betaFeedback

WorksPhysics and Astronomy

Powerful Radio Jets and Galaxy Assembly Align at Cosmic Noon

Abstract

Powerful radio galaxies at cosmic noon (1<z<21<z<2) offer a unique laboratory for investigating jet-driven supermassive black hole (SMBH) feedback on their host galaxies and surrounding environments. Unlike broad-lined quasars, radio galaxies lack a dominant nuclear point source in the rest-frame optical–UV, enabling direct constraints on host-galaxy structure and its relation to the radio jet. We analyze Hubble Space Telescope (HST) WFC3/IR (rest-frame optical) and WFC3/UVIS (rest-frame UV) imaging of the z>1z>1 3CR radio galaxies. Beyond confirming the classical rest-frame UV alignment effect, we find that the radio jets preferentially align with the major axes of the smooth rest-frame optical hosts, extending the alignment to the older stellar component. The projected axes of candidate mergers in eight systems preferentially follow the radio axes. This multiscale coherence suggests a common projected axis linking the radio source to host assembly, perhaps established by repeated mergers or anisotropic infall. We test scattered quasar light and jet-triggered star formation as origins of the UV emission along this axis. The much shallower-than-expected radial decline in UV surface brightness disfavors simple scattering as the dominant population-wide origin, while ionized-gas masses further disfavor electron scattering. The required dust masses are small relative to global reservoirs, so dust scattering remains feasible in individual regions. Comparing stellar-population ages and jet-propagation times, we identify seven UV-emitting regions as strong candidates for recently initiated jet-triggered star formation. Together, these alignments reveal an assembly-linked direction along which jets may shock or compress material and hidden-quasar radiation may illuminate it.

Introduction

The robust correlations observed between supermassive black hole (SMBH) mass and the stellar properties of galactic bulges have long suggested that SMBH growth is closely linked to the assembly and evolution of the galaxies that host them (e.g., Magorrian et al. [77]; Ferrarese & Merritt [57]; Gebhardt et al. [60]). Feedback from an active galactic nucleus (AGN) is often invoked as a possible mechanism coupling SMBH growth to the evolution of its host galaxy (see Kormendy & Ho [71], for a review). In radiatively efficient quasars, radiation pressure and associated accretion-disk winds can heat or expel cold gas, thereby reducing the fuel available for star formation and producing “negative feedback” (see reviews by Fabian [53]; Heckman & Best [64] (2023)). Under some conditions, however, shocks driven by quasar outflows propagating through a dense, clumpy interstellar medium (ISM) may instead produce “positive feedback” by promoting star formation. Although the shock initially heats and compresses the ambient gas, dense clouds with sufficiently short cooling times can radiate much of the deposited energy while remaining confined by the pressure of the outflow. If the compressed gas cools and fragments before it is dispersed, it may become gravitationally unstable and collapse to form stars (e.g., Nayakshin & Zubovas [86]; Zubovas et al. [135]).

In radio-loud AGNs, jets provide another means of driving shocks into the host-galaxy ISM. As a jet propagates through the ISM it drives a bow shock and inflates an overpressurized cocoon whose pressure can compress cool, dense clouds and trigger star formation (e.g., Rees [100]; Silk & Norman [110]; Silk [109]). Numerical simulations show that this interaction can enhance star formation locally even as gas dispersal and turbulence suppress it elsewhere in the same galaxy. The balance depends on the gas density, structure, and cooling efficiency, as well as the jet power, orientation, and interaction timescale (e.g., Gaibler et al. [59]; Mukherjee et al. [85]; Mandal et al. [78]). Cloud size is another important factor: smaller clouds are more efficiently ablated and dispersed, whereas larger clouds are more likely to remain intact and become susceptible to pressure-triggered collapse (Wagner et al. [126]).

Observations likewise suggest that both outcomes occur. In 3C 326 N, a massive molecular disk forms stars at least a factor of 60 below the Kennicutt–Schmidt relation. Its broad, shock-excited H2_2 emission suggests that jet-driven turbulence keeps the gas gravitationally unbound, thereby suppressing star formation (Nesvadba et al. [87]). At the population level, far-infrared-derived star formation rates were found to be ∼3\sim3–55 times higher in radio-detected than X-ray-only AGNs. Together with the higher inferred jet powers of the radio-detected samples, this association was interpreted as possible evidence for jet-enhanced star formation (Zinn et al. [133]).

Spatially resolved evidence consistent with localized triggering includes Centaurus A, where young stars lie near the edge of a radio lobe within the host-galaxy halo (Mould et al. [84]). Minkowski’s Object provides an external example: it is a young starburst galaxy located near the outer end of the jet from the neighboring radio galaxy NGC 541 (Croft et al. [34]). In another possible external case, the star-forming companion of HE 0450–2958 is aligned with a quasar radio lobe and exhibits shock-like emission-line ratios (Feain et al. [56]). Higher-resolution VLA imaging subsequently resolved a bipolar jet from the quasar and revealed spectral flattening at its projected intersection with the companion, providing further evidence of jet–ISM interaction (Molnár et al. [83]).

Studies of the relative orientations of radio galaxies and their optical host galaxies have a long history (see Miley [81]). A longstanding phenomenon that may provide evidence for positive jet feedback is the alignment effect, in which the rest-frame optical/UV emission of radio galaxies at z≳0.6z \gtrsim0.6 tends to be elongated along the radio axis, as first shown by Chambers et al. [29] and McCarthy et al. [80] (see also the review by Miley & De Breuck [82]). The effect appears to be more prominent at higher radio luminosities (Dunlop & Peacock [49]). The rest-frame optical/UV morphology also varies with radio extent: smaller radio sources exhibit more strongly aligned, knotty emission, whereas larger sources contain fewer and more diffuse components (Best et al. [13]). This dependence on radio size and luminosity, together with the more disturbed gas kinematics and shock-like line ratios observed in smaller sources, has been interpreted as evidence that young radio sources interact strongly with the dense ISM of their host galaxies (Best et al. [14]).

A recent HST study similarly found radio-aligned UV knots in five of seven compact steep-spectrum radio galaxies at z<0.6z < 0.6 attributable to recent star formation, while neither of two galaxies with larger radio sources showed UV emission attributed to star formation (Duggal et al. [48]). Although the small sample and uncertain relationship between the stellar and radio-source ages preclude a firm evolutionary interpretation, these results suggest that jet-induced star formation may be favored while the radio source is propagating through the dense host-galaxy ISM.

Jet-driven shocks and the pressure of the expanding radio cocoon may compress the host-galaxy ISM and promote star formation along the radio axis, making jet-triggered star formation a leading explanation for the alignment effect. A well-studied example is 4C 41.17 at z=3.8z = 3.8, where combined stellar, polarization, emissionline, and radio evidence points towards star formation triggered by jet activity [42, 16].

A second leading hypothesis for the alignment effect is scattered light from an obscured quasar. In the orientation-based scheme for type 2 radio galaxies, nuclear optical/UV light from an obscured quasar escapes preferentially along the polar directions and is scattered into our line of sight by dust or free electrons [4, 9]. This can produce extended UV emission along the radio axis with high linear polarization and an electric-vector position angle approximately perpendicular to the radio and UV axes. Detection of this signature in several high-redshift radio galaxies demonstrates that scattered AGN light can make an important contribution to the aligned rest-frame optical/UV continuum [43, 41, 124].

