Hostname: page-component-6898cfcb55-kwdml Total loading time: 0 Render date: 2026-11-11T16:38:01.721Z Has data issue: false hasContentIssue false
  • English
  • Français

Finding He II: Testing Novel Models of Binary Populations across Cosmic Time

Published online by Cambridge University Press:  09 November 2026

Beryl Hovis-AfflerbachOpen the ORCID record for Beryl Hovis-Afflerbach[Opens in a new window],
Allison L. Strom,
Alberto Saldana-Lopez and
Sophia R. Flury
Beryl Hovis-Afflerbach*
Affiliation:
Department of Physics and Astronomy, Northwestern University, 2145 Sheridan Rd, Evanston, IL 60208, USACenter for Interdisciplinary Exploration and Research in Astrophysics (CIERA), Northwestern University, 1800 Sherman Ave, Evanston, IL 60201, USA
Allison L. Strom
Affiliation:
Department of Physics and Astronomy, Northwestern University, 2145 Sheridan Rd, Evanston, IL 60208, USACenter for Interdisciplinary Exploration and Research in Astrophysics (CIERA), Northwestern University, 1800 Sherman Ave, Evanston, IL 60201, USA
Alberto Saldana-Lopez
Affiliation:
Department of Astronomy, Oskar Klein Centre, Stockholm University, SE-106 91 Stockholm, Sweden
Sophia R. Flury
Affiliation:
Institute for Astronomy, University of Edinburgh, Royal Observatory, Edinburgh EH9 3HJ, UK
*
email: beryl@u.northwestern.edu
Rights & Permissions [Opens in a new window]

Abstract

Our understanding of massive stars remains incomplete. Many high-z galaxies and nearby analogs exhibit strong He II emission, indicating an abundance of photons with energies $\gt54.4$ eV that standard single-star population models cannot explain. Recent studies show that binary evolution and non-solar abundance patterns are required to explain the distinct spectra of high-z galaxies observed by JWST. However, treatments of these properties vary drastically between models. We present the first results from a comparison of models with different treatments of binaries, including BPASS and novel stripped star models, with rest-UV-optical spectra of high-z galaxies’ local analogs. This type of investigation can provide insights into which aspects of binary evolution are important to reproduce observations and identify priorities in ongoing efforts to improve models. By constraining the properties of massive stars at high redshift, we can learn about the processes at play in high-z galaxies and massive star evolution more broadly.

Keywords

stellar populationshigh-redshift galaxiesinteracting binary starsmetallicity

Information

Type
Contributed Paper
Information
Proceedings of the International Astronomical Union , Volume 21 , Symposium S402: Massive Stars Across Redshifts in the Era of JWST and Large Surveys , January 2025, pp. 218 - 223
Check for updates
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2026. Published by Cambridge University Press on behalf of The International Astronomical Union

1. Introduction

Metallicity is one of the most significant factors in determining how a star evolves and shapes its environment. It is therefore crucial to understand how the metal-poor stars that populate the high-redshift universe differ from those with Solar metallicity, in order to understand how those stars, and their host galaxies, evolved. Ideally, it would be possible to test models of metal-poor massive stars with observations of individual stars from resolved populations in the nearby Universe, but due to many factors including shorter lifetimes and the scarcity of nearby low-metallicity environments, the population of such stars nearby is very small. Studying metal-poor massive stars therefore requires comparisons with high-redshift star-forming galaxies and their local analogs. In these galaxies, the ultraviolet (UV, tilde 100 minus 3000 $\sim 100-3000$ Å) part of the spectrum should be dominated by massive stars.

Approximately 1 divided by 3 $1/3$ of O stars are expected to have their hydrogen envelopes stripped by interaction with a binary companion (Sana et al. Reference Sana, de Mink and de Koter2012). The resulting hot, intermediate-mass “stripped stars” have been predicted for many years, but the first sample of 16 was only recently observed in the Magellanic Clouds (Drout et al. Reference Drout, Götberg and Ludwig2023; Götberg et al. Reference Götberg, Drout and Ji2023). Stripped stars are UV-bright, with the most massive potentially having spectra similar to those of Wolf-Rayet stars, and have long been considered as a potential source of the hard ionizing galaxy spectra that are ubiquitous at high redshift. In optical wavelengths, however, they are often obscured by a binary companion, which likely explains why they evaded detection for so long.

