Across dozens of extremely metal-poor galaxies, astronomers are collecting some of the cleanest emission-line spectra yet assembled for one precise purpose: to measure how much helium the universe produced in its first minutes. The Large Binocular Telescope project is designed to push the uncertainty in the primordial helium abundance from about 1.3% toward roughly 0.5%, using carefully selected galaxies, broader wavelength coverage, new atomic calculations and a more tightly controlled analysis.
Primordial helium is one of the quantities that can be used to test what happened in the early universe. Standard Big Bang nucleosynthesis predicts a helium mass fraction of 0.2467 ± 0.0002 when the measured baryon density and neutron mean life are used. That theoretical value has an uncertainty of better than 0.1%, far smaller than the uncertainty of direct astronomical measurements.
The reason astronomers care about closing that gap is that the amount of helium produced during Big Bang nucleosynthesis depends on several physical quantities, including the baryon density, the neutron mean life and the number of relativistic species present during nucleosynthesis. The latter can be characterized by the number of neutrino flavors. A sufficiently precise measurement of primordial helium can therefore provide a sensitive test of the conditions prevailing in the early universe.
The current observational situation is considerably less precise. A recent determination based on helium emission from metal-poor galaxies gave Yp = 0.2448 ± 0.0033, corresponding to about 1.3% precision. Measurements based on the cosmic microwave background have larger uncertainties.
That difference has not simply been a matter of collecting more spectra. Measuring helium from a galaxy’s glowing gas requires several physical properties of that gas to be estimated at the same time. Earlier work found that these parameters can be strongly degenerate, making apparently precise measurements vulnerable to systematic uncertainties.
The LBT Yp Project is built around the idea that both the observations and the analysis need to improve together.
Why extremely metal-poor galaxies are the targets
The project is not trying to measure primordial helium directly from a pristine cloud that has never formed stars. Instead, it observes H II regions, where gas is ionized and produces strong emission lines. The researchers seek galaxies with very little oxygen and other heavy elements because their helium abundance should be closer to the primordial value.
There is an important complication. The helium abundance measured in a galaxy is not necessarily exactly the primordial abundance because stars can alter the chemical composition of their surroundings. Conventional approaches therefore fit helium abundance against metallicity and extrapolate the relationship to zero metallicity. But that method assumes a particular relationship between helium and oxygen. The researchers argue that such a strictly linear relationship introduces an unquantifiable systematic uncertainty because helium and oxygen have different stellar origins.
The project’s preferred solution is to obtain enough extremely metal-poor galaxies that the primordial abundance can instead be determined from the galaxies themselves without depending on that extrapolation.
The target selection consequently has three main requirements. The galaxies should have very low oxygen abundance, strong Hβ emission and large Hβ equivalent widths. Strong emission lines make it possible to measure the weak helium and hydrogen lines needed to constrain the physical conditions of the gas. A strong continuum, by contrast, can bury weak emission lines or make them more vulnerable to underlying stellar absorption.
For the project’s observations, the researchers found that Hβ fluxes above 10−15 erg s−1 cm−2 were needed for the desired precision. They also established a guideline of at least 100 angstroms for the Hβ equivalent width. These criteria were developed from the team’s accumulating observations, although not every target selected from earlier literature met them.
The telescope campaign was built for consistency
The project uses the Large Binocular Telescope’s MODS spectrographs for optical observations and its LUCI instruments for near-infrared spectroscopy.
The optical MODS observations cover approximately 3200 to 10,000 angstroms across their blue and red channels. The broad wavelength coverage allows the researchers to measure a large collection of hydrogen and helium emission lines rather than relying on only a few lines.
The observing strategy was deliberately controlled. The team restricted most observations to airmasses below 1.5 and oriented the slit at the parallactic angle to reduce wavelength-dependent losses caused by atmospheric refraction. New targets generally received one hour of open-shutter MODS exposure, divided into three 20-minute exposures. Standard calibration observations were obtained during each observing run.
In total, the project obtained new high-quality MODS spectra of 56 targets, amounting to 70 hours of LBT observing time. The researchers also analyzed seven archival MODS targets and incorporated earlier observations of Leo P and AGC 198691.
The near-infrared observations address one of the project’s most important measurement problems.
