Albert Einstein postulated in his 1905 theory of special relativity that the speed of light in a vacuum is constant. Ever since, that’s been one of the fundamental assumptions of physics.
Now Enbang Li, a physicist at the University of Wollongong in Australia, has challenged this idea by building a machine he says is capable of detecting changes in the speed of light as it crosses Earth’s surface. The findings suggest that light is, in fact, sped up by gravity, which could have implications for Earth science applications ranging from climate monitoring to mineral resource exploration.
An Old Conundrum
The idea that light is influenced by gravity is not new. Einstein’s ideas, which were further developed with his theory of general relativity in 1915, predicted massive objects in space would bend light with their gravitational grab. This theory was famously proven in 1919 when two independent teams measured starlight passing a solar eclipse at two different points on Earth’s surface and found the results matched Einstein’s predictions.
This bending of light’s path, according to general relativity, is achieved by a warping of the space-time fabric. Under this scenario, the speed of light remains constant—it just has to travel farther as it navigates the warped space-time around celestial bodies, so to a distant observer, it appears to have been slowed.
But what if light doesn’t navigate warped space-time and actually is slowed down or sped up by the gravity of large objects?
Li pointed out that Einstein himself was not always convinced the speed of light was constant. In 1911, he wrote a paper postulating that light speed changed depending on the gravity of objects it passed by. However, “when he published his general theory,” said Li, “he just abandoned this model.”
If the movement of light can be affected by gravity, Li reasoned, it might be possible to detect variations in its speed on a local level—such as an elevator shaft in a building on the campus of the University of Wollongong.
Raising the Big Issues
Gravity on Earth varies locally, depending on altitude, underground density, and topography. Gravity at the top of a tall building, for example, is measurably weaker than it is at the bottom.
With these variations in mind, Li installed an experiment in an elevator. It consisted of a coil of fiber-optic cable that if stretched out in one direction, would be 10 kilometers (6.2 miles) long. Laser beams were fired through the cables and then reflected back, thus traveling 20 kilometers (12.4 miles) before reaching an ultrafast photodetector. An oscilloscope measured the time it took for the beam to travel that distance. The experiment was run at the top of the shaft and at the bottom.
The biggest challenge, Li said, was filtering out all the surrounding environmental “noise,” such as changing temperature and humidity, electromagnetic disturbance, and building vibrations. Li designed a temperature control system, and the experiment was sealed in an enclosure with electromagnetic shielding to isolate air flows. Li ran the experiment and found light moved minutely faster at the bottom of the shaft than at the top.
Gravity Sensing on the Go
Next, Li took his research a step further by building a small, portable machine he claims can detect changes in the speed of light as it nears more gravitationally dense objects.
In this second experiment, Li positioned a moveable 72-kilogram (159-pound) weight near the machine. Light, he found, moved faster when the weight was near the machine than when it was farther away.
The results, which were published in Scientific Reports, are consistent with the variable speed of light model Einstein proposed in 1911, although Li’s preliminary results are much larger than that model predicts.
If proven, the findings would present a fundamental challenge to our understanding of both general and special relativity.
In the world of Earth sciences, they could lead to greatly improved gravity-sensing technologies. Because of their sensitivity to changes in mass, gravity sensors are used to map the seafloor and to locate underground mineral reserves. Gravity sensing can also improve our understanding of Earth’s climate as variations in the gravity field can be linked to factors like changes in ice mass and shifts in groundwater.
Currently, gravimeters are vulnerable to vibrations and movement, whereas Li’s machine, which has no moving parts, could even be used on board a plane or submarine.
“A Striking Claim”
Chris Stevens, a numerical relativist with the University of Canterbury in New Zealand, called the work “intriguing and ambitious.” While Stevens, who was not involved in the research, said that Li’s work is “well founded,” he noted that any observable effects of gravity on light on Earth would be “extraordinarily small” and therefore these results must be treated with caution.
“In my own research on observable gravitational phenomena,” he explained, “I usually require a few black holes colliding somewhere in the universe. Separating genuine gravitational signatures from environmental and instrumental noise will therefore be exceptionally demanding.”
“The work is exciting because it pushes precision photonic measurement techniques into a regime where relativistic effects may become practically useful for geophysics and sensing applications.”
Stevens said the implications of Li’s research, if validated, would be far-reaching. “The work is exciting because it pushes precision photonic measurement techniques into a regime where relativistic effects may become practically useful for geophysics and sensing applications.”
John Norton, an historian of physics at the University of Pittsburgh who was also not involved in the research, called the findings a “striking claim.” He was, however, skeptical of them, saying “if there is a coupling between light and gravity of magnitude greater than general relativity predicts, it is hard to see how the 1919 eclipse test and later studies of gravitational lensing would not have found it.”
Li acknowledged there is a long way to go before his device finds everyday use. Disentangling the intricacies of space and time, he said, is a vast challenge. “In physics, people still say gravity is a mystery. Light is another mystery. So if you put these two mysteries together, that’s going to be a giant mystery.”
—Bill Morris, Science Writer
Citation: Morris, B. (2026), How Einstein’s lost theory could help us find minerals, Eos, 107, https://doi.org/10.1029/2026EO260189. Published on 12 June 2026.
Dust and water ice clouds are ubiquitous on Mars; they regulate the planet’s climate and can affect measurements of other atmospheric components. Constraining their spatial and temporal variability is also essential for improving Martian general circulation models.
Fedorova et al. [2026] use solar occultation measurements from the SPICAM infrared spectrometer on board the Mars Express orbiter to characterize nine Martian years (MY 28 through 36) of dust and water ice clouds. Because the spectrometer could not distinguish between these particles’ types, the researchers employ a new method integrating Mars Climate Sounder data and general climate model predictions to identify them.