Warm ionized gas can also contribute to the aligned structures through strong rest-frame UV emission lines and free–free, free–bound, and two-photon nebular continuum emission. In three powerful radio galaxies at 0.31≤z≤3.80.31 \le z \le3.8, nebular continuum contributed up to ∼60%\sim60\% of the rest-frame UV emission in individual extended regions [44]. The gas may be ionized by the obscured AGN, young stars, or jet-induced shocks.

The Radio Galaxy Sample

The primary goal of this study is to investigate how powerful radio jets interact with and potentially shape the structures, stellar populations, and environments of their host galaxies during the peak epoch of galaxy assembly. To this end, we analyze HST WFC3/IR F140W imaging, which primarily traces the older stellar component, together with WFC3/UVIS F606W imaging, which is more sensitive to recently formed stars, for the 35 type 2 radio galaxies in the z>1z > 1 3CR sample. We intentionally restrict our analysis to radio galaxies because the obscured nucleus minimizes contamination from a bright quasar point-spread function (PSF), enabling a more reliable characterization of the host-galaxy morphology and spatially resolved stellar populations. We combine these HST observations with radio maps that define the jet axes and projected radio-source extents. Together, these data allow us to quantify the relationships among the radio jets, the rest-frame optical host galaxies, the UV emission, and nearby candidate companions, and to test whether jet-triggered star formation or scattered quasar light can account for the aligned UV structures.

The parent z>1z > 1 3CR sample comprises 64 radio-loud AGN drawn from the low-frequency, flux-limited Third Cambridge Revised catalog (F178 MHz>9F_{178\ \mathrm{MHz}} > 9 Jy; δ>−5∘\delta> -5^\circ) (Spinnrad et al. 1985). These powerful radio sources predominantly exhibit edge-brightened FR II morphologies [55]. All but two of the z>1z > 1 3CR sources reside at 1≲z≲21 \lesssim z \lesssim2, within the epoch commonly referred to as cosmic noon, when both AGN activity and cosmic star formation were near their peak [76, 1, 2]. Consequently, the z>1z > 1 3CR sample is well suited for studying the interplay between radio-jet feedback and host-galaxy evolution.

Throughout this paper, unless otherwise specified, the term host galaxy refers to the smooth rest-frame optical stellar component traced by the WFC3/IR F140W imaging and dominated by older stellar populations. We refer separately to the rest-frame UV emission detected in the WFC3/UVIS F606W images, whose physical origin and relationship to the radio-source geometry are central to this work.

The HST and radio observations are described in § 2, and the image analysis, stellar-population modeling, scattering analysis, and alignment measurements are presented in § 3. We give the results in § 4, discuss their implications in § 5, and summarize our conclusions in § 6. Throughout this paper we adopt a Λ\LambdaCDM cosmology with H0=67.74 km s−1 Mpc−1H_0 = 67.74\ \mathrm{km\ s^{-1}\ Mpc^{-1}}, ΩΛ=0.69\Omega_\Lambda= 0.69, and Ωm=0.31\Omega_m = 0.31 [97]. With this cosmology, an angular separation of 3′′3'' corresponds to a projected physical separation of approximately 25 kpc, accurate to within ∼5%\sim5\% over the entire redshift range 1<z<2.51 < z < 2.5. All magnitudes are in the AB system [91] and are corrected for Galactic extinction following the reddening maps of Schlafly & Finkbeiner [104].

HST and Radio Imaging

HST Imaging

The radio galaxies analyzed in this work were observed as part of the HST-3CR program under SNAP-13023 and GO-16281 (PI: M. Chiaberge), which together obtained WFC3/IR F140W and WFC3/UVIS F606W imaging of the entire z>1z > 1 3CR sample. The GO program completed the observations of the sources that were not imaged during the SNAP program. The WFC3/IR observations used the F140W filter, which has a pivot wavelength of 1392.3 nm. At the median redshift of the sample (z≈1.4z \approx1.4), this corresponds to a rest-frame wavelength of approximately 580 nm and primarily traces the optical stellar continuum of the host galaxies’ older stellar populations. The accompanying WFC3/UVIS observations used the F606W filter, which has a pivot wavelength of 588.9 nm and probes the rest-frame near-UV, making it sensitive to much younger stellar populations.

The observational characteristics of the SNAP-13023 and GO-16281 imaging are summarized in Table 1, including the filters, pivot wavelengths, exposure times, and surface-brightness sensitivity limits. The F140W exposure times were uniform within each program, while the individual F606W exposure times varied between visits. We therefore report the range of F606W exposure times and calculate a conservative sensitivity limit for the shortest total integration in each program with the online HST WFC3 exposure time calculator (ETC).

HST FilterPivot Wavelength (nm)Project IDExposure† Time (s)3σ3\sigma Surface Brightness Limit‡ (mag arcsec−2^{-2})
F140W1392.28SNAP-130232×2502 \times25024.5
F606W588.74SNAP-130232×(490–530)2 \times(490\text{--}530)23.7
F140W1392.28GO-162813×3003 \times30024.8
F606W588.74GO-162813×(443–568)3 \times(443\text{--}568)23.8

Table 1. HST-3CR Imaging Observations

All WFC3/IR flt files and WFC3/UVIS flc files were downloaded from the Mikulski Archive for Space Telescopes (MAST) after processing through the standard calwf3 reduction pipeline [47]. The individual dithered exposures were combined using AstroDrizzle [61], which performs sky subtraction, cosmic-ray rejection, geometric-distortion correction, and image combination. For the GO-16281 observations, the WFC3/UVIS F606W images were drizzled with final_pixfrac=0.8 and final_scale=0.03 arcsec pixel−1^{-1}, while the WFC3/IR F140W images used final_pixfrac=0.9 and final_scale=0.06 arcsec pixel−1^{-1}. For the SNAP-13023 observations, we used final_pixfrac=1.0 and final_scale=0.04 arcsec pixel−1^{-1} for F606W, and the corresponding values of 0.9 and 0.06 arcsec pixel−1^{-1} were used for F140W.

Although both filters were obtained for all 35 radio galaxies, 3C 173 and 3C 294 were excluded from the HST-based morphological and photometric analyses because reliable host-galaxy measurements could not be obtained. In 3C 173, a prominent unresolved point-source component is present in both the F140W and F606W images, preventing a robust separation of the extended host-galaxy emission. In 3C 294, the very faint host is strongly contaminated in both bands by the PSF wings of a nearby bright star. Consequently, we did not perform galfit modeling or F606W/F140W photometry for either source, and neither galaxy was included in the HST-based alignment analyses. The usable HST analysis sample therefore comprises 33 radio galaxies.

Radio Imaging

We used the radio images to measure the radio-axis position angles, nucleus-to-hotspot distances, and total source extents for the alignment and jet-timing analyses. Radio images were assembled from three sources: the NRAO VLA Archive Survey (NVAS; https://www.vla.nrao.edu/astro/nvas/), archival radio images distributed by the HST-3CR collaboration (https://h3z3c.stsci.edu/Observations.html; [67]), and archival VLA observations of 3C 222 and 3C 249 that we reduced for this work. The radio datasets we reduced for this work were calibrated, imaged using CLEAN deconvolution, and iteratively self-calibrated following standard radio continuum data reduction procedures. For 3C 194, 3C 210, 3C 322, 3C 324, 3C 356, and 3C 368, we corrected the HST–radio registration by shifting the radio images to align the radio cores with the host-galaxy centers determined from GALFIT modeling of the F140W images. Table 2 summarizes the radio imaging data and source properties for the 33 galaxies retained in the HST analyses, including their redshifts, radio-axis position angles, and full projected radio-source extents.

