Skip to main content

Ultimate Angular Resolution Astrophysics with kernel-phase and full-aperture interferometry

Periodic Reporting for period 2 - KERNEL (Ultimate Angular Resolution Astrophysics with kernel-phase and full-aperture interferometry)

Reporting period: 2018-04-01 to 2019-09-30

The angular resolution of a telescope, that is its capability to distinguish
the fine details of an astrophysical scene, is fundamentally limited by the
phenomenon of diffraction. This constraint is one of the two major reasons why
astronomers keep wishing for larger and larger telescopes (the other being
sensitivity) or that of using multiple telescopes in concert using a special
recombining technique called optical homodyne interferometry.

While the phenomenon of diffraction dominates the general allure and properties
of high-angular resolution astronomical observations, it is by no means a hard
limit. The aforementionned technique of optical interferometry is regularly
able to produce high-quality measurements that lead to insights into angular
sizes that are smaller than this diffraction limit: a capability sometimes
refered to as super-resolution.

A special case of diffraction-dominated astronomical observation of emblematic
importance concerns the direct imaging and spectral characterization of
extrasolar planetary systems. Since the discovery of the first extrasolar
planet in 1995 using the indirect method of radial velocity, the direct imaging
of exoplanetary systems has been listed amongst the top three ambitions of
modern astrophysics. Each discovery of a new extrasolar planet brings us closer
to answering the fundamental question of the place of humanity in the grand
scheme of the Universe.

By going beyond indirect detection means, direct imaging of extrasolar planet
light will enable a major jump in our ability to characterize them: it will
open the doors of spectroscopy. The angular resolution of the current 8 to
10-meter class of telescopes observing in the near-infrared is theoretically
sufficient to be able to directly detect extrasolar planets orbiting the
majority of nearby stars. Yet only a handful of such systems has been directly
observed to this day, despite the deployment of specialized instruments used
over extended dedicated observing campains. In this scenario, it is the
diffraction of the light of the star hosting these planets that dominates by
several orders of magnitude, the scene and literally blinds the
instruments. Even if observing conditions were perfect,

The ability to reliably isolate the light of extrasolar planets remains an
operational challenge. Two sources of noise do dominate the high-contrast
imaging game: the photon noise of the star and the phase noise induced by the
atmosphere. Conceptual solutions of coronagraphy have existed for over a
decade. The ideal coronagraph suppresses the static diffractive pattern of a
bright source. Residual aberrations induce second order starlight
leakage. Because this leakage follows a quadratic law, it is difficult to
diagnose and remedy in real time. KERNEL tackes the problem from the opposite
view: from interferometry, we know it is possible to form observables that are
robust to phase errors but are strongly affected by photon noise. Direct
transposition of kernel-phase to coronagraphy is not possible but one can
envision modifying high-contrast imaging solutions to turn this idea into reality.

Can we apply or translate some of the tools and ideas developed in the context
of optical interferometry so as to apply them when interpreting astrophysical
images dominated by diffraction effects? This is the simple but powerful idea
at the core of the KERNEL project. By looking at the formation of high-angular
resolution images from an interferometric standpoint, KERNEL enables new ways
of overcoming the traditionnally accepted limit of diffraction. This has
applications ranging from the processing of existing archival data to the
definition of completely new types of high-performance instruments.
The entire project relies on a custom data analysis pipeline that implements
the recipes of kernel- and eigen-phase introduced in earlier papers by the
project PI. The first major task outlined in the project description of the
action (DoA) was therefore to produce an upgraded version of the pipeline. The
Python code was migrated to a public Github repository named XARA (for eXtreme
Angular Resolution Astronomy, XARA now
offers a series of modules designed to facilitate the implementation of planned
upgrades resulting from the findings by the different KERNEL staff members in
the context of the different work-packages.

To facilitate the use of XARA as a generic, distributed tool, a data exchange
format for kernel-phase and kernel-phase metadata was devised, using a standard
multi-extension FITS file format. Software documentation and examples of uses
are available on the KERNEL PI webpage (
The work carried out in the context of KERNEL also takes advantage of a
simulation package named XAOSIM also available on a Github public repository
( Open-source, both XARA and XAOSIM which
were initially written in Python 2.7 were updated and are now fully Python 3
compliant while still Python 2 compatible.

XARA was used as the major analysis tool by four peer reviewed publications
originating either from our group and our direct collaborators including one
technical (wavefront metrology). Each of these publication stretches the reach
of the KERNEL framework, providing the ability to sense new wavefront modes
(N'Diaye et al, 2018), to work around saturation effects (Laugier et al, 2019),
to better calibrate noisy data (Kammerer et al, 2019), and to control the
statistics involved in computing contrast detection limits (Ceau et al, to be
accepted). Several other papers are already in preparation.

