Australian PhD Team Uses AI to Restore James Webb’s Clarity

The cosmos is deep and vast. It is hauntingly beautiful and full of mystery. Understanding this universe is key to understanding our own origins and place within it.

To help with this, scientists have built the James Webb Space Telescope (JWST), a powerful tool designed to capture light from the first galaxies and reveal just how our world came to be. It is the largest telescope in space, enabling the study of astronomical objects using infrared radiation.

This form of light has longer wavelengths than visible light, and it can penetrate cosmic dust clouds that block visible light. And because Webb observes longer wavelengths, it needs a large collecting area to achieve high resolution.

Webb’s main mirror consists of 18 hexagonal mirror segments made of gold-plated beryllium, forming a 6.5-meter-diameter mirror. This large mirror, combined with high-resolution, high-sensitivity instruments, enables Webb to detect objects far more distant than previous space telescopes could.

To make these observations possible, Webb must be kept at extremely low temperatures, below 50 K (−370 °F; −223 °C). This prevents the telescope’s own heat from interfering with the faint infrared signals it’s trying to capture. A multi-layer sunshield protects the telescope from external heat sources like the Sun, Moon, and Earth.

With these powerful instruments, Webb has enabled investigations across various fields of astronomy and cosmology, including the observation of the first stars, the formation of the first galaxies, and detailed atmospheric science of potentially habitable exoplanets.

However, some recent images from Webb have been rather blurry due to distortions in one of its key instruments. To solve this issue, two PhD researchers at the University of Sydney, Max Charles and Louis Desdoigts, turned to AI. Impressively, they did it from Earth without any hardware-related changes.

But why does this matter? What’s Webb been finding out there that makes sharper images so important? Let’s take a look.

James Webb Space Telescope: Surprises From the Early Universe

Since going operational, Webb has found galaxies with complex structure and black holes when the universe was still very young, challenging what we knew about how quickly the first galaxies could form and grow.

Webb has revealed clear atmospheric signatures—including methane, carbon dioxide, sulfur dioxide, ammonia, and even hydrogen sulfide—on multiple worlds. Some of these molecules had only tentative evidence (or none at all) before JWST; Webb delivered the first unambiguous detections in several cases, dramatically advancing exoplanet chemistry.

The telescope has observed features such as Jupiter’s auroras and characterized small bodies embedded in the asteroid belt between Jupiter and Mars, which are difficult for other telescopes to detect, demonstrating the flexibility of JWST. Several massive black holes have also been discovered by the telescope.

With these and many more discoveries, JWST has been delivering on its promise. It is showing us that the early universe is much more active and complex than we thought it to be. Its power is further helping us go beyond our galaxy and solar system with more detailed and accurate insights into the nature of the cosmos.

Most recently, astronomers spotted¹ “one of the most puzzling discoveries” to date while going through the images from the Webb. 

These very bright and puzzling objects could be the earliest known galaxy in the universe, emerging 100 million years after the Big Bang, or they may be a brown dwarf, aka a failed star, which is bigger than the largest gas giant planets but not big enough to sustain nuclear fusion in its core. The exact identity of Capotauro remains uncertain, but it is certainly “interesting and promising.”

Capotauro was spotted by a team of astrophysicists at the National Institute of Astrophysics in Italy in a previous study, when they were trying to identify old galaxies in JWST observations. The lack of fine-grained data, however, made it impossible for them to specify its identity. But then JWST released more data on Capotauro earlier this year, which helped the team narrow down what it could be.

The team used the images taken by JWST’s Near Infrared Camera (NIRCam) at seven wavelengths to measure Capotauro’s brightness and limited but more fine-grained data from JWST’s Near Infrared Spectrograph (NIRSpec) to get an idea of its age and temperature.

Data from the NIRCam and NIRSpec were combined, and then models were used to test three possible galaxy configurations. A scenario in which Capotauro might be a brown dwarf on the Milky Way’s outer rim was also tested, along with a range of other possibilities, such as the object being a peculiar exoplanet or a very odd young galaxy.

While the results were inconclusive, the team has identified the two most likely options.

One possibility is that Capotauro was formed around 100 million years after the Big Bang, which pushes the age of the oldest known galaxy back by about 200 million years. The other possibility is that Capotauro is an unusual brown dwarf, which is the coldest and farthest in our galaxy.

Both are “very exciting” possibilities as they would challenge what we thought we knew about our galaxy and how it evolves, said study co-author Giovanni Gandolfi, an astrophysicist at INAF.

In another new observation using JWST, astronomers have discovered five carbon-based complex compounds² trapped in the ice around a star outside the Milky Way.

The organic molecules have been detected around a protostar in a dwarf galaxy, the Large Magellanic Cloud, which is 160,000 light-years from Earth and orbits close to our galaxy. This small galaxy is filled with hot, luminous stars and has fewer elements heavier than helium than the Milky Way does. By gaining an understanding of this galaxy, the astronomers aim to apply that to understanding more distant galaxies from when the universe was much younger.

“The harsh conditions tell us more about how complex organic chemistry can occur in these primitive environments where much fewer heavy elements like carbon, nitrogen, and oxygen are available for chemical reactions.”

– Study co-author Marta Sewilo, an astronomer at the University of Maryland and NASA’s Goddard Space Flight Center

The researchers pointed the Webb at ST6, a developing star in the Large Magellanic Cloud, and, with the help of tools that measure infrared light, they found the following complex molecules in the ice around it: methanol, ethanol, acetaldehyde, acetic acid, and methyl formate.

Out of these, methanol is the only one that’s been “conclusively detected” in protostars, which makes the new observations “exceptional.”

