Tuesday, February 13, 2018

Pair-based Analytical model for Segmented Telescopes Imaging from Space (PASTIS): Introduction to the model

Earth-like planets imaging and spectroscopy requires the ability of imaging objects 10^10 times fainter than their stars at very close angular separations. The solution found to reach such goals is to combine a coronagraph and wavefront sensing and wavefront control tools.
However, the telescope and the coronagraph design phases have to address stability studies, particularly in presence of segmented telescopes such as LUVOIR, to understand the impact of segment-level errors on coronagraphic PSF quality.
The PASTIS model provides a fast way to compute the contrast in the dark hole generated by a coronagraph, and affected by local errors on the segments. In this post, we introduce our very first results using this model and compare them with the outputs of a traditional end-to-end simulation.

For this study, we use a 36-segment pupil, combined with an apodized Lyot coronagraph (APLC) providing a contrast better than 10^10 between 4 and 9 lambda/D. An example of local aberrations on this pupil are shown in Fig. 1.

Fig. 1. Local aberrations applied on the pupil chosen for the study.

First of all, PASTIS provides the contrast in the dark hole, as a function of the aberrations on the different segments. In the following figure, we compare the mean contrasts in the dark hole computed from the images of the end-to-end simulation and from PASTIS.

Fig. 2. Plot of the output contrasts computed by the end-to-end simulation and from the matrix-based analytical model for piston aberrations from 1pm to 10nm rms on the segments. For each rms value, we select 250 random phases and compute the mean, minimum, and maximum contrasts over the 250 output contrasts.

On these curves, there is a 3% error on the computation of contrast between the PASTIS computation (continuous lines) and the end-to-end simulation computation (dotted lines). However, the PASTIS values have been 10^7 times faster to compute than the end-to-end simulation values.

Secondly, PASTIS can also provide a good estimation of the morphology of the speckle in the dark hole, depending on the aberrations applied on the different segments.

Fig. 3. Comparison of PSFs when local aberrations are applied on the pupil segments. First column of each table: type of Zernike polynomial being applied. Second column: phase applied on the pupil. Third column: deteriorated PSFs (in the dark hole only, which explains the donut shape) issued from the end-to-end simulation. Fourth column: deteriorated PSFs issued from PASTIS.

To conclude, PASTIS is adaptable to any segmented pupils, such as the ELTs, TMT, JWST, LUVOIR, or even to non-hexagonal-segment pupils such as the GMT. 

Tuesday, February 6, 2018

HiCAT: Iris AO flatmap calibration

After the successful installation of the second Boston deformable mirror (DM) on HiCAT two weeks ago and the first dark zone results with the carbon nano-tube (CNT) apodizer last week (article coming soon), the next step for the testbed is the implementation of the Iris AO segmented mirror (see fig. 1). As opposed to the Boston DMs, this mirror consists of 37 individual segments that can be controlled in piston, tip, and tilt (PTT) by using electrostatic actuation on each individual segment.

Figure 1: [left] Schematic diagram of a 700um diameter mirror segment and [right] close-up photograph of a PTT111 DM composed of 37 hexagonal piston/tip/tilt (PTT) segments with 3.5 mm inscribed aperture. (Source: Iris AO, Inc.)

For precise steering of the mirror segments, a calibration of the Iris AO is needed during which the flattest position of the mirror is defined. This was done by using a 4D Fizeau Interferometer and the mirror was controlled with the Iris AO GUI. Using the central segment as a default for piston, the goal was to set the segments up without any residual tip and tilt, and piston values are as similar as possible and make the mirror surface as smooth as possible.

A start was done by comparing three segments to each other and iteratively calibrate all segments of the inner ring. Once that was done, a mask showing only the inner ring segments was used to calibrate the second and third rings. A final iteration of fine tuning was done by comparing all the 37 segments to each other to further reduce the rms.

Figure 2: Calibration flatmap of the 37 segment Iris AO segmented mirror. In its flattest position, the mirror surface has an sms of 9nm and 66nm peak to valley (PV). At this precision, the individual segment flatness can be resolved.

After a couple of hours of PTT adjustment of the Iris AO, we got down to 9nm sms over the entire mirror surface and around 66nm PV. At this precision, we can resolve the individual segments' surface aberrations - the flattest segments, showing the most uniform color in the surface diagram (see Fig. 2) are the ones in the top right area of the mirror. There are two segments in the middle that have a distinct focus on them and a lot of segments show some astigmatism. The segment at the very bottom seems to have a more complex aberration on it, but this will not significantly influence our use of this Iris AO on HiCAT. We hope to install it on the testbed this week.

Wednesday, January 17, 2018

Segmented Coronagraph Design and Analysis (SCDA): Apodized Pupil Lyot Coronagraphs (APLC) designs for future segmented telescopes

In support of the community’s assessment of the scientific capability of a Large UV/Optical/IR Surveyor (LUVOIR) mission, the Exoplanet Exploration Program (ExEP) has launched a multi-team technical study: Segmented Coronagraph Design and Analysis (SCDA).The goal of this study is to develop viable coronagraph instrument concepts for a LUVOIR-type mission.The apodized pupil Lyot coronagraph (APLC) is one of several coronagraph design families that the SCDA study is assessing.This poster reports on the APLC team’s progress. 

Parameter Space Survey with One Dimension Approximation

To fully understand the joint optimization of the aperture and Lyot Stop, we have completed a massive survey of 65,100 axi-symmetrical designs, varying the central obscuration ratio and lyot stop inner and outer diameter parameters for each aperture geometry. APLC can handle a central obstruction up to about 25% before severely degrading the throughput.

APLCs for proposed LUVOIR architecture

Eight possible apertures for LUVOIR, along with their corresponding APLC anodizers. The 14.6-meter monolithic primary comprising Aperture 8 is used strictly as a performance reference point.

