An X-ray-emitting protocluster at z ≈ 5.7 reveals rapid structure growth

Abstract

Galaxy clusters are the most massive gravitationally bound structures in the universe and serve as tracers of the assembly of large-scale structure¹. Studying their progenitors, protoclusters, sheds light on the earliest stages of cluster formation. However, detecting protoclusters is demanding: their member galaxies are loosely bound and the emerging hot intracluster medium (ICM) may only be in the initial stages of virialization^2–4. Recent James Webb Space Telescope (JWST) observations located several protocluster candidates by identifying overdensities of z ≳ 5 galaxies^5–9. However, none of these candidates was detected by X-ray observations, which offer a powerful way to unveil the hot ICM. Here we report the combined Chandra and JWST detection of a protocluster, JADES-ID1, at z ≈ 5.68, merely one billion years after the Big Bang. We measure a bolometric X-ray luminosity of $L_{\mathrm{bol}}=\left(1.5_{-0.6}^{+0.5}\right)\times 10^{44}\mathrm{erg}{\mathrm{s}}^{-1}$ and infer a total gravitating mass of $M_{500}=\left(1.8_{-0.7}^{+0.6}\right)\times 10^{13}{M}_{\odot }$, making this system a progenitor of today’s most massive galaxy clusters. The detection of extended, shock-heated gas indicates that substantial ICM heating can occur in massive halos as early as z ≈ 5.7. Also, given the limited survey volume, the discovery of such a massive cluster is statistically unlikely¹⁰, implying that the formation of the large-scale structure must have occurred more rapidly in some regions of the early universe than standard cosmological models predict.

Discovery of a protocluster at z = 5.68, merely one billion years after the Big Bang, suggests that large-scale structure must have formed more rapidly in some regions of the early universe than previously thought.

Main

Protoclusters are the earliest phase in the assembly of large-scale cosmic structures. These systems mark a transition from initial galaxy overdensities to fully virialized galaxy clusters. X-ray and Sunyaev–Zel’dovich (SZ) observations have been widely used to identify clusters and protoclusters in the early universe, up to z ≈ 2.5, that are either gravitationally collapsed or in the process of collapsing^11–19. However, detecting protoclusters within the first billion years of the universe (z ≳ 5.5) remains a great challenge. At these early epochs, halos may not yet have experienced substantial virial heating (that is, shocks that bring the gas up to the virial temperature of the halo) or may only be in their earliest stages^1,3. This makes observing the faint hot ICM, the key signature of the onset of virial heating, difficult. Therefore, identifying the first protoclusters undergoing virialization necessitates a multi-wavelength effort. Although overdensity measurements reveal dense galaxy environments, unambiguous confirmation of the onset of gravitational collapse comes from the detection of ICM^20–22. In this work, we extend this approach by using JWST and Chandra data to investigate an even earlier protocluster, merely one billion years after the Big Bang.

JWST observations have revolutionized the study of high-redshift structures by identifying substantial populations of faint and distant galaxies^23–25. Its near-infrared imaging capability and accurate photometric redshifts allow the identification of galaxy overdensities, paving the way for detecting the earliest protoclusters. In the JADES field²⁶, one of JWST’s premiere regions, six protocluster candidates were identified in the redshift range z = 5–7 (ref. ⁸). Of these protoclusters, JADES-ID1 is the most compelling candidate, exhibiting high galaxy overdensity and cluster membership likelihood. A total of 66 galaxies are identified as potential members, with an inferred halo mass of $\log \left(M_{\mathrm{h}}/M_{\odot }\right)=13.28_{-0.34}^{+0.37}$, the highest among the JADES fields⁸. Because only JADES-ID1 has the richness and halo mass to produce a detectable ICM signal in the JADES field, our X-ray analysis targets this protocluster candidate. The Chandra Deep Field South (CDFS)/JADES field uniquely combines deep JWST imaging with the deepest Chandra exposure over a single footprint, making it at present the only survey volume in which X-ray emission from the ICM of high-redshift protoclusters could, in principle, be detected. Further details on the identification of JADES-ID1 and quantitative tests ruling out the presence of projected foreground galaxy groups and clusters are presented in Methods section ‘JWST characterization of the JADES-ID1 protocluster’.

Within a projected radius of 42″ (≈250 kpc) around the JADES-ID1 centroid, we measure the local galaxy overdensity of δ_gal = 3.9 relative to the mean galaxy density across the JADES field at the same redshift slice⁸. In the inner 21″ (≈125 kpc) region (coinciding with the detected X-ray emission), the overdensity increases to δ_gal = 4.5. This corresponds to a 4.2σ overdensity detection (for details, see Methods section ‘JWST characterization of the JADES-ID1 protocluster’). At z ≈ 5.7, this overdensity is exceptionally rare. Above z >5 , the expected amplitude of galaxy overdensities falls off sharply owing to the declining galaxy number density, increasing photometric incompleteness and the fact that large-scale structure has had less time to collapse. Although the large galaxy overdensity associated with JADES-ID1 indicates a nascent protocluster, X-ray data can confirm whether the virialization has begun. Indeed, detecting hot ICM, best accomplished through sensitive X-ray imaging, can trace the gravitational potential of the protocluster and allows a more precise determination of its centroid, luminosity and total mass.

