New research led by Craig Walton, a postdoc at the Centre for Origin and Prevalence of Life at ETH Zurich, and ETH Zurich professor Maria Schönbächler shows that these elements must already be available in the right amounts when a planet’s core forms. “During the formation of a planet’s core, there needs to be exactly the right amount of oxygen present so that phosphorus and nitrogen can remain on the surface of the planet,” explains Walton, lead author of the study. On Earth, that appears to have happened about 4.6 billion years ago, giving our planet an unusually fortunate chemical starting point. The result could influence how scientists search for life beyond Earth.

How Planet Core Formation Affects Habitability

Planets begin as bodies of molten rock. As they form, their materials separate by weight. Heavy metals such as iron sink inward and create the core, while lighter material remains above and eventually becomes the mantle and later the crust.

Oxygen levels during this stage are critical. If there is too little oxygen when the core forms, phosphorus bonds with heavy metals such as iron and gets pulled down into the core. Once that happens, it is no longer available in the parts of the planet where life might develop. If there is too much oxygen, phosphorus stays in the mantle, but nitrogen becomes more likely to escape into the atmosphere and be lost.

The Chemical Goldilocks Zone

Using extensive modeling, Walton and his co-authors found that both phosphorus and nitrogen remain in the mantle in large enough amounts only within a very narrow range of moderate oxygen conditions. They describe this as a chemical Goldilocks zone.

“Our models clearly show that the Earth is precisely within this range. If we had had just a little more or a little less oxygen during core formation, there would not have been enough phosphorus or nitrogen for the development of life,” says Walton.

The team also found that other planets, including Mars, formed under oxygen conditions outside this Goldilocks zone. On Mars, that meant more phosphorus in the mantle than on Earth, but less nitrogen, producing difficult conditions for life as we know it.

A New Way to Search for Life Beyond Earth

These findings may change how scientists think about habitability. So far, much of the focus has been on whether a planet has water. Walton and Schönbächler argue that this is not enough.

A planet may have water and still be chemically unfit for life from the very beginning. If oxygen levels were wrong while the core was forming, the planet may never have kept enough phosphorus and nitrogen in the places where life could use them.

Why Sun-Like Stars May Matter Most

Astronomers may be able to estimate these chemical conditions by studying other solar systems with large telescopes. The oxygen available during planet formation depends on the chemical makeup of the host star. Because planets form mostly from the same material as their star, the star’s composition helps shape the chemistry of the entire planetary system.

That means solar systems whose chemistry is very different from ours may be poor candidates in the search for life. “This makes searching for life on other planets a lot more specific. We should look for solar systems with stars that resemble our own Sun,” says Walton.

The findings, published in The Astronomical Journal, come from an international team led by Caleb Cañas of NASA’s Goddard Space Flight Center, with contributions from Carnegie Science’s Shubham Kanodia and others.

A Giant Planet Orbiting a Small Star

TOI-5205 b is about the size of Jupiter but orbits a much smaller star, one that is roughly four times Jupiter’s size and only about 40 percent as massive as the Sun. When the planet passes in front of its star in an event known as a “transit,” it blocks about six percent of the star’s light.

During these transits, astronomers used spectrographs to break the starlight into its component colors. This technique allows them to identify the chemical makeup of the planet’s atmosphere and gain insight into how it formed and evolved alongside its host star.

A Puzzle for Planet Formation Theories

Planets typically form within a rotating disk of gas and dust surrounding a young star. While this process is widely accepted, systems like TOI-5205 b challenge existing models. Massive planets orbiting small, cool stars at close distances are difficult to explain using current theories.

To investigate these unusual systems, Kanodia, Cañas, and Jessica Libby-Roberts of the University of Tampa are leading JWST’s largest Cycle 2 exoplanet program, Red Dwarfs and the Seven Giants. This project focuses on rare worlds like TOI-5205 b, often referred to as GEMS (for giant exoplanets around M dwarf stars).

JWST Detects Unexpected Atmospheric Chemistry

TOI-5205 b was first confirmed in 2023, when Kanodia led follow-up observations based on data from NASA’s Transiting Exoplanet Survey Satellite (TESS). Now, researchers have used JWST to examine its atmosphere in detail for the first time.