It is possible that their elusiveness is not only the product of observational bias but also a reflection of true scarcity. Studies using a variety of stellar evolution codes have predicted that low-metallicity massive stars should not expand until very late in their lifetimes, delaying binary interaction and therefore envelope stripping (e.g., Klencki et al. Reference Klencki, Nelemans, Istrate and Pols2020; Farrell et al. Reference Farrell, Groh, Meynet and Eldridge2022). The resulting stripped stars would be very short-lived or perhaps only partially-stripped. Hovis-Afflerbach et al. (Reference Hovis-Afflerbach, Götberg and Schootemeijer2025) used a binary population synthesis model initially developed by (Götberg et al. Reference Götberg, de Mink, Groh, Leitherer and Norman2019, hereafter G19) to model the mass distribution of stripped stars at different metallicities and found that this late expansion results in a “helium star desert” where massive ( upper M Subscript normal s normal t normal r normal i normal p Baseline tilde 12 minus 30 upper M Subscript circled dot Baseline $M_{\rm strip} \sim 12-30 {\ M_\odot}$ ) stripped stars are missing at low metallicity ( upper Z less than or equivalent to 0.002 $Z \lesssim 0.002$ ).

Using these distributions, Hovis-Afflerbach et al. (Reference Hovis-Afflerbach, Götberg and Schootemeijer2025) predict the impact of the helium star desert on ionizing photon production rates. Figure 1 (Hovis-Afflerbach et al. Reference Hovis-Afflerbach, Götberg and Schootemeijer2025, Figure 8) shows that while stripped stars have a non-negligible impact on hydrogen-ionizing photons ( upper Q 0 $Q_0$ , left), they dominate the production of helium-ionizing photons ( upper Q 2 $Q_2$ , right), and the upper Q 2 $Q_2$ prediction is therefore significantly impacted by the helium star desert. The He II lamda Baseline 1640 $\lambda 1640$ emission line and its weaker optical counterpart are therefore excellent probes of the helium star desert and of stripped stars in general.

Figure 1.

Estimates for hydrogen- ( upper Q 0 $Q_0$ , left) and helium-ionizing ( upper Q 2 $Q_2$ , right) emission rates from Hovis-Afflerbach et al. (Reference Hovis-Afflerbach, Götberg and Schootemeijer2025). The total rates are shown as a sum of rates for single stars (light yellow), low-mass stripped stars ( upper M less than 8.5 upper M Subscript circled dot Baseline $M \lt 8.5\ M_\odot$ , medium yellow), and higher-mass stripped stars ( upper M greater than 8.5 upper M Subscript circled dot Baseline $M \gt 8.5\ M_\odot$ , dark yellow). Stripped stars dominate upper Q 2 $Q_2$ , resulting in a significantly higher rate if there is no helium star desert (left column) than if there is (middle column).

Bar graph comparing hydrogen- and helium-ionizing photon emission rates from different star types.

More efforts are being made to identify and study nearby stripped stars. Ludwig et al. (Reference Ludwig, Drout, Gotberg, Lang and Laroche2025) present over 800 more candidate stripped stars in the Magellanic Clouds, and finding stripped stars is one of the goals of the Ultraviolet Explorer (UVEX, Kulkarni et al. Reference Kulkarni, Harrison and Grefenstette2021) mission, but it will be years before a sufficient sample of nearby stripped stars are detected to test for the helium star desert. In the meantime, high-z, star-forming galaxies and their local analogs provide a unique testbed to study stripped stars and binary evolution at low metallicity.

Here, we compare three different population and spectral synthesis models with observations of local analog galaxies to test their ability to reproduce these galaxies’ hard ionizing spectra. This is the first observational test of the G19 stripped star models.

2. Models of massive star populations

We compare three sets of models. For each, we use a grid of spectra for instantaneous 10 Superscript 6 Baseline upper M Subscript circled dot Baseline $10^6 {\ M_\odot}$ starbursts with ages 1, 2, 3, 4, 5, 8, 10, 15, 20 and 40 Myr and metallicities of approximately upper Z equals 0.014 comma 0.006 comma 0.002 $Z = 0.014,\ 0.006,\ 0.002$ , and 0.001 $0.001$ .

Starburst99 (Leitherer et al. Reference Leitherer, Schaerer and Goldader1999) is one of the most widely-used population synthesis codes. It does not include binaries, so we use it as a control with which to compare the two binary population models. We use the same Starburst99 models described in Section 2.3 of G19, with stellar metallicities upper Z equals 0.014 comma 0.008 comma 0.002 $Z = 0.014,\ 0.008,\ 0.002$ , and 0.001 $0.001$ .

