Professional Documents
Culture Documents
This content has been downloaded from IOPscience. Please scroll down to see the full text.
(http://iopscience.iop.org/0963-9659/7/3/005)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 130.239.20.174
This content was downloaded on 07/04/2015 at 08:51
Guy Artzner†
Institut d’Astrophysique Spatiale (IAS), Unité Mixte de Recherche (UMR) 120, Centre National
de la Recherche Scientifique (CNRS), Université Paris XI, Bâtiment 121, 91405 Orsay Cédex,
France
Introduction
Astrophysical knowledge depends on, among other factors, the quality of the optics
used for observations. For example, the recent discovery of a planet outside the solar
system is claimed from observations made with the 1.93 m Haute Provence observatory
telescope polished by optical expert Jean Texereau; soundings of Jupiter’s atmosphere
during comet impact were performed at the observatory at Pic du Midi using the 1 m
planetary telescope, also polished by Jean Texereau. Direct observation of the source of
coronal mass ejections with the EIT (extreme ultraviolet imaging telescope) instrument
[1] on board the SoHO observatory stems from progress made in producing multilayer
† E-mail address: artzner@iaslab.ias.fr
coatings that reflect coronal EUV light at normal incidence. In contrast, users of the
Hubble Space Telescope (HST) before and after correction of its primary mirror’s spherical
aberration [2] are aware of the fact that making correct measurements of optical systems
during production and before observations is useful, despite progress made in the fields of
active and adaptive optics. Double or triple checking by independent methods is the usual
means of safeguarding accurate measurements free from systematic errors at the production
stage, before shipment to distant ground- or space-based observing sites. Several methods
for laboratory measurements of a variety of systems have been proposed in books [3] and
papers [4–9].
Here we consider the centre-of-curvature testing of large aspherical mirrors. In table 1
we apply analytical formulae to five typical examples, ordered by increasing asphericity, and
in section 1.2 we study an extreme case, where ray tracing indicates that the nominal Shack–
Hartmann method does not apply because of interlacing rays. In section 2 we describe a
laboratory experiment employing such interlacing rays. We discuss the extrapolation of
the results of section 2 to other aspherical wavefronts, extending the operating range of
the Shack–Hartmann method to absolute laboratory measurements of a large astronomical
mirror.
Deviation
from best-fit Distance
Focal sphere Exact from C to Longitudinal
Diameter length r 4 /256f 3 value beam waist Half-waist aberration
Mirror (mm) (mm) (µm) (µm) (mm) diameter (mm)
We consider a 3.6 m diameter liquid mirror and choose a 000 1 value as an intermediate
criterion between the 1.22λ/D = 000 03 diffraction limit at 0.5 µm wavelength and 100
atmospheric turbulence. A 0.100 slope on a converging wavefront translates into a 4.3 µm
lateral aberration, compared to a 39 000 µm minimum blur circle. Obtaining 000 1 noise
on wavefront slopes demands 10−4 relative precision in matching actual measured lateral
and nominal lateral aberrations for Shack–Hartmann measurements. Similar requirements
are expected for Ronchi and Foucault methods relying on lateral or axial aberration global
measurements.
In contrast, interferometric measurements with correctors [2, 8] disentangle accuracy
and precision: accuracy relies on computing, constructing and correctly using a wavefront
compensator, whereas high precision is obtained from null interferometric testing.
Without further discussion of the relative merits of different methods, here we investigate
Shack–Hartmann testing [11, 17], noting that locating a point source as an optical fibre
directly on axis becomes practical for large mirrors because shadows from the supporting
devices cover only a small fraction of the useful pupil, thus alleviating the need for a half-
Figure 1. Shack–Hartmann testing of a 3.6 m diameter f/1.2 paraboloid using a single lens as
a collimator: computed distortion of the mirror’s image. X-axis, incident height on the liquid
mirror for the current ray emitted from a point source close to the centre of curvature of the
mirror, mm units, 100 mm ticks. Y -axis, distance from the symmetry axis of the reflected ray
incident on the best imaging plane, 0.01 mm units, 5 mm large ticks.
