Astrobites RSS

Illustration of two bright blue bodies colliding and emitting jets of matter in the process.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: Continued radio observations of GW170817 3.5 years post-merger
Authors: Arvind Balasubramanian et al.
First Author’s Institution: Texas Tech University
Status: Published in ApJL

More than 3 years since GW170817, astronomers have reported the latest updates from the post-merger kilonova, as seen through X-ray and radio telescopes.

Illustration of a radio overlaid on top of an image of colliding neutron stars.

What can we learn from the radio emission produced years after two neutron stars collide? [University of Warwick/Mark Garlick/ESO; modified by Sumeet Kulkarni]

On August 17, 2017, astronomers worldwide sprang to their feet following an alert from LIGO and Virgo. These gravitational-wave observatories had already made a series of groundbreaking detections involving pairs of black holes plunging into each other, but this new trigger was what their astronomer colleagues had been waiting for: a merger involving two neutron stars. Unlike binary black hole mergers, this event, labelled GW170817, was expected to light up and provide signatures of an explosion — a kilonova — in the electromagnetic (EM) spectrum. Indeed, just 1.3 seconds after the LIGO/Virgo trigger, the Fermi and Swift space-based observatories recorded a short-duration gamma ray burst (sGRB). What followed was a frantic hunt using optical telescopes, and the EM counterpart was finally located in the galaxy NGC 4993.

The wealth of science that came out of this multi-messenger observation was immense — including explaining how sGRBs occur, exploring the nucleosynthesis of heavy elements such as gold and platinum, and verifying that gravitational waves travel close to the speed of light.

Anatomy of a Kilonova

As the first-ever kilonova observed in association with gravitational waves, GW170817 also helped us understand the astrophysical processes that emit radiation in different parts of the EM spectrum after a neutron star merger. Within 24 hours of the merger, early light in the optical and UV bands showed signs of emission from the radioactive decay of heavy elements from the tidal tails of the disrupting neutron stars. This optical and UV light dimmed out within a couple of days, and was followed by a brightening of the signal in X-rays a week later. Combined with radio emission that emerged two weeks after the merger, this afterglow indicated that matter was ejected from the merger as a structured jet, where the velocity of the ejected material varied away from the jet axis.

Soon after the initial observations, astronomers were able to confirm or constrain various structured-jet kilonova models and predict how the EM radiation — particularly in radio and X-rays — would evolve over time. This is shown as the solid black line in Figure 1, below.

Plot showing the flux density of different wavelengths for the kilonova over time. The final few data points deviate from the model.

Figure 1: X-ray and radio observations of the flux density of the kilonova emission over time. The predicted evolution of the radiation is shown by the solid black line, and all observations confirmed this — until now. The kilonova emission has recently seen a re-brightening in X-ray emission (purple points), while most sensitive radio observations reported in this paper (yellow points) do not see a corresponding increase. [Adapted from Balasubramanian et al. 2021]

What New X-ray Scans Show

Recent X-ray observations (purple data points in Figure 1) show evidence of signals in excess of the afterglow predicted by the structured-jet model. While the X-ray emission used in the model is due to ejected particles moving at relativistic speeds (speeds close to the speed of light), this re-brightening could be a signature of ejecta moving at non-relativistic speeds interacting with the surrounding interstellar medium. Alternatively, it could be the initial sGRB that was seen a few seconds after the merger, scattering off interstellar dust! The uncertainties in the X-ray measurements are too large to definitively say which of these is correct, but continued combined observations in both the X-ray and radio spectra can help us figure it out.

Results from ‘Deep’ Radio Observations Using VLA

Today’s authors follow-up the new kilonova observations in the radio spectrum, by reporting the latest set of ‘deep’ observations made using the Very Large Array (VLA) of radio telescopes. They increased the sensitivity of of the search by increasing the observing time up to 32 hours. Previous observations of GW170817 were ‘shallow’, taken only over a duration of a few hours at a time.

The new radio observations (yellow points in Figure 1) show no radiation in excess to what is expected from the structured-jet afterglow model (black line). The radio emission is still following the model and not showing the flattening observed in X-rays, and astronomers are trying to find out why. If the X-rays are just a back-scatter from previous emission, re-brightening is not expected to be seen in radio. But if the new X-ray emission is from a transition to slower, non-relativistic particles, the radio emission should follow suit; whether it happens now or is delayed by a period of time remains to be seen.

More than 3.5 years since the neutron stars first chirped in gravitational waves, the ensuing fireworks continue to excite us!

Original astrobite edited by Roan Haggar.

About the author, Sumeet Kulkarni:

I’m a third-year PhD candidate at the University of Mississippi. My research revolves around various aspects of gravitational wave astrophysics as well as noise characterization of the LIGO detectors. It involves a lot of coding, and I like to keep tapping my fingers on a keyboard even in my spare time, creating tunes instead of bugs. I run a science cafe featuring monthly public talks for the local community here in Oxford, MS, and I also love writing popular science articles. My other interests include reading, cooking, cats, and coffee.

Illustration of a star enshrouded by material and emitting jets from its poles.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: Zooming into the Collimation Zone in a Massive Protostellar Jet
Authors: Carlos Carrasco-González et al.
First Author’s Institution: UNAM: Radio Astronomy and Astrophysics Institute (IRyA-UNAM), Mexico
Status: Published in ApJL

massive star formation

This Hubble image reveals N159, a nursery for massive star formation within one of the Milky Way’s satellite galaxies, the Large Magellanic Cloud (LMC). [ESA/Hubble & NASA]

Stars do not form quietly. As clumps of gas collapse in on themselves within the densest, coldest depths of interstellar molecular clouds, the precursors to stars — protostars accrete mass from their surroundings, but they also launch matter away at incredible speeds (up to hundreds of kilometers per second!) in beam-like or conical formations of interstellar wind. These protostellar jets, also known as outflows or bipolar winds, are powerful influences on the surrounding interstellar medium, and they are thought to be powered by the interaction of the matter falling onto the protostar and the magnetic fields surrounding the protostar. The exact nature of how protostellar jets are launched remains an area of active research. A particularly elusive mystery is whether the physical mechanism responsible for launching these jets might act differently depending on the mass of the protostar. Today’s paper takes the closest look yet at the outflows from a massive protostar, a critical step in understanding the intricacies of the early stages of star formation.

Low Mass vs. High Mass Jets

Studying the origin of protostellar jets is really tricky. These jets travel enormous distances and can affect their surroundings up to parsecs away, but if we want to understand the physics driving them, we must carefully observe the region where they originate, in close proximity to their parent protostar. Unfortunately, even with the most powerful radio telescopes, we can’t observe down to those kinds of physical scales for all but a handful of nearby star-forming regions. Recent high-resolution surveys of star-forming regions have revealed many of the details of protostellar outflows around the precursors of lower-mass stars (M < 8 solar masses or so), as described in this recent bite. More massive O and B type stars, however, are considerably more rare — which means there are very few examples of massive protostars nearby. The further away the protostar is, the better the angular resolution required to resolve the fine details that tell us how these jets are launched.

However, the outflows from massive (O/B type) protostars are thought to be notably different from the winds from their low-mass counterparts. It seems the the outflows from massive protostars are commonly less beam-like, (or collimated) and it is thought that they might have an entirely different physical mechanism responsible for the large-scale parallel structure of their jets. Instead of the protostar’s local magnetic field being responsible for the beam-like linearity of the outflow, it’s possible that massive stars eject mass wildly in nearly all directions, and the ambient magnetic fields of the protostar’s surroundings are responsible for collimating the beam. It’s impossible to determine the truth of the matter without high-resolution observations of the immediate surroundings of massive protostars.