The analysis reveals that the particles can reach altitudes up to 80 kilometers during perihelion, while their size remains relatively uniform with height. This suggests that Martian dust distribution is driven more by atmospheric dynamics and horizontal transport, capable of lifting and moving particles over vast distances, rather than by turbulent mixing against gravity alone.
The study also provides a detailed seasonal and spatial climatology of major Martian atmospheric features, including the Polar Hood Clouds, the Aphelion Cloud belt, and the Mesospheric Clouds. The detection of high-altitude clouds (70–90 km) during dust events confirms enhanced transport of water vapor into the upper atmosphere during both global and regional storms. These findings are consistent with simultaneous observations from the Atmospheric Chemistry Suite on the Trace Gas Orbiter.
These observations show that large-scale atmospheric dynamics, rather than local mixing alone, control how aerosols are distributed vertically on Mars, with important implications for the transport of water to the upper atmosphere and the planet’s climate evolution.
The figure shows how the water ice cloud layers vary with latitude and season (Ls), based on SPICAM observations. (a) altitude of the cloud layer in kilometers; (b) thickness of the cloud (optical depth); (c) average size of the ice particles in micrometers; and (d) number of particles within the layer (number density. The background color is the amount of dust in the atmosphere from Montabone et al. [2015]: red areas indicate high dust levels, while dark blue areas indicate low dust. Black open circles mark locations where no clear water ice clouds were detected. Credit: Fedorova et al. [2026], Figure 12
Citation: Fedorova, A. A., Luginin, M., Montmessin, F., Korablev, O. I., Bertaux, J.-L., Stcherbinine, A., & Lefèvre, F. (2026). Multiyear monitoring of aerosol vertical distribution on Mars by SPICAM IR/MEX. Journal of Geophysical Research: Planets, 131, e2025JE009388. https://doi.org/10.1029/2025JE009388
—Arianna Piccialli, Associate Editor, and Beatriz Sanchez-Cano, Editor, JGR: Planets
Orbital imaging has hinted that Mars may have carbon-containing rocks called carbonates on its surface. Carbonates on Mars could offer new insights into how water interacted with rock on the Red Planet, helping scientists learn more about its past. In addition, because carbonates on Earth are primarily produced by living organisms, these rocks are high-value targets in the search for signatures of past life on Mars.
NASA’s Perseverance rover has been traversing Mars since 2021, covering more than 41 kilometers, much of it within Jezero Crater in the Nili Fossae region. Previous orbital data indicated the crater contains carbonates, as well as abundant olivine, which can change to carbonate in the presence of water and carbon dioxide. Now Clavé et al. have analyzed spectroscopic data from Perseverance’s SuperCam instrument suite from multiple locations within Jezero Crater, providing clear evidence of carbonates on Mars, as well as detailed information on how the mineralogy varies between locations.
The authors confirmed the presence of both carbonates and olivine-bearing rocks throughout Jezero Crater and found a generally inverse relationship between the two minerals. By contrast, carbonates were generally positively correlated with the presence of hydrated silica. The researchers hypothesize that an ancient lake in the crater, along with potential hydrothermal activity, played a role in transforming olivine to carbonate. The varying amounts of carbonate and different alteration states seen today may have been caused by changing lake levels on Mars billions of years ago, the researchers suggest.
Amounts of carbonate by weight vary between locations, from 1%–3% in the Séítah unit to 6%–16% in the Eastern Margin Unit. Extrapolating to the entire regional olivine-rich unit, the researchers calculated it could contain as much as 1.1 × 1014 kilograms of carbon, or up to 0.4% of the current total mass of the Martian atmosphere. Overall, Mars’s crust could contain significant amounts of carbon, implying that widespread carbon sequestration may have cooled the planet significantly in the past. (Journal of Geophysical Research: Planets, https://doi.org/10.1029/2025JE009107, 2026)
Citation: Scharping, N. (2026), Carbon-rich rocks may have cooled the ancient Martian atmosphere, Eos, 107, https://doi.org/10.1029/2026EO260170. Published on 28 May 2026.
Solar activity affecting Earth and its planetary neighbors encompasses a wide range of phenomena, from the steady solar wind and the interplanetary magnetic field to extreme events like solar flares, coronal mass ejections (CMEs), and solar energetic particle (SEP) events. These space weather phenomena interact in complex ways with planetary magnetospheres and atmospheres. On Earth, we see the results in the dancing lights of stunning auroras and in less frequent but sometimes severe disruptions to telecommunications, navigation, and energy infrastructure.
Forecasting conditions throughout the heliosphere (the region influenced by the solar wind), understanding the variety of Sun-Earth interactions, and predicting arrivals of space weather events—both benign and potentially hazardous—are a grand challenge.
The Sun-Earth challenge requires tracking and predicting conditions—from routine and quiet to rare and extreme—across tens of millions of kilometers of interplanetary space.
Solar flares emit electromagnetic radiation that spreads in all directions. In contrast, the propagation of CMEs and SEP events depends on their source location on the Sun and on the heliospheric magnetic field, which is carried outward by the solar wind. The impacts these events have on magnetosphere systems further vary depending on particle energies and intensities in SEPs and on particle speeds and the magnetic field orientation in CMEs. The Sun-Earth challenge thus requires tracking and predicting conditions—from routine and quiet to rare and extreme—across tens of millions of kilometers of interplanetary space.
This tracking and prediction is powered by petabyte-scale datasets from solar observatories and spacecraft measurements that provide rich observational archives. Researchers use these data to deduce physically meaningful quantities describing the heliosphere and to identify patterns to distinguish quiet from active conditions. The resulting insights not only answer fundamental science questions but also provide critical prediction time frames needed by space weather forecasters.
Even with all these data, the enormity of space between the Sun and Earth presents a major obstacle to our predictive capabilities. Another obstacle is that the data are obtained by different instruments operating at different locations and times. These factors combine to create a unique data sparsity challenge that complicates large-scale analysis.
