretched vertically by a factor v that moves the altitude of a layer from hN to hH. The maximum value for v is limited by the ability of the high-altitude DM

pupil A1 to contain the meta-pupil Dm,H:

v = hH hN = Dm,H − DH hN · tan (F OVH) = DH(A1/A0− 1) hN · tan (F OVH) = 11. (2.3)

The Fried parameter then becomes

r0,H(λH) = v · θ0,H(λH) ·

r0,N(λN)

θ0,N(λN)

= 0.75 m, (2.4)

much closer to pH = 0.92 m, and therefore limiting the fitting error. In Table 2.3 I

report relevant parameters of the NFIRAOS and HeNOS designs.

The main requirement of the HeNOS bench is to perform the simulation of a full MCAO run in conditions similar to NFIRAOS. To meet this demand, the bench has to be set with the the two deformable mirrors and the three phase screens in the optical path. In addition, a non-common path aberration (NCPA) correction has to be linearly added to any other command applied on the DMs (see Section 2.6.1). A MCAO loop has then to go through these steps in order:

• Run a sequence of DM0 defocus commands while modulating the LGS sources to simulate elongated spots on the SHWFS due to the thick sodium layer, as described in Section 2.6.4.

• Stack individual exposures from the wavefront sensor camera taken during the previous step to produce a measurement of elongated LGSs, as in Figure 2.15.

(37)

NFIRAOS HeNOS

Fried parameter 0.75 m 0.75 m

Isoplanatic angle 9.400 0.8500 Operational wavelength 1.6 µma _{0.67 µm}

Aperture diameter 30 m 8 m

Actuators on DMO diameter 61 8.7

Actuators pitch on the aperture 0.49 m 0.92 m

DM0 conjugated altitude 0 km 0 km

DM11 conjugated altitude 11.2 km 123 km LGS asterism diameter 7000 6.400

a_{Representative of the IRIS filter range (Walth et al., 2016).}

Table 2.3: Comparison of NFIRAOS and HeNOS design parameters.

• Measure the slopes from the SHWFS spots with methods including matched filtering, center of gravity (CoG), thresholded CoG, weighted CoG, quad cell or correlation (for a review, see Thomas et al. (2006) and Gilles & Ellerbroek (2006)).

• Calculate the command for the deformable mirrors from the slopes, either by using a simple command matrix (Section 2.6.3) or by measuring the turbulence profile and using a tomographic reconstructor.

• Update the DM0 and DM11 commands using the calculated correction modified by the system gain factor.

• Advance the turbulence by one time step rotating all the three phase screen by an angle determined by the wind profile and the system frame rate.

• Turn on the natural stars and take an exposure with the science camera. The system must be able to store the telemetry—both slope measurements and DM commands—and the science camera “long exposure” image obtained by stacking all the individual exposures taken in the last step of each loop.

(38)

2.3 Optical Design

To simulate NFIRAOS, the HeNOS bench has been designed with the following prin-cipal components:

• Multiple light sources conjugated at infinity (“natural stars”) to be used for tip-tilt measurements and with the truth wavefront sensor.

• Multiple LGSs to measure the atmospheric turbulence in different directions for the tomographic reconstructor.

• Two deformble mirrors conjugated at different altitudes.

• Multiple phase screens to simulate a realistic atmospheric turbulence profile extended in altitude and concentrated in layers.

• A wavefront sensor for each LGS.

• A “science camera” focused on the natural stars to evaluate the system perfor-mance.

These elements and constraints were incorporated into the optical design of P. Span`o (Figure 2.1a). The figure also includes the optical path of a pyramid wavefront sensor designed by E. Mieda that has been recently implemented in HeNOS (Mieda et al., 2016) to replicate the truth wavefront sensor on NFIRAOS (V´eran et al., 2014). This subsystem is outside the scope of this thesis and will not be discussed here.

2.3.1 Natural Stars

HeNOS needs natural stars spanning the science field of view to test the bench’s ability to correct the global tip-tilt, measure the distortion, and detect how changes in the sodium layer can induce wavefront errors in radial modes. These natural sources are also useful to measure the photometric and astrometric performance of HeNOS across the field. Our design solution to generate a dense and uniform field of natural stars was to inject the light of a laser diodes carried by an optical fiber, collimate it and illuminate a microlens array. The lenslets create a regular grid of sources across the focal plane that—once collimated—reproduce point-like sources at infinity. The natural stars generator was designed to produce about 20 × 15 objects in the science field of view (see Figure 2.2).

(39)

LGSs

TTSs

PS0

Telescope

Collimator

PS2

DM11

PS1

DM0

SHWFS

Science camera

PWFS

(a) Optical design

(b) Candid photograph

Figure 2.1: The HeNOS bench at NRC Herzberg. In the top panel, the “science” light rays are in blue, the LGS SHWFS path is in red and the PWFS path is in green. The photo in the lower panel has been taken without arranging the bench, to show the typical difference between ideal planning and real implementation of an experiment.