438 G Artzner
Figure 2. Shack–Hartmann testing of 3.6 m diameter f/1.2 paraboloid using a single lens as a
collimator: ray tracing in a meridional plane for rays emitted from a point source close to the
centre of curvature of the mirror and reflected by the mirror. X-axis, distance from the vertex
of the liquid mirror, mm units. Microlens array is located at 9230. Y -axis, distance from the
symmetry axis, mm units. For clarity, interlaced rays downstream from 9270 and 9290 mm are
not drawn.
Aspherical wavefront measurements 439
from centre to edge. Bundles 1–35, 36–45 and 46–57 have quasi-point subimages in three
different imaging planes with 20 mm spacing as indicated in figure 2.
The laboratory testing at the centre of curvature for large aspherical mirrors differs from
the sky testing of low-f -number telescopes in at least three ways:
• as indicated in table 1, the focus is no longer a point but a caustic surface;
• as indicated in figure 2, some rays impinge on the microlens arrays far from the normal,
and some bundles of rays that impinge on the microlens array are no longer parallel,
but converge or diverge according to the location of a subpupil within the pupil, i.e.
subimages in the detector plane are no longer simultaneously focused over the complete
pupil.
Here we investigate these effects by actual Shack–Hartmann diagram recordings from a
strongly aspherical wavefront.
Figure 3. Aspherical lens testing: optical schematic: D, diffuser, illuminated from a white light
source; P, 0.15 mm diameter diaphragm as a point source; L, aspherical lens; SHA, Shack–
Hartmann analyser (see figure 4).
The resulting caustic surface of the incident beam B (figure 4) is wider than the observer’s
pupil. A high-aperture collimating objective C images the lens being tested with a ratio
g = 1/14.2 and with no loss of light. A p = 0.213 mm pitch 8 mm focal length array
of square contiguous lenslets, A, is located on the pupil image and divides the incident
light beam into 35 × 47 = 1645 bundles of rays. A wide-angle lens relay optics, R, on a
6 cm × 6 cm camera body, is fitted with a red filter. This optics relay produces real images
of these real or virtual bundles as it moves along the optical axis mounted on a precision
translation system, T. The final images are recorded on photographic panchromatic roll-film,
F, at a constant enlargement ratio of G = 3.78.
Figure 5. Calibration: partial negative image of microlens array. Measured rays from lenslets
separated by five periods are located within areas indicated by square hand-written marks.
Figure 6. Calibration: real partial negative image of virtual cross section 9 mm upstream from
the microlens array. Each set of the main four black square dots within areas indicated by square
hand-written marks correspond to the corners of a measured lenslet.
Aspherical wavefront measurements 441
Figure 7. Calibration: real partial negative image of virtual cross section 9 mm upstream
from the microlens array, as seen in the measuring microscope 0.76 mm field of view. The
characteristic size of the photographic grain is around 2 µm, corresponding to 0.5 µm along the
rays.
the focus of the collimating lens C in order to obtain a plane wavefront incident on array A.
Subbundles of rays originating from the lenslets of array A are easily followed as virtual
or real images in the viewfinder V (figure 4). We recorded 11 32 mm × 28 mm enlarged
real images corresponding to 11 cross sections located, respectively, on the microlens array
(figure 5), 9, 6 mm (figures 6 and 7) and 3 mm upstream from the microlens array, and also
3, 6, 9, 11, 13, 16 and 24 mm downstream. A typical size of 3 µm for the photographic
grain on these images as seen in figure 7, allows for a 5 µm measurement precision on
negatives, corresponding to a precision of better than 2 µm along the rays. The coordinates
are measured visually on negatives under a microscope fitted with X and Y 0.5 mm pitch
measuring screws and 0.002 mm graduated thimbles.
From 9 mm upstream to 3 mm downstream the size of subbundle cross sections is
larger than the field of view of the measuring microscope. We thus measure the position
of each corner of a subbundle, as indicated by figures 5 and 6, before averaging the results
in order to locate the global position of a subbundle. Traces further down are directly
measured. Taking into account the duration of this procedure we applied it to 38 out of
1645 subbundles to located every fifth column (figures 5 and 6), and every tenth row.
