The quiescent intracluster medium in the core of the Perseus cluster

X-ray observations of the core of the Perseus cluster reveal a remarkably quiescent atmosphere in which the gas has a line-of-sight velocity dispersion of about 164 kilometres per second in the region 30–60 kiloparsecs from the central nucleus; turbulent pressure support in the gas is four per cent of the thermodynamic pressure, necessitating only a small correction to the total cluster mass determined from hydrostatic equilibrium. The Hitomi collaboration reports X-ray observations of the core of the Perseus cluster of galaxies — the brightest X-ray-emitting cluster in the sky. Such clusters typically consist of tens to thousands of galaxies bound together by gravity and are studied as models of both small-scale cosmology and large-scale astrophysical processes. The data reveal a remarkably quiescent atmosphere, where gas velocities are quite low, with a line-of-sight velocity dispersion of about 164 kilometres per second at a distance of 30–60 kiloparsecs from the central nucleus. Clusters of galaxies are the most massive gravitationally bound objects in the Universe and are still forming. They are thus important probes1 of cosmological parameters and many astrophysical processes. However, knowledge of the dynamics of the pervasive hot gas, the mass of which is much larger than the combined mass of all the stars in the cluster, is lacking. Such knowledge would enable insights into the injection of mechanical energy by the central supermassive black hole and the use of hydrostatic equilibrium for determining cluster masses. X-rays from the core of the Perseus cluster are emitted by the 50-million-kelvin diffuse hot plasma filling its gravitational potential well. The active galactic nucleus of the central galaxy NGC 1275 is pumping jetted energy into the surrounding intracluster medium, creating buoyant bubbles filled with relativistic plasma. These bubbles probably induce motions in the intracluster medium and heat the inner gas, preventing runaway radiative cooling—a process known as active galactic nucleus feedback2,3,4,5,6. Here we report X-ray observations of the core of the Perseus cluster, which reveal a remarkably quiescent atmosphere in which the gas has a line-of-sight velocity dispersion of 164 ± 10 kilometres per second in the region 30–60 kiloparsecs from the central nucleus. A gradient in the line-of-sight velocity of 150 ± 70 kilometres per second is found across the 60-kiloparsec image of the cluster core. Turbulent pressure support in the gas is four per cent of the thermodynamic pressure, with large-scale shear at most doubling this estimate. We infer that a total cluster mass determined from hydrostatic equilibrium in a central region would require little correction for turbulent pressure.

were pinned to an absolute scale via extrapolation of a subsequent calibration of the whole array 10 days later using illumination by another 55 Fe source mounted on the filter wheel. (For more details, see Methods.) We used a subset of the Perseus data closest to that calibration to derive the velocity map. For the line-width determination, we used the full dataset to minimize the statistical uncertainty, and applied a scale factor to force the Fe xxv Heα complex from the cluster to have the same energy in all pixels. This minimizes the gain uncertainty in the determination of the velocity dispersion, but also removes any true variations of the line-of-sight velocity of the intracluster medium across the field.
A 5.5-8.5-keV spectrum of the full 3 arcmin × 3 arcmin field is shown in Fig. 1. This spectrum shows a thermal continuum with line emission 10  We adopt a minimally model-dependent method for spectral fitting, and represent the Fe xxv Heα, Fe xxv Heβ and Fe xxvi Lyα complexes in the spectrum with a set of Gaussians with free normalizations and energies fixed either at redshifted laboratory values (for He-like Fe) or theoretical values (for H-like Fe); see Extended Data Table 1. Figure 2 shows the profiles of these lines in a spectrum obtained from the outer region of the Perseus core, which excludes the active galactic nucleus (AGN) and prominent inner bubbles (Fig. 3). To measure the line-of-sight velocity broadening (Gaussian σ), we fitted the high-signal-to-noise, Fe xxv Heα line complex using nine Gaussians associated with lines known from atomic physics and obtain 164 ± 10 km s −1 (all uncertainties are quoted at the 90% confidence level). The widths of the 6.7008-keV resonance line and the 6.617-keV blend of faint satellite lines are allowed to be separate from the rest of the lines. The effect of the thermal broadening expected from the observed 4-keV plasma has been removed (alone it corresponds to 80 km s −1 ). Conservative estimates of the uncertainty in energy resolution (FWHM of 5 ± 0.5 eV) result in a systematic uncertainty range in the turbulent velocity of only ±6 km s −1 , because the total measured line width is roughly twice the instrumental broadening, which adds to the astronomical broadening in quadrature. Uncertainties in plasma temperature add only a further ±2 km s −1 . The statistical scatter in the energy-scale alignment of co-added pixels results in an overestimate of the true broadening by not more than 3 km s −1 . The finite angular resolution of the telescope in the presence of a velocity gradient across the cluster results in a small artificial increase in the measured dispersion (see Methods) that is difficult to quantify at this stage.