To examine the X-ray-emitting properties of JADES-ID1, we used deep Chandra Advanced CCD Imaging Spectrometer (ACIS-I) X-ray observations of the CDFS. The CDFS overlaps with the JADES field and is the site of the deepest X-ray observation ever conducted^27,28. The data were analysed with CIAO tools²⁹. First, the 99 individual observations (Extended Data Table 1) were reprocessed with the chandra_repro tool. These observations were then merged into a single event file, which had a total exposure time of 6.55 Ms. Next we generated images and their corresponding exposure maps in the 0.5–2.0-keV and 3–7-keV energy ranges. Bright X-ray point sources were excluded from the analysis. A complete description of the data analysis is presented in Methods section ‘Chandra data analysis’.

Extended Data Table 1.

List of the analysed Chandra observations

Figure 1 presents a multi-wavelength view of the JADES-ID1 protocluster and its surroundings. Although the JWST data reveal a galaxy overdensity at z ≈ 5.7, an initial inspection of the exposure-corrected 0.3–2.0-keV Chandra image does not show bright X-ray emission cospatial with JADES-ID1. It also indicates that none of the high-redshift galaxies are detected as individual X-ray point sources and a stack of the galaxy positions likewise shows no statistically significant signal (Methods section ‘Excluding alternative origins of the X-ray emission’). To better assess the presence of faint diffuse emission expected from a z ≈ 5.7 protocluster, we carry out an in-depth imaging analysis. To do this, we first filled the location of the excluded point sources using the dmfilth tool using the ‘global’ method. Next we subtracted the background components using an annulus with 40″–110″ radii around JADES-ID1 (Methods section ‘Chandra data analysis’). Finally, we applied Gaussian smoothing to the image with a kernel size of 15 pixels.

Fig. 1. Multi-wavelength view of the JADES-ID1 protocluster and its surroundings.

a, Composite JWST image of the JADES field with a 45″ × 45″ (265 × 265 kpc) box marking JADES-ID1. Scale bar, 30″. b, Zoom-in on that region. Scale bar, 10″. c, The exposure-corrected 0.3–2.0-keV band Chandra image of the same region. Scale bar, 10″. d, A JWST/Chandra overlay of JADES-ID1, in which the Chandra image has been processed by filling point sources, subtracting the background and applying Gaussian smoothing with a kernel size of 15 pixels. Scale bar, 10″, 59 kpc at z = 5.68. e, The same processed Chandra image, with the locations of likely cluster member galaxies highlighted. The Chandra image reveals extended X-ray emission that is spatially coincident with the galaxy overdensity identified by JWST. Scale bar, 10″.

The resulting image, shown in Fig. 1e, indicates that large-scale diffuse X-ray emission is present near the JWST-derived centroid. We find that the X-ray centroid (RA = 3 h 32 min 31.75 s, dec = −27° 46′ 51.5″) is offset by about 8″ (about 47 kpc) from the galaxy overdensity peak. Such separations are commonly observed in dynamically young or merging systems. Systematic studies of disturbed clusters find mean X-ray/optical offsets of about 75 kpc, with offsets reaching and exceeding roughly 100 kpc in the most unrelaxed cases^30–32. Even high-redshift systems, which are still in the early stages of collapse or undergoing mergers, exhibit comparable offsets^13,19. Furthermore, owing to the broad Gaussian smoothing kernel (15 Chandra pixels or 7.4″) applied to enhance faint emission, our X-ray centroid is uncertain at the few arcsecond level. The 0.3–2.0-keV Chandra image nonetheless clearly shows extended emission spatially coincident with the galaxy overdensity, suggesting the presence of a large-scale hot ICM. Motivated by the detection of extended emission, we quantify this emission through a multi-faceted approach: (1) we establish its extended nature by constructing a surface brightness profile; (2) we examine its presence in different X-ray energy ranges; and (3) we examine its spectral properties. We note that the unsmoothed, point-source-masked image was used for the further quantitative analysis.