These fundamental issues—the massive yet still insufficient supply of data available, the extreme differences in the scales of the processes we must illuminate, and the need for actionable predictions—suggest opportunities for artificial intelligence (AI) and machine learning (ML) to complement traditional physics-based analytical approaches [Camporeale, 2019]. In a series of workshops—insights from which inform the discussion below—scientists explored such opportunities and how they can advance heliophysics research and operational space weather forecasting.
The Need for Space Weather Forecasting
Space weather events can have significant impacts on infrastructure and humans. They can disrupt satellite operations (e.g., by enhancing atmospheric drag on satellites), damage electronics in space, interfere with radio communications and GPS, and even affect power grids (e.g., through geomagnetically induced currents) during the most severe events. They can also pose risks to people, especially astronauts beyond the protection of Earth’s atmosphere and airline crews and passengers on long-distance polar flights, during which exposure to energetic particles is elevated. Forecasting offers a first line of defense in preparing for or preventing damaging and hazardous effects of space weather.
In assessing major CMEs, forecasters consider whether and when events will reach Earth and whether they will trigger geomagnetic storms and substorms. For SEP events, predictions must include arrival times, peak intensities, durations, and energy characteristics.
Predicting extreme space weather phenomena is vital, but equally important is forecasting periods when no significant activity is expected, which is critical information for satellite operators and other stakeholders. Making such predictions requires understanding physics spanning 8 orders of magnitude in space and time, from subsecond processes in Earth’s magnetic environment to multiday solar eruptions propagating across the 150 million kilometers between the Sun and Earth (Figure 1) and long-term interactions at scales associated with the 11-year solar cycle.
Fig 1. Length scales and Sun-to-Earth transit times vary greatly for different types of space weather (SW), including solar flares, solar energetic particle (SEP) events, coronal mass ejections (CMEs), and interplanetary coronal mass ejections (ICMEs). High-speed particles are the first to arrive, usually within minutes of a flare, whereas CMEs arrive in 2–4 days. Credit: Georgoulis et al. [2026], CC BY-NC-ND 4.0
In addition to operational forecasting, these challenges are fundamental in heliophysics research. Such research includes work to reveal how the Sun generates its magnetic field, how solar wind accelerates and evolves, how planetary magnetospheres respond to external forcing, how particles are accelerated, and how energy transfers across multiple scales and regimes.
Unique Challenges in Heliophysics
Modern AI and ML algorithms excel at analyzing well-curated, extensive datasets that include millions of training examples. For example, AI-aided terrestrial weather forecasting relying on continuous, high-resolution coverage from thousands of ground stations, weather balloons, and satellites has advanced dramatically in recent years.
Fewer than a dozen spacecraft monitor Earth’s magnetosphere, a region spanning tens of Earth radii. Solar wind observations are even sparser.
Heliophysics, however, presents a unique and somewhat opposite scenario. Fewer than a dozen spacecraft monitor Earth’s magnetosphere, a region spanning tens of Earth radii (about 6,371 kilometers). Solar wind observations are even sparser, with just a handful of monitors scattered across the space between the Sun and Earth. This fundamental scarcity poses a challenge for data-driven approaches, which typically depend on abundant observations that are well distributed in space and time to produce trustworthy (i.e., generalizable and reproducible) models.
Data sparsity is further compounded by the relative rarity of intense space weather phenomena such as CMEs, major geomagnetic storms, and extreme substorms, which occur only a few times per solar cycle. Most heliophysical observations capture quiet, low-activity conditions when the solar wind is steady and magnetospheres are calm. Standard ML approaches trained on such imbalanced datasets may achieve high statistical accuracy by simply predicting a “nothing-will-happen” outcome but completely fail when extreme events occur.
Although solar eruptions and geomagnetic storms are relatively rare, they exhibit recurring patterns and consistency in their physical drivers. This regularity suggests that historical observations, when properly clustered and analyzed, can be used to enhance prediction capabilities. The challenge therefore lies in extracting meaningful patterns from sparse measurements of rare events while avoiding models that work well for average conditions but fail when they matter most [Chu et al., 2025].
AI Solutions for Data Sparsity
Heliophysics research employs clever approaches to extract maximum information from the limited available observations. One strategy is to mine multidecade observational records from various satellites and to match and group together measurements collected at times with similar solar wind and geomagnetic activity conditions.
Another, more universal approach is to embed fundamental physical laws directly into ML models through physics-informed neural networks [Raissi et al., 2019], ensuring that predictions respect physical reality even when training data are limited. Data assimilation techniques used in weather forecasting similarly blend sparse observations with physics-based simulations and update models as new measurements arrive.
This animated model shows Earth’s magnetosphere during a powerful May 2024 geomagnetic storm that involved strong solar flares and multiple CMEs. The visualization uses the Multiscale Atmosphere-Geospace Environment (MAGE) model from the Johns Hopkins Applied Physics Laboratory to depict wind rushing toward Earth and disturbing its magnetic field (orange and purple lines). The green cloud represents electric field current intensity; the blue squiggles are tracers of solar wind velocities. Credit: NASA Scientific Visualization Studio and NASA DRIVE Science Center for Geospace Storms
These methods converge on a common theme: building gray box models (so named because they’re less opaque than black box models) that are data driven but grounded in physically real constraints. For data-starved applications, hybrid approaches can outperform purely data-driven or purely physics-based methods [Liu et al., 2025].
Satellite instruments are generating increasingly large solar wind datasets. However, the variables obtained (e.g., solar wind speed and pressure) are highly intercorrelated [Borovsky, 2018], making it difficult to identify which ones truly drive magnetospheric responses. New algorithms are helping to distill datasets without losing critical scientific information [e.g., Camporeale, 2025]. Meanwhile, advanced statistical and ML methods can cut through dataset complexity by reducing dimensionality, identifying causal relationships among variables, and providing clues about dominant drivers.