Multipanel plot showing the observations of the innermost regions of Cepheus A HW2's jet, as well as a model that breaks down different components of the jet, including a conical outflow and a collimated one.

Figure 1: The authors’ modeling of the innermost hundred au of the massive protostellar jet Cepheus A HW2. Upper left: the VLA radio image of the protostellar jet. The right three panels show the piecewise construction of the outflow model for a collimated jet, a conical wind, and the combination of both. The bottom left panel shows the model adjusted to fit the outflow angles and mass-loss rates of the observed source. [Carrasco-González et al. 2021]

The Closest Look Yet…

Comparison of Cepheus A HW2 observations vs. a model.

Figure 2: A comparison of the VLA image of Cepheus A HW2 (upper) with a cartoon schematic showing how a disordered distribution of protostellar winds might be collimated into a beam-like jet of outflowing mass. [Carrasco-González et al. 2021]

Today’s paper takes the highest-resolution look yet at the massive protostar Cepheus A HW2, one of the nearest massive protostars with a known outflow. Using the Very Large Array (VLA), the authors are able to resolve the inner workings of the protostellar jet’s origin, on scales down to 20 astronomical units (au). The inner 100 au of this protostar has some key differences relative to the morphology of its low-mass counterparts (as visualized in Figure 1, above). By modeling the observed jets, the authors characterize the outflow as having both a cone-like component nearby to the protostar as well as a collimated, beam-like component that kicks in further out.

The authors suggest a couple interpretations for this fascinating system. Firstly, it’s possible that the same physics is responsible for launching highly collimated jets for both high- and low-mass stars, but the high-mass stars tend to become collimated further away from the protostar. Secondly, it might be that the high-mass protostars produce more disordered winds on their own, blowing away mass in wide cones or even spherically, and the magnetic ambient environment is responsible for turning the cone-like winds into a nice beam-like jet. Since so many massive protostars seem to have disordered outflows, it might be that a particularly opportune magnetic field structure in the surrounding cloud is needed to produce collimated jets so commonly seen for their low-mass protostars. While this is only one example of such an outflow, it brings us one big step closer to understanding the mysterious and elusive jets from massive protostars.

Original astrobite edited by Mitchell Cavanagh.

About the author, H Perry Hatchfield:

I’m a PhD candidate in Physics at the University of Connecticut, where I study star formation and gas structure in the Milky Way’s Galactic Center. I do this using radio observations of molecular clouds as well as hydrodynamic simulations, and I’m all about trying to find ways to compare these two exciting means of exploring the universe.

Photograph of a towering molecular cloud with streams of dust extending from cloud columns.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: A Revised Description of the Cosmic Ray-Induced Desorption of Interstellar Ices
Authors: Olli Sipilä, Kedron Silsbee, Paola Caselli
First Author’s Institution: Max Planck Institute for Extraterrestrial Physics, Germany
Status: Accepted to ApJ

Cosmic Ray Desorption

To first understand the origins of life, we must first understand the origins of life-sustaining molecules. Complex Organic Molecules (COMs) are a common group of carbon-based species that are often considered to be the first steps of sustaining molecules. COMs have been observed in the gas phase in space, but chemical models indicate that COMs have to form in the ice layers of dust grains, rather than in the gas. A common question astronomers seek to answer is: How do molecules, like COMs, go from the ice phase to the gas phase?

Diagram illustrating various behaviors of molecules and atoms on grain surfaces.

Figure 1: On grain surfaces, molecules and atoms can move around (diffuse), react, stick to the surface (accrete), and desorb into the gas phase. [van Dishoeck 2014]

This process, known as desorption (Figure 1), can occur several different ways, but for today’s bite we’ll focus on cosmic ray desorption. Cosmic rays are high energy particles (mostly hydrogen and helium) that can strike dust grains and deposit energy. The deposited energy is then converted to thermal energy, thus heating the grain. Dust grains will then shed the heat to “kick off” molecules (like COMs) from the ice into the gas phase. The rate at which desorption occurs can be directly measured by the grain cooling time, which is determined by the composition of the ice on the grain. If the ice is made of species that strongly bind to the grain (known as a high binding energy), then it takes longer for the grain to cool. Oppositely, if the ice is made of species that loosely bind to the grain (have low binding energies), then cooling and desorption occur more quickly.

Modeling Grain Heating

Two diagrams illustrating the two-phase and three-phase chemical models.

Figure 2: (a): Two-phase chemical model. For cosmic ray calculations, the ice is typically assumed to be CO. (b) Three-phase chemical model consisting of two layers of ice, where mixing between the mantle and core can occur. This allows for a diverse (and more realistic) ice mixture. [Astrobites]

Cosmic ray desorption was first modeled by Hasegawa & Herbst (1993), and their technique and assumptions are still widely adopted by computational astrochemists. This model used a two-phase chemical model consisting of a gas and ice phase (Figure 2a), with the assumption that the outer layer of ice is completely made of carbon monoxide (CO). Because of this assumption, the model uses a single binding energy, and it therefore has a single grain cooling time and a constant desorption rate of one molecule every 10-5 seconds.

Today’s paper challenges the assumption that ice layers are dominantly CO and introduces a dynamic cool-down rate that is dependent on the ice composition. The authors seek to understand whether a more realistic representation of ice could affect the rate at which molecules desorb to the gas phase, or whether the previous assumptions instead hold.

To test this, today’s authors ran 4 different models, including a two-phase and three-phase fiducial (Figure 2b) model based on Hasegawa & Herbst (1993) and a two-phase and three-phase dynamic model, with a variable grain cooling rate. The ice layers used in these models included species commonly known to exist in the ice layers on dust grains, not just CO. For example, their model included carbon, nitrogen, and oxygen, as well as ammonia, methane, and water, among other species.

Plot of the cooling rates for 4 different models.

Figure 3: This plot shows the different cool-down rates (y-axis) for the 4 models over time (x-axis). Since the dynamic models have variable cool-down rates, the cool-down rates change over time. Note that the fiducial models (F2 and F3) have constant cooling rates, the two-phase dynamic model is faster (D2), and the three-phase dynamic model is slower (D3). [Sipilä et al. 2021]

Cooling Down

As expected, the fiducial models yielded a constant cool-down rate of 10-5 sec. However, the two-phase dynamic model had faster cooling rates, while the three-phase dynamic model had slower cooling rates (Figure 3). The different cooling rates are reflected in the desorption rates of different molecules, where the three-phase fiducial model typically had the slowest desorption rates (Figure 4).

The dynamic models yielded different results from the fiducial models because molecules and atoms frozen out are not “stuck” in place in the dynamic models. In reality, species frozen on grains can migrate, which causes a dynamic and variable set of binding energies to be used in the grain cooling rate calculation. Because of this, the dynamic models produce variable cooling rates compared to the fiducial model, which assumes that the ice is stagnant.

The two-phase fiducial model yields faster cooling times, likely because species with low binding energies, like CO, dominate the ice layer. The three-phase fiducial model, with two layers of ice, allows for species with higher binding energies, like ammonia, to slow down the grain cooling rate.

5-panel plot showing desorption rates for 5 different species in each of the 4 different models.