To quantify the large-scale diffuse ICM emission associated with JADES-ID1, we constructed an exposure-corrected, azimuthally averaged surface brightness profile from the unsmoothed, point-source-masked 0.3–2.0-keV band Chandra image, centred on the X-ray centroid. The background was measured in an annulus with 40″–110″ (235–646 kpc at z = 5.68) radii. For the background subtraction, we accounted separately for vignetted (Milky Way foreground plus unresolved cosmic X-ray background) and non-vignetted (particle background) components by generating two exposure maps: one includes mirror vignetting and detector effects and the other only includes detector effects. We applied each map to its corresponding background components (see Methods section ‘Chandra data analysis’ for details). We also verified that the resulting surface brightness profile is insensitive to the exact choice of background region. The background-subtracted surface brightness profile, presented in Fig. 2, reveals extended X-ray emission within about 21″ (≈125 kpc) and a declining trend with radius from the centre. Beyond this radius, the signal-to-noise ratio declines and the emission becomes consistent with the background. Fitting this surface brightness profile with a β-model with fixed β = 0.6 yields a core radius of r_c = 7.1″ ± 3.9″ (42 ± 23 kpc). The observed 0.3–2.0-keV band emission is much more extended than the Chandra point spread function (90% of the encircled is contained within 1″), confirming the truly extended nature of this emission.

Fig. 2. X-ray surface brightness profile of the JADES-ID1 protocluster.

The 0.3–2.0-keV band profile was extracted in concentric annuli centred on the X-ray peak. The dashed line shows the best-fit β-model. The profile is corrected for exposure variations using the exposure map. The background is subtracted on the basis of the 40″–110″ (235–646 kpc at z = 5.68) annulus around the protocluster. The resulting background level is 8.03 × 10⁻¹⁰ photons s⁻¹ cm⁻² arcsec⁻². Extended emission is detected out to approximately 21″ (≈125 kpc). The error bars represent statistical uncertainties derived using the Gehrels approximation⁵⁸.

Within a 21″ aperture (≈125 kpc) of the JWST-derived centroid, in which the signal-to-noise ratio peaks, we measure 142 ± 45 net counts and 1,858 background counts, supporting the detection of an extended ICM. If this emission originates from plasma with a few keV temperature, we do not expect to detect X-ray emission in the 3–7-keV band. Indeed, at z ≈ 5.68, the observed 3–7-keV band corresponds to the 20.0–46.8-keV band in the rest frame of the cluster, in which such thermal emission is negligible. In agreement with this, the 3–7-keV band surface brightness profile does not show statistically significant X-ray emission in the same aperture. We detect only 51 ± 55 net counts and 2,947 background counts, consistent with the interpretation that the detected 0.3–2.0-keV band emission originates from a hot ICM.

Because of the relatively low count rate and the limited signal-to-noise ratio, extracting and fitting an X-ray spectrum of JADES-ID1 is not feasible. Instead, we investigate the spectral properties of the diffuse X-ray emission by calculating an X-ray hardness ratio HR = S/H, in which S and H correspond to the counts in the 0.5–1.2-keV and 1.2–2.0-keV bands, respectively (for details, see Methods section ‘X-ray hardness ratios’). Within the same 21″ source region, we obtain $\mathrm{HR}=1.84_{-1.84}^{+1.96}$. Although the large uncertainties only allow a broad estimate, these values imply an ICM temperature of at least about 2.5 keV, consistent with a hot protocluster.

We next combine the key X-ray observables—the 0.3–2.0-keV band detection, the 3–7-keV band non-detection and the declining surface brightness profile—into a combined likelihood map. This map quantifies the likelihood of random fluctuation mimicking all of these observational signatures simultaneously. For details on this map, see Methods section ‘Detection significance map’. In this map, shown in Fig. 3, JADES-ID1 stands out as the highest likelihood detection. Specifically, we find that the combined likelihood of detecting a statistical fluctuation with these parameters is 2.6 × 10⁻⁷, which corresponds to a 5.0σ detection. Although a few other regions also show elevated values on this map, those may trace structures at lower redshifts or they could represent statistical fluctuations. Thus, we conclude that the detected diffuse emission most likely originates from hot ICM associated with the JADES-ID1 protocluster. Finally, we combine the JWST-based overdensity significance (≈4.2σ)⁸ and Chandra X-ray detection likelihood (≈5.0σ) to establish the overall confidence level for JADES-ID1. We thus obtain a joint P-value of 3.4 × 10⁻¹², which corresponds to a 6.9σ detection.

Fig. 3. Combined detection likelihood of JADES-ID1 and its surroundings.

White contours show the 3–7-keV band emission. The map incorporates the probability of having a 0.3–2.0-keV band detection, a 3–7-keV band non-detection and a radially declining surface brightness profile. The position of the protocluster is highlighted with a magenta circle with 7″ radius. Scale bar, 1′ = 354 kpc.

