Discovering Exoplanets with the James Webb Space Telescope (JWST)

By Klaus Subbotina Stephenson ('25 Astrophysics), advised by Professor Andrew Skemer

What are exoplanets? Why are they cool?

Why is imaging planets difficult?

Imaging exoplanets in the era of the James Webb Space Telescope (JWST)

Koret project: Overview and goals

Koret project: Exoplanet vetting 101

1. Is it shaped like a planet?

2. Is it colored like a planet?

3. Does it move like a planet?

4. Position angle calculations are important!

Koret project: conclusion and looking forward

Acknowledgments

What are exoplanets? Why are they cool?

'Exoplanets,' also called 'extrasolar planets,' are planets outside of our solar system orbiting stars other than our sun. There are all sorts of exoplanets out there, including ones very unlike Earth. Below are a few examples of real exoplanets astronomers have discovered.

PSO J3118.5-22

A 'free-floating' planet that does not orbit any star. Astronomers believe free-floating planets originate from failed stars, or are exoplanets that 'went rouge,' and were ejected out of orbit around their original stars.

Epsilon Indi A b

This planet is ~3x the size of Jupiter, takes 45 years to complete an orbit, and is one of the coldest exoplanets ever imaged directly (~40 degrees Fahrenheit)!

2M1207 B

The first-ever directly imaged exoplanet! This planet does not orbit around its star but instead is apart of a binary system, meaning both the exoplanet and its star orbit each other! Also, 2M1207 B's gravity is so strong that it supports temperatures of ~2,000 degrees Fahrenheit.

Since the first exoplanet was discovered in 1995, we've discovered more than 5,500 exoplanets (as of early 2025). The field of exoplanetary science seeks to answer many questions, such as what sizes of exoplanets exist, what material exoplanets are made out of, and ultimately, do any of them have life?

Across the world, hundreds of exoplanet scientists ask questions and seek to answer them by discovering and characterizing new exoplanets. There are multiple methods astronomers use for uncovering new exoplanets, each with their own advantages and disadvantages. Examples include:

The transit method, which looks for periodic differences in brightness of nearby stars to indicate exoplanets obscuring/eclipsing their host stars as they periodically orbit in front of them.
The radial velocity method, which looks for periodic 'wobbling' of the spectra (colors) emitted by stars to denote the gravitational pull of a planet lightly tugging on its host star.

But unlike these methods, there is one exoplanet discovery method that actually allows us to see planets head-on, called direct imaging.

Direct imaging doesn't just provide the aesthetic pleasure of seeing exoplanets visually, it also allows us to study exoplanet atmospheres. Exoplanet atmospheres are extremely important to understand as astronomers work to seek habitable planets. Observing habitable atmospheres will play a key role in the search for alien life (for example, observing chemicals like Methane in Earth's atmosphere could indicate the existence of biomatter).

However, directly imaging exoplanets is an extremely difficult task and a technology that's largely still in development. Specifically, it is hard to directly image small, rocky planets close to their host stars. This is bad news for us astronomers since our only example of a habitable world is the small, rocky world called Earth; to image habitable exoplanets in the future, we need to work on imaging small rocky exoplanets in the present.

This page and project is a small part within the bigger picture of effort towards this goal of directly imaging rocky planets close to their host stars.

Why is imaging planets difficult?

Direct imaging is difficult for multiple reasons. Below are the key difficulties in directly imaging exoplanets, as well as the real methods we use to deal with these challenges.

Problem: Stars are too bright

If you were to directly-image our solar system from afar, the Sun is so bright that its light would cover up any of the light emitted by planets (particularly close in planets such as Mercury, Venus, Earth, and Mars).

This is for two reasons:

Stars are nuclear powerhouses that generate light at a surface level, whereas planets primarily produce light from tectonic activity deep within a planet.
Stars are much bigger than planets, meaning they produce significantly more light than planets, which have much smaller surface areas.