After observing three transits, the team encountered an unexpected result. The planet’s atmosphere contains significantly fewer heavy elements compared to hydrogen than Jupiter does. Even more surprising, its metallicity is lower than that of its own host star, making it unlike any giant planet studied so far.

The data also revealed the presence of methane (CH₄) and hydrogen sulfide (H₂S) in the atmosphere.

Heavy Elements May Be Hidden Deep Inside

To better understand these findings, researchers Simon Muller and Ravit Helled at the University of Zurich used advanced models of planetary interiors. Their results suggest that the planet as a whole is about 100 times more metal rich than its atmosphere appears to be.

“We observed much lower metallicity than our models predicted for the planet’s bulk composition, which is calculated from measurements of a planet’s mass and radius. This suggests that its heavy elements migrated inward during formation and now its interior and atmosphere are not mixing,” Kanodia explained. “In summary, these results suggest a very carbon-rich, oxygen-poor planetary atmosphere.”

The GEMS Survey and Future Research

This work is part of the broader GEMS Survey, which aims to study transiting giant planets around M-dwarf stars to better understand their formation, internal structure, and atmospheres. The research team includes Carnegie astronomers Peter Gao, Johanna Teske, and Nicole Wallack, along with former Carnegie postdoctoral fellow Anjali Piette, now at the University of Birmingham.

Additional contributors include researchers from institutions such as Johns Hopkins University’s Applied Physics Laboratory, the Academia Sinica Institute of Astronomy and Astrophysics, Catholic University, the University of Maryland, Caltech, NASA Goddard, the University of St. Andrews, Penn State University, the University of California Irvine, the Tata Institute of Fundamental Research, and the University of Amsterdam.

Correcting for Starspots Improves Accuracy

The team also accounted for interference caused by starspots on the host star. These dark, active regions can distort observations by brightening certain wavelengths and hiding parts of the atmospheric signal.

By correcting for these effects, the researchers improved the accuracy of their measurements. Wallack and Kanodia are now refining this approach in a newer JWST project focused on the same system. Their work could help future studies of planets orbiting active stars produce more reliable results.

This rapid growth has raised concerns about sustainability. In response, researchers at a School of Engineering have created a proof-of-concept AI system designed to be far more efficient. Their approach could reduce energy use by up to 100 times while also improving performance on tasks.

A Hybrid Approach Called Neuro-Symbolic AI

The research comes from the laboratory of Matthias Scheutz, Karol Family Applied Technology Professor. His team is developing neuro-symbolic AI, which combines traditional neural networks with symbolic reasoning. This method mirrors how people approach problems by breaking them into steps and categories.

The work will be presented at the International Conference of Robotics and Automation in Vienna in May and will appear in the conference proceedings.

Teaching Robots to See, Understand, and Act

Unlike familiar large language models (LLMs) such as ChatGPT and Gemini, the team focuses on AI systems used in robotics. These systems are known as visual-language-action (VLA) models. They extend LLM capabilities by incorporating vision and physical movement.

VLA models take in visual data from cameras and instructions from language, then translate that information into real-world actions. For example, they can control a robot’s wheels, arms, or fingers to complete a task.

Why Traditional AI Struggles With Simple Tasks

Conventional VLA systems rely heavily on data and trial-and-error learning. If a robot is asked to stack blocks into a tower, it must first analyze the scene, identify each block, and determine how to place them correctly.

This process often leads to mistakes. Shadows may confuse the system about a block’s shape, or the robot may place pieces incorrectly, causing the structure to collapse.

These errors are similar to the problems seen in LLMs. Just as robots can misplace blocks, chatbots can generate false or misleading outputs. Examples include fabricating legal cases or producing images with unrealistic details such as extra fingers.

How Symbolic Reasoning Improves Accuracy and Efficiency

Symbolic reasoning offers a different strategy. Instead of relying only on patterns from data, it uses rules and abstract concepts such as shape and balance. This allows the system to plan more effectively and avoid unnecessary trial and error.

“Like an LLM, VLA models act on statistical results from large training sets of similar scenarios, but that can lead to errors,” said Scheutz. “A neuro-symbolic VLA can apply rules that limit the amount of trial and error during learning and get to a solution much faster. Not only does it complete the task much faster, but the time spent on training the system is significantly reduced.”