One infrared helium line helps untangle the gas
The He I λ10830 emission line is particularly valuable because it provides a strong constraint on the density of the ionized gas. That helps break a degeneracy between density and temperature that can otherwise make the helium abundance difficult to determine accurately.
The project therefore used LUCI to observe this line for most of its sample, with values from the literature used for 11 targets. The LUCI observations covered the 0.9–1.35 micrometer region containing He I λ10830 and the hydrogen Pγ line.
The team obtained new high-quality near-infrared spectra for 48 targets, totaling 72 hours of LBT time, and included an archival observation of Leo P. Each target normally received one hour of integration, collected as six 600-second exposures in a dither pattern that shifted the target along the slit. The arrangement was designed to improve subtraction of the many bright atmospheric OH emission lines that crowd this part of the infrared spectrum.
For the near-infrared calibration, the researchers observed telluric standard stars at nearly the same airmass as the galaxies. Those stars were used to correct atmospheric absorption and establish the relative flux around the helium line. The team also observed the same Hubble Space Telescope primary calibrators used for the optical work during two observing runs and found that their assumptions about the LUCI response were consistent with those measurements.
Together, the optical and infrared observations gave the project 130 hours of high-quality, consistent spectrophotometric data. The observations were carried out remotely, beginning during the COVID-19 pandemic.
Calibration became part of the measurement
The researchers also revisited a source of uncertainty that had previously been treated more simply.
In earlier work, the team had assigned a wavelength-independent 2% flux-calibration uncertainty to all emission-line measurements. That was conservative, but it did not reflect the fact that two emission lines close together in wavelength can have a much smaller relative calibration uncertainty.
For the new project, the researchers examined archival MODS observations of two Hubble Space Telescope primary flux standards, G191B2b and GD153. Their final calibration sample contained 50 spectra of the first star and 19 of the second, covering observations from 2021 through 2024.
That long record allowed the team to track changes in the telescope and instruments rather than treating calibration as a single fixed uncertainty. They identified throughput changes associated with mirror washing and recoating and with changes involving the adaptive secondary mirrors. Despite those changes, the basic shapes of the MODS spectral response functions remained sufficiently stable for the researchers to construct wavelength-dependent calibration uncertainties.
The resulting relative-flux calibration is typically accurate to about ±0.5% across most of the MODS blue channel and ±0.6% across most of the red channel. The uncertainties rise toward the ends of the wavelength ranges, where detector, grating and optical transmission effects become more important. The red channel also contains regions affected by atmospheric absorption.
That is a different approach from simply assigning the same uncertainty to every line. The project is attempting to track how the uncertainty actually varies with wavelength.
The atomic data have also changed
The analysis uses updated calculations of the emission-line strengths produced by hydrogen and helium.
For hydrogen, the researchers use a fine grid of emissivities extending to the n = 15 level. They interpolate within the temperature and density grid rather than relying on a more limited set of values. The team assumes that the uncertainty in these hydrogen emissivities is negligible compared with other sources of uncertainty. That assumption is supported by comparisons in which the newest calculations differed from earlier hydrogen calculations by only about 0.10% to 0.55% for Hβ and 0.10% to 0.20% for three of the strongest Balmer-line ratios.
The helium calculations are less straightforward. The project adopts new He I emissivities from Del Zanna and Storey, whose calculations produced differences of more than 1% compared with earlier studies. The largest change relevant to the project occurs for He I λ6678, where the disagreement grows as density and temperature increase.
The researchers nevertheless assume no uncertainty in the helium emissivities because there are no quantified estimates available for that uncertainty. They explicitly note that this assumption is less secure than the corresponding assumption for hydrogen.
The project has also adopted a new grid of helium radiative-transfer corrections. Those corrections are designed to work with modern helium emissivities and address problems identified in earlier corrections.
More emission lines mean more constraints
The value of the broad wavelength coverage is not simply that it produces more measurements. Different emission lines constrain different properties of the gas, allowing the analysis to disentangle parameters that can otherwise mimic one another.
The project incorporates hydrogen and helium lines across the optical and infrared range to constrain quantities including temperature, density, reddening, optical depth, underlying stellar absorption and the neutral hydrogen fraction. The near-infrared He I λ10830 line is especially important for density.