SourcezzFreq. (GHz)Project CodeObs. Date YYYY-MM-DDRMS (Jy beam−1^{-1})Beam Size (arcsec)OriginPAradioP A_{\mathrm{radio}} (deg)DradioD_{\mathrm{radio}} (kpc)
3C 131.354.86AC2001987-10-065.69×10−55.69 \times10^{-5}0.376×0.3570.376 \times0.357NVAS145.1245.8
3C 361.304.86AC1141985-03-112.45×10−42.45 \times10^{-4}0.391×0.3320.391 \times0.332NVAS21.290.8
3C 651.181.46PERL1982-02-282.73×10−42.73 \times10^{-4}1.284×1.1251.284 \times1.125NVAS98.0189.4
3C 68.21.584.86AV1641990-05-106.46×10−56.46 \times10^{-5}0.529×0.3920.529 \times0.392NVAS154.8221.0
3C 1241.084.86AM2131987-08-175.37×10−55.37 \times10^{-5}0.445×0.3940.445 \times0.394NVAS7.825.9
3C 1941.184.86AV1641990-05-102.56×10−42.56 \times10^{-4}0.535×0.3760.535 \times0.376NVAS159.3136.9
3C 2101.174.86AO2302009-04-195.09×10−55.09 \times10^{-5}1.214×1.0161.214 \times1.016HST-3CR164.1198.6
3C 2221.348.46AB9782000-12-093.42×10−63.42 \times10^{-6}0.256×0.2360.256 \times0.236This work62.218.1
3C 225A1.574.86AV1641989-04-201.29×10−41.29 \times10^{-4}1.233×1.1361.233 \times1.136NVAS42.260.0
3C 2301.498.44AK4031995-07-276.95×10−56.95 \times10^{-5}0.316×0.2290.316 \times0.229HST-3CR160.877.4
3C 2381.414.71AB9782000-12-099.07×10−59.07 \times10^{-5}0.459×0.3900.459 \times0.390NVAS162.084.9
3C 2391.794.86AM2051987-07-181.36×10−41.36 \times10^{-4}0.398×0.3310.398 \times0.331NVAS76.6107.4
3C 2411.628.44AA1491992-11-101.00×10−41.00 \times10^{-4}0.204×0.1990.204 \times0.199NVAS81.88.4
3C 2491.554.86AM2131987-08-175.44×10−55.44 \times10^{-5}0.474×0.4180.474 \times0.418This work138.7425.8
3C 2501.264.86AV1331986-04-164.50×10−54.50 \times10^{-5}0.362×0.3560.362 \times0.356NVAS54.5437.2
3C 2551.368.44AV1571988-12-222.88×10−52.88 \times10^{-5}0.295×0.2300.295 \times0.230HST-3CR151.75.9
3C 2561.828.44AV1641990-05-101.57×10−41.57 \times10^{-4}0.316×0.2080.316 \times0.208NVAS133.841.1
3C 2572.478.44AV1641990-05-112.13×10−52.13 \times10^{-5}0.412×0.2260.412 \times0.226HST-3CR133.8101.5
3C 2661.274.86AL4541998-04-091.30×10−41.30 \times10^{-4}0.397×0.3480.397 \times0.348NVAS176.751.5
3C 2671.148.44AL3221994-02-276.90×10−56.90 \times10^{-5}0.227×0.2210.227 \times0.221NVAS78.8322.9
3C 2971.418.44AV1641990-05-101.91×10−51.91 \times10^{-5}0.357×0.2330.357 \times0.233HST-3CR134.553.6
3C 300.11.168.44AK4031995-07-273.32×10−53.32 \times10^{-5}0.291×0.2400.291 \times0.240HST-3CR13.6103.1
3C 305.11.1314.94AM1411985-02-182.67×10−42.67 \times10^{-4}0.168×0.0960.168 \times0.096HST-3CR10.419.6
3C 3221.684.86AV1331986-04-165.92×10−55.92 \times10^{-5}0.487×0.2970.487 \times0.297HST-3CR2.6309.8
3C 3241.218.21AB7551995-07-111.25×10−51.25 \times10^{-5}0.200×0.2000.200 \times0.200HST-3CR70.7103.4
3C 326.11.8314.94AV1331986-04-161.35×10−41.35 \times10^{-4}0.173×0.1170.173 \times0.117HST-3CR93.663.4
3C 3561.084.85AF1861990-04-222.48×10−52.48 \times10^{-5}0.350×0.3500.350 \times0.350HST-3CR164.4625.4
3C 3681.134.71AC4201995-07-311.13×10−41.13 \times10^{-4}0.518×0.4340.518 \times0.434NVAS16.5149.9
3C 4371.484.86AV1641990-05-102.89×10−52.89 \times10^{-5}0.575×0.4300.575 \times0.430NVAS160.4621.8
3C 454.11.8414.94AM2131987-08-171.33×10−41.33 \times10^{-4}0.122×0.1060.122 \times0.106HST-3CR167.017.0
3C 469.11.348.46AO1702003-06-171.01×10−41.01 \times10^{-4}0.321×0.1870.321 \times0.187NVAS171.2656.8
3C 4701.658.44AL3221994-02-278.18×10−58.18 \times10^{-5}0.300×0.2010.300 \times0.201NVAS38.0225.7
3C 13.661.458.44AL04011996-12-311.17×10−41.17 \times10^{-4}0.242×0.2190.242 \times0.219NVAS7.715.8

Table 2. Radio Imaging Data and Source Properties. Note—NVAS = NRAO VLA Archive Survey; HST-3CR = HST-3CR archive; This work = radio data reduced and imaged for this study. Beam sizes give the restoring-beam FWHM major axis × minor axis. DradioD_{\mathrm{radio}} is the full projected radio-source extent. The table is also available in machine-readable form online.

The primary goal of the radio analysis is to determine the projected orientation of the radio jets. Most of the z>1z > 1 3CR radio galaxies in our sample exhibit classical FR II morphology. For these sources, the jet position angle was measured from the axis connecting the hot spot to counter hot spot. In the majority of cases the radio structures are relatively straight on the scales considered here, resulting in a well-defined jet axis.

Position angles are measured from north through east, with 0∘0^\circ corresponding to north and 90∘90^\circ corresponding to east. All images are displayed with north up and east to the left. Because the radio structures considered here are orientation axes rather than directed vectors, the radio position angles were folded into the range 0∘≤PA<180∘0^\circ\le PA < 180^\circ.

Data Analysis

This section describes the primary analyses performed on the HST and radio imaging data described in § 2. We first measure the structural properties of the host galaxies from their rest-frame optical emission and the photometric and morphological properties of the UV-emitting regions. These measurements are then used to constrain the physical origin of the UV emission through stellar population synthesis modeling. We then compare the host-galaxy, UV-emission, and merger geometries with the radio jets.