We use Binary Population And Spectral Synthesis v2.3 (BPASS, Eldridge et al. Reference Eldridge, Stanway and Xiao2017) models with Solar-scaled metal abundances and stellar metallicities upper Z equals 0.014 comma 0.006 comma 0.002 $Z = 0.014,\ 0.006, 0.002$ , and 0.001 $0.001$ . Because stars at higher redshifts are expected to be alpha $\alpha$ -enhanced (e.g., Steidel et al. Reference Steidel, Strom and Pettini2016), a treatment including alpha $\alpha$ -enhancement would be more accurate. However, our other models use Solar-scaled abundances, and this version of BPASS focuses on alpha $\alpha$ -enhancement only in low-mass stars. As model spectra including alpha $\alpha$ -enhanced massive stars become available (e.g., Park et al. Reference Park, Conroy and Johnson2024; Byrne et al. Reference Byrne, Eldridge and Stanway2025), more realistic tests will be possible.

Finally, we use the same G19 stripped star population models used by Hovis-Afflerbach et al. (Reference Hovis-Afflerbach, Götberg and Schootemeijer2025) to predict the helium star desert. These include only stripped stars, so we add them to the single star population spectra from Starburst99. The contribution of stripped stars to the ionizing spectrum is significant only after 10 Myr Gotberg2019, so the UV spectra of younger bursts can be approximated using purely the single stars. We take the same approach as G19 and add stripped star models with upper Z equals 0.014 comma 0.006 comma 0.002 $Z = 0.014,\ 0.006,\ 0.002$ , and 0.0002 $0.0002$ to Starburst99 models with the closest metallicities, upper Z equals 0.014 comma 0.008 comma 0.002 $Z = 0.014,\ 0.008,\ 0.002$ , and 0.001 $0.001$ , respectively.

3. Methods

We compare the model spectra with CLASSY (Berg et al. Reference Berg, James and King2022), a sample of nearby star-forming galaxies with similar properties as seen at high redshift. We include the 19 galaxies with He II lamda Baseline 1640 $\lambda 1640$ emission detected at a signal-to-noise ratio greater than 3 $\gt 3$ (Mingozzi et al. Reference Mingozzi, James and Arellano-Córdova2022).

To simulate the nebular continuum and emission lines, we process all three grids of models, each spanning a range of stellar metallicities and ages, through Cloudy 23.00 (Chatzikos et al. Reference Chatzikos, Bianchi and Camilloni2023) using identical input parameters, so any differences should arise from only differences in the models themselves. For each model, the gas-phase metallicity is set equal to the stellar metallicity of the input spectrum, using Solar-scaled abundances (however, see Section 2 for notes on alpha $\alpha$ -enhancement). We assume spherical geometry and an ionization parameter of log upper U equals negative 2.5 $\log U = -2.5$ . We use an isobaric density profile with constant pressure 6.0 and a hydrogen density at the illuminated face of n Subscript normal upper H Baseline equals 10 squared normal c normal m Superscript negative 3 $n_{\rm H} = 10^2 \ {\rm cm}^{-3}$ . The calculation stops when the hydrogen ionization fraction drops to 10 Superscript negative 2 $10^{-2}$ , and each model is iterated to convergence.

Once we process each model through Cloudy, we add the resulting nebular continuum to the incident spectrum. We then use the FiCUS codeFootnote † (Saldana-Lopez et al. Reference Saldana-Lopez, Schaerer and Chisholm2023) to fit the UV continuum ( 1120 minus 1760 $1120-1760$ Å) of each observed galaxy. We do so independently with each of the three grids of models, which have been implemented in FiCUS as ssp_models and can be used by choosing sb99, bpass, or sb99stripped.

4. Comparison of model predictions with observations

4.1. Properties derived from the nebular continuum

From FiCUS, we derive a light-weighted population age, metallicity, and E(B-V). The two implementations of binaries yield different best-fit values for metallicity and age (Figure 2). Including binaries using BPASS gives similar results to Starburst99, though overall slightly lower metallicities and some scatter in both properties, consistent with Chisholm et al. (Reference Chisholm, Rigby and Bayliss2019).

Figure 2.

Comparison of metallicity (left) and age (right) for Starburst99 continuum fits compared with BPASS (purple) and Starburst99 with stripped stars (green). While the BPASS results generally agree with those from Starburst99, the stripped star results are virtually identical.

Two scatter plots compare metallicity and age of starburst99 continuum fits with BPASS and stripped stars.

However, accounting for binaries by adding the G19 stripped star models yields almost identical properties to Starburst99 alone. This similarity is because, while the stripped star models alone exhibit strong wind features for bursts of ages tilde 10 minus 25 $\sim 10-25$ Myr, the overall magnitude of their spectra in the non-ionizing UV is a factor of tilde 100 $\sim 100$ less than that of the Starburst99 spectra alone. When added together, the stripped star wind features get washed out, and the overall contribution of stripped stars to the non-ionizing UV continuum is negligible. The G19 models, therefore, predict that stripped stars may be able to harden a population’s ionizing flux without significant changes to the non-ionizing UV continuum.

