t, 17, Bâtiment B5c, B‐4000 Liège, Belgium; ABSTRACT
The James Webb Space Telescope (formerly the Next Generation Space Telescope) will be capable of unprecedented science, owing to its large diameter, its low background, and its dedicated IR instruments. One of these instruments is a mid‐IR facility called MIRI (Mid‐IR Instrument), which is being studied through a collaboration of US and European teams. Our group at Observatoire de Meudon has developed a coronagraphic device for the MIRI camera. Taking advantage of the reduced star‐planet contrast in this spectral range, the prime goal of this coronagraph is the direct detection of extrasolar planets. Differential imaging in several well‐chosen spectral windows is foreseen. In this paper, we evaluate the manufacturing process of the coronagraphic masks and compare the intrinsic performance, as measured in the laboratory, with the scientific requirements.
Export citation and abstract BibTeX RIS
1. INTRODUCTION
The James Webb Space Telescope (JWST) is a key mission in the NASA Origins program. It has the potential to address many fundamental questions, from the origin of the universe to the formation of stars and planetary systems. By 2013, the JWST will be the largest telescope operating in space, providing a wide and continuous spectral coverage from the visible (0.6 μm) to the mid‐IR (28 μm) wavelengths, using its imaging and spectroscopic facilities. With a diameter of 6.57 m, JWST will provide an unprecedented sensitivity at all wavelengths. The observatory will be equipped with a payload of four instruments: NIRSPEC, the near‐IR spectrograph, NIRCAM, the near‐IR camera, MIRI, the Mid‐IR Instrument, and FGS, the fine guidance sensor.
Our group at Observatoire de Meudon has proposed implementing a coronagraphic capability in MIRI in order to detect and characterize extrasolar giant planets (EGPs) around nearby stars (Boccaletti et al. 2005). The mid‐IR wavelengths appear to be an attractive spectral range for at least two reasons: (1) unlike stars, giant planets emit their maximum flux in this spectral region, so the contrast is more favorable, and (2) the phase aberrations arising from optical defects are less critical at longer wavelengths. One drawback is that the angular resolution is poor, and hence a coronagraph providing a small inner working angle is required.
MIRI is actually being developed through a NASA‐led partnership with a European consortium sponsored by the European Space Agency (Wright et al. 2003). MIRI includes a spectrograph and an imager. A French group is responsible for the MIRI camera (Dubreuil et al. 2003), and Observatoire de Meudon/LESIA is responsible for studying, characterizing, and delivering the coronagraphic device (Boccaletti et al. 2005).
In this paper, we briefly review (in § 2) our previous results regarding science requirements and signal‐to‐noise ratio (S/N) calculations. From the expected performance in terms of science, we derive the main technical requirements of the manufactured masks (§ 3). Section 4 is the core of this paper and presents the experimental results we obtained with a cryogenic coronagraphic test bed to evaluate the intrinsic characteristics of phase masks manufactured using several techniques and various materials. Finally, a conclusion and an extrapolation to an actual observation are given in § 5.
2. THE MIRI CORONAGRAPH
MIRI is made up of two modules, a camera where the coronagraph is installed and an integral field unit spectrograph (Wright et al. 2003). The optical concept of the camera is fully described in Dubreuil et al. (2003), and a description of the coronagraph implementation is given in Boccaletti et al. (2005). The coronagraph masks are located at the JWST focal plane (the entrance of the MIRI camera). The Lyot stops are set in the filter wheel, and each is associated with a single filter and a single coronagraph.