Rest-Frame Optical Host Galaxy Surface Brightness Modeling

To constrain the structural parameters of the host galaxies’ older stellar populations, we performed two-dimensional surface brightness modeling of the rest-frame optical emission using the HST WFC3/IR F140W

images and GALFIT ([94]). Following the same methodology employed in Breiding et al. [22], we used single Sérsic models to fit the host galaxy light profiles. Prior to fitting, we created custom bad-pixel masks to exclude nearby sources and localized irregular features, such as tidal structures and star-forming regions, so that the fits were driven by the underlying undisturbed stellar components of the host galaxies. In occasional cases where a massive companion lay in very close projection to the radio galaxy, we modeled the companion simultaneously with an additional Sérsic component rather than masking its pixels. The radio-galaxy host itself was still represented by a single Sérsic component. The Sérsic model has the following radial dependence ([106, 107]):

I(R)=Ieexp⁡[−κ((RRe)1/n−1)].(1)I(R) = I_{\mathrm{e}} \exp\left[-\kappa\left(\left(\frac{R}{R_{\mathrm{e}}}\right)^{1/n} - 1\right)\right]. \tag*{(1)}

In this expression, I(R)I(R) is the surface brightness at radius RR, ReR_{\mathrm{e}} is the effective radius enclosing half of the total light, IeI_{\mathrm{e}} is the surface brightness at ReR_{\mathrm{e}}, nn is the Sérsic index, and κ\kappa is a constant determined by the value of nn. The Sérsic index governs the overall shape of the radial light profile, with larger values corresponding to more centrally concentrated light distributions.

For each galaxy in the usable HST analysis sample, we fit a single Sérsic component to the HST/WFC3 F140W image, allowing the sky position, total magnitude, effective radius, Sérsic index, axis ratio, position angle, and sky background to vary freely. The best-fitting model was determined through χ2\chi^2 minimization, and residual images were visually inspected to assess each fit. Table 3 lists potential F140W emission lines from the SDSS MaNGA data analysis pipeline reference list ([11]), redshifted into 1.185–1.613 μ\mum, the range defined at 1% of peak throughput. Although localized irregular features were masked, diffuse line emission overlapping the smooth host could remain and affect the fitted structural parameters.

Sourcezzm140m_{140}b/ab/annReR_{\mathrm{e}}PAhostPA_{\mathrm{host}}Δθhost\Delta\theta_{\mathrm{host}}Potential F140W lines
3C 131.3519.510.762.638.52144.50.6Hα\alpha, [N II], [S II], [O I], He II, He I, [Ar III], [N I], Na I
3C 361.3019.620.773.034.3265.844.6Hα\alpha, [N II], [S II], [O I], He II, He I, [Ar III], [N I], Na I
3C 651.1819.260.703.0710.6127.770.3Hα\alpha, [N II], [S II], [O I], He II, [O II], He I, [Ar III], Na I, [Ar IV]

Table 3. Rest-Frame Optical Host-Galaxy Properties. Note—The complete table is available in machine-readable form.

UV Morphology from Image Moments

The rest-frame UV emission observed in the HST/WFC3 UVIS F606W images is generally clumpy and, in most sources, is considerably less spatially extended than the rest-frame optical emission detected in the F140W images. We therefore used the image-segmentation tools in photutils ([21]) to identify discrete regions of UV emission and assign pixels to each region. The background level and RMS were estimated using a median background model computed over the full image, and UV-emitting regions were identified as contiguous groups of at least 60 pixels exceeding a detection threshold of 1.5σ1.5\sigma relative to the background RMS. Pixels affected by image artifacts or unrelated sources were masked prior to source detection.

Because the UV emission is often irregular and poorly described by simple parametric profiles, we characterized the morphology of each detected region using the intensity-weighted second moments of its surface brightness distribution. We used the SourceCatalog routine in photutils to calculate the intensity-weighted centroid and second moments for each region. The second moments describe the spatial distribution of the emission about its centroid and define an equivalent ellipse whose semimajor axis, semiminor axis, and position angle characterize the extent, elongation, and orientation of the UV emission. We constructed these ellipses with semimajor and semiminor axes equal to three times the corresponding second-moment dispersions. These moment-defined ellipses are only used to characterize the extent and orientation of the UV emission and are not used for the photometric measurements described in § 3.3. We adopted the position angle of the semimajor axis, PAUVPA_{\mathrm{UV}}, as the orientation of each extended UV-emitting region for comparison with the radio jet axis (§ 3.5). These UV-region PAs are listed in Table 4.

SourceRegionm140obsm_{140}^{\mathrm{obs}}m140resm_{140}^{\mathrm{res}}m606m_{606}μ606\mu_{606}RprojR_{\mathrm{proj}}m606−m140resm_{606}-m_{140}^{\mathrm{res}}b/ab/aPAUVPA_{\mathrm{UV}}ΔθUV\Delta\theta_{\mathrm{UV}}
(AB mag)(AB mag)(AB mag)(AB mag arcsec−2^{-2})(kpc)(mag)(deg)(deg)
3C 13122.6523.4424.6823.679.91.240.64131.613.5
3C 13220.6523.3322.9322.830.7-0.400.44159.914.8
3C 36120.4522.7922.7721.570.5-0.020.9553.732.5

Table 4. Photometric and Morphological Properties of the Rest-Frame UV Regions

Aperture Photometry

In order to measure rest-frame optical and UV fluxes from the WFC3/IR F140W and WFC3/UVIS F606W images for the UV-defined emission regions identified in § 3.2, we performed aperture photometry using the photutils package. These measurements were used to test competing emission mechanisms for the UV emission, primarily distinguishing ongoing or recent star formation from electron or dust scattering of radiation from an obscured quasar.

Photometric apertures were defined from the original, unconvolved F606W segmentation maps used for the image-moment analysis (§ 3.2). Before measuring the fluxes, we convolved the F606W images to match the F140W PSF; the construction and validation of the PSF-matching kernels are described in Appendix § 7. The F140W images were reprojected onto the F606W pixel grid so that identical apertures could be applied to both filters. For the PSF-matched F606W images, a local background was estimated independently for each region from a surrounding annulus. The median count rate after 3σ3\sigma clipping within this annulus was subtracted on a per-pixel basis from the aperture flux. For the F140W images, a global sky background was estimated from the full image after masking the host galaxy and other significant sources using a 3σ3\sigma-clipped estimate of the remaining sky pixels. The corresponding sky contribution within each aperture was subtracted from the observed aperture flux. We then measured the flux of the best-fitting GALFIT host-galaxy model (§ 3.1) within the same aperture and subtracted it from the sky-subtracted aperture flux. This subtraction removes the contribution from the smooth, older stellar population of the host galaxy, allowing us to isolate any excess rest-frame optical emission above that expected from the underlying host galaxy. We estimated the statistical uncertainty in each aperture-integrated F140W residual from the sigma-clipped background RMS and the equivalent number of native F140W pixels in the aperture. Residuals with signal-to-noise ratios below 3 were treated as nondetections, for which the corresponding 3σ3\sigma flux limit was converted to a lower limit on the F140W magnitude, mF140W>mF140W,limm_{\mathrm{F140W}} > m_{\mathrm{F140W,lim}}.