On the basis of the detection of hot ICM associated with JADES-ID1, we derive the X-ray luminosity and total mass of the protocluster. For this, we use the net count rate within a 21″ radius, along with the Chandra exposure maps to calculate the X-ray flux. We assume an optically thin thermal plasma emission model with kT = 2 keV temperature, Z = 0.3 Z_⊙ metallicity^33,34 and a galactic hydrogen column density of N_H = 6.7 × 10¹⁹ cm⁻² (ref. ³⁵). This yields an absorption-corrected 0.3–2.0-keV band X-ray flux of f_X = (4.6 ± 1.5) × 10⁻¹⁶ erg s⁻¹ cm⁻². Using the redshift of z = 5.68 and applying a bolometric correction, we obtain a bolometric (0.01–100 keV) luminosity of $L_{\mathrm{bol}}=\left(1.5_{-0.6}^{+0.5}\right)\times 10^{44}\mathrm{erg}{\mathrm{s}}^{-1}$. To estimate the total mass and ICM temperature of the protocluster, we assume self-similar evolution and apply the L_bol–M₅₀₀ and L_bol–kT scaling relations obtained for ‘all clusters’ in the sample of ref. ³⁶. These result in a gas temperature of $kT=2.7_{-0.7}^{+0.5}\mathrm{keV}$, which is consistent with the lower limit from the analysis of the X-ray hardness ratio. Finally, on the basis of the luminosity, we estimate a total cluster mass of $M_{500}=\left(1.8_{-0.7}^{+0.6}\right)\times 10^{13}{M}_{\odot }$, which corresponds to R_500,c ≈ 80 kpc at z = 5.68. We note that the uncertainties include the statistical errors on the flux measurements and the uncertainties on the scaling relation parameters. However, these estimates carry several caveats. First, the assumption of self-similar evolution may not fully capture deviations from the scaling relations at high redshift, in which protoclusters are still assembling and undergoing early-stage accretion. Second, if JADES-ID1 is experiencing mergers or shocks, its ICM temperature and luminosity could be elevated^37–42. However, such merger-driven boosts only add a modest bias in the scaling relations and do not qualitatively alter our conclusions^38,43,44. Third, if the ICM temperature or metallicity differs from the assumed values, it could introduce a small change in the inferred parameters, although it would not alter our conclusions in any notable way. Finally, deviations from hydrostatic equilibrium, which are expected in dynamically young systems, could also affect the mass estimates, as non-thermal pressure support from turbulence or bulk flows may bias the derived masses low^45,46.

The JWST-Chandra detection of a protocluster and its hot ICM at z = 5.7 provides insights into the formation of the first galaxy clusters and constrains the evolution of the hot ICM. JADES-ID1 is a rare example of a protocluster caught in the early phases of virial heating^34,47. Theoretical studies suggest that protocluster shock heating (virialization) typically begins at lower redshifts (z ≈ 2–3) (ref. ²). However, the presence of a hot ICM in JADES-ID1 demonstrates that, at least in some of the most massive protoclusters, this process began merely one billion years after the Big Bang. Although the detection of a hot ICM indicates the onset of gravitational collapse, the protocluster is unlikely to be fully virialized at this early stage. The signatures of ICM heating in JADES-ID1 may be linked to its exceptional richness. Indeed, with 66 potential members, it is by far the richest protocluster candidate in the sample of ref. ⁸, suggesting that its deep gravitational potential could have accelerated both its collapse and the heating of its ICM.

The detection of a massive protocluster with $M_{500}=\left(1.8_{-0.7}^{+0.6}\right)\times 10^{13}{M}_{\odot }$ at such a high redshift is rather surprising. This detection provides an important data point for examining the abundance of large-scale structure systems in the early universe. To quantify how rare such systems are, we used the Tinker halo mass function and derived the expected number of massive halos within the z = 5–7 redshift range¹⁰. At these redshifts, the observable universe is predicted to host only a few dozen systems with M₅₀₀ = 10¹³ M_⊙ and less than one with M₅₀₀ = 2 × 10¹³ M_⊙ mass. Considering the small volume examined by the JADES/CDFS field, with the Chandra pointings covering a footprint of 16′ × 16′ (or a comoving volume of about 1.3 × 10⁶ Mpc³), the probability of detecting a 10¹³ M_⊙ protocluster in this region is about 4 × 10⁻⁵ under a ΛCDM cosmology⁴⁸. The probability drops even further to about 2 × 10⁻⁷ for a protocluster with a mass of 2 × 10¹³ M_⊙. To further illustrate the improbability of ≳10¹³ M_⊙ halo, we note that the most massive halo in our survey volume is expected to be about 10¹² M_⊙, roughly an order of magnitude below our X-ray-inferred mass. Although these halo abundance probabilities assume a cosmic average, the matter–density variance on the scales of the JADES field only modestly alters the detection likelihood (Methods section ‘JWST characterization of the JADES-ID1 protocluster’). Overall, the detection of the JADES-ID1 protocluster challenges our understanding of early-structure formation. This finding is analogous to recent JWST discoveries revealing an overabundance of unexpectedly luminous galaxies at z = 9–12 (refs. ^{23,24,49–51}). However, we note that the presence of extended, shock-heated gas in an approximately 10¹³ M_⊙ halo is the direct consequence of gravitational collapse and virial shocks, offering a probe of rapid halo assembly. Taken together, these results provide further important clues that, in some regions, structure may have formed more rapidly than previously thought.

