🔭 astrophysics

Evidence for elemental diffusion in the eclipsing binary star AI Phoenicis

This paper presents high-precision spectroscopic measurements of the eclipsing binary AI Phoenicis that reveal clear signatures of elemental diffusion in its F7V component, demonstrating the system's potential as a testbed for single-star diffusion and mixing models.

Original authors: Pierre F. L. Maxted, Nicholas Storm, Andreas J. Korn, Marília Carlos, Matthew R. Gent, Sviatoslav B. Borisov, Paula Jofré, Maria Bergemann, Hans-Günter Ludwig, Nicola J. Miller

Published 2026-07-21

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Pierre F. L. Maxted, Nicholas Storm, Andreas J. Korn, Marília Carlos, Matthew R. Gent, Sviatoslav B. Borisov, Paula Jofré, Maria Bergemann, Hans-Günter Ludwig, Nicola J. Miller

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, cosmic kitchen where stars are the chefs. For billions of years, these chefs have been cooking up the elements that make up everything around us, from the iron in your blood to the magnesium in your bones. But here's the tricky part: stars aren't just static lumps of gas; they are dynamic, churning balls of plasma where gravity is constantly trying to pull heavy ingredients down toward the center, while the heat and pressure try to keep them mixed up. This tug-of-war is called "elemental diffusion." Think of it like a bowl of soup where, if you let it sit too long without stirring, the heavy vegetables sink to the bottom and the lighter herbs float to the top. In stars, this process can change the chemical recipe we see on the surface, making the star look different than it actually is inside. Scientists care deeply about this because if we don't understand how these ingredients settle or mix, our recipes for how stars are born, live, and die will be wrong. To figure out the truth, we need to find stars where we can measure the ingredients with extreme precision, essentially taking a perfect snapshot of the soup before the heavy stuff has had a chance to sink too far.

This is exactly what the authors of this paper did with a star system called AI Phoenicis (AI Phe). Imagine AI Phe as a cosmic dance duo: a hotter, brighter partner (an F7 V dwarf star) and a cooler, slightly larger partner (a K0 IV subgiant star) locked in a tight embrace, orbiting each other every 24.6 days. Because they are so close, they eclipse each other, meaning one passes directly in front of the other from our point of view on Earth. The team used this celestial event like a natural filter. When the cooler star passed in front of the hotter one, it blocked the hotter star's light, allowing the astronomers to take a "clean" picture of the hotter star's atmosphere without any glare from its partner. They also used a clever computer trick called "spectral disentangling" to separate the mixed-up light of the two stars from other observations, effectively untangling a knot of light to see each star individually.

By analyzing the light from both stars, the team measured the amounts of iron and magnesium on their surfaces. They found a clear difference: the hotter star (AI Phe A) had less iron and magnesium on its surface compared to its cooler partner (AI Phe B). This isn't a mistake; it's a signature of diffusion. The heavier elements in the hotter star have slowly sunk deeper into its interior, leaving the surface "depleted" or poorer in these metals. The cooler star, which has a different structure with a deep convection zone, has had its original composition restored by this mixing, so it hasn't lost as much. To confirm this wasn't just a fluke, the researchers compared their findings to stars in a nearby star cluster called M67, which is roughly the same age and has a similar chemical makeup. The pattern they saw in AI Phe matched the pattern seen in M67 perfectly: stars that are still on the main sequence (like the hotter partner) show signs of these elements sinking, while stars that have evolved slightly (like the cooler partner) show the original mix.

The paper suggests that AI Phe is a fantastic "benchmark" or test case for scientists. Because we know the mass, size, and temperature of these two stars with incredible precision (down to about 0.1%), they provide a strict test for computer models that try to simulate how stars evolve. The authors found that the difference in iron abundance between the two stars is about 0.1 dex (a specific logarithmic unit used in astronomy), and the magnesium difference is even more pronounced. These numbers strongly suggest that elemental diffusion is happening right now in this system. While the paper doesn't claim to have solved every mystery about how stars mix their ingredients, it provides solid, measured evidence that diffusion is real and significant. It implies that future models of single stars must include this "settling" effect to be accurate, and AI Phe is now the go-to place to check if those models are getting the physics right.

