It has been known for decades that the observed number of baryons in the local Universe falls about 30–40 per cent short1,2 of the total number of baryons predicted3 by Big Bang nucleosynthesis, as inferred4,5 from density fluctuations of the cosmic microwave background and seen during the first 2–3 billion years of the Universe in the so-called ‘Lyman α forest’6,7 (a dense series of intervening H i Lyman α absorption lines in the optical spectra of background quasars). A theoretical solution to this paradox locates the missing baryons in the hot and tenuous filamentary gas between galaxies, known as the warm–hot intergalactic medium. However, it is difficult to detect them there because the largest by far constituent of this gas—hydrogen—is mostly ionized and therefore almost invisible in far-ultraviolet spectra with typical signal-to-noise ratios8,9. Indeed, despite large observational efforts, only a few marginal claims of detection have been made so far2,10. Here we report observations of two absorbers of highly ionized oxygen (O vii) in the high-signal-to-noise-ratio X-ray spectrum of a quasar at a redshift higher than 0.4. These absorbers show no variability over a two-year timescale and have no associated cold absorption, making the assumption that they originate from the quasar’s intrinsic outflow or the host galaxy’s interstellar medium implausible. The O vii systems lie in regions characterized by large (four times larger than average11) galaxy overdensities and their number (down to the sensitivity threshold of our data) agrees well with numerical simulation predictions for the long-sought warm–hot intergalactic medium. We conclude that the missing baryons have been found.
Access optionsAccess options
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work is based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and NASA. F.N. and M.E. acknowledge support from NASA grant NNX17AD76G. S.M. acknowledges NASA grant NNX16AF49G. S.B. acknowledges financial support through agreement ASI-INAF n.2017-14-H.0 and an INFN INDARK grant. Y.K. thanks INAOE for the support offered during a sabbatical visit in 2017 and acknowledges support from grant DGAPA-PAPIIT 106518 and from programme DGAPA-PASPA. R.C. acknowledges support from NSF grant AST-1515389.Reviewer information
Nature thanks R. Davé and T. Fang for their contribution to the peer review of this work.
Extended data figures and tables
Broadband, unfolded RGS1 (black points; 1σ error bars) and RGS2 (red points; 1σ error bars) spectra (in bins with a signal-to-noise ratio per bin ≥20) and best-fitting models (black and red histograms) for the blazar 1ES 1553+113, in physical units. Blue arrows mark Galactic absorption lines; green and magenta arrows indicate absorption lines from our WHIM System 1 and System 2, respectively. FE, source flux (power per unit area and energy).
Normalized raw RGS1 (black points; 1σ error bars) and RGS2 (red points; 1σ error bars) data (in bins with a signal-to-noise ratio per bin ≥30) of the blazar 1ES 1553+113, in the wavelength interval λ = 26−32 Å. Thick dashed curves are the RGS1 (black) and RGS2 (red) best-fitting models, folded through the response functions of the RGSs. Thin solid curves at the bottom of the graph are RGS1 (black) and RGS2 (red) effective areas (in arbitrary units), showing instrumental features due to cool pixels in the dispersing detectors. Of the five absorption lines shown, only the weak O vii Heβ at zX = 0.4339, and only in RGS1, can be affected by the presence of an instrumental feature.
Outcome of a Monte Carlo procedure consisting of 10,000 evaluations of the single-line statistical significance of an unresolved (that is, broadened to the instrument line spread function) Gaussian added to the best-fitting continuum-plus-line model of the RGS spectrum of 1ES 1553+113 (see Methods). At each run, the line position is frozen at a random position in the range 8–33 Å, and its EW is allowed to vary freely from negative (emission) to positive (absorption). The data distribution is symmetric but slightly flatter than the red curve that shows the expected normal distribution for a standard deviation of unity. The normal distribution that best fits our Monte Carlo results has a standard deviation of 1.15 (green curve).
a, b, RGS spectra of 1ES 1553+113 obtained in the 2015 (black) and 2017 (red) observations for 800 ks and 950 ks, respectively, centred around the O vii Heα (a) and Heβ (b) transitions of System 1. The two lines are consistent with no variability between the two epochs, within their 1σ errors (Extended Data Table 2).
Histogram of photometric redshifts of r′ > 23.5 galaxies within cylindrical volumes with base radii of 500 kpc at each redshift interval and a line-of-sight depth of Δz = 0.075 (at 1σ redshift accuracy), centred on the line of sight to the blazar 1ES 1553+113. The black curve is the average number of galaxies with r′ > 23.5 expected within the explored volumes at each redshift bin, based on the galaxy logN–logS from ref. 11 (where N is the number of galaxies with surface brightness ≥S and we use a common but conservative B−r′ = 1 for all galaxy types, where B is the galaxy’s B magnitude). Vertical dashed green and blue lines indicate the redshifts of System 1 and System 2, respectively.
OSIRIS-GTC mosaic (in band r′) of the 12′ × 12′ field surrounding the line of sight to 1ES 1553+113 (indicated by the magenta circle). Green circles have radii of 500 kpc and 1.75 Mpc at z = 0.4125. Thin yellow circles highlight the positions of galaxies with photometric redshift estimates within the z = 0.375–0.450 interval, whereas thicker circles highlight spectroscopic redshifts (solid, confirmed; dashed, unconfirmed).
Same as Extended Data Fig. 5, but for System 2, in the z = 0.300–0.375 interval.