Continued multi-wavelength synergy is essential for mapping the first protoclusters. Although JWST identifies high-redshift galaxy overdensities, next-generation X-ray missions could detect their extended ICM^52–54, whereas SZ experiments could reveal the thermal SZ imprint of the earliest protoclusters^55–57.

Methods

JWST characterization of the JADES-ID1 protocluster

Here we overview several key aspects of the JWST analysis. Specifically, we summarize the method used to define the JADES-ID1 centroid, quantify the statistical significance of its galaxy overdensity and present tests that rule out the presence of any substantial low-redshift foreground structures.

To define the centre of JADES-ID1 (ref. ⁸), we construct two-dimensional galaxy overdensity maps in narrow-redshift slices using the DETECTIFz algorithm⁵⁹, which uses Monte Carlo realizations of the redshift probability distribution function of each galaxy. We then identify the slice with the highest overdensity peak, z = 5.68 for JADES-ID1, and use the coordinates of that peak as the protocluster centre.

Next we quantify the rarity of the JWST-identified overdensity around JADES-ID1 by comparing it with field fluctuations. Within a projected radius of 42″ (≈250 kpc) around the JADES-ID1 centroid, the local galaxy overdensity is measured as δ_gal = 3.9 relative to the mean density across the JADES field in the same redshift slice⁸. Focusing on the inner 21″ (≈125 kpc) region, which is coincident with our X-ray aperture, the overdensity rises to δ_gal = 4.5. This is comparable with or exceeds those of previously confirmed protoclusters at similar redshifts⁷. To assess the statistical significance of this overdensity, we compared these values to the mean field density over 5.44 < z < 6.08 within a spherical volume of radius 410 kpc. Accounting for an approximately 30% cosmic variance, which is appropriate for the typical ultraviolet luminosities of candidate members^8,60, we performed 10⁶ Monte Carlo realizations. The chance of obtaining the observed overdensity of the observed galaxies by random fluctuation is 1.4 × 10⁻⁵, corresponding to a roughly 4.2σ detection. We note that our field baseline includes cluster members, so both δ_gal and its significance are slight underestimates relative to a purely field reference. This confirms that, at z ≈ 5.7, such a strong overdensity is exceptionally rare.

Although ref. ⁸ identifies a clear overdensity at z ≈ 5.68, we further verify that no other substantial structures exist at lower redshifts. To do this, we investigated the photometric galaxy catalogues based on JWST and Hubble Space Telescope (HST) observations in the JADES field. We measured the galaxy surface density within a 40″ × 40″ box centred on the X-ray emission peak, corresponding to the extent of the detected emission from the JADES-ID1 protocluster. We binned galaxies in redshift slices with width of Δz = 0.3 in the redshift range z = 0–6.6 and derived the surface density in each bin. To correct for redshift-dependent selection biases (most notably the increased completeness at lower redshifts), we carried out the same measurement in a large background region within the JADES footprint, while excluding the 40″ × 40″ region with the X-ray detection. The difference between the galaxy surface density at the position of the X-ray excess and the field average exhibits a single significant peak in the z = 5.25–6.23 redshift bin. There are no comparable galaxy overdensities at any other redshift bins. This result demonstrates the absence of any substantial foreground structure along the line of sight of the JADES-ID1 protocluster.

The halo detection probabilities presented in the main part of this paper are based on the cosmic mean density field. To estimate how local density fluctuations could bias these results, we estimate the variance of the matter density over a JADES-sized volume. We find σ(R) ≈ 0.059, which corresponds to a 6% typical fluctuation between different patches of this size. Although a 6% overdense field would allow structure to grow more rapidly, the chance of finding a 10¹³ M_⊙ halo would increase only from about 4 × 10⁻⁵ to about 3 × 10⁻⁴.

Chandra data analysis

To examine the presence and physical properties of the hot ICM associated with the protocluster JADES-ID1, we analysed 99 Chandra ACIS-I observations that cover the CDFS. The CDFS represents the deepest X-ray field ever observed. The list of analysed Chandra observations is given in Extended Data Table 1. We performed most of the data analysis using standard CIAO tools, specifically, we used the latest version of CIAO (4.17) and the Calibration Database (CALDB 4.11.6). The main steps of the X-ray data analysis followed those outlined in our previous studies^61,62. Below, we outline the main steps of the X-ray analysis.