Figure 4: This figure shows the desorption rates of the indicated species over time, for each of 4 models (defined in Figure 3). Note how each species desorbs at different rates between each model. [Adapted from Sipilä et al. 2021]

While we observe a diverse number of chemical species in the gas phase, the majority of these species, such as COMs, form in the ice layers of dust grains. To best understand the past of life-sustaining molecules, we must first understand how complex and diverse ices are desorbed into the gas. Today’s paper introduces a new, easy to apply dynamic cosmic ray desorption that is likely more accurate to real astronomical conditions.

Original astrobite edited by Jason Hinkle.

About the author, Abygail Waggoner:

I am a second year chemistry graduate student at the University of Virginia and NSF graduate fellow. I study time variable chemistry in protoplanetary disks. When I’m not nerding out about space, I’m nerding out about fantasy by reading or playing games like dungeons and dragons.

Illustration of a large, watery planet in the foreground and an additional 5 planets in the background around a bright star.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: K2-138 g: Spitzer Spots a Sixth Planet for the Citizen Science System
Authors: Kevin K. Hardegree-Ullman et al.
First Author’s Institution: Caltech / IPAC-NExScI
Status: Published in AJ

What do you get when you have five exoplanets that sing and add a sixth? Why, the K2-138 system of course!

Discovered in 2018 through the Exoplanet Explorers program, K2-138 was the first system found by citizen scientists with K2, the extension mission of the original Kepler program. By spotting regular dips in K2 light curves, the citizen scientists were able to find four sub-Neptune exoplanets, with an additional super-Earth discovered after further analysis. All the planets were found to be in a near perfect 3:2 resonance chain, meaning their orbital periods follow successive ratios of each other — as discussed in this Astrobite.

But, the K2-138 system had more to offer! The analysis that identified the super-Earth also spotted two additional dips in the K2 light curves, roughly 42 days apart. Dips like these, shown by the dark blue lines, and the letter g, in Figure 1, indicate that a sixth planet might transit K2-138, waiting to be confirmed by the authors of today’s paper.

Plot showing the raw light curve for the system K2-138.

Figure 1: The top panel shows the raw K2 light curve, while the bottom panel shows the same light curve flattened to highlight the planetary transits. In both panels transits of each planet in the system are shown with colored lines. The potential planet g is represented by the darkest blue lines. [Adapted from Hardegree-Ullman et al. 2021]

To determine the origins of these mystery dips, the authors used the Spitzer Space Telescope to stare at K2-138 for 11 hours, centered around the predicted transit time of the proposed sixth planet. By fitting the original K2 and new Spitzer data, a clear transit event, shown in Figure 2, was found in the Spitzer observations within an hour of the expected time, confirming the existence of a sixth planet, K2-138g.

Two plots show the folded transit light curves of K2-138.

Figure 2: The transit light curves from the K2 and Spitzer observations. In the left panel, yellow circles and red triangles show each of the two transits seen by K2. In the right panel, grey points show the Spitzer observations. The red circles show the data binned to 20 minute intervals, showing the drop in flux caused by the transiting planet. In both panels, the blue line gives the fitted transit model. [Adapted from Hardegree-Ullman et al. 2021]

Orbiting at over twice the distance of planet f, the sub-Neptune K2-138g is something of a loner compared to its tightly packed siblings. With its 42 day orbit, K2-138g is not only one of the longest period K2 planets found to date, but it also makes K2-138 the K2 system with the most discovered planets yet.

While the transit durations of the two light curves in Figure 2 are nearly identical, the Spitzer data show K2-138g to have a slightly larger transit depth, and hence radius. As the two transit lengths are consistent within one sigma, the authors note that the limited number of data points in the K2 transits mean that any outliers could skew the results, causing the slight discrepancy with Spitzer.

The More the Merrier

While planets b, c, d, and e are in near 3:2 resonance with their respective neighbours, the outer planets f and g are not. Given this fact, along with the sizeable gap in orbital period between f and g, could there be additional planets in the system yet to be discovered? It seems possible. If the pattern of resonances continued beyond planet f, resonant orbits would be expected at periods of around 20 and 30 days, but more observations are needed to confirm whether any such planets exist.

Diagram showing the semimajor axes for planets within 8 different planetary systems.

Figure 3: The orbital spacings of a selection of multiplanet systems, in order of the size of their stellar hosts from largest at the top. Each system is shown by a coloured line with a width corresponding to the size of the host star. Transiting planets are represented by circles scaled to the line width, enlarged 10x for clarity. Non-transiting planets are shown in blue. The large separation between the two outer planets of K2-138 is similar to that seen in the Kepler-11, Kepler-20, HD 219134, and Kepler-80 systems. [Hardegree-Ullman et al. 2021]

K2-138g isn’t unique in its socially distanced orbit, however. Around half of the 9 other exoplanet systems with 6 or more planets also have a large gap between their outermost planets, as seen in Figure 3. While this apparent trend could be the result of planet formation processes, planets at large orbital radii can be harder to detect, so observational biases might be at play.

A Benchmark System

With its tightly packed resonant inner planets and abundance of sub-Neptunes, the authors argue that the K2-138 system is a more than worthy target of follow-up observations. The inner planets provide an excellent opportunity to study their potential transit-timing variations (TTVs) — discrepancies in the regular periods of planets — and observations have already been scheduled. Alongside radial velocity (RV) data, this could enable precise mass measurements and see the potential discovery of additional planets. While the planets have atmospheric signals too small to be studied with the James Webb Space Telescope, they are prime targets for the European Space Agency’s upcoming ARIEL mission. The system’s five sub-Neptunes could provide a key testbed for comparative studies of the atmospheres of a planet category not seen in our solar system.

Whatever the future holds, it certainly seems likely that we’ll be hearing more from K2’s most musical system in the years to come!

Original astrobite edited by Brent Shapiro-Albert.

About the author, Lili Alderson:

Lili Alderson is a first year PhD student at the University of Bristol studying exoplanet atmospheres with space-based telescopes. She spent her undergrad at the University of Southampton with a year in research at the Center for Astrophysics | Harvard-Smithsonian. When not thinking about exoplanets, Lili enjoys ballet, film, and baking.

Illustration of the top portion of a watery exoplanet with a thin atmosphere, with a dim red star in the background.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: How to identify exoplanet surfaces using atmospheric trace species in hydrogen-dominated atmospheres
Authors: Xinting Yu (余馨婷) et al.
First Author’s Institution: University of California Santa Cruz
Status: Published in ApJ

Of the 4,400 (and counting!) exoplanets, the population of intermediate-sized planets is one of the most interesting. With sizes between Earth and Neptune not seen in our solar system, the most commonly occurring type of planet can be a confusing one. A planet in this category could be a giant terrestrial planet, with a solid surface and thin atmosphere (a “super-Earth”), or it may be more like a shrunken down version of the solar system’s ice giants (a “sub-Neptune”), with a surface located deeper within the planet at high-pressure levels, if there is one at all. Even though many intermediate-sized exoplanets have been discovered, the internal structure of any one planet isn’t always clear. Large uncertainties in the masses and radii of these planets, and hence in their densities, can make understanding their precise compositions a challenge, and even the most sensitive upcoming telescopes like JWST and ARIEL cannot directly probe surfaces, leaving many exoplanets in composition limbo.

However, JWST and ARIEL will be capable of precisely measuring the atmospheres of such planets — so what if there was a way to find out how deep the surface of an exoplanet lies by studying its atmosphere? The authors of today’s paper investigate whether there is a relation between the abundances of species found in an exoplanet’s atmosphere and the location of the exoplanet’s surface.