There are actually four coronagraphs inside the MIRI camera: one standard Lyot mask operating at 23 μm, optimized for cold objects, such as circumstellar disks, plus three monochromatic four‐quadrant phase masks (FQPMs; Rouan et al. 2000) at 10.65, 11.40, and 15.50 μm, respectively, optimized for the detection and characterization of EGPs. These filter wavelengths are designed to derive some physical parameters of the EGPs, such as the temperature of the planet and the abundance of ammonia in the atmosphere. This information is essential for comparing evolutionary models (Burrows et al. 1997; Allard et al. 2001) with actual data. To measure the ammonia absorption of the EGPs, the spectral resolution of the filters must be larger than about 20. For such filters, the typical star‐planet contrast of EGPs with temperatures between 300 and 500 K varies from 2000 to 12,000 around an M2 V star, and from 8000 to 275,000 around a G2 V (calculated from the theoretical spectra of Allard et al. 2001).
For comparisons with the actual performance measured in the lab and presented in § 4, it is important to review the results we obtained in the past with numerical simulations (Boccaletti et al. 2005). The hypotheses are briefly recalled here. The simulation includes the segmented telescope pupil with gaps (15 mm), the wave‐front aberrations (145 nm rms), including the dynamic jitter (7 mas rms), the defocus between JWST and MIRI (1–3 mm), the pupil shear with respect to the Lyot stop (1.5%–5% of the telescope diameter), the differential pointing between the target star and the calibrator star (5 mas), the chromaticity of the FQPM for a spectral resolution of R = 20, and the radial transmission of the coronagraph.
The results of the simulation are presented here in a different way than in Boccaletti et al. (2005). We have identified three conditions—an optimistic, an average, and a pessimistic case—for several levels of pupil shear and defocus. Table 1 gives the expected attenuations (ratio of the maximum peak intensity with and without the FQPM) and contrasts (ratio of the maximum peak without the FQPM to the FQPM intensity at a given angular separation). The contrast at a separation of 3 λ/D (typically 5 AU at 10 pc) is about a few thousand on the raw coronagraphic image. In that simulation, we considered a calibration of the diffraction residuals either by using a well‐chosen reference star or by rolling the observatory at different angles. As already mentioned, the simulation includes a differential pointing error of 5 mas between the target star and the calibrator to take into account subtraction errors. In Figure 1, we plot the S/N against the angular separation for several planet temperatures and for the average case. A 300 K planet could be detected at 3 σ at a separation of 1'' in 1 hr of integration, but the left panel of Figure 1 shows that a 10 σ detection is more reliable.
 |
The following sections are devoted to the comparison between these simulation results, which assume a perfect FQPM, and the performance expected and measured with an actual manufactured FQPM.
3. DERIVED REQUIREMENTS OF MASKS
The simulation results presented in Boccaletti et al. (2005) and recalled in § 2 already take into account the limitations of the observing conditions (pupil shape, pupil shear, and defocus). However, to reach the performance expected from simulations, the manufactured FQPM must be as close to a theoretically perfect FQPM as possible. The simulation includes the chromaticity of the monochromatic FQPM but does not take into account other intrinsic limitations of the FQPM, such as the precision for manufacturing the optimized wavelength or the width of the phase transition. In the following subsections, we evaluate the effect of these defects on the performance of the FQPM.
3.1. Chromaticity
The FQPMs that are chosen for MIRI are monochromatic. The effect of monochromatic FQPMs has already been defined in Riaud et al. (2003). The attenuation of the coronagraph is equal to the total rejection rate (ratio of the total integrated intensity with and without the FQPM) and is defined by τ = (48/π2)R2. A filter resolution of R = 20 has been chosen so that the attenuation that reaches 2000 with a monochromatic FQPM is not a limiting factor, compared to telescope defects (Table 1). As mentioned above, it is also compatible with the measurement of ammonia absorption in the EGP atmosphere.
3.2. Optimized Wavelength Precision
As described in Rouan et al. (2000), a monochromatic FQPM is manufactured by a deposition or engraving of two opposite quadrants on an optical medium. The thickness of the FQPM step directly defines the optimized wavelength λ0 for which the attenuation is the best. A difference between the optimized wavelength λ0 and the working wavelength λ reduces the attenuation of the FQPM. The attenuation is equal to the total rejection rate τ, as defined in Riaud et al. (2003):
A precision of less than 2% for the optimized wavelength must be achieved to reach an attenuation larger than 1000. The manufacturing error depends on both the precision of the manufacturing FQPM step thickness and the precision with which the optical index of the FQPM is known at the operating temperature.