The optical spectra are also used to determine the oxygen abundance and the physical conditions of the gas. Direct measurements of the [O III] temperature are used as a weak prior on the helium-emitting gas temperature in the project’s MCMC analysis. The oxygen abundances themselves have typical uncertainties of about 4%.
The overall goal is to bring the uncertainty of individual helium abundance measurements to around 2% or less, while ultimately using the low-metallicity sample to obtain a primordial helium abundance with about 0.5% precision.
The sample is designed to avoid an extrapolation
The project’s eventual sample contains 41 objects with oxygen abundances of O/H ≤ 14.5 × 10−5. Fifteen of those have O/H ≤ 4 × 10−5.
According to the researchers, those 15 objects occupy a sufficiently low-metallicity regime that evolutionary effects are minimal. That allows the primordial helium abundance to be obtained from their weighted mean without assuming a linear relationship between helium and oxygen. The individual helium abundances, reliability screening and final Yp determination are presented in Paper IV of the project series rather than in this project-description paper.
This distinction is important because the present work does not itself report the project’s final new value of Yp. It establishes the observations and methodology intended to produce that measurement.
The strategy represents a deliberate change in emphasis from earlier large samples. Some previous analyses increased the number of galaxies, but many of those spectra had relatively low signal-to-noise ratios, lacked the crucial near-infrared helium line and included objects at substantially higher metallicity. The LBT team argues that increasing the number of lower-quality measurements can reduce statistical uncertainty while leaving greater vulnerability to assumptions about the helium-to-oxygen relationship and other systematic effects.
The analysis is designed to expose bad fits
The project also treats the statistical fit itself as part of the quality control.
Earlier analyses using MCMC methods found that some observations were statistically inconsistent with the models used to interpret their emission lines. In one example discussed by the authors, a starting set of 93 observations was reduced through successive quality cuts until only 22 remained, with additional concerns identified among those. Later work using improved emissivities did not substantially resolve the problem, suggesting that the observational data themselves were likely an important source of the inconsistencies.
The addition of He I λ10830 helped in earlier work. Among 31 targets with that infrared measurement, the number surviving the analysis cuts increased from 11 without the line to 16 with it, illustrating how an additional independent constraint can improve the modeling.
The LBT project carries that approach further by seeking observations with enough wavelength coverage and signal quality to constrain the model parameters simultaneously.
The researchers also retain a distinction between the best-fitting model and the distribution of MCMC samples. In the appendix, they point out that parameters restricted to positive values, such as density, stellar absorption and optical depth, can have strongly asymmetric MCMC distributions when their best-fit values are close to zero. In those cases, taking the median of the MCMC distribution can produce a value substantially higher than the maximum-likelihood solution. The project’s approach retains the positive restrictions but uses the maximum-likelihood value rather than the MCMC median for the final parameters.
The authors also discuss differences in how earlier studies decided whether a galaxy’s model fit was good enough to include in a Yp determination. Their preferred approach uses the minimized χ² as a direct statistical test of the fit, while another methodology required individual emission lines to agree with the model within 2σ. The authors say the latter approach can admit more objects, but the effect on the resulting Yp is not clear.
They further note that some earlier samples contained many galaxies without He I λ10830 observations. The potential systematic effects of that inhomogeneity, beyond the larger uncertainties for individual objects, remain uncertain.
The observations are meant to become a reusable dataset
The project is not intended to produce only a single helium number.
The researchers plan to release machine-readable measurements of the optical and infrared emission-line fluxes, along with wavelength- and flux-calibrated one-dimensional MODS and LUCI spectra in FITS format. The data will include relevant observation metadata.
The project is organized as six papers. The present work describes the motivation, sample selection, observations, emissivities, database and overall methodology. The other papers cover the optical reductions and oxygen abundances, the infrared observations, the individual helium abundances and Yp determination, the implications for the number of relativistic species, and additional chemical abundances in the observed galaxies.
The full project therefore combines a carefully selected low-metallicity sample with high-quality optical and infrared spectroscopy, detailed calibration work and an analysis intended to account for the physical degeneracies that have complicated helium measurements in the past.
For the 15 galaxies at the lowest oxygen abundances, the intended payoff is a primordial helium measurement that does not need to be obtained by extrapolating a helium-versus-oxygen relation. The resulting helium abundance and its implications for early-universe physics are left to the subsequent papers in the series.
The study was published in The Astrophysical Journal.