Under Pressure

As it turns out, the existence of a solid surface plays a key role in the makeup of the atmospheres within our own solar system. Both Jupiter and Saturn’s moon Titan contain very little ammonia (NH3) within their upper atmospheres, as it gets destroyed by photochemical reactions that occur there. But while Jupiter contains significant amounts of NH3 deep within its atmosphere, Titan does not. The difference here is Titan’s surface. In Jupiter, the lack of a solid surface means the constituent parts of NH3 are transported into the hot, high-pressure lower atmosphere where they can reform into ammonia via thermochemical reactions, whereas Titan’s surface prevents its atmosphere from reaching high enough temperatures and pressures for the recycling reactions to occur. Titan instead has a larger abundance of nitrogen, left over from the destroyed NH3.

The authors propose that a similar situation could occur with other species within the atmospheres of exoplanets. To test this theory, they modelled the atmospheric evolution of sub-Neptune K2-18b under varying surface assumptions: first with no surface, and then with a surface at one of three different pressure levels.

two diagrams illustrate chemical pathways in the atmosphere of an exoplanet for different surface conditions.

Figure 1: Diagrams describing the main chemical pathways within the atmosphere of K2-18b for a deep surface or no surface (left) and a shallow surface (right). Arrow thickness indicates the importance of each pathway, with dashed arrows being the least important. In both cases, UV photons impacting the upper atmosphere cause photochemical reactions that break down sensitive molecules such as NH3, HCN, H2O, and CH4. In the deep/no surface model, thermochemistry in the deep, hot atmosphere recreates the molecules lost to photochemistry. In the shallow surface case, atmospheric temperatures are never hot enough for thermochemistry to be effective, causing a decrease in abundances of the species in blue compared to the no surface case, and an increase for the red species. [Yu et al. 2021]

Much like within the solar system, if K2-18b has a shallow surface, the atmosphere is never hot enough for thermochemical reactions to take place, meaning the abundances of photochemically fragile species such as ammonia decrease compared to when the surface is much deeper or non-existent, as shown in Figure 1. For each version of the model, the changes in the volume mixing ratios of key chemical species within the observable atmosphere demonstrate the impacts of surfaces at different pressure levels.

four plots showing volume mixing ratios under four different surface conditions

Figure 2: Plots showing how the volume mixing ratios of key chemical species change with pressure through the atmosphere of K2-18b with different surfaces. Higher pressures indicate deeper into the atmosphere. The pale blue shaded region indicates the observable part of the atmosphere. [Yu et al. 2021]

When the planet has no surface or a very deep surface, large amounts of hydrocarbons and nitrites such as hydrogen cyanide (HCN) are produced, while significant quantities of ammonia are found deep in the atmosphere just like in the case of Jupiter. When a surface exists at 10 bars, key nitrogen species can no longer be replenished and produced as easily, leading to decreasing volume mixing ratios for HCN and NH3. For the shallowest, Earth-like surfaces, thermochemistry is prevented for the majority of species, and the atmosphere is now also depleted in water (H2O) and methane (CH4). As Figure 2 shows, changing the presence or depth of a planet’s surface will change the abundances of a whole host of species — but are these changes significant enough to distinguish between surfaces?

A New Tool For Observers?

Using the finding that a variety of species are uniquely sensitive to the presence of different surfaces, the authors are able to use the abundance ratios between a species when a surface is and isn’t present, and between different pairs of species to tentatively outline a way to distinguish where a surface could be.

flowchart diagramming possibilities to identify the pressure depth of the planet surface from species abundance ratios

Figure 3: Flowchart to aid in the possible determination of the pressure level of a surface within an exoplanet similar to K2-18b using the observed abundance ([X]) ratios of different species. [Adapted from Yu et al. 2021]

With the flowchart shown in Figure 3 at hand, it could be possible to determine where surfaces lie within exoplanets, provided accurate abundance measurements are available. Unfortunately, current measurements of K2-18b’s Neptunian atmosphere aren’t precise enough to make a prediction about any potential surface, but upcoming JWST observations of this planet could provide further information.

So, does this mean the mystery of intermediate planet surfaces can finally be resolved? Not completely. More modelling is needed to extend the range of planetary parameters and scenarios. In the future, the flowchart could be expanded to include exciting but less well-studied species such as phosphine (PH3). The current study also does not consider the potential impacts of processes that occur on the surface, such as volcanic activity and reactions with oceans or rocks, or the potential escape of gases from the top of the atmosphere — all processes that could change the observed abundance ratios in an exoplanet. Nevertheless, today’s paper outlines an exciting new concept that extends our toolkit as we continue to try to understand the growing number of strange new worlds waiting to be explored.

Original astrobite edited by Alice Curtin.

About the author, Lili Alderson:

Lili Alderson is a first year PhD student at the University of Bristol studying exoplanet atmospheres with space-based telescopes. She spent her undergrad at the University of Southampton with a year in research at the Center for Astrophysics | Harvard-Smithsonian. When not thinking about exoplanets, Lili enjoys ballet, film, and baking.

VISTA view of the newly discovered globular cluster VVV CL001

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: VVV CL001: Likely the Most Metal-Poor Surviving Globular Cluster in the Inner Galaxy
Authors: José G. Fernández-Trincado et al.
First Author’s Institution: University of Atacama, Chile
Status: Published in ApJL

Globular clusters are dense, spherical collections of hundreds of thousands of stars. They’re some of the oldest and most metal-poor parts of the galaxy. As such, they provide clues to the Milky Way’s past, including the galaxy merger events that may create them. Today’s paper examines the globular cluster VVV CL001 to determine its metal content and possible origins. We’ll look at how the cluster compares to others like it to see if it makes the most metal-poor podium in the Globular Cluster Olympics.

Determining the Athletes

VVV CL001 was discovered by the Vista Variables in the Via Lactea (VVV) Survey, which has increased the number of known globular clusters in the Milky Way to over 300. Think of this like the Olympic trials — this globular cluster and many others were undiscovered until they jumped out and caught the VVV Survey’s eye.

Selecting who makes an Olympic team is a difficult process, and determining which stars belong to a cluster is no different. Today’s authors used the Apache Point Observatory Galactic Evolution Experiment (APOGEE-2) and archival data to obtain spectra for stars near VVV CL001. There are a lot of stars in the field of view (or athletes in the running), so the authors had to separate out members and non-members of the globular cluster using radial velocities. Figure 1, below, shows the radial velocities in the field of view with distance from the estimated center of VVV CL001. The center of the cluster is based on proper motions from the Gaia mission. Gray points show APOGEE-2 targets, mostly near zero. Zooming into the bottom of the figure, the blue points show probable VVV CL001 stars from past studies. Those are like the athletes who have been to the Olympics before — they’re likely to qualify again. The distinct, negative radial velocities compared to other stars in the sample allowed the authors to identify two stars within the APOGEE-2 data that are also very likely members of the cluster. Those are shown in black squares.

Plot of radial velocity vs. distance from VVV CL001.

Figure 1: Radial velocities with distance to the center of VVV CL001. Stars within the cluster have a much lower radial velocity than those outside of it, making them easy to distinguish. [Fernández-Trincado et al. 2021]

Start the Match

The authors of today’s paper use a Markov Chain Monte Carlo (MCMC) method to determine the most likely age and distance of VVV CL001 based on the colors of its member stars. It turns out to be 11.9 Gyr old and 8.22 kpc from the Sun. That makes it an old, inner-galaxy cluster. And although age might not be the best criteria for judging Olympic athletes, old globular clusters are helpful because they let astronomers look back in time at what conditions were like in the galaxy in the past.