3.3. Precision of the FQPM Transition
To reach a complete nulling, the transition between the four quadrants must in principle be infinitely small. Departure from this ideal case decreases the attenuation capability of the real FQPM. To compare the effect of different cosmetic defects, the attenuation is calculated from simulated images. The attenuation varies as a power law of the width defects (power‐law index of −2 or −4, depending on the errors). The most critical effect is the misalignment of the FQPM axis, as shown in Figure 2. For this defect, the attenuation is equal to the total rejection rate τ, and it can be empirically estimated with the formula
where Wd is the defect width given in λF/D.
Fig. 2.— Example of an image of one of our Ge FQPMs recorded with an electron microscope.
Assuming that the FQPM shown in Figure 2 (2 μm defects) were placed in a F/D = 20 beam, the attenuation would be limited to 3400 at 10.65 μm (monochromatic case; everything else perfect otherwise). As a comparison, it would only reach 34 for a transition of 20 μm.
In MIRI, the spectral resolution of R = 20 increases the theoretical peak attenuation of FQPM to 2000. At this level, the efficiency of the FQPM will also be limited by its intrinsic defects (quadrant thickness error, transition quality). The impact of both defects, combined with chromaticity, are estimated in Table 2. A peak attenuation of 1000 can be expected, with a transition quality similar to that shown in Figure 2 (2 μm) and a precision of 1% on the optimized wavelength.
 |
4. LABORATORY TESTS OF MID‐IR FQPMs
Prior to implementing a FQPM on a space telescope, it has to be fully tested and qualified for a space environment. We developed a full facility to test, compare, and finally to choose the most reliable technique for manufacturing the FQPMs. This facility consists of classical optical testing procedures (microscope and mechanical tests), a visible spectrometer that provides an accurate measure of the thickness of the quadrant step, and a cryogenic bench to test the manufactured FQPMs in the thermal IR spectral range and at low temperature.
In order to choose the optimal FQPM for the MIRI camera, several FQPMs were manufactured with different processes, using the most favorable materials in the mid‐IR: germanium (Ge), zinc selenide (ZnSe), and chemical vapor deposition diamond (Table 3). In the following sections, we describe the approach we took to qualify the FQPMs for MIRI. We especially emphasize the tests of the FQPM manufactured with Ge, which was finally chosen for MIRI.
 |
4.1. FQPM Thickness Measurements
A dedicated visible spectroscopic bench is used to measure the thickness of the FQPM step. As described in Riaud et al. (2003), an unresolved beam is focused on the FQPM before being filtered by a Lyot stop in a pupil plane. The beam is then sent to a visible spectrometer that can record spectra between 450 and 850 nm. Both transmissive and reflective measurements of the FQPM are possible. In both cases, the ratio between spectra of the beam centered and not centered on the FQPM shows a series of minima and maxima (Fig. 3) corresponding to destructive and constructive interference. In the transmissive case, the extrema follow the equation
Fig. 3.— Low‐resolution spectroscopy of an IR FQPM manufactured by deposition of ZnSe on ZnSe (polycrystalline) and recorded by transmission. Four orders can be identified from k = 7 to 11 (11 being noisy). Gray curve: Fit of the theoretical variation of the nulling (the thickness of the curve takes into account the precision of the fit, ±10 nm). The FQPM step is 1707.6 nm, and the optimized wavelength is 4.80 μm at 12 K.
The thickness of the FQPM step e introduces the π dephasing. The optical index of the material n(λ,T) depends on the wavelength λ and temperature T. The order of the extrema k is odd for minima, and even for maxima.