For instance, information theory provides tools to detect dependencies in complex systems, establish causality, and rank variables that most effectively predict space weather outcomes [Wing et al., 2022]. Such techniques can be paired with other “explainable” tools, such as SHAP (SHapley Additive exPlanations) values, a method inspired by game theory, to pinpoint physical variables (e.g., solar wind speed or magnetic orientation) that drive a prediction [Ma et al., 2023].
Distilling datasets and improving model interpretability help make ML more practical and more scientifically trustworthy and its predictions more robust. But fully trusting ML models in operational environments requires rigorous validation and uncertainty quantification. These models must not only make predictions but also indicate their confidence levels for operational decisionmaking.
When a model forecasts a major geomagnetic storm, operators need to know whether that prediction carries 60% or 95% confidence, for example.
When a model forecasts a major geomagnetic storm, operators need to know whether that prediction carries 60% or 95% confidence, for example. Ensemble approaches, in which multiple models provide a range of outcomes, help quantify this uncertainty, while using standardized, well-documented datasets enables fair model intercomparisons.
The research community is developing ML-ready benchmark datasets with consistent formatting and clear metadata to establish such validation procedures [e.g., Angryk et al., 2020]. These resources allow researchers to test new algorithms against common baselines, accelerating progress while ensuring that advances are robust and reproducible rather than artifacts of specific data processing choices.
Notably, one domain in heliophysics that is not affected by severe data sparsity is solar imaging. Decades of continuous, high-resolution observations from the Solar Dynamics Observatory (SDO), which delivers 1.5 terabytes of data every day, have created enormous data archives. Because the Sun drives space weather throughout the heliosphere, these datasets offer an ideal opportunity for use in foundation models, large-scale ML systems trained to learn comprehensive internal representations that can then be easily adapted to specific scientific tasks with minimal additional training.
Surya, a foundation model designed to construct a digital representation of the Sun, represents one such effort. It is still in early development and has yet to be validated, but this approach illustrates how data-rich domains can be leveraged with modern AI techniques to create tools that broadly benefit heliophysics research and space weather forecasting.
Advancing Research and Operational Forecasting Together
In addition to the needs for data and model development and validation, applying AI to address the challenges of heliophysics requires sustained, multidisciplinary collaborations. Fostering those collaborations has been the focus of a series of workshops, with the most recent being 2025’s Machine Learning, Data Mining and Data Assimilation in Geospace (LMAG25) meeting at the Johns Hopkins University Applied Physics Laboratory. The workshops have brought together heliophysicists, machine learning experts, data scientists, and specialists from weather forecasting and applied mathematics to exchange knowledge and establish community standards.
Space weather forecasters need models that are accurate and interpretable and that provide not just statistical metrics but also actionable predictions.
The LMAG forums also serve as gathering spaces for scientists to validate models against diverse datasets, compare physics-based and data-driven approaches, develop performance benchmarks, and discuss how to bridge research and operational requirements. Space weather forecasters need models that are accurate and interpretable and that provide not just statistical metrics but also actionable predictions with known limitations and reliability. Of course, researchers also benefit. These conversations allow them to gain insight into operational constraints that shape how modeling approaches become practical in real-world settings.
LMAG and similar initiatives facilitate direct exchanges among adjacent communities, including by making meeting presentations openly available. These efforts are helping translate cutting-edge AI and ML techniques into practical tools that help protect critical infrastructure and human well-being. They are also deepening our understanding of how the Sun shapes space weather throughout the solar system and its effects—both mundane and major—on Earth.
Borovsky, J. E. (2018), The spatial structure of the oncoming solar wind at Earth and the shortcomings of a solar-wind monitor at L1, J. Atmos. Sol. Terr. Phys., 177, 2–11, https://doi.org/10.1016/j.jastp.2017.03.014.
Camporeale, E. (2019), The challenge of machine learning in space weather: Nowcasting and forecasting, Space Weather, 17(8), 1,166–1,207, https://doi.org/10.1029/2018SW002061.
Chu, X., et al. (2025), Imbalanced Regression Artificial Neural Network Model for Auroral Electrojet Indices (IRANNA): Can we predict strong events?, Space Weather, 23(5), e2024SW004236, https://doi.org/10.1029/2024SW004236.
Georgoulis, M. K., et al. (2026), Prediction of solar energetic events impacting space weather conditions, Adv. Space Res., in press, https://doi.org/10.1016/j.asr.2024.02.030.
Liu, Y., et al. (2025), Data-driven modeling of electrostatic turbulence by physics-informed Fourier neural operator, Mach. Learn. Sci. Technol., 6(4), 045050, https://doi.org/10.1088/2632-2153/ae19cd.
Ma, D., et al. (2023), Opening the black box of the radiation belt machine learning model, Space Weather, 21(4), e2022SW003339, https://doi.org/10.1029/2022SW003339.
Raissi, M., P. Perdikaris, and G. E. Karniadakis (2019), Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, J. Comput. Phys., 378, 686–707, https://doi.org/10.1016/j.jcp.2018.10.045.
Stephens, G. K., et al. (2019), Global empirical picture of magnetospheric substorms inferred from multimission magnetometer data, J. Geophys. Res. Space Phys., 124(2), 1,085–1,110, https://doi.org/10.1029/2018JA025843.
Wing, S., et al. (2022), Modeling radiation belt electrons with information theory informed neural networks, Space Weather, 20(8), e2022SW003090, https://doi.org/10.1029/2022SW003090.
Author Information
Savvas Raptis (savvas.raptis@jhuapl.edu), Manolis K. Georgoulis, Mikhail Sitnov, Anthony Sciola, and Simon Wing, Johns Hopkins University Applied Physics Laboratory, Laurel, Md.