4.2. Predicted emission line fluxes

In addition to providing the nebular continuum, photoionization modeling with Cloudy yields predictions for nebular line emission. In Figure 3, we show how He II lamda Baseline 1640 $\lambda 1640$ /H beta $\beta$ changes over the life of a starburst, for all four metallicities and all three model grids.

Figure 3.

He II lamda Baseline 1640 $\lambda 1640$ /H beta $\beta$ as a function of time after a starburst, from Cloudy runs using incident spectra from Starburst99 (circles, dashed line), Starburst99 plus G19 stripped stars (stars, solid line), and BPASS (diamonds, dotted line). The four metallicities in the model grid are indicated by color. Both BPASS and stripped stars boost He II lamda Baseline 1640 $\lambda 1640$ /H beta $\beta$ significantly beginning at 10 Myr.

A line graph showing He II λ1640 /H β ratios over starburst age with different models and metallicities.

For the Starburst99 models with single stars, this line ratio is strongest at lower metallicity and significant only at earlier times, becoming negligible by tilde 5 $\sim 5$ Myr. The primary source of the nebular He II lamda Baseline 1640 $\lambda 1640$ emission line in these models is from massive Wolf-Rayet stars, which die out at these early times. With the addition of the G19 stripped star models starting at 10 Myr (stars, solid line), the line ratio value increases significantly, reaching values tilde 3 minus 30 $\sim 3-30$ times higher than the Wolf-Rayet stars did at their peak.

The BPASS models (diamonds, dotted line) show an overall similar trend in He II lamda Baseline 1640 $\lambda 1640$ /H beta $\beta$ , with initially high values due to emission from Wolf-Rayet stars that decrease as they die out, followed by even higher values at times greater than or equals 10 $\geq 10$ Myr due to older binary products that produce hard ionizing radiation at later times. There is one significant difference between the two implementations of binaries: for bursts with ages greater than or equals 10 $\geq 10$ Myr, the highest He II lamda Baseline 1640 $\lambda 1640$ /H beta $\beta$ values are produced by the BPASS models with the lowest metallicities but by the G19 models with the highest metallicities. This may be another predicted consequence of the helium star desert at low metallicity in the G19 models—a dearth of stripped stars should decrease the hardness of the ionizing spectrum. However, we note that the stripped star atmosphere models used by G19, first presented by Götberg et al. (Reference Götberg, de Mink and Groh2018), were made prior to the first observations of intermediate mass stripped stars. Their recent discovery has revealed weaker winds, which should affect their temperatures and, in turn, impact the He II lamda Baseline 1640 $\lambda 1640$ line.

The difference between BPASS and G19 in Figure 3 may also be related to the sources predominantly producing their ionizing radiation. While the G19 models contain only stripped stars, BPASS considers a range of other binary and stellar physics, including quasi-chemically homogeneous evolution (QHE). Though this phenomenon has not been conclusively observed, BPASS assumes that QHE can occur for stars at metallicities less than or equals 0.004 $\leq 0.004$ and masses greater than 20 upper M Subscript circled dot Baseline $\gt 20 {\ M_\odot}$ (Eldridge et al. Reference Eldridge, Stanway and Xiao2017). This metallicity cut-off corresponds with the jump in He II lamda Baseline 1640 $\lambda 1640$ /H beta $\beta$ between the BPASS models with upper Z equals 0.014 comma 0.006 $Z = 0.014,\ 0.006$ and upper Z equals 0.002 comma 0.001 $Z = 0.002,\ 0.001$ , suggesting that QHE stars may be the most significant source of He II lamda Baseline 1640 $\lambda 1640$ photons in BPASS at later times.

5. Conclusions and future work

Different implementations of binaries in models have a significant impact on the predicted ionizing spectrum, and in turn, on the conclusions drawn from comparisons with data. Because it requires photons with energies in excess of 54.4 eV, the nebular He II lamda Baseline 1640 $\lambda 1640$ emission line is sensitive to hard ionizing sources such as stripped stars, which can harden a population’s ionizing flux at later times after a starburst without significant changes to the non-ionizing spectrum. We can use this line as a diagnostic for discriminating between models and studying binary populations, including testing whether a “helium star desert” exists at low metallicity.