Measurements by transmission in the visible are not possible for Ge but were used for all other materials. Figure 3 shows a low‐resolution spectrum of a mask built by lift‐off deposition techniques on polycrystalline ZnSe. Orders from k = 7 to 11 can be identified on the curve (for 750, 670, 610, 565, and 530 nm, respectively). Fitting the theoretical variation of the nulling versus the wavelength enables us to define the thickness of the dephasing quadrants. The fitting takes into account the variation of the theoretical nulling with classical optical aberrations and defects (defocus, chromatism, and centering error) and is calculated using an estimated index value for the deposited ZnSe materials between 550 and 840 nm (Fig. 3). The thickness of the step for this FQPM is 1707.6 ± 10 nm, which corresponds to an optimized wavelength (λ corresponding to k = 1 in eq. [1]) of 4.80 μm at 12 K.
To avoid the uncertainty in the optical index of the material, we performed a spectral analysis of the light reflected by the FQPM. Similarly, as with the transmission case, the ratio between spectra centered and not centered on the FQPM shows a series of minima and maxima (Fig. 4) corresponding to an increasing nulling order following an equation that is independent of the optical index:
where e is the thickness of the FQPM step and k is the order of the extrema (k is odd for minima and even for maxima).
Fig. 4.— Same as Fig. 3, but for a FQPM manufactured in Ge and recorded by reflection. Four orders can be identified from k = 4 to 7 (7 being noisy). The thickness of the FQPM step is estimated to be 787.6 ± 5 nm, corresponding to an optimized wavelength of 4.71 μm at 12 K.
Figure 4 shows a low‐resolution spectrum of a mask built by reactive ion etching on Ge. We can identify the orders k = 4, 5, 6, and a very noisy k = 7 (for 790, 630, 530, and 465 nm, respectively). To calculate the thickness of the FQPM step, we compare the data to a fit of the theoretical nulling with classical aberrations. The thickness of the FQPM step is found to be 787.6 ± 5 nm, corresponding to an optimized wavelength of 4.71 μm at 12 K.
The two examples shown above were related to a FQPM manufactured to be tested at 4.8 μm and at low temperature on our IR bench test (see § 4.3). The MIRI FQPMs will operate at longer wavelengths (10.65, 11.4, and 15.5 μm). The spectra obtained for these cases show more extrema, because the thickness of the FQPM step is larger (Fig. 5).
Fig. 5.— Same as Fig. 4, but with an optimized wavelength corresponding to one of the MIRI filters. Orders from 13 to 23 can be identified. The thickness is estimated to be 2654 nm, corresponding to an optimized wavelength of 15.55 μm at 12 K.
Measurements of the step thickness were done for the different FQPMs (ZnSe, Ge, and diamond), and the precision with respect to the specified thickness value was found to be better than 1% for the different manufacturers. The chemical vapor deposition (CVD) diamond and the polycrystalline ZnSe give the best reproducibility (better than 0.5%). The measurement of both the transmissive and reflective spectra for all the materials but Ge is a new, indirect way to measure the optical index.
4.2. Quality of the FQPM Transition
The lift‐off techniques are well suited for FQPMs operating at visible wavelengths (Riaud et al. 2003) but were found to be inappropriate for MIRI. Transitions as large as 20 μm were measured for the FQPMs manufactured with the lift‐off deposition techniques. For the other techniques, the transition width proved to be between a few microns and a few tenths of a micron in some cases. An example of typical defects of etched FQPMs is shown in Figure 2. As described in § 3.3, such a transition implies an attenuation of 3400 at 10.65 μm, which is thus fully compatible with the expected attenuation of MIRI.
4.3. FQPM Performance in the Mid‐IR
4.3.1. Description of the IR Bench
We developed an IR cryogenic test bench (Fig. 6) to verify that the FQPM was working as expected in theory in the mid‐IR and at a temperature close to the nominal MIRI temperature (9–15 K). This facility was developed at Observatoire de Meudon and consists of a roughly 1.5 m long bench located in a clean‐room.
Fig. 6.— Optical layout of the mid‐IR bench, which can test the FQPM at low temperature.