Strong Results in Puzzle Tests

The researchers tested their system using the Tower of Hanoi puzzle, a classic problem that requires careful planning.

The neuro-symbolic VLA achieved a 95% success rate, compared with just 34% for standard systems. When given a more complex version of the puzzle that it had not encountered before, the hybrid system still succeeded 78% of the time. Traditional models failed every attempt.

Training time also dropped sharply. The new system learned the task in only 34 minutes, while conventional models required more than a day and a half.

Massive Energy Savings in Training and Use

Energy consumption was reduced dramatically as well. Training the neuro-symbolic model required only 1% of the energy used by a standard VLA system. During operation, it used just 5% of the energy needed by conventional approaches.

Scheutz compared this inefficiency to everyday AI tools. “These systems are just trying to predict the next word or action in a sequence, but that can be imperfect, and they can come up with inaccurate results or hallucinations. Their energy expense is often disproportionate to the task. For example, when you search on Google, the AI summary at the top of the page consumes up to 100 times more energy than the generation of the website listings.”

The Growing Strain of AI on Power Infrastructure

As AI adoption accelerates across industries, demand for computing power continues to climb. Companies are building increasingly large data centers, some of which require hundreds of megawatts of electricity. That level of consumption can exceed the needs of entire small cities.

This trend has sparked a race to expand infrastructure, raising concerns about long-term energy limits.

A More Sustainable Path for AI

The researchers suggest that current approaches based on LLMs and VLAs may not be sustainable in the long run. While these systems are powerful, they consume large amounts of energy and can still produce unreliable results.

In contrast, neuro-symbolic AI offers a different direction. By combining learning with structured reasoning, it may provide a more efficient and dependable foundation for future AI systems.

Scientists at UC San Francisco have now pinpointed a protein that appears to drive much of this decline.

FTL1 Emerges as a Key Driver of Brain Aging

To understand what changes with age, the researchers tracked shifts in genes and proteins in the hippocampus of mice over time. Among everything they examined, only one stood out as consistently different between young and old animals. That protein is called FTL1.

Older mice showed higher levels of FTL1. At the same time, they had fewer connections between neurons in the hippocampus and performed worse on cognitive tests.

How FTL1 Alters Brain Function

When the team boosted FTL1 levels in young mice, the effects were striking. Their brains began to look and function more like those of older mice, and their behavior reflected this shift.

Lab experiments revealed more detail. Nerve cells engineered to produce high amounts of FTL1 developed simplified structures, forming short, single extensions instead of the complex, branching networks seen in healthy cells.

Reversing Memory Decline by Lowering FTL1

The most surprising result came when researchers reduced FTL1 in older mice. The animals showed clear signs of recovery. Connections between brain cells increased, and their performance on memory tests improved.

“It is truly a reversal of impairments,” said Saul Villeda, PhD, associate director of the UCSF Bakar Aging Research Institute and senior author of the paper, which was published in Nature Aging. “It’s much more than merely delaying or preventing symptoms.”

Metabolism Link Points to New Treatments

Further experiments showed that FTL1 also affects how brain cells use energy. In older mice, higher levels of the protein slowed cellular metabolism in the hippocampus. However, when researchers treated these cells with a compound that boosts metabolism, the negative effects were prevented.

Hope for Future Brain Aging Therapies

Villeda believes these findings could pave the way for treatments that target FTL1 and counter its effects in the brain.

“We’re seeing more opportunities to alleviate the worst consequences of old age,” he said. “It’s a hopeful time to be working on the biology of aging.”

Authors and Funding

Other UCSF authors are Laura Remesal, PhD, Juliana Sucharov-Costa, Karishma J.B. Pratt, PhD, Gregor Bieri, PhD, Amber Philp, PhD, Mason Phan, Turan Aghayev, MD, PhD, Charles W. White III, PhD, Elizabeth G. Wheatley, PhD, Brandon R. Desousa, Isha H. Jian, Jason C. Maynard, PhD, and Alma L. Burlingame, PhD. For all authors see the paper.

This work was funded in part by the Simons Foundation, Bakar Family Foundation, National Science Foundation, Hillblom Foundation, Bakar Aging Research Institute, Marc and Lynne Benioff, and the National Institutes of Health (AG081038, AG067740, AG062357, P30 DK063720). For all funding see the paper.