To estimate the projected distances of the UV-emitting regions from the nucleus, we measured the separations between the UV-region centroids and the host-galaxy centroid determined from the best-fitting F140W GALFIT model. We emphasize that these projected distances represent lower limits to the true three-dimensional separations. Consequently, the inferred surface brightness of the quasar radiation field at the locations of the UV-emitting regions is an upper limit. This choice is therefore conservative when evaluating scattering scenarios, since scattered quasar light is more easily produced at smaller distances where the incident nuclear radiation field is strongest.

Modeling Stellar Populations

To investigate whether the colors of the UV-emitting regions are consistent with emission from young stellar populations, we compared the measured rest-frame UV–optical colors with stellar population synthesis (SPS) models generated using the Flexible Stellar Population Synthesis code (FSPS; [33, 32]) through the python-fsps interface ([69]). The models assume a Chabrier initial mass function ([28]), solar metallicity, a constant star-formation history, and the Calzetti et al. [24] dust attenuation law. Nebular emission and intergalactic absorption were not included, so the models represent stellar-continuum-only predictions.

Synthetic photometry was calculated in the observer-frame HST/WFC3 F606W and F140W filters over a redshift grid spanning 1.0≤z≤2.51.0 \le z \le2.5 in steps of Δz=0.1\Delta z = 0.1. The duration since the onset of constant star formation was sampled at 80 logarithmically spaced values between 1 Myr and 3 Gyr. At each redshift, durations exceeding the age of the Universe were excluded. We considered dust attenuations of AV=0.0A_V = 0.0, 0.3, 0.5, 1.0, 1.5, 2.0, and 3.0 mag.

For each UV-emitting region, the observed color was defined as Cobs=mF606W−mF140WC_{\mathrm{obs}} = m_{\mathrm{F606W}} - m_{\mathrm{F140W}} using the Galactic extinction-corrected photometry. The F140W magnitudes correspond to the residual flux remaining after subtraction of the smooth host-galaxy model (§ 3.3). These measurements therefore test whether the UV emission is accompanied by a spatially coincident rest-frame optical excess consistent with a young stellar population.

For every combination of redshift, model duration, and dust attenuation, we calculated the corresponding synthetic color, Cmodel=mF606WSPS−mF140WSPSC_{\mathrm{model}} = m_{\mathrm{F606W}}^{\mathrm{SPS}} - m_{\mathrm{F140W}}^{\mathrm{SPS}}. Each source was compared with the model grid evaluated at the nearest sampled redshift. For regions detected in both filters, models were retained when their predicted colors agreed with the observed color to within 0.20 mag, such that ∣Cmodel−Cobs∣≤0.20|C_{\mathrm{model}} - C_{\mathrm{obs}}| \le0.20 mag.

We used this deliberately inclusive model-comparison tolerance because the host-subtracted F140W knot fluxes are sensitive to small uncertainties in the smooth host-galaxy model. The residual knot flux is the difference between the total observed aperture flux and the host-model flux. Consequently, a small fractional change in the host model can produce a much larger fractional change in the residual when the host dominates the aperture flux. As an illustrative sensitivity test, we rescaled the aperture-integrated GALFIT host-model flux by ±1%\pm1\% and recomputed the residual colors. The larger absolute color change had a median of 0.035 mag and a 90th percentile of 0.125 mag. Thus, the 0.20-mag threshold encompasses the response to a small host-normalization perturbation. This tolerance is an operational model-consistency criterion rather than a source-specific uncertainty or formal confidence interval.

We also assessed whether finite sampling of the model grid materially affected the predicted colors. The 80 model durations have a logarithmic spacing of Δlog⁡(t)=0.044\Delta\log(t) = 0.044 dex. Repeating the analysis with 800 duration values and an interpolated attenuation spacing of ΔAV=0.01\Delta A_V = 0.01 mag produced a maximum coarse-grid color offset of 0.071 mag. Evaluating the models directly at each source redshift, rather than at the nearest Δz=0.1\Delta z = 0.1 grid point, changed the colors by at most 0.026 mag. These effects, together with the small residual systematic uncertainties from PSF matching (Appendix § 7.§ 1), are smaller than the 0.20-mag tolerance.

For the F140W nondetections defined in § 3.3, the constraint mF140W>mF140W,limm_{\mathrm{F140W}} > m_{\mathrm{F140W,lim}} yields an upper limit on the color, Cobs<ClimC_{\mathrm{obs}} < C_{\mathrm{lim}}, where Clim=C_{\mathrm{lim}} = mF606W−mF140W,limm_{\mathrm{F606W}} - m_{\mathrm{F140W,lim}}. For these regions, we retained models satisfying Cmodel≤Clim+0.20C_{\mathrm{model}} \leq C_{\mathrm{lim}} + 0.20 mag. For each region, we recorded the minimum and maximum model durations and dust attenuations among the accepted models. Because only one UV–optical color is available, these ranges are neither unique solutions nor formal confidence intervals. Instead, they describe the parameter space permitted under the stellar-continuum model assumptions.

Jet–Host, Jet–UV, and Jet–Merger Alignments

To quantify the projected relationships among the radio jets, the stellar structures of the host galaxies, the rest-frame UV emission, and the nearby companion environment, we compared the position angle of each component with the radio-axis position angle measured from the radio images described in § 2.2. For the predominantly FR II radio galaxies in our sample, the radio axis was defined by the line connecting the terminal hot spot and counter-hot spot. The only exception is the highly curved “C-shaped” jet in the radio galaxy 3C 297. For this source, we define the radio jet axis to be the line connecting the host galaxy center to the hot spot on the galaxy side relevant to the UV emission.

The rest-frame optical host-galaxy position angle and axis ratio were obtained from the best-fitting Galfit Sérsic model in the WFC3/IR F140W image and are listed in Table 3. The position angle and axis ratio of each UV-emitting region were derived from its light-weighted second moments (§ 3.2). Components with axis ratio b/a≥0.8b/a \geq0.8 were excluded from the alignment tests because their position angles are poorly constrained.

The host-galaxy, UV-region, and merger-axis position angles were expressed using the same astronomical convention as the radio position angles (measured from north through east) and folded into the same range, 0∘≤PA<180∘0^\circ\leq PA < 180^\circ. For each component, we calculated the acute projected misalignment angle relative to the radio axis as

Δθ=min⁡[∣PAcomp−PAradio∣,180∘−∣PAcomp−PAradio∣],(2)\Delta\theta= \min\left[ \left|PA_{\mathrm{comp}} - PA_{\mathrm{radio}}\right|, 180^\circ- \left|PA_{\mathrm{comp}} - PA_{\mathrm{radio}}\right| \right], \tag*{(2)}

where PAcompPA_{\mathrm{comp}} denotes the host-galaxy, UV-region, or merger-axis position angle, and PAradioPA_{\mathrm{radio}} denotes the radio-axis position angle. The resulting misalignment angles lie in the interval 0∘≤Δθ≤90∘0^\circ\leq\Delta\theta\leq90^\circ, with smaller values indicating closer alignment with the radio axis.

We examined the UV alignment at both the region and galaxy levels. In the region-level analysis, each retained UV-emitting region was treated as independent and thus assigned equal weight. Galaxies containing more UV regions therefore contributed proportionally more measurements to this analysis. To obtain one galaxy-level UV statistic per source, we also calculated the mean misalignment angles of all individual UV regions in that galaxy. Host-galaxy alignments were represented by one misalignment angle per galaxy derived from the Galfit major-axis position angle.