2.2.2. Reduction. Each of the 11 sets of 38 pairs xik , yik (i = [1, . . . , 38], k = [1, . . . , 11])
is converted to a common Xik , Yik , Zk (Zi = −9 mm, −6 mm, . . . , 24 mm) coordinate
system using the three adjustment parameters xok , yok and 2k .
Using straightforward formulae, each of the 38 sets of 11 Xik , Yik , Zk triplets is then
adjusted to a fit
Xik = ai + bi Zk Yik = ci + di Zk .
X Y
Figure 8. Aspherical wavefront measurement: global, real, positive image seen in the viewfinder
of figure 4 when focused on a virtual cross section 3 mm upstream from the microlens array.
X Y
to the microlens array. We thus use, as indicated in table 3, a shorter baseline for each
subbundle. This, in turn, induces a lower precision for the slope determination. In order to
partly compensate for this effect we measured a larger number of subbundles.
In order to further compensate for this shorter baseline each subbundle is actually located
by averaging measurements of four rays at the corners of corresponding lenslets, as shown
in figures 10 and 11.
2.3.2. Reduction. Applying the procedure described in section 2.2.2 to these 11 sets of 57
x, y averaged coordinate measurements we obtain the results presented in table 3.
3. Discussion
Figure 10. Aspherical wavefront measurement: enlarged part of the negative obtained for the
camera setting of figure 8. For each lenslet taken into account we measure the position of the
four black dots corresponding to its corners.
Figure 11. Aspherical wavefront measurement: enlarged part of the negative obtained for the
camera setting of figure 9. For each lenslet taken into account we measure the position of the
four crosses corresponding to its corners.
446 G Artzner
the respective calibration values. This demonstrates that although the measured aspherical
wavefront deviates noticeably from a spherical wavefront over a lenslet, our procedure still
properly defines light rays.
The internal consistency for the measured slopes has been estimated by a statistical
test for each measured ray. We show in tables 2 and 3 an average slope uncertainty
for an aspherical wavefront measurement which is much larger than for calibration. This
difference results from two additional effects. Firstly, due to overlap of aberrating light rays
the baseline for aspherical wavefront measurements is shorter than the calibration baseline.
Secondly, calibration rays are straighter than aberrating rays.
The expected uncertainty ratio amounts to
3.1.2. Accuracy. The wavefront deformations apparent in figures 8 and 9 are due to the
fabrication tolerances of a moulded lens. There are no nominal values for comparison with
the measured values that allow a determination of the accuracy of measurements reported
in table 3.
In contrast, the values of table 2 allow us to discriminate precision from accuracy in the
calibration measurements. We specifically note in table 2 that the estimated slopes of the
rays, or collimation errors, are clearly larger than their estimated uncertainty. This undesired
systematic effect could, in principle, be attributed to any component of the system (figure 4)
from the collimating objective C to the film F, inclusive. As extreme 164 µrad collimation
errors correspond to a 35 nm wavefront error per 213 µm square lenslet we believe that it
is due to the substrate of the microlens array [17], which is made from window glass.
In addition to measuring the local deviations from straight line propagation of light as a
precision indicator, and the global collimation errors as an accuracy indicator, we consider
the microlens array’s image distortion in tables 2 and 3 and figure 12. The distortion in
calibration measurements is low on each axis, but presents a regular pattern apparent in
figure 12. This pattern disappears in aspherical wavefront measurements as the dispersion
increases by a factor of two, according to a less precise definition of light rays as noted
in subsection 3.1.1. We estimate that distortion effects from relay lens R in figure 4 are
negligible.
Due to the fact that some potential sources of systematic errors such as field aberration
of microlenses or shrinkage of film have been overlooked in our estimation of a 45 nm local
wavefront accuracy, we speculate only briefly in the next section about the measurement of
larger optics.