We intend to continue this work by comparing observed and predicted He II lamda Baseline 1640 $\lambda 1640$ line fluxes. The observed emission line fluxes will be measured by subtracting the FiCUS best fit continuum from the observed spectrum and fitting the residual flux. The predicted emission line fluxes will be calculated by adding the emission line values output by Cloudy with the same weights from the FiCUS best fit. By comparing the observed and predicted line fluxes, we can assess which models are able to self-consistently reproduce high-z galaxies’ spectra.

Footnotes

† FItting the stellar Continuum of Uv Spectra, https://github.com/asalda/FiCUS

References

Berg, D. A., James, B. L., King, T., et al. 2022, ApJS, 261, 31 CrossRefGoogle Scholar
Byrne, C. M., Eldridge, J. J., & Stanway, E. R. 2025, MNRAS, 537, 2433 CrossRefGoogle Scholar
Chatzikos, M., Bianchi, S., Camilloni, F., et al. 2023, RMxAA, 59, 327 CrossRefGoogle Scholar
Chisholm, J., Rigby, J. R., Bayliss, M., et al. 2019, ApJ, 882, 182 CrossRefGoogle Scholar
Drout, M. R., Götberg, Y., Ludwig, B. A., et al. 2023, Science, 382, 1287 CrossRefGoogle Scholar
Eldridge, J. J., Stanway, E. R., Xiao, L., et al. 2017, PASA, 34, e058 CrossRefGoogle Scholar
Farrell, E., Groh, J. H., Meynet, G., & Eldridge, J. J. 2022, MNRAS, 512, 4116 CrossRefGoogle Scholar
Götberg, Y., de Mink, S. E., Groh, J. H., et al. 2018, A&A, 615, A78 Google Scholar
Götberg, Y., de Mink, S. E., Groh, J. H., Leitherer, C., & Norman, C. 2019, A&A, 629, A134 Google Scholar
Götberg, Y., Drout, M. R., Ji, A. P., et al. 2023, ApJ, 959, 125 CrossRefGoogle Scholar
Hovis-Afflerbach, B., Götberg, Y., Schootemeijer, A., et al. 2025, A&A, 697, A239 Google Scholar
Klencki, J., Nelemans, G., Istrate, A. G., & Pols, O. 2020, A&A, 638, A55 Google Scholar
Kulkarni, S. R., Harrison, F. A., Grefenstette, B. W., et al. 2021, arXiv e-prints, arXiv:2111.15608 Google Scholar
Leitherer, C., Schaerer, D., Goldader, J. D., et al. 1999, ApJS, 123, 3 CrossRefGoogle Scholar
Ludwig, B., Drout, M. R., Gotberg, Y., Lang, D., & Laroche, A. 2025, arXiv e-prints, arXiv:2505.18632 Google Scholar
Mingozzi, M., James, B. L., Arellano-Córdova, K. Z., et al. 2022, ApJ, 939, 110 CrossRefGoogle Scholar
Park, M., Conroy, C., Johnson, B. D., et al. 2024, arXiv e-prints, arXiv:2410.21375 Google Scholar
Saldana-Lopez, A., Schaerer, D., Chisholm, J., et al. 2023, MNRAS, 522, 6295 CrossRefGoogle Scholar
Sana, H., de Mink, S. E., de Koter, A., et al. 2012, Science, 337, 444 CrossRefGoogle Scholar
Steidel, C. C., Strom, A. L., Pettini, M., et al. 2016, ApJ, 826, 159 CrossRefGoogle Scholar
Figure 0

Figure 1.Estimates for hydrogen- (Q0$Q_0$, left) and helium-ionizing (Q2$Q_2$, right) emission rates from Hovis-Afflerbach et al. (2025). The total rates are shown as a sum of rates for single stars (light yellow), low-mass stripped stars (M<8.5 M⊙$M \lt 8.5\ M_\odot$, medium yellow), and higher-mass stripped stars (M>8.5 M⊙$M \gt 8.5\ M_\odot$, dark yellow). Stripped stars dominate Q2$Q_2$, resulting in a significantly higher rate if there is no helium star desert (left column) than if there is (middle column).

Figure 1

Figure 2.Comparison of metallicity (left) and age (right) for Starburst99 continuum fits compared with BPASS (purple) and Starburst99 with stripped stars (green). While the BPASS results generally agree with those from Starburst99, the stripped star results are virtually identical.

Figure 2

Figure 3.He II λ1640$\lambda 1640$/Hβ$\beta$ as a function of time after a starburst, from Cloudy runs using incident spectra from Starburst99 (circles, dashed line), Starburst99 plus G19 stripped stars (stars, solid line), and BPASS (diamonds, dotted line). The four metallicities in the model grid are indicated by color. Both BPASS and stripped stars boost He II λ1640$\lambda 1640$/Hβ$\beta$ significantly beginning at 10 Myr.