To investigate whether the radio axes are preferentially aligned with the projected geometry of nearby candidate companions, we constructed source-segmentation maps from the F140W images using a 5σ5\sigma detection threshold and a minimum detection area of 20 connected pixels. Overlapping detections were deblended using 16 levels and a contrast parameter of 0.01. The host was identified as the segmented source closest to its coordinates. Candidate companions were required to lie at host-centered projected distances 5<Rproj<255 < R_{\mathrm{proj}} < 25 kpc, have a segmented area of at least 100 pixels, and have an F140W AB magnitude within two magnitudes of the host magnitude. The magnitude threshold corresponds to a minimum companion-to-host F140W flux ratio of approximately 1 : 6.3. We refer to these systems as candidate mergers and characterize their projected geometry using the merger axis defined below. This photometric selection does not establish stellar mass ratios or confirm physical association.

When exactly one candidate companion satisfied these criteria, the merger axis was defined by the line connecting the host and companion. When multiple companions were present, the merger axis was determined from a flux-weighted principal-component analysis of the host and companion positions. The eigenvector corresponding to the largest weighted positional variance defined the dominant projected orientation.

Under the null hypothesis of random axial orientations, individual misalignment angles are uniformly distributed between 0∘0^\circ and 90∘90^\circ. We therefore used one-sided one-sample Kolmogorov–Smirnov tests to compare the individual UV-region, host-galaxy, and 25-kpc merger-axis misalignment distributions with this uniform distribution. The one-sided alternative tested specifically for an excess of small misalignment angles.

Because an average of multiple uniformly distributed angles is not itself uniformly distributed, the galaxy-level mean UV misalignments were evaluated using a Monte Carlo random-orientation test. In each of 10610^6 trials, an independent misalignment angle was drawn from a uniform distribution over 0∘0^\circ–90∘90^\circ for every retained UV region. The observed assignment and number of UV regions per galaxy were preserved, and the simulated angles were averaged within each galaxy in the same manner as the observations. The test statistic was the mean of the resulting galaxy-level misalignments, with smaller values representing stronger alignment. The one-sided Monte Carlo probability was calculated from the fraction of trials producing a test statistic no larger than the observed value.

Results

Rest-Frame Optical Host-Galaxy Properties

We obtained successful single-Sérsic profile fits to the HST/WFC3 F140W images of all 33 radio galaxies in this study. Table 3 reports the source redshift, F140W apparent magnitude, m140m_{140}, axis ratio (semiminor-to-semimajor axis), b/ab/a, Sérsic index, nn, effective radius, ReR_{\mathrm{e}}, host-galaxy major-axis position angle, PAhostPA_{\mathrm{host}}, and acute host–radio misalignment angle, Δθhost\Delta\theta_{\mathrm{host}}. Host PAs use the north-through-east convention defined in § 3.5; the radio PAs are listed in Table 2. The complete table retains all 33 hosts, including those excluded from the alignment tests by the b/a<0.8b/a < 0.8 requirement.

To place these measurements in context, we compared the radio galaxies studied here with the 25 resolved z>1z > 1 3CR quasar hosts analyzed in Breiding et al. [22]. The two samples are well matched in redshift, with median values of 1.36 and 1.38 for the radio galaxies and quasars, respectively. A two-sided two-sample Kolmogorov–Smirnov (KS) test does not indicate a difference between their redshift distributions (p=0.416p = 0.416).

We converted the F140W apparent magnitudes to rest-frame VV-band absolute magnitudes, MVM_{\mathrm{V}}. For both radio galaxy and quasar samples, we used the solar-metallicity stellar-population grid from the 2016 update to the Bruzual–Charlot models distributed with Bagpipes, assuming a single-burst population formed at z=5z = 5 and evolved passively ([23, 27]).

Figure 1 presents box-and-whisker comparisons of MVM_{\mathrm{V}}, nn, and ReR_{\mathrm{e}}. The fitted quasar hosts are brighter than the radio-galaxy hosts by 0.537 mag at the median. Their MVM_{\mathrm{V}} distributions differ according to a two-sided KS test (p=4.58×10−3p = 4.58 \times10^{-3}). A comparison using the apparent F140W magnitudes gives a similar median difference of 0.636 mag, with the radio-galaxy hosts again fainter than the resolved quasar hosts.

Figure 1. Comparison of the fitted rest-frame optical properties of the 33 radio galaxies analyzed in this study and the 25 resolved z>1z > 1 3CR quasar hosts from Breiding et al. [22]. Individual objects are shown as horizontally jittered points. Boxes span the 25th to 75th percentiles, horizontal lines mark the medians, and whiskers extend to the most extreme values within 1.5 times the interquartile range. Left: rest-frame VV-band AB absolute magnitude inferred from the F140W host apparent magnitude. Middle: Sérsic index; the quasar points and box include only the eight freely fitted indices. Right: semimajor-axis effective radius.

Seventeen of the 25 resolved quasar hosts were fitted with n=4n = 4 fixed. Because these indices were assumed rather than measured, we exclude them from the middle panel of Figure 1 and the statistical comparison of Sérsic indices. All 33 radio galaxies were fitted with nn free to vary. Among the eight quasar hosts with freely fitted indices, the median is n=2.0n = 2.0, compared with n=1.72n = 1.72 for the radio galaxies. A two-sided KS test does not indicate a difference between these distributions (p=0.255p = 0.255), although the quasar sample is small.

The median effective radius of the resolved quasar hosts is 5.80 kpc, compared with 7.80 kpc for the radio galaxies. The full distributions differ according to a two-sided KS test (p=4.94×10−3p = 4.94 \times10^{-3}). Fixing nn could contribute to the smaller fitted quasar-host radii, but excluding these marginally resolved hosts may bias the comparison against compact quasar hosts (Breiding et al. [22]). Using only the freely fitted quasar hosts gives a median effective radius of 10.45 kpc, and a two-sided KS test does not detect a significant difference from the radio-galaxy distribution (p=0.162p = 0.162). We therefore interpret the apparent size difference with caution.

The greater fitted luminosities identified for the quasar host galaxies in comparison with the radio galaxies should not be interpreted as secure evidence that the quasar hosts are intrinsically more luminous. The quasar images were dominated by a much brighter unresolved quasar point source superposed on a much fainter extended host, whereas the radio galaxies lack a comparable unresolved nuclear component. Degeneracy between the quasar PSF and Sérsic components may assign some nuclear light to the host model, biasing the recovered host magnitude brightward. In addition, restricting the comparison to resolved quasars may preferentially retain quasars with brighter or more readily detected hosts. Establishing an intrinsic luminosity difference between quasar and radio-galaxy hosts would require a matched nucleus-injection-and-recovery analysis.

Figure 2 illustrates the F140W host-galaxy modeling for 3C 324, including the best-fitting single-Sérsic model and its residual image.