Technical Summary: Evidence for Elemental Diffusion in the Eclipsing Binary Star AI Phoenicis

Problem Statement
Stellar evolution models often struggle to account for changes in surface and core composition caused by atomic diffusion (gravitational settling) and mixing processes. While diffusion signatures are well-documented in A-type stars and metal-poor globular clusters, their effects in solar-metallicity F-, G-, and K-type stars are subtler and often obscured by uncertainties in stellar parameters (mass, radius, surface gravity, and effective temperature). Previous studies of the open cluster M67 have revealed abundance differences of approximately 0.15 dex between turn-off stars and red giants, attributed to diffusion, but these measurements are complicated by systematic errors in stellar parameter determination. The binary system AI Phoenicis (AI Phe) offers a unique laboratory to test these effects because its fundamental parameters are known with exceptional precision (mass and radius to ~0.1%, effective temperature directly measured via bolometric flux), effectively removing stellar-parameter uncertainties that plague single-star studies.

Methodology
The authors obtained high-quality spectroscopic data for AI Phe to measure the iron ([Fe/H]) and magnesium ([Mg/H]) abundances of its two components: a K0 IV subgiant (AI Phe A) and an F7 V dwarf (AI Phe B).

Observations: Four spectra were acquired during the total eclipse of the F7 V companion by the K0 IV star using the UVES spectrograph on the ESO VLT. Additionally, archival HARPS spectra were utilized.
Spectral Disentangling: To isolate the individual spectra of both stars from the combined light, the authors applied a spectral disentangling algorithm (based on Simon & Sturm 1994) to the HARPS data. Crucially, they incorporated the UVES eclipse spectra into the disentangling vector. Since the F7 V star is completely eclipsed during these observations, the contribution of its flux to the K0 IV spectrum is negligible, allowing for an unambiguous determination of the flux ratio and the reconstruction of pure component spectra.
Abundance Analysis: The disentangled spectra were analyzed using three independent methods to ensure robustness:
- webSME: A 1D LTE radiative transfer code using the SME fitting engine.
- TSFitPy: A wrapper for the Turbospectrum NLTE radiative transfer code.
- q2: A differential line-by-line analysis relative to the Sun, using the MOOG code.
  The analysis focused on Fe II and Mg I lines, employing 1D MARCS model atmospheres. The authors also applied 3D corrections to the 1D abundances to assess modeling assumptions.

Key Results
The analysis consistently reveals a systematic depletion of heavy elements in the photosphere of the hotter, main-sequence component (AI Phe B) compared to the cooler subgiant (AI Phe A).

Iron Abundance: The difference in iron abundance is measured as $\Delta$ [Fe/H] = [Fe/H] $_A$ - [Fe/H] $_B \approx -0.10$ to $-0.12$ dex. Specifically, the webSME analysis yields a difference of $-0.117 \pm 0.012$ dex (1D LTE), which remains significant ( $>8\sigma$ ) even after applying 3D corrections ( $-0.097 \pm 0.012$ dex).
Magnesium Abundance: A similar trend is observed for magnesium, with AI Phe B showing a lower abundance than AI Phe A by approximately 0.13 dex.
Comparison with M67: When plotted on a Kiel diagram alongside stars from the open cluster M67 (which has a similar age and metallicity), AI Phe follows the same pattern: the dwarf component (AI Phe B) is metal-depleted relative to the subgiant (AI Phe A). This mirrors the behavior seen in M67, where diffusion depletes metals in turn-off stars, which are then restored in the photosphere of subgiants as the convective envelope deepens.

Significance and Claims
The paper asserts that the measured abundance differences provide clear evidence for elemental diffusion in the F7 V component of AI Phe. The primary significance of this work lies in the precision of the system's fundamental parameters. Because the mass, radius, and effective temperature of AI Phe are known to very high accuracy, the observed abundance differences cannot be attributed to errors in stellar parameter determination, a common source of bias in single-star studies.

Consequently, the authors conclude that AI Phe serves as a robust benchmark for testing single-star evolution models that include diffusion and mixing. The system allows for a direct test of whether models including diffusion can reproduce the observed abundance stratification between a dwarf and a subgiant of the same initial composition. While the authors note that previous modeling efforts (Valle et al. 2023) found the convective overshooting parameter ( $\beta$ ) to be poorly constrained when diffusion is included, they argue that the high-quality abundance data presented here, combined with the precise physical parameters, offers a superior constraint for future model calibrations than previously available. The paper does not claim to solve the overshooting problem but establishes AI Phe as a critical test case for models incorporating diffusion.

More like this