(40)

Figure 2.2: Grid of natural stars on the HeNOS science camera full field.

2.3.2 Laser Guide Stars

NFIRAOS will measure the high-order aberrations using six LGSs in an asterism with αN = 7000 diameter on sky (Boyer et al., 2008), at an altitude of about hs,N = 90 km.

Since the sodium layer is not at infinity, the images of the LGSs are out of focus at the science camera focal plane. The LGSs appear out of focus on the HeNOS science camera too, at a simulated altitude of hs,H = v · hs,N = 990 km in the stretched

atmosphere. The size of the asterism changes too, shrinking to αH = αN/v = 6.400.

For HeNOS we use a simple square asterism, where 6.400 of diagonal corresponds to 4.500 of side. We can take the slight performance reduction on HeNOS associated with having just four LGSs into account when comparing results to those estimated for NFIRAOS.

On the bench, the LGS sources are produced by four laser diodes each coupled to an optical fiber. The other ends are held in a metal plate by holes in a square metal

(41)

plate. A series of three lenses injects light from the LGSs into the optical path at the correct altitude from the telescope (Figure 2.3). The light from the natural and laser stars is combined using a wedged plate beamsplitter. During bench operations (Section 2.6), the two sources are never turned on at the same time, to avoid confusion in the science and wavefront detectors.

Figure 2.3: LGS generator on HeNOS. On the right is the plate with holes holding in a square configuration the four optical fibers. A series of lenses delivers the light to the telescope.

The LGS asterism is designed to be centered with respect to the optical axis of the system to minimize the static aberrations and distortions. The science camera field of view is also centered on the LGSs in order to verify that the MCAO performance is symmetric.

2.3.3 Deformable Mirrors

The collimated light from the natural sources is focused by a group of lenses that represent the telescope and it is then recollimated in a path that includes the DMs. The reason for these extra optics is to create two pupils. One placed on the ground phase screen, the other located at the ground deformable mirror. The two pupils have different sizes to accommodate the individual scales of the Fried parameter of the phase screen and the actuator pitch. NFIRAOS will have two DMs conjugated at 0 km and 11.2 km of altitude (Herriot et al., 2014). On HeNOS they are stretched by the factor v = 11. To keep the on-axis collimated beam normal to the reflecting

(42)

surfaces of the mirrors and preserve the square pattern of actuators, a beamsplitter cube is in front of each DM to separate the paths of the incident and reflected light (Figure 2.4). DM0 works also as the optical stop of the system.

Figure 2.4: HeNOS DM0 on the left with a beamsplitter cube in front of it. On the right is the optical flat that substitutes PS1 during the calibrations.

2.3.4 Phase Screens

The turbulence profile is generated on HeNOS by employing three rotating phase screens representing ground (PS0), mid-altitude (PS1), and high-altitude (PS2) lay-ers. Their function is to introduce a wavefront distortion statistically similar to an atmospheric one to the collimated beams. The phase screens are on custom-built rotary stages (Figure 2.5) and the relatively small optical footprint of the collimated beam passes through them close to the edge. By doing so, the ratio between the size of the pupil and the distance to the center of rotation is small enough that when the phase screens rotate, the atmosphere seems to translate linearly, as in frozen-flow turbulence (see Section 1.1).

PS0 was manufactured at the University of California in Santa Cruz by pseudo-randomly spraying acrylic paint on a λ/10 plastic plate using a purpose-built robot (Rampy et al., 2010). The other two phase screens were produced by Lexitek Inc. (Ebstein, 2002) and are produced using the index matching technique: two glass plates

(43)

Figure 2.5: Two of the three phase screens used on HeNOS, complete with rotary stages and motors. On the left is one of the two Lexitek, on the right is the UCSC phase screen.

with close refraction indices are machined in the desired profile with complementary shapes and then are glued together. The Lexitek phase screens can be produced from a digital phase map with the r0 size required to be placed in the same conjugated

space of the DMs. While these are advantages to the Lexitek phase screens, they are more expensive and made of thick glass. The thinner and cheaper UCSC phase screens are produced with an r0 too small for the DMs pitch. But since the ground

phase screen can not share the same physical space of the ground DM, it has to be put in another conjugated space where the collimated beam can be compressed to an appropriate size. It is positioned in the pupil in front of the telescope (Figure 2.1a), where the beam size is 10 mm instead of the 13.5 mm of PS1 and PS2.

