Mapping the Stellar Kinematics in the Central 240 Parsecs of M87 with the James Webb Space Telescope

TL;DR

Using JWST/NIRSpec IFU, the study maps the stellar kinematics in the central 240 pc of M87 at high spatial resolution ($0.05''$ spaxels) and high $S/N$, deriving the LOSVD up to Gauss-Hermite order $h_8$. The authors implement a robust data-reduction and cube-construction workflow, including manual outlier masking and a union-based data-cube merge, to extract $V$, $\,\sigma$, and $h_3$–$h_8$ from CO bandheads while masking Ca I. They find $V \sim \pm 45\ \mathrm{km\,s^{-1}}$, a central $\,\sigma \sim 420\ \mathrm{km\,s^{-1}}$ at $r=0.45''$, and a mild central $h_4$ enhancement, with results consistent with large-scale KCWI kinematics, providing a seamless kinematic transition from parsec to kiloparsec scales. The combined NIRSpec and KCWI data establish a comprehensive multiscale kinematic map that will enable rigorous stellar-dynamical modeling to constrain the SMBH mass, mass-to-light ratio, and orbital distribution, reinforcing M87 as a cornerstone of SMBH–host galaxy relations.

Abstract

The supermassive black hole (SMBH) in the giant elliptical galaxy M87 is one of the most well-studied in the local universe, but the stellar- and gas-dynamical SMBH mass measurements disagree. As this galaxy is a key anchor for the upper end of the SMBH mass$-$host galaxy relations, we revisit the central $3''\times 3''$ ($\sim 240\times240$ pc) region of M87 with the Near Infrared Spectrograph (NIRSpec) integral field unit (IFU) on the James Webb Space Telescope (JWST). We implement several improvements to the reduction pipeline and obtain high signal-to-noise spectra ($S/N \sim 150$) in single $0.''05 $ spaxels across much of the NIRSpec field of view. We measure the detailed shape of the stellar line-of-sight velocity distribution, parameterized by Gauss-Hermite moments up to $h_8$, in $\sim 2800$ spatial bins, substantially improving upon the prior high angular resolution studies of the M87 stellar kinematics. The NIRSpec data reveal velocities with $V \sim \pm 45$ km s$^{-1}$, velocity dispersions that rise sharply to $\sim$$420$ km s$^{-1}$ at a projected radius of 0.$''$45 (36 pc), and a slight elevation in $h_4$ toward the nucleus. We comprehensively test the robustness of the kinematics, including using multiple velocity template libraries and adopting different polynomials to adjust the template spectra. We find that the NIRSpec stellar kinematics seamlessly transition to recently measured large-scale stellar kinematics from optical Keck Cosmic Web Imager (KCWI) IFU data. These combined NIRSpec and KCWI kinematics provide continuous coverage from parsec to kiloparsec scales and will critically constrain future stellar-dynamical models of M87.

Figures (8)

• Figure 1: Images of M87 taken at two slices of the data cube when no additional bad pixels are flagged in the 2D detector images before building the cube (left) and the same two slices of the cube after applying our supplemental masking of bad pixels in each of the calibrated detector images prior to assembling the data cube (right). The red boxes in the left images denote positive and negative artifacts that remain the final cube, but that do not appear in the data cube on the right when using our approach. The wavelength of each slice is given at the bottom of the left images, the spaxels are 005 in side, and the images are oriented with north up. The feature to the northwest of the M87 nucleus is a jet knot.
• Figure 2: A comparison of the spectrum extracted from a single 005 spaxel in the final M87 data cube when using the standard method (red) and our union method (black) for building the cube. With the typical approach, artifacts are commonly seen adjacent to the NRS1/NRS2 detector gap. The detector gap is the white space with missing data, at $\sim$$2.45\ \mu$m above. With the union method, the spectra are clean and enable the secure measurement of the stellar LOSVD from the CO bandheads near the detector gap.
• Figure 3: Top: A spectrum from a single 005 spaxel in the M87 data cube constructed using the default JWST pipeline (light blue) and in the final data cube generated using the modifications to the pipeline described in Sections \ref{['sec:obs_reduction']}, \ref{['subsec:removing_outliers']}, and \ref{['subsec:cube_building']} (black). The latter spectrum exhibits noticeable improvements relative to the former. Middle: An example spectrum from our final data cube extracted from a single 005 spaxel just outside the nucleus, with the emission lines labeled in dark blue and the $^{12}$CO bandhead absorption features labeled in teal. The horizontal dashed gray line indicates the wavelength range over which the CO bandheads were fit, as described in Section \ref{['sec:stellar_kin']}. A zoom-in of this spectral region for the same spectrum is shown in the top right inset of the panel, along with a second spectrum extracted from a single spaxel at near the edge of the data cube. Even in single spaxels, just outside the nucleus and at the edge of the FoV, the $S/N$ is very high. Bottom: Typical nuclear spectra from single spaxels in the final M87 data cube. Wiggles are clearly seen, despite our 16 exposure, small-cycling dither pattern. The wiggles improve with increasing distance from the center, and are no longer observed in the top two panels, at $r$$\sim$045.
• Figure 4: The first eight GH moments used to describe the LOSVD are shown across the NIRSpec FoV. There are 2180 spatial bins for which a kinematic measurement was made, and 2232 of the spatial bins are composed of a single 005 spaxel. The central 045 and the jet knot to the northest have been masked. The maps are oriented with north up and east to the left.
• Figure 5: Left: Panels (a) and (b) compare even GH moments derived using the K2 OSIRIS and Winge GNIRS libraries to those from the PHOENIX library, all with Ca1 masked. Panel (c) contrasts moments obtained with and without masking Ca1, using PHOENIX templates. The dotted line indicates one-to-one correspondence. When Ca1 is masked, the GH moments are consistent across all libraries. However, omitting the mask with PHOENIX introduces a systematic offset in the kinematics. Although not shown, when using K2 OSIRIS and Winge GNIRS consistent kinematics are found regardless of whether Ca1 is masked. Right: Panels (d) and (e) show spectral fits to the same M87 spectrum (black) using PHOENIX, K2 OSIRIS, and Winge GNIRS templates, with Ca1 masked (gray band) and unmasked, respectively. The red lines show best-fit pPXF models, with residuals in blue. All models are a good match to data when Ca1 is masked. When Ca1 is not masked, the PHOENIX templates cannot reproduce Ca1 well, whereas the two empirical libraries are able to do so. While the same single spaxel M87 spectrum is shown, the ending wavelengths differ due to the varying spectral coverage of the template libraries, with K2 OSIRIS cutting off earlier than the others.