The first step of the analysis was to reprocess each individual observation using the chandra_repro tool, thereby applying the latest calibration products. Next we identified and filtered high-background periods from the observations using the lc_clean routine by applying a 3σ threshold to remove fluctuations in the light curves. Because ACIS-I observations are not highly sensitive to solar flares, this step only reduced the exposure time by approximately 2%. The total cleaned exposure time of the dataset was 6.55 Ms.

Because we analyse and combine a large set of observations, small differences exist in the alignment between the individual Chandra observations. To account for this effect, we correct the absolute astrometry using the wcs_match and wcs_update tools. This step ensures that point sources are accurately aligned and exhibit a sharp point spread function, thereby minimizing contamination of the extended emission. To perform the astrometry correction, we cross-matched the positions of X-ray point sources in individual observations with the coordinates of guide stars in the same field. Using the coordinates of the X-ray–optical source pairs, we applied frame transformations for each Chandra observation, including transformations for rotation, scale and translation. We set the deepest observation, ObsID 8594, as the reference coordinate system. For the subsequent analysis, we used these astrometry-corrected event files.

The next step in the analysis was to combine the individual X-ray observations. For this, we used the merge_obs tool, which coadded the data, resulting in a merged event file with 6.55 Ms exposure time. This process also generated energy-filtered images in the 0.3–2.0-keV (soft) and 3–7-keV (hard) bands. In our analysis, we use the 0.3–2.0-keV range as our soft band because it maximizes sensitivity to a few-keV thermal plasma at z ≈ 5.7. Because most observations were taken relatively early in the Chandra mission, molecular contamination is minimal^63,64, allowing us to extend reliably down to 0.3 keV. For the hard band, we use the 3–7-keV band, thereby avoiding the 2–3-keV energy range in which the Au L fluorescence line complex dominates^65,66. This hard band is sensitive to unresolved active galactic nuclei (AGN), very hot ICM at high redshift or ICM emission from nearby clusters. We generated exposure maps for both energy ranges using this tool. These maps account for vignetting, molecular contamination, gaps between charge-coupled devices and the filtering of bad pixels. To construct the exposure maps, we assumed an optically thin thermal plasma model (APEC in XSPEC) with a galactic column density of N_H = 6.8 × 10¹⁹ cm⁻², a kT = 2 keV temperature and Z = 0.3 Z_⊙ solar abundance. The count images were then divided by the exposure maps, producing the exposure-corrected images used in the analysis.

To study the extended emission associated with JADES-ID1, we must ensure that bright X-ray point sources do not contaminate the extended emission from the protocluster. Therefore, we searched for resolved X-ray point sources (mostly originating from the cosmic X-ray background) using the wavdetect tool. We searched for sources using the wavelet scales of 1.0, 1.414, 2.0, 2.828, 4.0, 5.657 and 8.0, which allows the detection of sources on a wide range of spatial scales and a significance threshold of 10⁻⁶. Furthermore, we set the ellsigma parameter to 5. To carry out a comprehensive masking of all point sources, we cross-referenced our source list with the CDFS point source list presented in ref. ²⁸. We found that some of their faintest sources, detected using a lower significance threshold, fell below our initial detection criteria, so we added these to our point source catalogue. We then visually inspected the source regions and, for especially bright sources, enlarged the exclusion regions to fully encompass their point spread function wings. We note that, within the inner 21″ of JADES-ID1, only one moderately bright and two faint point sources are detected (Extended Data Fig. 1). The detected point source regions were excluded from the merged images. Given the sharp Chandra point spread function, our large exclusion regions account for ≳96% of the counts from each point source, implying that any residual counts have a negligible impact on the detected diffuse emission. Moreover, because spillover counts from point sources contribute similarly to both the source and the background regions, it does not bias our measurements in any notable way.

Extended Data Fig. 1. Chandra 0.3–2.0-keV band images of the JADES-ID1 field.

Left, merged Chandra image with detected X-ray point sources marked by cyan regions and candidate z ≈ 5.7 member galaxies indicated by white circles. The large white circle with 21″ radius denotes the aperture used for ICM analysis. Right, the same Chandra image after all X-ray point sources have been excised.

Because the X-ray emission from JADES-ID1 is relatively faint, precisely accounting for the background components is essential. The Chandra background comprises several components: the vignetted sky emission (Milky Way foreground and unresolved cosmic X-ray background) and the non-vignetted particle background. To separate these components, we generated two sets of exposure maps. One of them is a ‘full’ map including mirror vignetting and all ACIS detector corrections, whereas the other is a ‘detector-only’ map that does not include the vignetting term. We then split the total background accordingly and applied the full map to the sky component and the detector-only map to the particle component. In the deep CDFS field, Chandra resolves about 90% of the cosmic X-ray background, so we find that particle background accounts for about 77% of the 0.3–2.0-keV band and about 95% of the 3–7-keV band background.