The bench source is a tungsten lamp that feeds a multimode fiber transmitting wavelengths up to 5 μm. A pinhole (a 30 × 30 μm square) is added in front of the fiber output to keep the source sufficiently small compared to the angular resolution (≈λ/5D). A full optical train of ZnSe lenses carries the beam from this unresolved pinhole to the detector, going through the FQPM, which is mounted in a dewar that has been cooled down to 12 K using a cryocooler. A Lyot stop corresponding to 83.3% of the entrance pupil diameter plays the role of the cold stop in the detector cryostat, which is cooled down to 77 K.
The IR detector covers the 1 to 5 μm spectral range. In order to test the FQPM at different wavelengths, we inserted a tunable bandpass filter between the tungsten lamp and the input of the optical fiber. In fact, the central wavelength λ of this filter varies from 4.84 to 4.66 μm when tilting the filter from 0° to 45°, following the equation
where θ is the angle of the tilt. The width of this filter is 0.5 μm, and thus the spectral resolution is R = 9.5. Another filter is placed inside the detector cryostat to increase the blocking capability of the filter to 104 at short wavelengths (below 4 μm).
4.3.2. Laboratory Results
The exposure time is limited to 30 ms by the thermal background. Subtraction of a background image is thus mandatory, and the calibration is performed with the source occulted. Since most of the FQPMs were uncoated, they were tilted by about 20° to move the ghosts to a distance where they could be numerically subtracted from the images (Fig. 7).
In Figure 7, the rings of the diffraction pattern clearly appear on the image that is not centered on the FQPM (left), but these rings do not appear on the FQPM‐centered image (right), for two reasons. First, the FQPM decreases the intensity of the brightest pixel by a factor 400 so that the FQPM residual peak is hardly above the background noise (detected at only 10 σ in a single exposure of 25 ms). Second, the effect of the resolved source increases the size of the residual image core and blurs the rings.
For the images shown in Figure 7, the total rejection rate τ was measured by integrating the intensity up to 20λ/D, both with and without a FQPM (excluding the ghost images). This is reported in Table 4, as are the measured attenuation and contrasts, which can be compared to the values expected for MIRI (Table 1).
 |
4.3.3. Index of the Materials at Low Temperature and in the Thermal IR
As a by‐product of the cryogenic nulling tests, we determined the optical index of materials at low temperature, which is not well known. Using the tunable IR filter, we measured the attenuation on the peak for the FQPM at different wavelengths. The FQPM performs at its best when the IR filter is centered on its optimized wavelength (i.e., the wavelength that solves eq. [3] for k = 1 and for the temperature of the measurements). In Figure 8, a theoretical attenuation is fitted to the data to estimate the optimized wavelength of the FQPM for this temperature. For this case (i.e., a temperature of 54.5 K and a FQPM manufactured with Ge), the optimized wavelength of the FQPM is 4.735 μm. Using the thickness measured at a visible wavelength and this optimized wavelength found with the IR bench, we can determine the optical index for the temperature at which the IR measurements were taken. The variation of the effective FQPM thickness when tilted is taken into account in the calculation of the optical index. We compared our measurements for different temperatures to the theoretical variation of the optical index of the Ge (Fig. 9).
Fig. 8.— Measurements of peak attenuation vs. wavelength for a FQPM made of Ge at a temperature of 54.5 K. The curve is a fitted theoretical attenuation, giving an optimized wavelength equal to 4.735 μm.
Fig. 9.— Optical index for Ge as recorded with our experiment, compared to the fourth‐order polynomial function given in Hoffman & Wolfe (1991) but rescaled for our working wavelength.