Citation: Raptis, S., M. K. Georgoulis, M. Sitnov, A. Sciola, and S. Wing (2026), Vast space, sparse data: An AI answer to twin space weather challenges, Eos, 107, https://doi.org/10.1029/2026EO260188. Published on 11 June 2026.
The Sun continuously blasts charged, magnetic field–carrying particles, or plasma, in all directions. This solar wind interacts with the magnetic fields and atmospheres of several of our solar system’s planets and other bodies, sculpting long magnetic tails of charged particles—magnetotails—that stretch into space behind them.
Magnetotails contain thin layers of electric current–carrying plasma sheets, which sometimes “flap” in an up-and-down waving motion. Spacecraft observations have revealed that flapping in Earth’s magnetotail can be driven by a process called magnetic reconnection, in which magnetic field lines rapidly break and then snap together in a new configuration, releasing stored energy. However, whether reconnection plays this same role beyond Earth has thus far been a mystery.
Wen et al. report the first evidence that magnetic reconnection may also trigger magnetotail flapping at Mars.
Unlike Earth, Mars lost its global magnetic field billions of years ago. But it still sports a magnetotail, thanks in large part to interactions between the solar wind and charged particles in its upper atmosphere. Strong magnetic fields embedded in certain patches of the Martian crust—remnants of its lost planet-wide field—also influence the magnetotail.
Until recently, Mars’s magnetotail could only be studied using observations from NASA’s Mars Atmosphere and Volatile Evolution (MAVEN) spacecraft. MAVEN showed that the Martian magnetotail is highly dynamic, with a structure that twists, shifts, and flaps—and from which charged particles may escape into space. But because MAVEN can observe only one part of the magnetotail at a time, it couldn’t identify what processes might trigger flapping.
Another spacecraft, China’s Tianwen-1 orbiter, has now provided a second set of eyes. The researchers analyzed simultaneous observations from the two spacecraft, finding that signatures of magnetic reconnection detected by MAVEN in the upstream part of the magnetotail tended to coincide with flapping events detected downstream by Tianwen-1.
Before or during flapping, the spacecraft also detected temporary, twisted plasma structures known as flux ropes. A similar link has previously been observed on Earth, and it suggests that flux ropes generated by magnetic reconnection upstream might propagate downstream, driving instabilities in the magnetotail’s plasma sheets and triggering flapping.
Though more research is needed to confirm these findings, they shed new light on how energy moves and is released in space around Mars—and possibly other planets and celestial objects. (AGU Advances, https://doi.org/10.1029/2026AV002343, 2026)
In low Earth orbit (typically below about 700 kilometers altitude), atmospheric drag is the primary source of uncertainty when predicting the trajectories of satellites. These prediction errors largely arise from limitations and inaccuracies in the models used to estimate the density of the upper atmosphere, particularly within the thermosphere.
Mutschler et al. [2026] introduce a new method for estimating atmospheric density along the path of an individual satellite by using Energy Dissipation Rates (EDRs). The derived single-satellite density measurements provide valuable insight into variations in thermospheric density and can help characterize how the upper atmosphere responds to disturbances such as geomagnetic storms. Incorporating these observations can contribute to ultimately improving the accuracy of satellite orbit predictions.
Effective density and Space Force effective density estimated by the Kosmos 1508 satellite (plotted on the right-hand y axes) compared to estimates from satellites Swarm-A and Swarm-C (plotted on the left-hand y-axes). Credit: Mutschler et al. [2026], Figure 17a
Citation: Mutschler, S., Pilinski, M., Zesta, E., Oliveira, D. M., Delano, K., Garcia-Sage, K., & Tobiska, W. K. (2026). First results of a new inversion tool for thermospheric neutral mass density computations during severe geomagnetic storms. AGU Advances, 7, e2025AV002079. https://doi.org/10.1029/2025AV002079
Research & Developments is a blog for brief updates that provide context for the flurry of news that impacts science and scientists today.
To date, astronomers have confirmed the existence of just under 6,300 exoplanets. New research could more than double that number, adding a potential 10,000 new planets in one fell swoop.
Yes, that’s right. A 1 with 4 zeros.
The T16 project has announced the discovery of 10,091 exoplanet candidates observed by NASA’s Transiting Exoplanet Survey Satellite (TESS). Since 2018, the all-sky survey has been monitoring more than 200,000 nearby stars using the transit method, which detects the faint dip in a star’s light when a planet crosses in front of it. Astronomers typically require 3 dips to be sure that what they’re seeing is actually a planet and not a one-off event such as an asteroid or comet in that distant star system.
The T16 project analyzed the light curves of more than 54 million stars observed during the first year of the TESS mission. The project’s analysis technique allowed it to search for planets around stars up to 16 times fainter than TESS typically searches, drastically increasing the field of discovery.
That’s more than were detected in the entirety of NASA’s Kepler mission and its follow-on K2.
Their pipeline detected 11,554 planet candidates. Of those, 1,052 of those had been detected previously and 411 only had one transit—not enough to confirm a planet.
That leaves 10,091 potential new planets. That’s more than were detected in the entirety of NASA’s Kepler mission and its follow-on K2 and more than double the existing planet candidates from TESS that await confirmation. These discoveries will be published in the Astrophysical Journal Supplement.
All of the new planet candidates orbit their stars quickly, with orbital periods between 12 hours and 27 days. Although most of the stars that TESS observes are smaller and cooler than the Sun, those close orbits likely mean that most of those planets are far too hot to be habitable.
The T16 project team confirmed the planet-hood of one of their candidates not using the transit method, but a different method that measures the gravitational tug a planet exerts on its host star. That planet, TIC 183374187, is hot and slightly larger than Jupiter.
The remaining 10,090 newly discovered planet candidates require additional verification to determine whether they truly are planets or not. But given the rigor of the team’s analysis and the requirement of at least 3 transits to even make this list, it’s likely that most of the new discoveries are indeed planets.