To verify that the X-ray detection associated with JADES-ID1 is not driven by a small subset of observations, we split the CDFS Chandra data into two groups in two complementary ways. First, we divided the observations chronologically into two sets with approximately equal total exposure times. Owing to the build-up of the molecular contaminant on the ACIS optical blocking filter (which absorbs soft X-rays^63,64) and the soft, redshifted spectrum of the JADES-ID1 protocluster, the earlier dataset contains approximately 65% of the net counts, whereas the later one includes about 35%. Second, we randomly split the observations into two groups once again with approximately equal exposure times. In this case, as both subsets have the same average molecular contamination level, the net counts split roughly evenly between them. These tests demonstrate that the detected X-ray signal is not driven by a handful of observations or by temporal variation but arises uniformly across the Chandra dataset. This supports the conclusion that the X-ray emission associated with JADES-ID1 originates from a genuine, extended ICM.

X-ray hardness ratios

Owing to the relatively low number of net X-ray counts and the limited signal-to-noise ratios associated with the JADES-ID1 protocluster, it is not feasible to fit a full X-ray spectrum. Instead, we measure a hardness ratio to gain insight into the spectral properties of the emission. We define HR = S/H, in which S and H are the net counts obtained in the 0.3–1.2-keV and 1.2–2.0-keV bands, respectively.

We estimate the hardness ratio and its associated uncertainties using the Bayesian Estimation of Hardness Ratios (BEHR) code⁶⁷, which uses a Bayesian framework designed for low-count data. For JADES-ID1, this yields $\mathrm{HR}=1.84_{-1.84}^{+1.96}$. We note that the BEHR code evaluates the uncertainties through 10⁶ draws of Gibbs sampling that take into account the background counts.

To explore how the measured hardness ratio constrains the ICM temperature, we generated synthetic hardness ratio values using APEC models and using the exposure-averaged response files. When deriving the hardness ratio values, we covered a temperature range kT = 1–10 keV and a metallicity range Z = 0–1 Z_⊙. Within the kT ≈ 1–6 keV range, the hardness ratio is only weakly sensitive to metallicity. On the other hand, the hardness ratio is more sensitive to temperature: hotter plasmas yield stronger emission above 1.2 keV (rest frame 8 keV) and hence produce lower hardness ratio values. The upper limit of the measured HR ≈ 3.8 suggests that the ICM temperature is at least kT ≈ 2.5 keV, although a higher temperature is also possible. This is consistent with our X-ray luminosity-based estimates, but the large hardness ratio uncertainties do not allow tighter constraints on either temperature or metallicity. Hotter ICM temperatures may imply additional heating from the continuing mergers and other non-thermal processes (such as turbulent motions), both of which are expected in dynamically assembling clusters.

Excluding alternative origins of the X-ray emission

Although the diffuse X-ray properties of JADES-ID1 consistently demonstrate that it originates from hot ICM, we have tested, and ruled out, potential non-thermal or point source origin of this emission.

First, we cross-matched all resolved X-ray point sources with the positions of the candidate JADES-ID1 members within the source aperture (Extended Data Fig. 1). None of these high-redshift galaxies coincides with a resolved Chandra point source, implying that these galaxies do not host luminous AGN. Next, to investigate whether a population of fainter, individually undetected AGN is responsible for the diffuse emission, we stacked the Chandra counts at the location of each z ≈ 5.7 galaxy within the source aperture. To do this, we extract the source counts using a 1″ radius aperture and a local 3″–6″ annulus for background. This source region encircles about 90% of the source counts at the position of JADES-ID1. This analysis results in stacked net counts of −0.2 ± 8.2 in the 0.3 – 2.0-keV band and −10.9 ± 9.2 counts in the 3–7-keV band. We note that the non-detection of high-redshift AGN is consistent with previous analyses showing that most high-redshift galaxies are X-ray faint^68–70, with only rare exceptions^62,71. Thus, the absence of detection both individually and in stack firmly rules out any substantial contribution from either resolved or unresolved AGN.

A recent study⁷² suggested that inverse-Compton (IC) scattering of cosmic microwave background photons off radio galaxy electrons can effectively mimic the X-ray appearance of a high-redshift galaxy cluster. To test this scenario for JADES-ID1, we examined the MeerKAT 1.28-GHz map of the CDFS field⁷³.

At the location of the X-ray peak, we find no notable radio emission above the noise level. Using a typical intracluster magnetic field^74,75 of 4 μG (2 μG), IC scattering sufficient to reproduce the observed X-ray flux would imply a 1.4 GHz radio surface brightness that exceeds the MeerKAT detection limit by more than an order of magnitude (>12σ (about 4σ) per MeerKAT beam), which is not observed. We do, however, identify a faint radio point source 11.5″ offset from the X-ray peak with a flux density of 13.7 ± 1.4 μJy. This radio source is coincident with a galaxy at z = 1.173 (CANDELS J033230.91-274649.5)⁷⁶ and not with any of the z ≈ 5.7 potential cluster members. At this position, a faint X-ray point source is detected, which has been excised from the analysis of the diffuse emission. We do not expect substantial X-ray IC emission from the radio lobes of this galaxy. Even if we (incorrectly) attribute the entire radio flux at z = 5.7, we expect only about 5 (about 15) IC X-ray counts in either the soft or the hard bands, far below the observed X-ray signal. Thus, IC emission cannot account for the observed soft X-ray emission.