The IR optical index for Ge is well known at temperatures higher than 150 K, using a Sellmeier equation that depends linearly on temperature (Barnes & Piltch 1978). However, measurements at 10.6 μm down to 20 K (Hoffman & Wolfe 1991) showed that the variation at low temperature is no longer linear. Hoffman & Wolfe (1991) defined a fourth‐order polynomial function to fit the temperature variation of their measurements (only done at 10.6 μm). We assumed that the optical index of the Ge shows the same temperature variation at 4.8 μm as it does at 10.6 μm (fourth‐order polynomial function) but is shifted vertically so that the index at room temperature matches the index calculated with the Sellmeier equation. The behavior of the optical index of the Ge at low temperature agrees with such a function (Fig. 9). Thus, we can rely on this optical index function to define the exact thickness of the FQPM step that will operate between 9 and 15 K in MIRI.
4.3.4. Comparison with Expected Performance
We calculated the mean radial profiles of both images shown in Figure 7 (numerically suppressing the ghosts) and report them in Figure 10 (solid lines). The increase in the size of the FQPM diffraction profile comes from the fact that the size of the source is resolved. The raw attenuation reached on the peak is 379, while a contrast of 5 × 104 is reached at 3λ/D (Table 4).
Fig. 10.— Radial profiles of the results with a FQPM (lower curves) and without a FQPM (upper curves). Dotted lines show the radial profiles calculated using the numerical code to simulate the actual bench, and the solid lines show the laboratory results.
In order to compare the IR cryogenic nulling obtained on the bench with the expected performance, we developed a numerical code that takes into account the different limitations of the bench. The parameters for the simulations are (1) the size of the source (30 μm wide square), (2) the different wavelengths, which are simulated by adding monochromatic images over the 4.5–5.2 μm range and using the transmission of the different bench elements (quantum efficiency, IR filter transmission, antireflective coatings, etc.), and (3) the sampling on the detector (2.55 pixels per element of resolution λ/D at 4.85 μm).
There is excellent agreement between recorded and simulated data (Fig. 10) at any angular distance. This confirms that the FQPM operates as simulated and behaves as expected at low temperature and in the thermal IR range. The peak attenuation, the contrast at a few angular distances, and the total rejection rate are also calculated with the simulation tool, and all agree with recorded data within 10% (Table 4).
4.4. Assessment of the Masks
Using all the tests done on the four different FQPM materials and techniques, we rated the masks depending on the precision of the manufactured thickness, the quality of the surface, and transition and technology readiness (Table 3). Polycrystalline ZnSe was discarded because of the poor quality of the transition. Both the monocrystalline ZnSe and the CVD diamond were discarded, mostly because of a poor knowledge of the manufacturing techniques and material properties compared to the project deadlines. However, diamond is a very promising material for future experiments. Ge was ultimately selected, since it fulfills the requirements and is a well‐known material that is already being used in space environments.
5. CONCLUSION
We have demonstrated that the behavior of the four‐quadrant phase mask in the mid‐IR and at a very low temperature (12 K) corresponds to the expected performance. The laboratory equipment we developed to compare several manufacturing processes for the MIRI FQPM enabled us to measure the optical index of our FQPM substrates at temperatures lower than 20 K, where only one measurement was available. The effective results of the FQPM in terms of contrast and attenuation (Table 4) are much better than what is specified for MIRI (Table 1), although the experimental results were obtained with a spectral resolution (R = 9.5) that limits the performance of the monochromatic FQPM to 433. In MIRI, the spectral resolution of R = 20 limits the theoretical peak attenuation of monochromatic FQPMs to 2000. However, as shown in Table 2, manufacturing defects can quickly decrease the attenuation. With a transition width of 1 μm and a precision of 0.5% on the optimized wavelength, the attenuation of a monochromatic FQPM will be limited to 1500. Aberrations and alignment errors in the telescope itself will add to these intrinsic limitations of the FQPM performance. In fact, the limitations from the James Webb Space Telescope (defocus, pupil shearing, jitter, aberrations, and pupil geometry) largely dominate the intrinsic defects of the manufactured FQPM. Thus, we can hope to reach the sensitivity estimated with simulations that take into account the limitations of the JWST. In that case, an extrasolar giant planet orbiting at 10 AU around its parent star should be detectable down to 400 K in 1 hr.