The main event in these Olympic Games is determining metal content. In this competition, the most metal-poor globular cluster is going to take home the actual medal. That’s because metal-poor clusters are usually really old and can show us what the very first generations of stars look like. Using model spectra, today’s authors fit the APOGEE-2 data for temperature, gravity, and metal content in the two VVV CL001 stars. Then they calculated the ratios of metal absorption lines in each star, like nitrogen to iron, [N/Fe], and iron to hydrogen, [Fe/H], which is used as the overall value for metal content in a star, or the metallicity.

Present the Me(t)dals

Within the orbit of the Sun, globular clusters have metallicities ranging from –2.37 to 0. Previous studies found ESO280-SC06 to be the most metal poor globular cluster in the entire galaxy, with [Fe/H] = –2.48, so that cluster is the reigning gold medalist. But VVV CL001 might give ESO280-SC06 some good competition this time around! Today’s authors found that the two stars in VVV CL001 have an average [Fe/H] = –2.45 ± 0.24. Based on the large error, it’s still unclear whether VVV CL001 has beaten ESO280-SC06 for the gold medal, but it’s certainly on the podium.

Figure 2 shows a comparison of metal ratios in the two VVV CL001 members (in black) compared to other metal-poor globular clusters (other athletes in various colors). It’s a pretty close race. In general, the clusters are similar, but VVV CL001 has a lower overall metallicity. Also, the big difference in values of [N/Fe] between the two stars in black might suggest that there are multiple populations of stars in VVV CL001, meaning that there have been multiple epochs of star formation within the cluster. It’s like VVV CL001 has some really solid older players, but the team might also include some great new up-and-comers. However, because there are only two data points, this is far from a certain conclusion.

plot showing metal abundances for various globular clusters

Figure 2: Metal abundances in metal-poor globular clusters. There is a slight horizontal offset between points for clarity. VVV CL001 has a low [Fe/H], even compared to other metal-poor clusters. [Fernández-Trincado et al. 2021]

Closing Ceremonies

The authors of today’s paper also simulated possible orbits for VVV CL001, finding that it may have been created from the Sequoia or Gaia–Enceladus–Sausage dwarf galaxies that played a big role in forming the Milky Way halo. They also found that the cluster is twice as massive as previously thought, which makes VVV CL001 one of the most massive, metal-poor globular clusters in the galaxy, and an excellent example of the extreme properties possible in a cluster. As an athlete in the Globular Cluster Olympics, it may not be the clear champion, but it is certainly unique.

Understanding globular clusters and how they are formed will provide clues that track the history of galaxies. As more and more of the clusters are identified and studied in depth by studies like the VVV Survey and APOGEE-2, we will learn more about the Milky Way and may be able to finally award the medals in the Globular Cluster Olympics!

Original astrobite edited by Sumeet Kulkarni.

About the author, Ashley Piccone:

I am a third year PhD student at the University of Wyoming, where I use polarimetry and spectroscopy to study the magnetic field and dust around bowshock nebulae. I love science communication and finding new ways to introduce people to astronomy and physics. In addition to stargazing at the clear Wyoming skies, I also enjoy backpacking, hiking, running and skiing.

Image of a field containing hundreds of visible galaxies of different colors, shapes, and sizes.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: New Determinations of the UV Luminosity Functions from z ~ 9 to z~ 2 Show a Remarkable Consistency with Halo Growth and a Constant Star Formation Efficiency
Authors: R.J. Bouwens, P.A. Oesch, M. Stefanon, et al.
First Author’s Institution: Leiden University, The Netherlands
Status: Submitted to ApJ

For decades, astronomers have used methods of “extragalactic census” to anchor our theoretical understanding of galaxy evolution. The idea is to simply measure the number of galaxies within a volume of the universe as a function of some physical variable (like dark matter halo mass, stellar mass, or luminosity, for instance). When measured at one epoch or point in time, these censuses tell us a great deal about the universe by simply recording how common galaxies of different types are. If this measurement is carried out at various redshifts (z), one obtains a timeline of the universe that encodes the changes in the underlying galaxy population according to the variable in question.

Photograph of a space telescope above the earth.

Photograph of the Hubble Space Telescope. [NASA]

Consider the ultraviolet (UV) luminosities of galaxies. The number of galaxies in a given volume as a function of UV luminosity has the catchy name of the UV luminosity function. The UV luminosity function is a particularly useful probe in galaxy evolution studies, since the UV luminosity of a galaxy is strongly correlated with its star formation rate (newly formed stars are bright in the UV). This means that if we can measure the evolution of the UV luminosity function across the history of the universe, we can investigate the growth of stars within galaxies across cosmic time. Since this extragalactic census is purely observational, it is completely agnostic about the underlying physics of star formation. However, these types of measurements are nonetheless invaluable for theorists. Census-like observations like the UV luminosity function provide bedrock constraints for theories of galaxy evolution, because at the end of the day, any working model of galaxy evolution must be capable of reproducing them. The goal then is to obtain the most precise censuses possible.

In today’s astrobite, we explore a recent effort to obtain the premier measurements of the evolution of the UV luminosity function using observations from the Hubble Space Telescope.

Tackling the UV Luminosity Function with Hubble

This work combines nearly all of the deepest survey observations conducted with the Hubble Space Telescope (HST) to obtain the largest selection of galaxies at z > 2 to date from this facility. Their primary sample contains some 24,000 galaxies, making it more than twice as large as previous selections of galaxies from HST surveys. To measure the evolution of the UV luminosity function, galaxies are first grouped into redshift slices of width 1 in redshift space, from z ~ 2 to z ~ 9 (for example, galaxies within 1.5 < z < 2.5 are grouped together at z ~ 2). To achieve this, the team first uses the well-tested “drop-out” method (see this astrobite for another detailed explanation). A strong spectral feature in star-forming galaxies is the Lyman break at 912 Angstroms, where flux at shorter wavelengths is absorbed by gas in the intergalactic medium. At a given redshift, it is possible to fairly accurately exclude low-z galaxies by selecting only sources with significant flux at wavelengths longer than the Lyman break (their flux “drops out” at shorter wavelengths). For an even cleaner sample, the team also models the spectral energy distributions of the galaxies in their sample, comparing their observed photometry to model galaxy templates. This yields a distribution of possible redshifts for each source and an associated probability, which is used to exclude the color-selected galaxies that may in fact be low-z interlopers.

With the approximate redshift of each galaxy known, it is possible to infer its absolute magnitude and finally measure the UV luminosity function. Before measuring the UV luminosity function, they add to their dataset the constraints from several other efforts, including ground-based observations and a recently selected sample of galaxies at z ~ 10 from another work. The final UV luminosity functions at each redshift slice is shown below in Figure 1, tracking its evolution from z ~ 10, when the universe was 0.48 billion years old, to z ~ 2, when the universe was 3.3 billion years old.

plot of the number of galaxies in a given volume vs. the UV luminosity

Figure 1: The UV luminosity function measured in this paper at each redshift between 2 < z < 9, plus that measured at z ~ 10 in another work (Oesch et al. 2018). Across this ~ 3 billion year time period, the number density of all galaxies increases by orders of magnitude, with fainter galaxies becoming less dominant over time. [Bouwens et al. 2021]

Extracting Extragalactic Information

As can be seen from Figure 1, interpreting the changes in the UV luminosity function between 2 < z < 10 is complicated. To simplify the matter, the authors model the UV luminosity function at each redshift with a parameterization known as a “Schechter function.” This model combines a power law at the faint end and an exponential decline on the bright end. It has three free parameters: α, the slope of the faint-end power law; M*, the absolute magnitude where the slope changes from power-law behavior to exponential; and φ, the normalization. The Schechter function fits to the data are also shown in Figure 1.

three plots showing the evolution with redshift of the three schechter function parameters.