Taken together, the absence of any X-ray point source coinciding with a JADES-ID1 cluster member, the non-detection from the stacking analysis and the inconsistency between predicted IC flux emission and the observed flux as well all demonstrate that neither AGN nor IC scattering can explain the extended soft X-ray emission. Therefore, the only viable explanation remains thermal bremsstrahlung from a few-keV ICM at z ≈ 5.7 associated with the JADES-ID1 protocluster.

Detection significance map

The detection of extended X-ray emission in the 0.3–2.0-keV band, along with the declining surface brightness profile and the 3–7-keV band non-detection all provide strong evidence for the presence of hot ICM associated with JADES-ID1.

Combining all of this information, we construct a likelihood map to quantify the probability of hot intracluster gas associated with JADES-ID1 (Fig. 3). We define the combined likelihood for a high-redshift cluster as follows,

$$\mathit{L}={\mathit{S}}_{\mathrm{s}}(r<6^{″})\times {\left({\mathit{S}}_{\mathrm{s}}(6<r<20^{″})\times {\mathit{S}}_{\mathrm{s}}(0<r<6^{″}\right)}^{0.5}\times {\mathit{C}}_{\mathrm{h}}(<6^{″}),$$ 1

in which $\mathit{S}$ is the Poisson survival function and $\mathit{C}$ is the cumulative distribution function. The subscripts s and h denote the soft and hard bands, respectively. The first part, ${\mathit{S}}_{\mathrm{s}}(r<6^{″})$, defines the probability of detecting counts within 6″ above the background level (50″–75″). The second part, ${\mathit{S}}_{\mathrm{s}}(6^{″}<r<20^{″})$ describes the detection of counts between 6″ and 20″ but the background is assumed to be the 20″–30″ region. The third part, ${\mathit{S}}_{\mathrm{s}}(0^{″}<r<6^{″})$, analogous to the second part, describes the excess counts in the inner 6″ but used the 6″–20″ region as background. Together, the second and third parts will detect a rising profile towards the centre. The last part, ${\mathit{C}}_{\mathrm{h}}(<6^{″})$ brings in the probability of having hard band counts within 6″ consistent with the background (50″–75″). We note that, for the β-model shown in Fig. 2, the choice of 0″–6″ and 6″–20″ bins maximizes the signal-to-noise ratio and also makes it equal in the two bins.

For plotting purposes, we show the combined likelihood converted into a significance, using the inverse survival function. This map reveals that the combined likelihood of a statistical fluctuation with these parameters is 2.6 × 10⁻⁷, which corresponds to a 5.0σ detection. This indicates that the diffuse X-ray emission most likely originates from hot ICM associated with JADES-ID1. In Fig. 3, JADES-ID1 is the most notable detection. Specifically, this is the only region that satisfies all of our X-ray criteria (a soft band detection, a hard band non-detection and a declining surface brightness profile) and is cospatial with the independently identified JWST galaxy overdensity. All other apparent ‘hotspots’ on this map either lie outside the z ≈ 5.7 redshift slice or fail one or more of the X-ray criteria required to identify a high-redshift galaxy cluster, indicating that these regions are statistical fluctuations or correspond to lower-redshift structures tracing the large-scale structure of the universe.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-025-09973-1.

Author contributions

Á.B. analysed the Chandra observations, led the overall analysis and drafted the manuscript. G.S. assisted with the Chandra analysis, provided statistical tests and methods, contributed to the interpretation and played a notable role in writing the manuscript. Q.L. led the analysis of the JWST data and contributed to the interpretation and text of the manuscript. C.J.C. contributed to the interpretation and text of the manuscript.

Peer review

Peer review information

Nature thanks Fabio Vito and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Data availability

This paper uses a list of Chandra datasets, obtained by the Chandra X-ray Observatory, contained in the Chandra Data Collection (CDC) 489 10.25574/cdc.489. The JWST data of the JADES field is publicly available at http://archive.stsci.edu.

Code availability

Data reduction and analysis used standard, publicly available software packages: CIAO (Chandra Interactive Analysis of Observations) for processing and analysing the X-ray data, Python with scientific libraries for handling the data and plotting and SAOImage DS9 for image visualization.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41586-025-09973-1.

Supplementary information

The online version contains supplementary material available at 10.1038/s41586-025-09973-1.