Solution 1: Cover up the star

To fix this, telescopes have a physical component called a 'coronagraph,' which is placed overtop the center of a star before taking a picture. Using coronagraphs allows our cameras be more sensitive to dimmer things (ie. exoplanets). We do this every day when we use our hands to cover up the sun to see our surroundings better. On the left is an example of a coronagraph being used to cover up our Sun. With most of the starlight removed, we can directly image much dimmer objects, such as comet Lovejoy (the small moving object near the bottom of the image).

Solution 2: Take images in the infrared

Because planets produce light from their internal activities, their emitted light is thermal, just like how our bodies generate thermal light because of internal processes like digestion and blood flow. This means that if you were to stare at an exoplanet with heat-googles, i.e., an infrared wavelength camera, it would shine brightly in contrast to the cold background of space. Additionally, stars do not shine as brightly at redder wavelengths as they do at bluer (optical) wavelengths, meaning not only does taking pictures of stars in the infrared help us uncover bright thermal light from exoplanets that may be orbiting nearby, but we also combat the blinding brightness of stars by avoiding imaging them at the wavelengths where they are brightest.

Problem: Earth's atmosphere is noisy

Stars twinkle in the night sky not because the stars themselves are rapidly changing, but instead, stars twinkle because of the turbulence in Earth's atmosphere. Stars themselves are relatively static in nature over long timescales and are considered 'point sources' of light. As the light from a star's surface passes through space, it's relatively undisturbed. But once starlight hits the Earth's atmosphere, it is met with a plethora of behavior like wind shearing, clouds, and different layers of atmospheric material, which act on extremely short timescales and persistently change the direction starlight takes to travel to your eye.

The same physics happens when astronomers image exoplanets with telescopes, except the stars we hunt for exoplanets around are much further and dimmer than those you can see with the naked eye on the night sky; meaning atmospheric turbulence can be detrimental to finding exoplanets.

Solution 1: Correct for the atmosphere (adaptive optics)

Luckily for us astronomers, Adaptive Optics (AO) was invented (in part by UCSC affiliated researchers) for correcting Earth's atmospheric effects. By using a series of lenses, mirrors, and complicated control algorithyms, AO works to adapt to a snapshot turbulence profile by detecting how starlight is being affected, analyzing how to correct/undo the detrimental effects of the atmsophere, and apply the corrections to starlight as it hits a telescope. What this looks like zoomed out is an 'unblurring' of atmospheric turbulence. The image on the right shows what pictures taken by a telescope look like before and after using AO. Before, the image looked like a series of blobs, with no discernable objects visible. After AO is applied, you can make out each individual star in the field of view.

AO is crucial since the spacing between a star and its exoplanet(s) is extremely small (much smaller than the space between objects within the image on the right); meaning the high resolution AO provides is necessary for discovering new exoplanets.

Solution 2: Take pictures from space!

Don't want to deal with Earth's atmosphere? Launch your telescope to space! That way, the only distortions that will affect your images are the small thermal changes space telescopes experience when they are first launched into space. This is, in large part, is why images taken by the Hubble space telescope are so much crisper than images taken by telescopes on Earth, even if those ground-based telescopes have cameras and technology much newer than Hubble's. Here is an example.

Imaging exoplanets in the era of the James Webb Space Telescope (JWST)

The James Webb Space Telescope (JWST), which launched on Christmas morning in 2021, is the largest space telescope ever made with a 6.6-meter diameter segmented mirror. It is not in orbit around Earth but instead at Lagrange point 2 (L2), just under a million miles away from Earth. The JWST hosts a slew of technologies from all over the world and features many instruments, including two infrared cameras (NIRCam and MIRI, at 'Near InfraRed' and 'Mid InfraRed' wavelengths). The JWST features a giant origami stack of sun-shielding layers that keep it cool enough to be sensitive to weak thermal light, and its primary mirror is gold-plated to aid in collecting infrared light.

Also, the MIRI and NIRCam are some pretty amazing high-resolution cameras. To the left is a comparison of the Hubble Space Telescope's (launched in 1990) and the JWST's (launched in 2021) picture of 'The Pillars of Creation.'

All of these features are excellent news for those of us astronomers interested in finding exoplanets, which require small resolution capabilities and sensitivity to the thermal/infrared light emitted by exoplanets.