The Fe xxvi Lyα complex alone (554 counts) yields a consistent velocity broadening of 160 ± 16 km s −1 . A search for spatial variations in velocity broadening using the Fe xxv Heα lines reveals that Letter reSeArCH we measured a ratio of fluxes in Fe xxv Heα resonant and forbidden lines of 2.48 ± 0.16, which is lower than the expected value in optically thin plasma (for kT = 3.8 keV, the current APEC 16 and SPEX 17 plasma models give ratios of 2.8 and 2.9-3.6) and suggests the presence of resonant scattering of photons 18 . On the basis of radiative transfer simulations 19 of resonant scattering in these lines, such resonance-line suppression is in broad agreement with that expected for the measured low line widths, providing independent indication of the low level of turbulence. Uncertainties in the current atomic data, as well as more complex structure along the line of sight and across the region, complicate the interpretation of these results, which we defer to a future study.
A velocity map (Fig. 3b) was produced from the absolute energies of the lines in the Fe xxv Heα complex, using a subset of the data for which such a measurement was reliable, given the limited calibration (see Methods). We find a gradient in the line-of-sight velocities of about 150 ± 70 km s −1 , from southeast to northwest of the SXS field of view. The velocity to the southeast (towards the nucleus) is 48 ± 17 (statistical) ± 50 (systematic) km s −1 redshifted relative to NGC 1275 (redshift z = 0.01756) and consistent with results from Suzaku CCD (charge-coupled device) data 20 . Our statistical uncertainty on relative velocities is about 30 times better than that of Suzaku, although there is a systematic uncertainty on the absolute SXS velocities of about 50 km s −1 (see Methods). all 1-arcmin-resolution bins have broadening of less than 200 km s −1 .
With just a single observation we cannot comment on how this result translates to the wider cluster core.
The tightest previous constraint on the velocity dispersion of a cluster gas was from the XMM-Newton reflection grating spectrometer, giving 11,12 an upper limit of 235 km s −1 on the X-ray coolest gas (that is, kT < 3 keV, where k is Boltzmann's constant and T is the temperature) in the distant luminous cluster A1835. These measurements are available for only a few peaked clusters 13 ; the angular size of Perseus and many other bright clusters is too large to derive meaningful velocity results from a slitless dispersive spectrometer such as the reflection grating spectrometer (the corresponding limit for Perseus 13 is 625 km s −1 ). The Hitomi SXS achieves much higher accuracy on diffuse hot gas owing to it being non-dispersive.
We measure a slightly higher velocity broadening, 187 ± 13 km s −1 , in the central region ( Fig. 3a) that includes the bubbles and the nucleus. This region exhibits a strong power-law component from the AGN, which is several times brighter than the measurement 14 made in 2001 with XMM-Newton, consistent with the luminosity increase seen at other wavelengths. A fluorescent line from neutral Fe is present in the spectrum ( Fig. 1), which can be emitted by the AGN or by the cold gas present in the cluster core 15 . The intracluster medium has a slightly lower average temperature (3.8 ± 0.1 keV) than the outer region (4.1 ± 0.1 keV). By fitting the lines with Gaussians,  NGC 1275 hosts a giant (80-kpc wide) molecular nebula seen in CO and Hα data with a total cold-gas mass of several 10 10 M ⊙ , which dominates the total gas mass out to a 15-kpc radius. The velocities of that gas 21,22 are consistent with the trend of the SXS bulk shear, suggesting that the molecular gas moves together with the hot plasma. (More details of the X-ray spectra and imaged region are provided in Extended Data Figs 1-8.) The large-scale bulk shear over the observed 60-kpc field is of comparable amplitude to the small-scale velocity dispersion that we derive for the outer region. The dispersion can be due to gas flows around the rising bubble at the centre of the field 23,24 , a velocity gradient in the cold front 25 contained in this region, sound waves 26,27 , turbulence 28 or galaxy motions 29 . The large-scale shear could be due to the buoyant AGN bubbles or to sloshing motions of gas in the cluster core that give rise to the cold front 25 .
