The star spreads across the camera
Imagine photographing a small lamp beside a car headlight at night. Covering the brightest part of the headlight helps, but scattered light still reaches the lamp. A telescope has a related problem. A star is concentrated into a small central image, yet diffraction and tiny optical errors spread some of its light into a broad halo.
A coronagraph reduces that stellar light before it reaches the science camera. In a Lyot coronagraph, a small mask blocks the bright central focus, and another aperture removes light redistributed near the edge of a reimaged pupil. A companion arrives from a slightly different direction, so much of its light follows a path around the central mask.
The remaining halo has bright knots called speckles. Slight differences in the distance light travels across a mirror change where waves reinforce each other. A knot can look remarkably like the compact image of a companion. Better angular resolution sharpens the knot as well as the planet.
Reading: Marois et al. · Soummer
Let the sky supply the motion
During pupil-tracking observations, the instrument keeps its pupil pattern in a nearly fixed orientation. Earth continues to rotate. The sky then turns across the detector, taking a companion around the stellar position. Over an observing sequence, the planet draws a short arc through a halo that changes more slowly.
Astronomers use the sequence to estimate the stellar background, subtract that estimate from each frame, then turn the residual frames back to a common sky direction. Companion light comes together at one position. Many instrumental features spread out. This procedure is angular differential imaging, usually shortened to ADI.
The motion happens because the observing coordinates turn. An apparent arc lasting half an hour is not half an hour of visible planetary orbital motion. Keeping that distinction makes the camera sequence much easier to read.
Reading: Marois et al.
The reference has seen the companion too
Suppose the companion scarcely moves between frames. The estimated background now includes a bright point in nearly the same place every time. Subtraction removes that point along with the halo. More rotation helps because each frame puts the source on different camera pixels.
This loss is called self-subtraction. It often leaves a positive core with negative shapes beside it. The negative patches mark places where the reference contained companion light; they are part of the processing response. Summing every pixel of that signed image can therefore give a very misleading measure of the source brightness.
One practical remedy is to add a faint source with known brightness to the original frames and process the enlarged sequence again. The recovered light tells us how the chosen procedure responds at that location. If the reference is learned from the data, the added source can change that reference as well. The calibration has to follow that change.
Reading: Pueyo
A real sequence from Beta Pictoris
On 10 December 2013, the Gemini Planet Imager observed the young giant planet Beta Pictoris b in the near-infrared J band. Bonnefoy and colleagues described nineteen exposures in their 2014 analysis. Each exposure was already a small stack of images at different wavelengths, so the source could be followed both around the star and through its colors.
An unrelated background star also follows the turning sky. Repeated astrometry tests whether a source belongs to the target system. Beta Pictoris b has a particular advantage here: its location can be compared with images from other instruments and dates. In the small public sample used by the companion tutorial, three exposures span about twenty degrees of field rotation. Their headers, satellite calibration spots and sky orientation can all be checked before extracting a spectrum.
The teaching camera uses its own synthetic observations so that source brightness, optical errors and noise are known. The real-data tutorial returns to the telescope files. These two paths meet at the same question: how much source light survives the procedure used to reveal it?
Reading: Bonnefoy et al. · Jason Wang and contributors
Keep the colors attached to the light
A spectrum measures brightness across wavelength. It can show broad changes caused by a cool atmosphere, but image processing also changes those brightnesses. Some methods compare wavelengths after resizing the images so that stellar speckles line up. A companion moves during that resizing, and light from one channel can influence several recovered channels.
The useful endpoint is a spectrum accompanied by its measured processing response and uncertainty. A striking point starts the investigation. Following its motion and accounting for its light lets the investigation continue.
Reading: Greco & Brandt · Bonnefoy et al.