Example imaging and structural decomposition for 3C 324

Figure 2. Example imaging and structural decomposition for 3C 324, which exhibits prominent knotty rest-frame UV emission aligned with its radio structure. Top left: HST/WFC3 F606W image with yellow equivalent ellipses derived from the intensity-weighted second moments of the segmented UV-emitting regions. The semimajor and semiminor axes of each ellipse are three times the corresponding second-moment dispersions. Regions are labeled by number when more than one region is detected. Top right: HST/WFC3 F140W image tracing the rest-frame optical host galaxy. Bottom left: best-fitting Sérsic galfit model of the F140W image. Bottom right: residual image after subtraction of the best-fitting model. Radio contours are overlaid on the F606W and F140W images. The circle is centered on the host galaxy and has a physical radius of 25 kpc, while the arrow indicates an angular scale of 3′′3^{\prime\prime}. The model and residual panels each cover a 10.1′′×10.1′′10.1^{\prime\prime} \times10.1^{\prime\prime} field of view around the host galaxy. The images are displayed with logarithmic intensity stretches spanning 24.7–31.5 AB mag per pixel in F140W and 27.4–32.4 AB mag per pixel in F606W. All images are rotated north up and east to the left. The F606W images are smoothed for display with a Gaussian kernel of σ=1.5\sigma= 1.5 pixels and a radius of 3 pixels. Equivalent panels for the other radio galaxies in our sample are available in the online figure set.

Rest-Frame UV Morphologies and Segmented Regions

The WFC3/UVIS F606W image-segmentation analysis identified 48 spatially resolved UV-emitting regions in 30 of the 33 radio galaxies analyzed. Eighteen of the 30 detected galaxies contain a single segmented UV region, while 12 contain multiple regions. The rest-frame UV emission is generally clumpy and occupies only a subset of the smoother and more extended rest-frame optical emission traced by the F140W images. Table 4 lists the region photometry, projected centroid distances from the host, intensity-moment axis ratios, major-axis position angles, PAUVPA_{\mathrm{UV}}, and acute radio-axis misalignment angles, ΔθUV\Delta\theta_{\mathrm{UV}}. The PAs follow the north-through-east convention defined in § 3.5.

The same figure illustrates the UV-morphology analysis: the F606W emission of 3C 324 contains several distinct components aligned with the radio source. The yellow moment-defined ellipses characterize their extent, axis ratio, and orientation, while the segmentation boundaries, rather than the ellipses, define the photometric apertures used in § 3.3.

We also tested whether the strength or multiplicity of the rest-frame UV emission depends on projected radio size. Among the 30 galaxies with the required measurements, Spearman rank tests show no significant correlation between radio size and either the summed luminosity of the segmented F606W regions (ρ=−0.17\rho= -0.17, p=0.36p = 0.36), their luminosity relative to the smooth F140W host (ρ=−0.10\rho= -0.10, p=0.59p = 0.59), or the number

of segmented UV regions (ρ=−0.19\rho= -0.19, p=0.30p = 0.30). We therefore do not recover evidence that the strongest or most multicomponent UV emission is preferentially associated with the most compact radio sources.

Each segmented component was treated as an independent UV-emitting region for the photometric and morphological measurements. We next quantify the orientations of the retained UV regions, the rest-frame optical host galaxies, and the candidate merger axes relative to the radio jets.

Alignment with the Radio Axis

Among the 33 galaxies with usable HST data, we apply the b/a<0.8b/a < 0.8 criterion (§ 3.5) separately to the host galaxies and to the UV regions detected in 30 of those galaxies. This leaves 26 host orientations and 38 UV-region orientations in 27 galaxies. A galaxy can contribute to one alignment analysis even if it lacks a usable orientation for the other. The misalignment-angle distributions for the individual UV regions and the rest-frame optical host galaxies are shown in panels (a) and (c) of Figure 3, respectively. Both histograms are strongly weighted toward small misalignment angles, indicating that the UV emission and the stellar major axes of the host galaxies preferentially follow the radio axis. The one-sided KS test comparing the UV-region misalignments with a uniform distribution gave p=7.86×10−8p = 7.86 \times10^{-8}. The same comparison for the rest-frame optical host galaxies gave p=9.01×10−5p = 9.01 \times10^{-5}. The median misalignment angles are 16.7∘16.7^\circ for the individual UV regions and 16.9∘16.9^\circ for the host galaxies, compared with 45∘45^\circ for random axial orientations.

Figure 3. Radio-axis misalignment distributions for (a) individual UV-emitting regions, (b) the mean UV-region misalignment within each galaxy, (c) host-galaxy major axes, and (d) 25-kpc merger axes. Gray histograms show the observed distributions. The UV and host samples in panels (a)–(c) include only components satisfying b/a<0.8b/a < 0.8. Dashed horizontal lines in panels (a), (c), and (d) show the expected number of objects per 10∘10^\circ bin for a uniform distribution over 0∘0^\circ–90∘90^\circ. In panel (b), open circles show the mean bin counts expected from the Monte Carlo random-orientation simulations, and vertical bars show the central 68% simulation intervals (16th–84th percentiles). The connecting line is included as a visual guide. These simulation intervals represent variation under the null hypothesis and are not measurement uncertainties.

Because galaxies containing multiple UV regions contribute more measurements to the region-level analysis, we also calculated one mean UV misalignment for each of the 27 galaxies (Figure 3, panel b). In the galaxy-level Monte Carlo test described in § 3.5, only one of the 10610^6 random-orientation realizations produced an overall mean as small as or smaller than the observed value. Including the finite-simulation correction gives p=2.0×10−6p = 2.0 \times10^{-6}, demonstrating that the UV alignment remains significant when each galaxy is given equal weight. The observed mean of the galaxy-level mean UV misalignments is 23.3∘23.3^\circ, compared with the null expectation of 45∘45^\circ. Holding the observed UV-region position angles fixed and drawing one uniformly random radio axis per galaxy in each of 10710^7 trials gave p≃1.3×10−6p \simeq1.3 \times10^{-6} for the same statistic.

We also compared the host–radio misalignment with the mean host–UV misalignment within individual systems to test whether the host galaxies tend to align more strongly with the rest-frame UV regions than with the radio jets. We would expect such a preference if emission lines from the same gas structures were the dominant inNote—The complete table is available in machine-readable form.

fluence on both the fitted F140W host orientations and the measured F606W region orientations. We used the 22 systems in which the host and at least one UV region satisfy b/a<0.8b/a < 0.8, averaging only over UV regions that meet this cut. Twelve of the 22 hosts are more closely aligned with the UV-region axes than with the radio axis. A one-sided sign test for an excess of hosts closer to the UV-region axes than to the radio axes gives p=0.42p = 0.42. This finding is consistent with the host–radio alignment tracing the older stellar component, although diffuse line emission could still affect the fitted host orientations.

Note—NcompN_{\mathrm{comp}} counts the companion candidates selected at 5<Rproj<255 < R_{\mathrm{proj}} < 25 kpc (§ 3.5).

Table 5 lists the eight galaxies containing at least one candidate companion satisfying the selection criteria in § 3.5 at 5<Rproj<255 < R_{\mathrm{proj}} < 25 kpc, together with their merger- and radio-axis position angles and acute misalignments. Panel (d) of Figure 3 shows the corresponding misalignment distribution. As with the UV-region and host-galaxy distributions, the merger-axis histogram is weighted toward small misalignment angles. Its departure from random orientations is significant (p=0.0115p = 0.0115 in a one-sided KS test). The eight merger axes have a median misalignment of 22.6∘22.6^\circ, with six of the eight lying within 30∘30^\circ of the radio axis.