If the observed dispersion is interpreted as turbulence driven on scales comparable with the size of the largest bubbles in the field (about 20-30 kpc), then it is in agreement with the level inferred 28 from X-ray surface brightness fluctuations. In this case, our measured velocity dispersion suggests that turbulent dissipation of kinetic energy can offset radiative cooling. However, assuming isotropic turbulence, the ratio of turbulent pressure to thermal pressure in the intracluster medium is low at 4%. Such low-velocity turbulence cannot spread far (<10 kpc) across the cooling core during the fraction (4%) of the cooling time in which it must be replenished, so the turbulent-dissipation mechanism requires that turbulence be generated in situ throughout the core. Another process is needed to transport energy from the bubbling region. The observed level of turbulence is also sufficient to sustain the population of ultrarelativistic electrons that give rise to the radio synchrotron mini-halo observed in the Perseus core 30 .
A low level of turbulent pressure measured for the core region of a cluster, which is continuously stirred by a central AGN and gas sloshing, is surprising and may imply that turbulence in the intracluster medium is difficult to generate and/or easy to damp. If true throughout the cluster, then this is encouraging for total mass measurements, which depend on knowledge of all forms of pressure support, and for cluster cosmology, which depends on accurate masses.
The Hitomi spacecraft lost its ground contact on 2016 March 26, and later the recovery operation by JAXA was discontinued.

MetHOds
Gain corrections and calibration. Gain scales for each pixel were measured in ground calibration using a series of fiducial X-ray lines at several detector heatsink temperatures (a single spectral energy reference is sufficient to determine the effective detector temperature and thus the appropriate gain curve to use). As the heat-sink temperature varies, the gain of each pixel tracks the gain change in the separate calibration pixel that is continuously illuminated by a dedicated 55 Fe source. However, time-varying differential thermal loading of the pixels changes their gains by different factors. Thus, use of the gain history of the calibration pixel alone can be insufficient to correct the gain scale of the main array.
The Perseus observation used for this work was performed in two parts, 7 days apart, during which the gain of the calibration pixel changed by 0.6%. Ten days after the final observation, a fiducial measurement for the full array was obtained with an on-board 55 Fe source mounted on a filter wheel. To relate this calibration to the two Perseus observations, a two-stage approach was used. First, a correction factor was applied to all pixels using the gain history of the calibration pixel. Second, the differential pixel-pixel gain error was removed using the science observation itself. To do this, the two Perseus observations were subdivided, and the He-like Fe complex was fitted for each pixel in each subset. The time-dependent relative gain of each pixel (compared to the gain correction of the calibration pixel) was then linearly fitted and extrapolated to the later full-array calibration. The full dataset was then corrected using this time-dependent gain function, and the fitting errors were incorporated into the error analysis. To validate this approach, we compared the first observation, which required a substantial gain correction, to the second, for which the instrument was much closer to thermal equilibrium and thus required much less correction. In the first case, the bulk velocity uncertainties are dominated by the uncertainties in the gain correction, whereas, in the second, the uncertainties are dominated by the fit to the He-like Fe complex. The results for the two datasets agree for both bulk velocity and velocity dispersion, indicating that this is a robust approach. For the absolute velocity maps, we are presenting only the result from the second observation of the two used in this work, which requires the least correction and thus has the smallest uncertainty. Note that the limited gain calibration results in pixel-to-pixel uncertainty of 50 km s −1 on the absolute velocities.
To derive the absolute velocities of the cluster, we applied the heliocentric correction, which was −26.4 km s −1 for the observation used for velocity mapping. The orbital motion of the satellite around Earth averages out. Our velocities are compared to the heliocentric velocity of NGC 1275 in Fig. 3 and Extended Data Fig. 6.
An additional validation of our calibration comes from a weak background line in the whole-array spectrum from stray 55 Fe X-rays, which, after the above procedure, is observed at the correct energy to ±1.8 eV (equivalent to ±90 km s −1 ). Although the line is not strong enough to verify the calibration of individual pixels (because there should be about 68 counts in this line, non-uniformly distributed across the array), it is a convincing check of the approach.