Koret project: Overview and goals

For this Koret project, I am performing exoplanet vetting, looking through images of stars taken by the JWST to see if there are any new exoplanets or background sources (such as stars or galaxies) visible within them. The rate of discovering a new exoplanet through direct imaging is less than one percent, so this project is mainly about learning how to perform exoplanet vetting rather than finding a new exoplanet (although finding a new exoplanet would be awesome).

I am a researcher a part of the JWST program GO 4050, a large group of astronomers consisting of over 30 members hailing from institutions all over the world (some US-specific institutions include NASA, STScI, University of Michigan, UCSC, and the University of Hawaii). As of the writing of this website, our program's data is still only available privately to program members, so this site is one of the first looks into some actively ongoing JWST research!

JWST candidate selection and data collection

For our data, our program looked at ~90 different stars that are all a part of a 'moving group,' meaning that all of these stars are close to each other in distance (which is important when we think about the time it takes to point the JWST at each star) and move in relatively the same direction. The JWST has an amazing sensitivity to faint thermal signals, and its Near InfraRed Camera (NIRCam) detects planets better the further away they are from their star, meaning that we should be able to image exoplanets that are around the size of Jupiter!

As mentioned briefly earlier, it's hard to directly image rocky planets close to their stars; astronomical technology is effectively working its way inwards from gas giants to the rocky planet regime, and sub-Jupiter mass planets are the smallest mass planets we are currently able to directly image; this is based on the capabilities of the JWST, our best telescope at the moment. In the future, we hope to be able to directly image Earth-sized exoplanets as we move towards the search for habitable worlds.

The proposal for this program was submitted in 2023, and our data was collected autonomously by the JWST during the summer of 2024. We used NIRCam to take our pictures with two different wavelength filters, F200W and F444W (the larger the number, the redder/more thermal the light imaged is), and used coronagraphs to cover up the star centered in each image. The data was then sent to Earth, received and stored by satellites briefly before being transmitted to the ground, and program members quickly began working on image reduction soon after receiving the data.

JWST image reduction

After taking images of a star with a coronagraph, there are, unfortunately, still some streaks of starlight that affect planet detection (this can be seen in the Comet Lovejoy example discussed earlier). To get rid of the remaining starlight, we use 'Reference Differential Imaging,' also called RDI subtraction. This is a post-processing method that allows us to remove the remaining starlight leftover after coronagraphic imaging, improving sensitivity to faint objects that would have otherwise been obscured.

On the right is a diagram of how RDI works. The leftmost image is what a raw/untouched image from JWST might look like; lots of light from behind the circular coronagraph is still visible in our image, and our suspected exoplanet(s) may be slightly visible (or not at all!). By using a 'reference target,' a completely unrelated, separate, lone star that is of similar color and brightness to the star in the 'science target' image, we can subtract away the similarities, effectively removing any lingering starlight.

This process is not 100% perfect, and many variables are involved, so you may end up with minimal residuals in your final image. The rightmost diagram portrays what an RDI-processed image might look like; almost all of the starlight is removed, and the other sources (in this example, an exoplanet in the bottom right) in the image shine brightly.

For our JWST data, several team members went through each image, understanding the PSF (Point-Spread-Function), i.e., the shape and intensity of each star within each NIRCam image, and performed a slew of reductions, including RDI subtraction.

The next stage in the process of discovering exoplanets, after the data has been collected, downloaded, and reduced, is for someone to go through and figure out which sources of light in each star mage are not exoplanets, about ~99% of them. For the 1% where it's uncertain whether or not a source of light is a planet, we perform follow-up observations to confirm that the first observation wasn't a fluke/error. It is then, after a confirmation made by a separate observation (imaged by JWST or a different telescope), that we can confidently say we've found a new planet.

For my Koret project, I started carrying out this process of exoplanet vetting for JWST program GO 4050. Below are the challenges and triumphs of learning how to search for exoplanets, as well as key takeaways and future plans for my research.

Koret project: Exoplanet vetting 101

1. Is it shaped like a planet?

The first question in determining whether or not something is a planet is 'is it shaped like one'? And to answer this we first need to know what a source/possible planet looks like.