Figure 2: The evolution of the Schechter function parameters as measured in this work. The remarkable result presented in this paper is demonstrated in the top panel: The evolution of alpha in the UV luminosity function almost matches perfectly with the evolution of the low-mass end of the dark matter halo mass function, supporting a fundamental link between dark matter halo mass and star formation in galaxies. [Bouwens et al. 2021]

The evolution of each of the Schechter parameters in the UV luminosity function tells us something physical about star formation in galaxies throughout cosmic time. The measured trends in these parameters are shown in Figure 2. As can be seen, the faint-end slope (α) follows a strongly linear relationship with z and becomes a weaker power law over time, indicating that brighter galaxies become more common from early times to later times. The turnover magnitude (M*) stays roughly constant until a dramatic change at z ~ 2.5, where it is thought that the most massive and bright galaxies begin to quench and fade in the UV. Finally, the normalization (φ) smoothly increases across cosmic time and flattens out near z = 0, indicating a smooth growth in the overall number density of galaxies.

What is most interesting, the authors note, is that the faint-end slope evolution almost identically matches the evolution of the dark-matter (DM) halo mass function (the number of galaxies per unit volume as a function of DM halo mass). In fact, if one assumes a constant efficiency of turning gas into stars within a given DM halo, and applies this assumption to the dark-matter halo mass function, the trends shown in Figure 1 are almost exactly reproduced. This supports an intimate link between the DM halo mass of galaxies and their ability to form stars.

By combining all of the existing survey fields observed by HST, the authors obtained a precise characterization of the UV luminosity function across a dramatic three billion years of our universe’s history, from 2 < z < 10. With the imminent launches of the Roman Space Telescope (and its enormous field of view, 100x that of HST!) and the James Webb Space Telescope (with its extreme infrared sensitivity), we can be confident that our understanding of the universe will only grow from here.

Original astrobite edited by Alex Pizzuto.

About the author, Lukas Zalesky:

I am a PhD student at University of Hawaii’s Institute for Astronomy. I am interested in understanding the way galaxies form and evolve over billions of years, as well as gravitational lensing by galaxy clusters. Outside of research I spend my time with animals, exercising, practicing Zen, and exploring the beautiful island of Oahu.

Two images — one wide-field and one zoomed in — show a small, compact dwarf galaxy.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: A recently quenched isolated dwarf galaxy outside of the Local Group environment
Authors: Ava Polzin, Pieter van Dokkum, Shany Danieli, Johnny P. Greco and Aaron J. Romanowsky
First Author’s Institution: Yale University
Status: Accepted to ApJL

Earth Orbit — Hubble

A mock-up of a daily newspaper front page named "Astrobites News". The headline of the article is "Teenage Galaxy Eludes Authorities" and an image of a galaxy is captioned "COSMOS-dw1 eluded authorities for far too long." A teen galaxy, going by the name COSMOS-dw1, refuses to testify to authorities (i.e., scientists) about its nature or, for that matter, about anything much at all.

The young specimen has been taken into custody (i.e., was discovered) recently and created more questions than it answered. Specifically, it appears to have quenched all star formation processes, feeding once again into unfounded prejudice about young “slacker” galaxies. Why it has done so, it refuses to imply directly. One would naturally assume that such staggering behavior must be a direct consequence of the environment of such a young, impressionable galaxy. But authorities were stunned when they learned that COSMOS-dw1 exists in isolation, acting as a lone wolf. An attempt to identify possible accomplices is underway, however none make convincing suspects thus far. Further investigation will be necessary to persuade the galaxy to cooperate and for us to learn more about this puzzling behavior.

An Addition to the Collection

Alright, this fictional newspaper snippet may simplify things a little. But it does provide a nice overview on the newly discovered young dwarf galaxy COSMOS-dw1, a (most likely) quenched galaxy outside of the Local Group, found by a stroke of luck. A galaxy is considered quenched if it has shut down all star formation. These small galaxies are fairly underrepresented in contemporary surveys, due to the bias of larger galaxies with high surface brightnesses being much easier to find.

In the context of galaxies, “small” or “low mass” refers to galaxies with masses below 10solar masses (the Milky Way has roughly 1012 solar masses). The galaxy discussed in today’s paper is named COSMOS-dw1 and was discovered in archival Hubble Space Telescope data in the so-called COSMOS-CANDLES field, one of the most observed regions of the sky. It showed up as a semi-resolved object within the data, which suggests that it cannot be too far away.

But COSMOS-dw1 is not only interesting because it adds to our short list of well-documented low-mass galaxies, but also due to a number of further notable features: it displays a rather asymmetric shape with a clump of blue stars (which are hotter, heavier, and much more short-lived than yellow or red stars) off-center to the north (see Figure 1). The rest of the stellar population appears to be much older — most likely post-main sequence — and is distributed more evenly.

Two-panel figure shows a color-magnitude diagram on left, with some data points in blue, and an image of a small, dense galaxy on the right.

Figure 1: Color–magnitude diagram of COSMOS-dw1 (left). The blue points refer to the bright blue stars marked in the RGB image of the galaxy (right). [Adapted from Polzin et al. 2021]

Missing Lines

In order to learn more about the galaxy, spectroscopic observations were taken by the authors with the Low Resolution Imaging Spectrograph (LRIS) on the Keck I telescope in Hawaii. A distance of around 22 Mpc was deduced from the measurements, and the galaxy’s radius was determined to be around 450 pc (around 127 times smaller than the Milky Way), with a mass of 2.4 x 106 solar masses.

Arguably the most important information obtained from the spectroscopy measurements is the absence of emission lines, specifically H-alpha emission lines. In galaxy research, the brightness of the H-alpha line indicates the number of massive stars within that galaxy, which can be used to infer the rate of star formation. Massive stars ionize the gas around them. The hydrogen recombination produces line emission, such as the well-known Balmer-series lines of H-alpha and H-beta. The lack of H alpha emission suggests that the galaxy is in fact quenched, i.e., it does not produce any new stars at the moment. However, strong Balmer absorption is visible, implying the presence of A-type stars at an age of roughly 1 Gyr. Figure 2 displays the spectrum of COSMOS-dw1 and the location of where H-alpha emission would be expected, as a function of wavelength.

Spectrum for COSMOS-dw1, showing several spectral lines but no evidence of H-alpha.

Figure 2: LRIS spectrum of COSMOS-dw1 in black with the best-fit model overplotted in pink. The gray area on the right shows the wavelength range where H-alpha emission would be expected. [Adapted from Polzin et al. 2021]

From the color–magnitude diagram (seen in Figure 1) it is evident that within COSMOS-dw1, a population of bright, very blue stars exists within a rather complex stellar population. The location of these stars provides an upper limit for the age of the galaxy. As stars age, they disperse throughout their host galaxy. Since these blue stars are quite close to the center, it can be inferred that the galaxy is young.