PSF throughput values are unites and normalized to the design for the Aperture 8: an obscured monolithic primary of diameter D = 12.7 meters, with FPM occulter radius 3.4 lambda/D, optimized for 1e10 contrast with a 10% bandpass and outer working angle 10 lambda/D, with PSF throughput 22%. For each aperture the FPM occulter radius (a reliable proxy for inner working angle) was optimized for high throughput.
The throughput is computed as the ratio of the energy inside the unoccupied PSF core within radius 0.7 lambda/D to the total incident energy on the obscured primary mirror.

Thursday, December 7, 2017

Makidon Lab team

Here is our almost complete team with, from left to right: (back) Iva Laginja, Johan Mazoyer, Laurent Pueyo, Tom Comeau, Christopher Moriarty, (front) Anand Sivaramakrishnan, Marshall Perrin, Keira Brooks, RĂ©mi Soummer, Lucie Leboulleux, Peter Petrone, and Greg Brady.

Wednesday, November 29, 2017

HiCAT: Speckle Nulling

Once we developed software interfaces to each of the hardware components, we started writing Python scripts to control the testbed safely for long periods of time.  Before diving right into a complicated algorithm for creating a dark zone, we decided to implement speckle nulling because of its simplicity. In a short development time, it would serve as an end to end test to confirm that the testbed is aligned and our software infrastructure is working. 
Our version of speckle nulling is implemented in Python, along with Wolfram scripts for sensing and control.  Prior to running, two other steps need to be completed to derive the plate scale and the control normalization.  Both of those steps are scripted to collect the data, and call a Wolfram script to generated the necessary output artifacts. 
Speckle nulling was first tested without an apodizer, and we relied on the DM’s satellite spots for image re-centering. 
PSF being corrected thanks to multiple Speckle Nulling iterations.
Results obtained on the HiCAT testbed, with a classical Lyot coronagraph.

Now that we have the WFIRST apodizer in, we no longer have satellite spots, and had to get creative.  We decided create sine wave command on the DM to inject two speckles as far away from the PSF as possible (17 lambda/d), and we use those speckles for image re-centering.  The injected speckles fall outside the dark zone, so speckle nulling never finds them.
Our speckle nulling algorithm is simple yet effective.  It identifies the brightest speckle in the dark zone, and senses the number of cycles, initial amplitude and angle. Then we collect a data set over a range of phases, and fit a sine wave to the speckle intensity to find the best phase value.  Using the new phase, we collect data over a range of amplitudes and fit a parabola to improve our amplitude value.  Once we have the new phase and amplitude, a sine wave is generated as a DM command to kill the speckle, and the whole process repeats. Each iteration takes about 2 minutes, and a nice dark zone can be obtained with 150-200 iterations. 

Speckle Nulling algorithm Flow Chart

Each iteration generates a three-frame diagram to show the speckle that was chosen and the result after the kill.
Iteration 5 of the Speckle Nulling process:
Left: Speckle chosen to be killed in the dark hole
Center: Same speckle has been killed, another is chosen to be killed
Right: Same, including the entire image 

Each iteration of speckle sensing and control generates plots to show the fits for phase and amplitude.

Photometry of the chosen speckle as a function of the phase (top)
and the amplitude (bottom) of the sine phase added on the DM

Tuesday, November 21, 2017

JOST: assembled testbed and its components

JOST, the James Webb Space Telescope (JWST) Optical Simulation Testbed, was designed to test wavefront sensing and control algorithms on segmented apertures. Its optical properties model the key optical aspects of the JWST, as it provides JWST-like optical quality over a field equivalent to a NIRCam module. It successfully produces images extremely similar to NIRCam images from cryotesting in terms of the point spread function (PSF) morphology and sampling relative to the diffraction limit. The components of JOST are detailed below.

  • Fiber launch & off-axis parabola (OAP): The OAP collimates the beam from a fiber launch.
  • Beam Capture Mirror: Capturesthe beam from the OAP and sends it onto the steering mirror.
  • Steering Mirror: Controls the field position of the laser beam.
  • Pupil aperture: JWST-like pupil, conjugated with the deformable mirror (DM) with struts and a hexagonal central obscuration.
  • Cooke Triplet of L1, L2, L3: The three custom lenses build a refractive analogue to the JWST three mirror anastigmat (TMA). L2 is motorized for movement in x, y, z translation and tip and tilt.
  • Fold mirror: Folds the beam to fit the testbed on the optical table.
  • Iris AO segmented DM: 37 hexagonal segments that can be individually controlled in piston, tip, and tilt. The pupil limits the beam to 19 active segments and the segment gaps are 10-12 um (~0.1% of the 1.4 mm segment size), which makes them to scale with the actual JWST geometry.
  • CCD camera: A CCD camera is mounted on a translation stage to allow for the acquisition of focused and defocused images.

Thursday, November 16, 2017

HiCAT: WFIRST Apodizer Test

We have developed a process to reliably swap an apodizer in and out of HiCAT, because there are several models we plan to test. The process uses Michelson interferometers and Theodolites to ensure the new apodizer is aligned as closely as possible to the previous. 

The first apodizer we used is designed for WFIRST, and while the apodizer is not designed for HiCAT, we were still able to test our swap process. The PSF of the apodized image is so different, there were several software updates required. Specifically, we no longer have our trusty satellite spots to use for image centering. Nevertheless we are close to finishing the software updates, and will soon be able to run speckle nulling to create a dark zone with the apodizer.

In the mean time, check out this beautiful coronographic image using the WFirst apodizer. There is also a sine wave placed on the DM to create speckles on the x-axis.

Left: PSF using the WFIRST apodizer (log scale).
Right: WFIRST apodizer, currently set up on HiCAT.