On the left are two pictures of TWA 34, one of the stars the JWST imaged for our dataset. The leftmost picture is the star system in our shorter (less red, less thermal) wavelength filter. On the right is the same star system, but at a redder wavelength. The difference in filters is what cause all of the sources in the image to move inwards towards the center of the image.

NIRCam F200W TWA 34 images(shorter wavelength)

NIRCam F444W TWA 34 images (longer wavelength)

At the center of the image above, there is a white pixelated circle surrounded by some dark currents. This is where our star TWA 34 was before it was RDI-subtracted away! The dark currents are leftover noise from the subtraction.

In the second row of images, I have circled all of the possible sources within them. these blobs of light are shaped like hexagonal flowers (particularly visible with source '2') because the mirrors of JWST are hexagonally shaped.

Now, a possible planet source would look round, or in this case, like a hexagonal flower. A star would look the same. But a galaxy would look smeared out like a line or oval of paint smeared on a canvas. This means we can likely rule out source number '3' which looks more like a blob than a pointed flower.

NIRCam F200W TWA 34 images (shorter wavelength)

NIRCam F444W TWA 34 images (longer wavelength)

2. Is it colored like a planet?

Now that we've ruled out the shapes of each source let's consider their color/magnitude. As discussed earlier, planets give off thermal or infrared light. So from the F200W -> F444W, an exoplanet should get brighter, whereas a star should get dimmer.

In our images, there are actually several sources (3-6) that aren't even visible in the shorter wavelength, meaning that they shine brighter further in the infrared! These are good exoplanet candidates. Unfortunately, these sources are so dim even with our current best telescope (JWST), we can barely see them. This means we cannot confirm or deny their possibly-planet nature unless we request the JWST (or a future telescope, such as the Roman) to reimage them.

Sources 1 and 2 are visible in both images, and they don't get significantly brighter or dimmer between filters. Meaning that we need to check their nature with a different method.

From now on, let's use source 2 as a case study since this object is bright enough that it should be visible with archival data from telescopes other than the JWST.

3. Does it move like a planet?

Stars move across the sky over time. Although this isn't visible to the naked eye, it is measurable via 'proper motion.' to calculate the proper motion of a star, you must observe it at two different times and make precise calculations on its change in position based on astronomical coordinate systems. Most stars have easily accessible public proper motion information associated with them, including TWA 34.

Knowing how a star moves across time is useful because exoplanets are bound to their host stars, meaning that they will move according to their host star's proper motion.

JWST image of TWA 34 (2024)

NIRC2 image of TWA 34 (2017)

We'll need to compare our JWST TWA 34 image to an archival one to do this astrometric proper motion calculation. Above on the right (teal picture) is archival data of TWA 34, as imaged by Keck Observatory's NIRC2 camera (located on Mauna Kea in Hawai'i). This image looks strikingly different from our JWST image because it was taken from the ground (where Earth's atmospheric turbulence affects our image quality) and with older technology. In our NIRC2 image, our source is just barely visible at the top near the image border.

Each data file comes with ancillary information that tells astronomers about not just what object they're looking at but also the condition of the instrument and telescope when the image was taken, including essential items such as:

RA/Dec: where the star/object is located in the sky
AO: What kind of Adaptive Optics was used, and what settings were selected
PA: Position Angle, how the image was taken relative to an orientation of 0 degrees where north is up.
Filter: what color/wavelength range the pixels in the image include.

With all of this information, as well as knowledge about TWA 34's proper motion, we can calculate how we expect the star to move through space and time, and observe whether or not the source does as well.

3.1 How does the star move?

Between 2024 and 2017 I found that TWA 34 should decrease in both it's RA (Right Ascension) and Dec(lination); If we were to map its motion on our 2017 NIRC2 image, it would appear as if the star moved down and to the right, as shown on the right.

For this diagram and all of the ones following, as well as all of the calculations for this project, I wrote Python code utilizing common astronomy packages such as Astropy, Numpy, and WebbPSF.