“Astronomers are a bit conservative when it comes to claims like this, and want to be sure they pass a bunch of tests to make sure everything was done correctly and these planets actually exist,” astronomer Phil Plait wrote in his Bad Astronomy Newsletter. “Having said that, the process the astronomers went through looks legit to me, and I would bet the majority of these new candidates are real. That’s amazing.”
These updates are made possible through information from the scientific community. Do you have a story about science or scientists? Send us a tip at eos@agu.org.
Wind-driven waves on Earth move sediments and shape shorelines. They transport energy between the atmosphere and planetary surface and also mix bodies of liquid, affecting both chemistry and biology. On other worlds with surface liquids, either now or in the past, wind waves would likely perform the same function and so would play a key role in climate and astrobiological potential.
“They’re basically the interface between how the atmosphere communicates with the landscape, especially at the coast.”
New research went back to the fundamentals and explored the conditions that can generate waves on worlds with different physical properties and different liquids, such as Titan, Mars, and select exoplanets.
“Wind waves are really interesting phenomena,” said Una Schneck, a planetary science doctoral student at the Massachusetts Institute of Technology (MIT) in Cambridge. “They’re basically the interface between how the atmosphere communicates with the landscape, especially at the coast.”
The Physics of Waves
Past models of wind generation on other planets struggled because they tended to start from preexisting models of Earth waves. Those models were developed to describe waves in Earth’s specific combination of gravity, atmosphere, and surface liquid, namely, water, said Schneck, who led the new research. Such models were sometimes tailored to describe a particular location and season. Adapting those models for conditions on other worlds, including other liquids like methane and sulfuric acid, always seems to leave traces of Earth behind.
However, the physics of what creates wind-driven waves should be universal, Schneck said, so the team went back to the basics of wave generation. They developed a wave model that explores the relationship between a world’s bulk properties, like gravity and air density, and liquid properties, like surface tension, to determine the wind strength needed to produce a wave.
The team “created this model that went back to the basic physics of waves, instead of just trying to fit to known wave conditions,” said Taylor Perron, an MIT geomorphologist and planetary scientist and coauthor of the research.
The Curiosity rover landed in Gale Crater on Mars (left) and has since found evidence—wavy bedforms—that this former crater lake had waves. Titan’s northern hemisphere hosts a sprawling lake district (right). The shores of one of the moon’s largest bodies of liquid, Ligeia Mare, shows evidence of wave activity. Credit: Left: NASA/JPL-Caltech/ESA/DLR/FU Berlin/MSSS; right: NASA/JPL-Caltech/ASI/USGS
The model showed that the threshold wind speed to generate a wave is lower for liquids with less surface tension, which makes it easier to change the liquid’s shape. Higher air density provides more force to push against a liquid’s surface, and lower gravity makes it easier for a wave to rise up—both factors allow a weaker wind to create a wave. The team published these results in the Journal of Geophysical Research: Planets in April.
Waves on Other Worlds
The team first tested their model on the only set of wind and wave data we have—Earth. They used 20 years of wave and weather data for Lake Superior. The model found, correctly, that it takes wind speeds of 2.2 meters per second to generate waves on the lake’s surface and accurately predicted the height of waves for different wind speeds.
They then used the model to predict wave conditions on other worlds. They started with Mars, which likely had ancient oceans and lakes. Winds of 1.2 meters per second would have created waves in the lake that filled Gale Crater millions of years ago. A wave in Gale Crater would have been taller than a wave on Earth produced by wind of the same strength owing to Mars’s lower gravity.
The story is similar on Titan, the largest moon of Saturn. Waves in Titan’s hydrocarbon lakes would swell with a mere 0.5 meter per second of wind and would rise higher than an Earth wave under similar wind conditions. But they would travel much more slowly than Earth waves and would be spaced farther apart.
“The paper represents our best theoretical understanding of how we expect for waves to behave in a variety of environments,” said Jason Barnes, a planetary scientist at the University of Idaho in Moscow who was not involved with this research. “The movie of Titan waves is particularly awesome—very slow moving for such large amplitudes! Although I don’t expect waves to get that high ever in Titan’s sluggish atmosphere, it’s fun to be able to visualize what they might look like if they did.”
“In theory, this is something that people could do.”
The team also explored wave-generating conditions on three Earth-sized exoplanets. The possible sulfuric acid lakes of the exo-Venus Kepler-1649 b would grow in winds of 5.3 meters per second but would grow to a height similar to that of Earth waves because of its Earth-like gravity. Water lakes on LHS 1140 b would grow in 2.7 meter winds, similar to those on Earth, but would not grow as high because of its higher gravity. And on 55 Cancri e, a lava world, it would take winds of 37 meters per second—a category 1 hurricane—to move tiny waves of molten rock.
“Would you be able to ever detect this? Is this a useful thing to think about, or is it just a fun thought experiment?” Schneck asked. “If the waves are tall enough, you should be able to detect a change in the polarization [of an exoplanet’s light curve] that would not only suggest that there is a liquid surface on that exoplanet, but that liquid surface has waves.…In theory, this is something that people could do.”
Will We See It? Not Soon
Right now, the only world known to have surface liquid other than Earth is Titan, but we don’t have the right observations of Titan to test the new model. The European Space Agency’s Huygens probe landed on the moon in 2005, but nowhere near the northern lake district. NASA’s Cassini mission (of which Huygens was a part) did not detect any waves but did observe a changing lake shore that hinted at wave activity.
It’s possible that Titan’s waves are seasonal and Cassini just didn’t have the right timing, Perron noted. Temperature changes during Saturn’s year could affect wind speeds and also the composition of Titan’s lakes, changing the conditions of wave generation.
Still, the wind speed needed to make a wave on Titan is so low that “it would be very surprising if waves never formed. It just may be difficult to catch them when they’re there,” he said.