A Lone Wolf

So, why has this young galaxy ceased all star formation, at least for the moment? The authors point out that usually dwarf galaxies are quenched due to environmental effects, such as ram pressure — the process of pressure from the environment stripping a small galaxy of the gas essential for star formation as the galaxy falls towards a larger mass.

For any kind of environmentally caused quenching, one would expect to see a bright companion galaxy close to the dwarf galaxy. Intriguingly, such a companion is nowhere to be found for COSMOS-dw1. The authors have searched the immediate surroundings of the dwarf galaxy and found only two galaxies that exceed the minimum mass to be considered a “luminous neighbor”. However, it appears these are too far away to be responsible for the quenching. Additionally, the complex stellar population within COSMOS-dw1 suggests that star formation started and stopped several times in the past. We also know that the quenching happened rather recently due to the population of rather young stars.

Supernovae to the Rescue!

The authors propose a different mechanism responsible for quenching: internal feedback from supernovae. This basically means that a violent process such as a supernova may inject energy and momentum into the interstellar medium — enough maybe to shut off star formation, at least temporarily. The clump of blue stars found within COSMOS-dw1 may be the location of this feedback event.

More surveys may help us to better understand these dwarf galaxies. The authors note that finding this quenched small galaxy in such a well-studied field in the sky suggests that they are fairly common.

While we are quite calmly and comfortably existing in a side arm of a spiral galaxy that does not currently experience any violence, it is exceedingly interesting to take a look at other galaxies, especially outside the Local Group, to gain a perspective on the many altering processes happening to other galaxies and the star formation within. Space is always dynamic; even if an object such as COSMOS-dw1 exists in isolation, rapid exchange of energy and great change can always come from within.

Original astrobite edited by Alex Pizzuto.

About the author, Jana Steuer:

I’m a second year PhD student at the LMU Munich, working for the University Observatory (USM), which owns the 2.1m Fraunhofer Telescope Wendelstein. My field of research is exoplanets. I hunt for traces of them in data from big surveys, like the TESS mission and then follow them up, using spectroscopy and photometry. Mainly, I focus on long period planets that may potentially harbor life. When I’m not planet hunting, I act as a DM for several Dungeons and Dragons groups and annoy people with facts from Tolkien’s Silmarillion. I enjoy kickboxing and learning about ancient human history.

Simulation still showing the formation of the cosmic web

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: Cosmic Velocity Field Reconstruction Using AI
Authors: Ziyong Wu et al.
First Author’s Institution: Sun Yat-Sen University, China
Status: Published in ApJ

Going with the (Hubble) Flow?

Hubble’s law is a beautifully simple statement: a galaxy caught in the Hubble flow, moving with the expansion of the universe, should be traveling away from us at a speed proportional to its distance. Unfortunately, however, this velocity–distance relation is too good to be true: due to the pesky influence of gravity, Hubble’s law is invalid in the vast majority of cases. In general, a galaxy’s net motion can be attributed to a combination of the Hubble flow, the galaxy’s motion within its galaxy cluster or group, and the motion of the cluster or group itself. We collectively refer to these deviations from the Hubble flow as “peculiar motions” or “peculiar velocities.”

While the presence of peculiar motions spoils the simplicity of Hubble’s law, these motions can be a blessing in disguise: since diversions from the Hubble flow are caused by gravitational interactions — and therefore by the presence of matter —  peculiar motions serve as excellent probes for the physics of structure in the universe. Peculiar velocities have been used to map the cosmic web — the vast network of filaments connecting matter on the universe’s largest scales (explored further here, here, and here) — and are linked to the dynamics of galaxy clusters and the cosmic microwave background via the kinematic Sunyaev–Zel’dovich effect. Peculiar motions are also the root cause of redshift–space distortions, and thus one requires precision measurements of peculiar velocities in order to test cosmological models using the Alcock–Paczynski effect (see here and here for deeper explanations of this technique).

One caveat, though: measuring peculiar velocities is hard. To decouple peculiar motions from the Hubble flow observationally, we need a means of measuring distances that doesn’t require redshifts. To this end, a distance ladder or the Tully–Fisher and Faber–Jackson relations are viable methods, but each carry significant measurement uncertainty. Alternatively, we can take a theoretical approach, using perturbation theory to infer cosmic velocities from cosmic density data. However, any attempts to fully model the nonlinear growth of large-scale structure by hand quickly become prohibitively complex, necessitating a number of approximations and simplifications. How, then, can we accurately and efficiently compute peculiar velocities on cosmological scales? The authors of today’s paper may have found a solution in the field of machine learning: convolutional neural networks.

From Convolutions to Cosmology

Artificial neural networks are, in essence, models with very many free parameters. As one trains the neural network by feeding it many input data sets and scoring its output against the expected results, the network adjusts its parameters, thus learning how best to map the given inputs to the desired outputs. Figure 1 shows a simple neural net with a fully connected three-layer “feed-forward” architecture; the data, in the form of an array of real numbers, is reprocessed as it’s transmitted from the “input layer” to a “hidden layer” and finally to the “output” layer. Each connection between layers bears a weight that dictates how a layer’s “neurons” should process their inputs — these weights are the free parameters in the neural network. Ultimately, neural nets produce models that are highly nonlinear, thus making them ideal for studying the complex dynamics of cosmic structure formation.

Diagram of an interconnected group of nodes

Figure 1: A schematic diagram of a fully connected three-layer feed-forward neural network, where each circle represents a neuron. Here, the data is fed into the input layer as an array, then transmitted to the hidden layer where it is mixed and reprocessed based on the weights of the connections leading into the hidden layer; the resulting values are sent to the output layer, where they are reprocessed one final time, ultimately producing a highly nonlinear model. []

Typically, neural networks contain many hidden layers, and thus possess an obscene number of parameters — in this paper, the authors use a network with 48,690,307 parameters! With this many parameters, neural nets run the risk of overfitting the data, using up a large amount of memory, and running extremely slowly. Fortunately, one can ameliorate these issues by adding one or more “convolution” layers to a network, filtering and contracting the data and preserving only the most salient features (for a more thorough explanation of this convolution process, see here); this is especially useful when processing detailed image data, such as the cosmic density maps that the authors use as their input data. The authors optimize their network by adopting a U-Net architecture, which employs a series of convolutions followed by a series of deconvolutions to quickly parse the input and highlight its key components.

To generate their training and testing data sets, the authors simulate the formation of large-scale structure up to the present day, retrieving both cosmic density and momentum maps; the density maps are used as inputs to the neural net, while the corresponding velocity maps — computed by dividing the momentum fields by the density fields — are used to evaluate the neural net’s output and to subsequently train, cross-validate, and test the resulting model.

Math vs. Machine

The authors assess the performance of their trained neural network by comparing its peculiar velocity predictions to those of linear perturbation theory. In nearly all cases, the neural net clearly outperforms the theoretical model. Perturbation theory performs well in regions of low density and velocity, occasionally yielding better predictions than the neural net. However, in regions of high density and velocity and in merger situations where two regions of opposing velocity collide with one another, perturbation theory fails completely, while the neural net still faithfully reconstructs the velocity field (see Figure 2). Over multiple testing data sets, the neural net is shown to be robust in all situations, while perturbation theory becomes practically useless in the presence of nonlinear dynamics.

six panel plot evaluating the neural net results

Figure 2: Comparison of a simulated velocity field (upper left) with a field predicted by the neural network (upper middle) and by perturbation theory (upper right); the lower left shows the underlying density field, while the lower middle and lower right show the residuals for the neural net predictions and the perturbation theory predictions, respectively. In regions of high density and velocity and in regions of converging flow, perturbation theory breaks down. [Wu et al. 2021]

While the neural network used in this paper can definitely be improved — perhaps by further optimizing its architecture or by using more training data — the authors have shown that neural nets can be valuable tools for predicting peculiar velocities. With such programs as DESI, EUCLID, the Rubin Observatory, and the Nancy Grace Roman Space Telescope promising to map out an unprecedented volume of the cosmos within the next decade, it is of utmost importance that we possess fast and accurate methods for parsing the new data — and neural networks are surely at the forefront of these methods. Maybe the rise of machines isn’t such a bad thing after all!