3.2 How 'should' the source move?

If the source is a star, it should stay in the same location for both images; it shouldn't move at all between our 2017 data and 2024 data (excluding parallax and other minor uncertainties).

However, if our source is bound to TWA 34, say, an exoplanet in orbit around its host star, it should follow the same proper motion as TWA 34. Meaning it should move down and to the right like the image immediately prior. To calculate this, I found the RA/Dec of the source in my JWST dataset, and then re-calculated the proper motion movement, but with respect to the source rather than TWA 34.

To map this, I marked the location of the source in 2024 to be a red 'background' circle; if the source in the NIRC2 image falls within this circle range, that means it didn't move at all between our datasets and is, therefore, a background star that is unrelated to TWA 34. There is also a purple 'planet' circle that denotes where my proper motion calculations predict the source to be if it were to be 'co-moving' with the star, aka an exoplanet.

The white circle in the image to the right is the real location of the source in my NIRC2 data. You can see it falls into neither the 'background' nor 'planet' prediction locations, meaning this is a null result, and there is an error somewhere.

4. Position angle calculations are important!

The reason my result was null is that I forgot to consider the fact that all of our telescopes take pictures at angles! Even the most minuscule angle can make a difference here, so it's important to hunt down and figure out what angle all of your images were taken at.

Originally, this was a big hurdle for me in this project because I had no familiarity with Keck data going into it; before starting this research, I had only used the JWST, and every telescope has their own unique methods for keeping track/reporting values like the Position Angle, or PA, of each image taken.

To fix this, I tried correcting my image rotation according to the PA value given in my data file and according to the official Keck observatory's website. I then also realized I needed to consider the PA of the original JWST data I used for my proper motion calculations. After making corrections, I mapped the image on the right.

It looks like our source now matches up with one of our predicted locations... the predicted 'planet' location if the source were to have moved with star TWA 34.

...Does this mean we found an exoplanet?

Koret project: conclusion and looking forward

Is it or is it not a planet? Probably not (99% no). In short, we have more evidence for this source being a background star rather than an exoplanet. More-in-depth reasons for this being a star are because:

1. After doing a magnitude calculation, it looks like this source gets slightly dimmer in the more infrared wavelengths (meaning it doesn't shine as brightly in thermal colors, which is expected of stars, which shine brighter in less thermal/more optical colors). Meaning the color/spectrum of this source is more like a star than a planet.

2. This system was well studied prior to my starting on the project, meaning many people have considered it an exoplanet candidate and found in their research that it's not planetary in nature. This source was meant to serve as a case study; I.e., I expected this source to be a background star before starting, and also...

3. There are uncertainties in my calculations that I did not delve into detail here on this website. Particularly, I struggled with calculating the position angle, which is largely what my results hinge on; I have yet to look at another source in our JWST campaign, so I cannot confirm/deny that my calculations and methodology are 100% correct yet.

4. Additionally, I ran some other tests to look for any other indication that the source could/couldn't be in orbit around TWA 34, including whether or not its radial separation changed (i.e., Earth does not get any closer to the Sun each year because it is in a stable orbit). I also used Kepler's 3rd law to see whether or not the source could have had orbital motion within the 7 years between these two datasets, and I will need a second opinion on both of these calculations before anything can be determined.

That being said, this result is interesting. My future plans are to look at an alternative angle of TWA 34 as taken by a different archival dataset, either from Keck's NIRC2 (as we looked at here) or from some other telescope, to see whether or not I can recreate my finding that this source is co-moving with TWA 34. Additionally, I plan to move on to other exoplanet candidates within our JWST data to see if I find I've forgotten to consider anything with this calculation, and I will likely do that sooner rather than later to get a refreshed look at the findings presented here.

If you have any questions or are interested in hearing more, my email is ucsc address 'ssubboti' at ucsc dot edu!

Acknowledgments

Thank you so much to my mentors who have given me great guidance and support over the years: Andrew Skemer and Aarynn Carter

And to the UC Santa Cruz's Physics (Astrophysics) Department, Koret Scholarship fund, NASA, and STScI who have funded this research financially and through academic accreditation.

Page updated

Report abuse