The three screens have von K´arm´an statistics (see Section 1.1) within the limits of their production processes, contributing—from the lowest—72.3%, 19.8% and 7.9% to the total aberration variance σ2_{. These values reflect the expectation that the}

ground layer is dominant and that turbulence decreases with height. PS0 is kept as close as possible to the pupil to simulate the ground layer (see Section 1.1) as well as turbulence included within the TMT dome itself. PS1 and PS2 represent the extended turbulence at higher altitudes (The TMT Site Selection Team, 2008) and are positioned respectively at v · 4 km and v · 13 km of altitude, distant enough from DM11 to prevent any physical interference. The combination of powers and

(44)

altitudes of the three phase screens was chosen to produce on the bench a generalized isoplanatic angle on HeNOS θ2,H (see Section 1.3) such that

θ2,H

θ0,H

= θ2,N θ0,N

. (2.5)

V´eran et al. (2012) have calculated that on NFIRAOS the generalized isoplanatic angle will be θ2,N(λN) = 34.600. By knowing that the isoplanatic angles are θ0,N(λN) =

9.400 on sky and θ0,H(λH) = 0.8500 on the bench (see Table 2.3), we can predict that

θ2,H(λH) = 3.100.

2.3.5 LGS Wavefront Sensors

The HeNOS bench uses a Shack-Hartmann wavefront sensor to measure the aberra-tions of the LGSs. The minimum number of lenslets across the pupil for a proper spatial sampling of the wavefront should be DH/r0,H(λH) ∼ 11. For this experiment

it is more useful to measure precisely the wavefront than to test the effects of spatial aliasing on the performance (Section 2.6.2). The SHWFS is therefore designed to have w0 = 30 subapertures on the pupil diameter and—as for the science camera—

the downside of the wavefront is mitigated on the bench by the ability of controlling the intensity of the LGSs and the exposure time of the LGS WFS.

Instead of having a separate SHWFS for each LGS, we opted for a simpler solution. A single microlens array can image all four laser stars at once because the field of view of each of the lenslets was chosen to be larger than 4.500. Four spots form behind each lenslet and–since they are separated from each other by enough pixels—their positions can be measured independently within the same frame, even after introducing spot elongation.

After DM0, an optical relay resizes the collimated beam into the conjugated mi-crolens array. A second optical relay reimages the spots formed by the lenslets onto the camera, scaling the field to fit the detector size.

2.3.6 Science Camera

A plate beamsplitter between DM0 and the first optical relay of the wavefront sensor diverts half of the light to the science camera. A simple system of two lenses focuses the light from the natural star sources onto the detector, with minimal aberrations and geometric distortions.

(45)

In recent months, the Grasshopper2 detector has been replaced with an Andor Zyla 4.2 PLUS sCMOS after the bias voltage of the former was found to vary dras-tically on a timescale of minutes. The latter detector also has the advantage of a substantially smaller read noise (0.9 e ) and dark current (0.1 e /s) that will allow to observe multiple diffraction rings on sources with high Strehl ratios. Note however that all the measurements on the bench for this thesis were taken with the Point Grey camera.

2.4 Optical Alignment

I began aligning the optics by defining the reference optical axis using a laser. Its height from the bench was set by the height of the centers of the phase screens mounted on their rotary stages. Where closely spaced optical elements had to be aligned sharing the same axis, we employed a “cage” system as illustrated in Fig-ure 2.3. The horizontal metal bars assFig-ure a reasonable transverse alignment and only the longitudinal positioning is left to be determined. I used retro-reflecting tech-niques with the laser to align most of the refractive optics. When mounting the refractive optics—one at a time—I checked the centering of each lens by returning to the aligning laser the beams reflected by the refractive or reflective surfaces.

Cage systems were used for both the natural and laser guide stars generators. The beamsplitter that combined the light from the two sets of sources was included in the cage of the natural stars. The lenses were positioned with respect to the optical fibers at the distances provided in the optical design. The axial alignment of the collimator doublet in front of PS0 was determined by the observation of a well-focused image of the natural stars on a small portable telescope focused to infinity. I used this same method for placing both the collimator in front of the deformable mirrors and the optical relay between DM0 and the microlens array. The telescope doublet was put at a distance from the collimator determined in the design.

Phase screens were not put in the optical path during the alignment and cali-bration of the bench because their aberrations would deteriorate the images used to position collimators and focal planes. Also, the defocus term that they introduce are characteristic of the particular sections of phase screens traversed and are not representative of the average. But because PS1 and PS2 are much thicker than PS0, they introduce a substantial optical path length that should be maintained in the system when they are removed, otherwise it would produce a shift in conjugated

Advancing next generation adaptive optics in astronomy: from the lab to the sky

Contents

List of Tables

List of Figures

Dalle stelle alle stalle

(From the stars to the stables)

Introduction: Diffraction-Limited

Imaging from the Ground

1.1

Atmospheric Turbulence

1.2

Adaptive Optics

1.3

Multi-Conjugate Adaptive Optics

1.4

Thesis Overview

Chapter 2

HeNOS: An MCAO Test Bench

2.1

Introduction

2.2

Experimental architecture

2.3

Optical Design

2.3.1

Natural Stars

LGSs

TTSs

PS0

Telescope

Collimator

PS2

DM11

PS1

DM0

SHWFS

Science camera

PWFS

2.3.2

Laser Guide Stars

2.3.3

Deformable Mirrors

2.3.4

Phase Screens

2.3.5

LGS Wavefront Sensors

2.3.6

Science Camera

2.4

Optical Alignment