Original astrobite edited by Pratik Gandhi.

About the author, Ryan Golant:

I am a first-year astronomy Ph.D. student at Columbia University. My current research involves the use of particle-in-cell (PIC) simulations to study magnetic field growth in gamma-ray burst afterglows and closely related plasmas. I completed my undergraduate at Princeton University, and am originally from Northern Virginia. Outside of astronomy, I enjoy playing violin, studying art history, reading Wikipedia, and watching cat videos.

Spitzer photograph of a dramatic nebula surrounding bright point sources.

Editor’s note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. As part of the partnership between the AAS and astrobites, we occasionally repost astrobites content here at AAS Nova. We hope you enjoy this post from astrobites; the original can be viewed at

Title: The consistency of chemical clocks among coeval stars
Authors: Francisca Espinoza-Rojas et al.
First Author’s Institution: Pontifical Catholic University of Chile
Status: Submitted to ApJ

Stellar age is an extremely valuable parameter to constrain because it introduces time into our study of astronomical objects. Pairing the observed properties of stars with time opens up a rich new dimension in the study of our galaxy and beyond. For example, when we pair stellar age with stellar kinematics, we can dynamically trace stars back to their birth locations to study things like galactic evolution and star formation in detail. When we consider stellar age in our study of exoplanets, we can peer into the planet formation and evolution process. When we pair stellar age with stellar chemical abundances, we can trace the evolution of specific elements over time in the galaxy. Weaving time into these various analyses opens up a new realm of insight that enhances our understanding of the universe. However, with this all said, stellar age is extremely difficult to constrain.

Stellar Ages Are Hard to Determine

Some methods of constraining stellar ages include using photometry, dynamics, gyrochronology, and the abundances of individual elements like lithium in stars. For example, the locations of stars on the color–magnitude diagram (CMD), which are determined by photometry, can hint at stellar age. Many stellar and galactic astronomers fit isochrones, lines of constant age in the CMD, to the photometric data of a single or group of stars to estimate their age. However, this method relies on very well-constrained dust parameters between the observer and the object. Gyrochronology, using stellar rotation to estimate age, is another effective method, but it requires knowledge of the inclination of the star, something that is often difficult to determine. We can also use lithium abundances to estimate stellar age. Lithium, however, is only an effective age indicator in young stars with convective envelopes. As you can probably tell, there are tons of ways to estimate stellar age, but they all suffer from various limitations and uncertainties.

Abundance Ratios of Certain Elements Track with Age

An interesting, and somewhat new, avenue for probing stellar age is through the use of chemical clocks. Chemical clocks are sets of elemental abundance ratios that have been observed to track with stellar age. The idea behind chemical clocks is rooted in the notion that different families of elements are expelled into the interstellar medium (ISM) on different time scales (see Figure 1). For example, elements like Mg, Al, and Ti are produced in dying massive stars, which live short lives that end in core-collapse supernovae. As a result, these elements follow very different timescales than, say, Ba and Y — elements that are produced primarily in low-mass stars, which have much longer lifetimes and subsequently take longer to spread their nucleosynthetic products out into the ISM. This means that the ratios of various abundances in the ISM are constantly changing. When a star is born, it traps with it the chemical abundances of the ISM at the time of its birth like a time capsule and carries them with it throughout most of its life. Thus, the ratios of certain elements in a star could probe at what point in the Milky Way’s chemical evolution (and thus in time) the star was born.

diagram showing sources of chemical enrichment over time

Figure 1: A cartoon depicting the different timescales of chemical enrichment from various sources, the concept behind chemical clocks. Core-collapse supernovae, which come from short-lived massive stars, for example, dominate the chemical enrichment of the Milky Way early on. Asymptotic Giant Branch (AGB) stars, which originate from long-lived low- and intermediate-mass stars, start contributing to galactic chemical enrichment later on. [Jacobson & Frebel 2014]

Testing Chemical Clocks in Wide Binaries

The authors of today’s paper set out to investigate just how reliable chemical clocks are at keeping time by testing their consistency in wide binaries. Wide binaries are pairs of stars that were born together and orbit a common center of gravity. As their name implies, wide binaries have large separations, making them easier to study observationally. These systems are a great way to test chemical clocks because they consist of two stars that share an age. Today’s authors investigate various chemical clock abundance ratios in 36 pairs of wide binaries to see which chemical clocks are most consistent among stars born at the same time.

The authors are first able to recreate the result found in previous studies that wide binaries are more chemically similar in their elemental makeup than random pairs of stars in the field. This makes sense. Stars born in the same place should share the same chemical composition because the interstellar medium is understood to be very homogeneous on small spatial scales. The chemical abundances of stars directly reflect the chemical abundances of the material from which they were born, so if the interstellar medium is well-mixed, and stars share a birth place and age, then they should share a similar chemical profile.

42-panel plot exploring different abundance ratios among the binary pairs

Figure 2: The consistency in the abundance of various chemical clocks between both components of wide binaries. The x-axis in each subplot is the abundance in the indicated chemical clock for one component of the binary (A), and the y-axis is the same for the other component (B). The tighter the 1-to-1 relationship in a subpanel, the more consistent a chemical clock between stars in the binary pair. [Sc/Ba], [Al/Ba], and [Ti/Ba] (all in the 4th row), among others, stand out as chemical clocks that appear to be promising age indicators. [Espinoza-Rojas et al. 2021]

The authors then make an interesting discovery: when they investigate chemical clocks among wide binaries, they find that components of wide binaries tend to be even more similar in chemical clock abundances than other elemental abundances, as seen in Figure 2. They find that even when components of a wide binary are quite dissimilar chemically in [X/Fe], as is the case in one particular pair in their sample (black box in Figure 2), they are still very consistent in chemical clock abundances. This result suggests that chemical clocks could be effective age indicators even when stars are extremely dissimilar in other elements. The authors highlight that three chemical clocks in particular, [Sc/Ba], [Al/Ba], and [Ti/Ba], seem to be the most consistent among wide binaries and thus the most promising indicators of age.

What is next for the field of chemical clocks? One new avenue involves calibrating chemical clocks using stars with ages derived through other means, such as gyrochronology. This way, we can create an empirical, observed relationship between a star’s abundance in a chemical clock and its age. These empirical relationships will likely vary with Milky Way location, but they will open up a new avenue of probing stellar age in stars with a variety of parameters. With chemical clocks, we can hopefully expand our stellar age toolbox and allow for more checks on stellar age, an important parameter in observational astronomy.

Original astrobite edited by Lili Alderson.

About the author, Catherine Manea:

Catherine is a 2nd year PhD student at the University of Texas at Austin. Her research is in galactic archaeology, the practice of using the kinematic and chemical information of individual stars to study the evolution of our Milky Way. She is particularly interested in pushing chemical tagging, the practice of tracing stars back to their birth sites, to new limits.

1 2 3 4 28