Aspherical wavefront measurements 447
Figure 12. Calibration: distortion of the microlens array’s image, residuals. Y values are
displaced by 20 µm from one line to the next. µm units along both axes. Corresponding values
across the microlens array are obtained by dividing the X and Y values measured here on the
negative by G = 3.78.
3.2. Extrapolation
3.2.1. Blue light testing. Moving from red to blue light should increase precision, at
least for reflecting systems, if the calibration errors remain stable. We repeated our
section 2.3 measurements with a blue filter without obtaining the better precision expected
from reduced diffraction effects because the increasing chromatic aberration made local
wavefront departures from spherical wavefronts larger, increasing the uncertainty in locating
the light rays.
3.2.2. Digitizing. Figure 7 and tables 2 and 3 indicate that the positions of the centroids of
details on the photographic film F in figure 4 are visually estimated to within 1/40 of their
size, or better, within a field of view comparable to the size of the digital detectors. We note
that the photographic film unit used here could be replaced by a large, 4096-level dynamic
CCD detector available from camera manufacturers in order to perform measurement of
small- or medium-size optics with a precision comparable to, or better than our experiment,
with a faster measurement time.
3.2.3. Centre-of-curvature testing for large mirrors. In contrast the measurement of large
mirrors would, according to figure 2, involve still larger unavailable digital detectors.
4. Conclusion
References
[1] Delaboudinière J P et al 1995 EIT: extreme-ultraviolet imaging telescope for the SOHO mission Solar Phys.
162 291–312
[2] Wilson R N 1990 Matching error (spherical aberration) in the Hubble Space Telescope (HST): some technical
comments Messenger 61 22–4
[3] Malacara D (ed) Optical Shop Metrology (SPIE MS18)
[4] Yatagai T 1984 Fringe scanning Ronchi test for aspherical surfaces Appl. Opt. 23 3676–9
[5] Malacarra D 1990 Analysis of the interferometric Ronchi test Appl. Opt. 29 3633
[6] Rezyapov R A 1990 Plotting the topography of an optical surface in the testing of large optical elements by
the Ronchi method Sov. J. Opt. Technol. 57 144
[7] Durand E and Bachhus J M 1992 Mesure des surfaces asphériques, sphériques et planes par déflectométrie
à détection de phase Compte-Rendus OPTO92 (Paris, 1992) pp 283–7
[8] Kühne C 1975 Production and testing of the optical elements of the first 2.2 m-telescope for MPIA Astron.
Astrophys. 41 345–53
[9] Espiard J 1991 Le polissage et le contrôle des miroirs du ‘Very Large Telescope’ de l’‘European Southern
Observatory’ Nouv. Rev. Opt. 22 61–72
[10] Artzner G, Bonnet R M, Lemaire P, Vial J C, Jouchoux A, Leibacher J, Vidal-Madjar A and Vite M 1977
The LPSP instrument on OSO8. I. Instrumentation, description of operations, laboratory calibrations and
pre-launch performance Space Sci. Instrum. 3 131–61
[11] Artzner G 1994 On the absolute calibration of Shack–Hartmann sensors and UV laboratory wavefront
measurements Pure Appl. Opt. 3 121–32
[12] Texereau J 1960 La Construction du Télescope d’Amateur 2nd edn, Société Astronomique de France, p 280
[13] Samain D 1995 High spectral resolution atalas of the balloon ultraviolet spectrum of the Sun: 1950–2000 Å
Astron. Astrophys. Suppl. 113 237–55
[14] Borra E F 1991 The case for a liquid mirror in a lunar-based telescope Astrophys. J. 373 317–21
[15] Borra E F 1992 The case for liquid mirrors in orbiting telescopes Astrophys. J. 392 375–83
[16] Wang M, Morettto G, Borra E F and Lemaı̂tre G 1994 A simple active corrector for liquid mirror telescopes
observing at large zenith angles Astron. Astrophys. 285 344–53
[17] Artzner G 1992 Microlens arrays for Shack–Hartmann wavefront sensors Opt. Eng. 31 1311–22