“The best way to test this work would be to send a sea probe to float or motor on one of Titan’s big 3 seas.”
“The best way to test this work would be to send a sea probe to float or motor on one of Titan’s big 3 seas—Kraken Mare, Ligeia Mare, or Punga Mare,” Barnes said. “Such a ‘buoy’ probe would be able to simultaneously measure both the sea conditions and the wind conditions, allowing for a comprehensive test of the model.”
Alas, no such mission is in the works, and the upcoming Dragonfly mission won’t travel near any lakes to test this theory either. A future Titan orbiter might provide that information, while a current or future Mars rover might yet gather evidence showing how lakes worked in that planet’s past.
“The improved understanding of waves from this paper might help to constrain the possibilities for wave erosion at the margins of bodies of water…thereby helping us to probe into the past climates of Mars and Titan,” Barnes said.
In late 2025, astronomers spotted an interstellar comet making a quick trip through the solar system. 3I/ATLAS was discovered in July when it was just inside Jupiter’s orbit. It’s now about halfway between Jupiter and Saturn and getting farther away every day.
The European Space Agency’s Jupiter Icy Moons Explorer (ESA JUICE) mission, on its way to Jupiter, imaged 3I/ATLAS on 5 November 2025 when the comet was 64 million kilometers from the spacecraft. Credit: ESA/Juice/JANUS, CC BY-SA 3.0 IGO
Astronomers have been observing 3I/ATLAS throughout its journey inward toward the Sun and back out again, compiling the most comprehensive and detailed view thus far of an interstellar object, including the chemistry of the gases that sublimated from its surface and formed its coma and tail.
In a first-of-its-kind observation of an interstellar object (ISO), researchers have discovered that the ratio of deuterium to hydrogen in 3I/ATLAS’s outgassed water is 30–40 times higher than in solar system objects. That suggests that the comet formed in a much colder environment than our own solar system did.
“It is always hard to really pinpoint where these objects form,” said Luis E. Salazar Manzano, the lead researcher on these observations and a doctoral student at the University of Michigan in Ann Arbor. “We know that they were formed in different parts of the galaxy, but it’s hard to connect what we measure with how they were formed. These types of measurements, such as the relative abundance of deuterium to hydrogen in water, are one of the best ways we have to actually [learn] about their forming conditions and their evolution.”
Coming In from the Cold
Water appears to be ubiquitous throughout the universe, sprinkled within distant galaxies and in star-forming nebulae. But there are different flavors of water: heavy, semiheavy, and plain old H2O. In the molecular clouds where stars form, the cold environment favors a chemical reaction that increases the amount of gaseous deuterium (D), an isotope of hydrogen, relative to regular hydrogen atoms. That deuterium then bonds with hydrogen and oxygen atoms to create semiheavy water, or HDO.
By measuring the quantity of semiheavy water relative to regular water in an object, scientists can infer the object’s ratio of deuterium to hydrogen, or D/H, and decode the physical conditions in which that water formed. Astronomers have made such measurements for baby stars, planet-forming disks, solar system comets, and meteorites, as well as Earth’s ocean.
“What is fundamentally important about ISOs is that they are physical leftovers of the process of forming another planetary system and they can give us clues to that process,” said Karen Meech, an astrobiologist at the University of Hawaiʻi at Mānoa who was not involved with this research.
“The conditions in the stellar system in which 3I/ATLAS formed may have been quite different from the one in the solar system.”
The team observed 3I/ATLAS with the Atacama Large Millimeter/submillimeter Array (ALMA) in Chile on November 2025 when the comet was 335 million kilometers (208 million miles) from Earth. It had just passed its closest approach to the Sun and was as bright as it was ever going to be. This timing was critical for the measurements the team wanted to make because the signal for HDO is very subtle, especially when it has to compete with the much more abundant H2O in the comet and within Earth’s atmosphere, Salazar Manzano explained.
Those measurements showed that for every 1,000 hydrogen atoms in 3I/ATLAS, there were about 5–7 deuterium atoms. While that’s not a lot, the ratio is still at least 40 times more than what’s found in ocean water and at least 30 times the average value in solar system comets.
“The conditions in the stellar system in which 3I/ATLAS formed may have been quite different from the one in the solar system,” said Paul Hartogh, a physicist and atmospheric science researcher at the Max Planck Institute for Solar System Research in Göttingen, Germany.
The first interstellar object, 1I/ʻOumuamua, did not outgas any material, and although the second object, 2I/Borisov, did, it was not bright enough to detect deuterium. 3I/ATLAS was the first opportunity astronomers had to measure the D/H ratio of an interstellar comet. Those measurements suggest that 3I/ATLAS formed in a much colder galactic environment than the solar system did, less than 30°C above absolute zero. The team published these results in Nature Astronomy in April.
Planning for the Next Interstellar Visitor
Hartogh, who was not involved with this research, said that on the one hand, 3I/ATLAS’s high deuterium enrichment is surprising because it is higher than that of any known comet. On the other hand, he added, some scientists predicted such high values for cometary water several decades ago.
Meech said she found these results “really interesting.” She never expected all other solar systems to have formed just like ours, and 3I/ATLAS fits with that idea.
“This gives us an intriguing look into the processes of planetary system formation—and that there are differences from our own solar system,” Meech said. “It is too early to tell what this implies for the formation of planets or habitable worlds. We are just at the beginning of an exciting story.”
“The fact that we were able to make this measurement with 3I will allow us to better prepare what to expect with the next generation of interstellar objects.”
3I/ATLAS is getting harder to see with telescopes, but astronomers still have a lot of data from when it was much brighter to go through, Salazar Manzano said. Teams around the world are working on creating a holistic picture of the comet’s chemistry and evolution.
What’s more, “the fact that we were able to make this measurement with 3I will allow us to better prepare what to expect with the next generation of interstellar objects,” Salazar Manzano said.