To determine velocity dispersion, we applied additional scale factors for each SXS pixel to match the apparent energies of the cluster Fe Heα complex in order to remove any residual gain errors at the relevant energy. This also removes the effect of true bulk shear. Pixels were then combined in physically relevant regions to minimize statistical uncertainties.
We presumed a fixed energy resolution of 5.0-eV FWHM in all the analyses. By comparing the line widths in the first and second parts of the observation to estimate the broadening from residual gain drift, and accounting for the variation in resolution of the calibration pixel in time over the observation and during the later calibration of the array, we estimate that the composite resolution of the array and of the separately analysed central and outer regions is bounded with high confidence between 4.5-eV and 5.5-eV FWHM. This 10% uncertainty in instrumental broadening produces a much smaller fractional uncertainty in velocity broadening because the instrumental broadening is roughly half as large as the astronomical broadening, and adds in quadrature with it.
The error from gain aligning the different pixels in a region is smaller than the uncertainty in instrumental broadening because of the small statistical errors in the determination of the scale factor at the Fe Heα complex (in an outer pixel, equivalent to 30 km s −1 at 90% confidence). Adding the spectra of multiple pixels with the same velocity uncertainty will add 30 km s −1 of noise in quadrature with the measured broadening, producing an overestimate by no more than 3 km s −1 .
Our velocity dispersion measurements exclude velocity variations across the field on scales of 20 kpc and greater because of the aforementioned self-calibration procedure, but integrate over all scales along the line of sight (weighted by X-ray emissivity, which essentially limits integration to the cluster core). Any comparison with simulations will have to take these into account. Effects of angular resolution. The point spread function (PSF) of the telescope has a 1.2′ half-power diameter as measured during ground calibration. This means that regions used for spectral extraction get photons not only from the corresponding cluster regions in the sky, but also from the surrounding regions. The PSF image is shown in the right panel of Extended Data Fig. 5, centred on the SXS pixel that contains the cluster peak. By comparing the PSF with the middle panel of Extended Data Fig. 5, which shows the image in the Fe Heα line (which comes mostly from the gas, as opposed to the central AGN), we see that the diffuse emission of the cluster is resolved. However, small regions in the detector, such as the 1′ × 1′ regions of the velocity map shown in Fig. 3b and Extended Data Fig. 6, are significantly correlated. The fraction of the emission that originates in a given 1′ cluster region and ends up in the corresponding 1′ detector region is 36%-37%, with the rest spreading over the surrounding regions. For example, for the region marked '−60' in Extended Data Fig. 6, the scattered contribution from the neighbouring region marked '78' is 23% of the flux that originates in region −60 itself; the contribution from −60 into 78 is a similar 22% of the flux that originates and stays in 78. Regions adjacent to the brightness peak (which is in region marked '48') are most affected-the region marked '94' has a ratio of photons scattered in from 48 to its own photons of 27%. This means that the true line-of-sight velocity gradients on a 1′ scale have to be steeper than what we measure, but not by much. Scattered flux from an adjacent region with a large velocity difference (for example, from region 78 to region −60) should contribute lines at a different velocity in the spectrum, but such contributions would be very small compared to the observed line-of-sight velocity dispersion of >160 km s −1 . Correction of the PSF effects is left for future work.
The PSF scattering also has the subtle effect of inflating our measured value of velocity dispersion. Although the self-calibration procedure that aligns the Fe Heα lines in each pixel (as described above) removes most of the velocity-gradient contribution from the measured velocity dispersion, it does so after the PSF scattering has occurred and mixed the photons from regions with different line-of-sight velocities, so that contribution remains. Pointing. For this early observation, accurate pointing direction of the spacecraft was not available. We therefore assumed that the observed brightness peak in the SXS image is the AGN in NGC 1275. The resulting uncertainty of the sky coordinates should be less than 15″. The peak of the source determined in short time intervals revealed a small drift of the source in the detector image, within the above coordinate uncertainty. It causes image smearing that is insignificant compared to the effect of PSF scattering. . This motivates the model-independent approach adopted in the manuscript for determining the line widths. Error bars are 1 s.d.