One of the most anticipated prospects of the project is the development of
adaptations of the kernel method suited to the high-contrast imaging
scenario. In order to reach contrast detection limits that would make
extrasolar planets directly accessible, it is essential to simultaneously
address the photon noise problem (traditionnally taken care of by a
coronagraph), and the phase noise (brought by the use of kernels). This
combined use is however no straightforward task.

Taking advantage of a collaboration with the SCExAO high-contrast imaging
instrument of the Subaru Telescope, we designed, fabricated and installed a
pupil apodizing mask with features that make it compatible with the kernel
analysis (one example of application that made it possible to refine the XARA
pipeline). We also deployed our fast low-readout IR detector and installed it
at the side port focus of SCExAO. We were able to use this new observing mode
on-sky over several engineering nights and are currently working on writing a
paper on what constitutes the first demonstration of kernel-phase with a

Independantly from this attempt, we were also able to outline a very different
way of approaching the high-contrast imaging problem, taking advantage of
long-baseline interferometry. We were able to show that a nuller (the
interferometric equivalent of coronagraph), can be modified to enable the
formation of kernels. The concept aptly called a kernel-nuller (Martinache &
Ireland, 2018) is be able to account for small cophasing errors in the
array. One such device at the focus of the the European VLT-Interferometer
would be able to directly detect a large number of planets.

The software developed in the context of the KERNEL project is tightly
integrated to the control of the SCExAO instrument. XARA is at the heart of the
focal plane wavefront control application developed described by N'Diaye et al
(2018) and XAOSIM provides direct access to the shared memory data structure
used by SCExAO (see for instance Lozi et al, 2019).
"With the increase in capabilities of the KERNEL pipeline, the publication phase
of astrophysical results based on the kernel-phase analysis of images has
begun. I expect more papers will follow and hope that the use of the technique
will eventually replace sparse aperture masking interferometry as the first
choice when super-resolved observations are required, whenever relevant. The
work that we are publishing these days (not accepted yet) on simulated JWST
data should contribute to this.

In order to get to this point, there is one question that needs to be clearly
answered, which concerns the so-called systematic errors reported in the
litterature. The framework theoretically places observables in a mathematical
space called ""the kernel"" that is protected against small instrumental
perturbations. As observers become more and more ambitious and attempt to reach
higher and higher performance using this technique, they become confronted to
imperfections in the kernel, that lets a small amount of perturbation go
through the filter... with each author attempting to account for these
systematic errors in a different way. Work done thus far in the context of
KERNEL has suggested several ways of improving on this limit, and reduce the
contribution of these systematic errors, using a more accurate representation
of the instruments. With this bread-and-butter type of work behind us, we will
be able to confidently tackle very ambitious observations that will make the
technique and the KERNEL project really shine.

With the PhD project that will start in the Fall 2019, and that will be further
cement the collaboration between KERNEL and the SCExAO instrument team
(co-supervision with SCExAO PI Olivier Guyon), we will take the wavefront
sensing approach that was initially applied to the sensing and control of
low-order aberrations and low-wind effect modes and turn it into a
fully-fledged focal-plane based adaptive optics control loop, with the
potential to solve in a definite way, the problem of non-common path
aberrations that affects all high-contrast imaging instruments currently in
operation in the world and limit their scientific potential. If successful,
this approach has the potential to really revolutionize the quest for the
direct imaging of extrasolar planets with a single telescope.

In addition to the pipeline, the other major project deliverable featured in
the DoA is a test-bench relying on state of the art components such as the fast
readout IR camera and a segmented aperture deformable mirror that can together
serve a wide variety of purposes. The availability of a visitor IR port of the
SCExAO instrument at the Subaru Telescope had been judged too good to pass
early into the project and the major hardware of KERNEL was temporarily added
to SCExAO (while leaving us exclusive access to the camera itself). With the
hardware back in the lab at the time of this writing, we are now looking at
finalizing the integration of the KERNEL test-bench and provide the environment
that will make it possible to achieve two important goals: (1) test a
polychromatic wavefront sensing / fringe tracking concept with a wide capture
range and (2) integrate the kernel-nulling device invented during the first
part of the project and whose fabrication we are currently investigating.

If successful, this lab demonstration of the announced properties of the
kernel-nulling device, and a thorough evaluation of its interferometric imaging
potential will be the stepping stone to a fully-fledged high-contrast
spectro-imaging instrument working in the mid-infrared for VLTI: the best
legacy KERNEL could hope for."