Scientists expect that the Vera C. Rubin Observatory could discover between 6 and 51 interstellar objects within the next 10 years. If objects are detected early enough in their journey through the solar system, “there may be enough time to coordinate observations with ground-based and spaceborne telescopes, taking advantage of the recent experience gained by the multiple 3I/ATLAS observations,” Hartogh said.
“These are rare opportunities to study another planetary nursery up close, and we have to take advantage of each new ISO to learn as much as we can,” Meech said. “It may be harder for a large number of individual teams to get all the data they want, so I think coordination and collaboration is needed more than ever.”
Citation: Cartier, K. M. S. (2026), Interstellar comet was born in a very cold place, Eos, 107, https://doi.org/10.1029/2026EO260141. Published on 7 May 2026.
The solar system is bathed in galactic cosmic rays: protons and atomic nuclei traveling, nearly at the speed of light, from all directions. Earth’s magnetic field and atmosphere shield us from most of this harmful radiation, but outside of that shelter, the bombardment is strong enough to prove a threat to astronauts.
But a new analysis of data from the Chang’e-4 lunar lander published in Science Advances revealed an extended cosmic ray shelter stretching from Earth at an unexpected angle at least as far as the Moon, though exactly how far is unclear. When the Moon passes through this shelter in its orbit of Earth, the lunar surface experiences a roughly 20% reduction in the galactic cosmic ray flux.
“We found Earth casts kind of a shadow in the galactic cosmic ray space,” said Robert F. Wimmer-Schweingruber, a space physicist at Kiel University in Germany. “This was unexpected, and to me that was the cool part of this paper.”
The surprise came in part because the shape of Earth’s magnetic field is well understood: It forms a strong protective region around the planet known as the magnetosphere, with a long “tail” shaped by the solar wind of charged particles streaming from the Sun.
If the magnetotail is like a person’s shadow cast behind them by sunshine, this newly discovered bubble would be like if that shadow extended to the front of the person as well.
“You would expect an effect inside the tail or as [the Moon goes] through the tail, but we find an effect of the tail ahead of the tail,” said Wimmer-Schweingruber. He noted that if the magnetotail is like a person’s shadow cast behind them by sunshine, this newly discovered bubble would be like if that shadow extended to the front of the person as well and tilted rather than lying along a line connecting Earth, the Sun, and the Moon.
“The observed region of reduced [galactic cosmic ray] flux on the sunward side of the Moon’s orbit outside the geomagnetic field where it is compressed by the solar wind is unexpected,” Brian Flint Rauch wrote in an email. Rauch, a cosmic ray physicist at Washington University in St. Louis who was not involved in the Chang’e-4 study, added that any reduction in cosmic ray exposure is noteworthy for potential astronauts on the Moon.
A 20% decrease in flux during part of the lunar orbit is unlikely to make a large difference in determining when it’s safest for astronauts go out onto the lunar surface. But it might help guide individual decisions in the moment because while spacesuits won’t protect astronauts from cosmic rays, the metal of a habitat or lander would.
Shelter from the Storm
The China National Space Administration’s Chang’e-4 spacecraft was the first successful mission to the lunar farside, landing in the Von Kármán crater on 3 January 2019. As part of its suite of scientific instruments, the probe carried the Lunar Lander Neutron and Dosimetry experiment (LND) developed by Wimmer-Schweingruber and collaborators at Kiel University in an astonishingly rapid 18 months. This detector was designed in part to gauge conditions for human exploration by measuring the radiation on the Moon’s surface, including cosmic rays.
LND collected data between January 2019 and January 2022. Though Apollo astronauts carried radiation dosimeters, those instruments did not provide detailed information about fluctuations in exposure, making LND the primary source for such information from the lunar surface. For that reason, it provided the best data on galactic cosmic rays, which consist mostly of protons accelerated to nearly the speed of light in the remnants of supernovas.
Measurements show the ambient radiation dose on the lunar surface is more than twice as high as on the ISS and nearly 200 times as high as on Earth.
These protons arrive in the solar system from every direction, often undeflected by the magnetic fields of stars or planets. However, Earth’s magnetosphere is strong enough to repel many galactic cosmic rays in low orbit, where the International Space Station (ISS) resides. Meanwhile, measurements show the ambient radiation dose on the lunar surface is more than twice as high as on the ISS and nearly 200 times as high as on Earth, which is a matter of concern for long-term human presence on the Moon.
All of these reasons are why everyone was surprised when LND data revealed Earth’s magnetic protection extends far beyond the magnetosphere and at an angle to the line connecting Earth and the Sun. Lead author Wensai Shang of Shandong University in Weihai, China, worked out that the angle corresponds to the twisting of the Sun’s magnetic field.
“As the Sun rotates, it pulls the solar wind along the solar magnetic field,” Wimmer-Schweingruber said. “That produces a spiral.” Apparently, an unanticipated interaction between this twist in the solar magnetic field and Earth’s magnetic field produces the cosmic ray shelter revealed by LND.
Wimmer-Schweingruber noted that he was extremely skeptical that such results were possible at first. He warned Shang, a graduate student he worked with, that he might be wasting his time looking for cosmic ray anomalies in the Chang’e-4 data. It was only after Shang provided ironclad analyses ruling out other possibilities that he was swayed.
With the LND instrument shut off, researchers need other sources of data to continue the work. Wimmer-Schweingruber expressed particular interest in understanding how cosmic rays produce secondary radiation—especially neutrons, which are very dangerous to humans—when they impact the lunar soil. In the meantime, the general understanding of the radiation environment provided by Chang’e-4 shows we still have some surprises in store as humans explore the solar system.
Citation: Francis, M. R. (2026), Moon mission data reveal unexpected cosmic ray “shadow,” Eos, 107, https://doi.org/10.1029/2026EO260137. Published on 4 May 2026.
Login