The researchers even found signals that may be caused by glycolaldehyde, which can react with other molecules to form ribose, a type of sugar that’s an important component of ribonucleic acid (RNA), which is essential for life.

Click here to learn how laser & 3d printing will build our future in space.

AMIGO: The AI Calibration That Sharpened JWST’s NIRISS-AMI

Swipe to scroll →

With all these discoveries coming from Webb, two PhD researchers at the University of Sydney, Max Charles and Louis Desdoigts, realized they could make the telescope even more effective. Their AI-driven software corrected the blurring in Webb’s images from Earth without requiring costly space repair missions.

The problem they solved was in the Aperture Masking Interferometer (AMI), the telescope’s only Australian-designed component. AMI on the NIRISS instrument is the highest-resolution infrared interferometer ever placed in space. It allows astronomers to take high-resolution images of stars and planets outside our solar system by combining light from different sections of Webb’s main mirror.

But when the telescope began operations, AMI’s performance was affected by electronic distortions in its infrared camera detector. With charge migration, or the Brighter-Fatter Effect, limiting AMI’s performance, the images captured by the telescope suffered subtle fuzziness.

Space telescopes have faced similar optical flaws before. Hubble had a comparable issue back in the 90s due to a lens-spacing error. Its main mirror had the wrong shape, a flaw called spherical aberration, which caused the telescope’s primary mirror to have more than one focal point, making its images blurry.

To bring the blurry data into focus, NASA redesigned the Wide Field and Planetary Camera 2 (WFPC2) and developed Hubble’s Corrective Optics Space Telescope Axial Replacement (COSTAR) instrument package, which was the size of a large refrigerator and acted like a pair of eyeglasses for Hubble’s Faint Object Camera (FOC), Goddard High Resolution Spectrograph (GHRS), and the Faint Object Spectrograph (FOS). Both WFPC2 and COSTAR were installed by astronauts during Hubble’s first servicing mission in late 1993.

This time, instead of sending astronauts to make physical repairs, the researchers created a software-based calibration method right here on Earth. Charles and Desdoigts (who is now a postdoctoral researcher at Leiden University in the Netherlands) developed what they called Aperture Masking Interferometry Generative Observations, or AMIGO³.

AMIGO is an open-source, calibration framework “that forward-models the full JWST AMI system – including its optics, detector physics, and readout electronics – using an end-to-end differentiable architecture implemented in the Jax framework and in particular exploiting the dLux optical modelling package.”

It uses neural networks and advanced simulations to replicate how Webb’s optics and electronics function in space. To design their algorithms to digitally correct the images, the researchers identified the problem where the electric charge spreads to adjacent pixels.

This way, “they managed to fix things with code” rather than “sending astronauts to bolt on new parts,” said Professor Peter Tuthill from the University of Sydney’s School of Physics and the Sydney Institute for Astronomy, who created AMI and worked on the solution. “It’s a brilliant example of how Australian innovation can make a global impact in space science.”

With the help of AMIGO, Webb delivered clear and sharper images, capturing obscure celestial objects in precise detail.

To test just how effective AMIGO is, Charles led a separate study⁴, where, using the improved calibration, Webb produced clear images of volcanoes on Jupiter’s moon Io, a black hole jet, and the dust-filled winds of WR 137.

“This work brings JWST’s vision into even sharper focus,” said Dr. Desdoigts. “It’s incredibly rewarding to see a software solution extend the telescope’s scientific reach – and to know it was possible without ever leaving the lab.”

Investing in Space Tech

In the realm of Space technology, Northrop Grumman (NOC ) is the leading name, which was the primary contractor for the James Webb Space Telescope.

Besides helping build JWST, the company has developed many other major space technologies, including Cygnus spacecraft, which is a cargo resupply vehicle for the International Space Station (ISS), Antares rocket for Cygnus missions, air-launched Pegasus rocket, Mission Extension Vehicle (MEV) to dock with satellites to extend their operational life, and HALO for the Artemis program.

Being an aerospace and defense technology company, Northrop Grumman also develops military aircraft systems, advanced tactical weapons, and missile defense solutions for US government agencies and international customers. Furthermore, it provides command, control, communications, and reconnaissance systems.

Recently, the company signed a deal with AI startup Luminary Cloud for accelerated, AI-enabled spacecraft design. 

The startup has developed an AI model, powered by NVIDIA’s PhysicsNeMo, for the defense contractor to design and build new spacecraft more quickly. With the partnership, engineers will gain access to this purpose-built tool called Physics AI that can generate high-fidelity simulations of subsystems in mere seconds, instead of the usual 12 hours it can take with computational fluid dynamics simulations to produce just one simulation. It is currently tailored for designing spacecraft thruster nozzles.

“Physics AI is the next level of complexity in AI, and Northrop Grumman is bringing this technology to our design engineers to dramatically speed up hardware development. Using AI to make something small, like a spacecraft thruster, puts us on a path to do much bigger things, like using AI to design larger components or even an entire spacecraft.”

– Han Park, vice president of AI integration at Northrop Grumman Space Systems

To build models that can generate new designs to predict the physical world, the startup didn’t use the data available on the internet but rather used the laws of physics to create spacecraft designs.

“We make up for the lack of data with the fact that we’re not trying to predict whether that image or set of pixels is a cat or dog or an elephant. Rather, we’re trying to predict whether the airflow flows in a particular way that obeys some physical principles.”

– said Luminary Cloud’s CTO, Juan Alonso

Now, when it comes to the company’s market performance, as of writing, the $85 billion market cap Northrop Grumman’s shares are trading at around $595, up 26.8% this year. Just earlier this month, NOC shares hit an all-time high (ATH) at $640.90.