Sensitivity to the Axis-Ratio Cut

In order to examine how sensitively our results depend on the choice of axis-ratio cut, we repeated the host–radio and UV–radio alignment analyses using the stricter cut b/a<0.7b/a < 0.7 in place of b/a<0.8b/a < 0.8. This leaves 15 host galaxies and 37 UV regions in 27 galaxies. The median host–radio misalignment angle is 7.8∘7.8^\circ, with a one-sided KS-test pp-value of p=1.4×10−5p = 1.4 \times10^{-5}. For the UV regions, the median misalignment angle is 16.5∘16.5^\circ, with a one-sided KS-test pp-value of p=2.0×10−8p = 2.0 \times10^{-8}. Thus, both the host–radio and UV–radio alignment signals persist under the stricter axis-ratio cut.

Stellar-Population and Jet-Timing Constraints

Of the 48 UV-emitting regions detected in the segmentation analysis described in § 3.2, 47 had at least one acceptable SPS model solution. The color limit for 3C 194 knot 2 lies blueward of every stellar-continuum model and therefore yields no accepted solution.

For the following discussion, we define a recent onset of star formation as an elapsed duration since the onset of constant star formation of ≤100\leq100 Myr. Recent onset is considered permitted when at least one accepted model satisfies this criterion and required when all accepted models satisfy it. Recent onset is permitted for all 47 regions with accepted stellar-continuum solutions, although this alone provides little constraint on their star-formation histories. More importantly, it is required for 12 regions: none of their accepted models has a constant-SFR duration exceeding 100 Myr. Both recent and longer-duration solutions remain allowed for the other 35 regions.

The extinction constraints further distinguish these two groups. Among the 35 regions that permit longer-duration solutions, 27 allow both zero and nonzero extinction, while eight require nonzero extinction. Of these eight, five have minimum permitted values of AV=0.3A_V = 0.3 or 0.5 mag, while three require AV≥1.5A_V \geq1.5 mag. The 12 recent-onset-required regions are more restrictive: all permit AV=0A_V = 0, none permits AV>1A_V > 1 mag, and two admit only models with AV=0A_V = 0. Thus, although the duration–extinction degeneracy remains substantial for most regions, the 12 UV-emitting regions for which a recent onset of star formation is required occupy a more restricted part of the model grid: none of their accepted models has a constant-SFR duration greater than 100 Myr or AV>1A_V > 1 mag. If their UV emission is stellar, they therefore constitute the strongest candidates for recently initiated, ongoing star formation.

We next used the SPS constraints to determine whether the permitted constant-SFR durations are temporally consistent with jet-induced star formation. In the SPS models, tSFt_{\mathrm{SF}} is the time elapsed since the onset of constant star formation. We restricted this comparison to UV-emitting regions located within the projected host galaxy.

For a localized UV-emitting region, we estimated the projected distance traveled by the advancing radio structure since passing the region as

Dtiming=Dhs−RUV,(3)D_{\mathrm{timing}} = D_{\mathrm{hs}} - R_{\mathrm{UV}}, \tag*{(3)}

where DhsD_{\mathrm{hs}} is the projected distance from the nucleus to the hotspot on the same side of the radio source as the UV region, and RUVR_{\mathrm{UV}} is the projected nuclear distance of the UV-region centroid. The corresponding available post-passage time was calculated as

tavail=Dtimingβhsc,(4)t_{\mathrm{avail}} = \frac{D_{\mathrm{timing}}}{\beta_{\mathrm{hs}}c}, \tag*{(4)}

where βhs=vhs/c\beta_{\mathrm{hs}} = v_{\mathrm{hs}}/c denotes the assumed mean axial advance speed of the lobe head, for which the terminal hotspot provides an observational tracer. This is distinct from both the relativistic bulk speed of the jet plasma and the backflow speed within the radio lobe. This calculation treats the projected radial nucleus-to-region distance as a proxy for the distance already traversed along the relevant side of the radio source. It is therefore a first-order geometrical approximation and does not establish that an off-axis region was physically crossed by the jet or lobe. Regions with RUV≥DhsR_{\mathrm{UV}} \ge D_{\mathrm{hs}} were excluded because the lobe head has not demonstrably propagated beyond the UV region in projection.

For systems in which the UV emission is distributed broadly across the host galaxy rather than confined to a localized component, we instead used one-half of the total projected hotspot-to-counter-hotspot extent for DtimingD_{\mathrm{timing}}. Assuming that the two lobe heads advance simultaneously at the same mean speed, this corresponds to the mean projected nucleus-to-hotspot distance and provides a host-wide estimate of the radio-source propagation time. These host-wide cases are identified by the geometry code H in Table 6.

SourceRegiontSF,mint_{\mathrm{SF,min}}tSF,maxt_{\mathrm{SF,max}}AV,min⁡A_{V,\min}AV,max⁡A_{V,\max}GeometryDtimingD_{\mathrm{timing}}tavailt_{\mathrm{avail}}Timing result
(Myr)(Myr)(mag)(mag)(kpc)(Myr)
3C 1314.573000.000.02.0L92.1230.04Some
3C 1322.0325.610.01.0L142.6946.54All
3C 3611.00143.450.01.0L37.7712.32Some

Table 6. Stellar-Population and Fiducial Jet-Timing Constraints for the UV-Emitting Regions

We evaluated Equation (4) for a range of βhs=0.005\beta_{\mathrm{hs} } = 0.005, 0.010.01, 0.020.02, 0.050.05, and 0.100.10. Resolved spectral-aging studies of mature FR II radio galaxies find characteristic lobe-head advance speeds of order 0.01c0.01c, while analyses of very powerful FR II sources generally indicate growth speeds of a few percent of cc and characteristic speeds below approximately 0.1c0.1c [103, 90, 63]. We use βhs=0.01\beta_{\mathrm{hs}} = 0.01 as the fiducial value and included βhs=0.10\beta_{\mathrm{hs}} = 0.10 as a deliberately fast limiting case. No correction was applied for projection effects. Deprojecting the distances would increase DtimingD_{\mathrm{timing}} and therefore tavailt_{\mathrm{avail}}. Our values of tavailt_{\mathrm{avail}} are consequently lower limits and are conservative when evaluating whether the timing can accommodate jet-triggered star formation.

We define tSF,mint_{\mathrm{SF,min}} and tSF,maxt_{\mathrm{SF,max}} as the minimum and maximum constant-SFR durations among the accepted SPS models. The timing permits jet-induced star formation for at least one accepted solution when

tSF,min≤tavail,(5)t_{\mathrm{SF,min}} \le t_{\mathrm{avail}}, \tag*{(5)}

and is consistent with every accepted solution when

tSF,max≤tavail.(6)t_{\mathrm{SF,max}} \le t_{\mathrm{avail}}. \tag*{(6)}

Table 6 presents the permitted SPS duration and attenuation ranges for each UV-emitting region, together with the timing geometry, DtimingD_{\mathrm{timing}}, tavailt_{\mathrm{avail}}, and the timing result for

with the timing geometry, DtimingD_{\mathrm{timing}}, tavailt_{\mathrm{avail}}, and the timing result for the fiducial βhs=0.01\beta_{\mathrm{hs}} = 0.01. The timing analysis includes 41 of the 48 UV-emitting regions in 28 galaxies. Of these, 33 use the localized post-passage distance in Equation (3), while eight use one-half of the total host-wide radio-source extent. Four regions were excluded because they lie outside the projected host galaxy. Three additional regions—3C 225A knot 1, 3C 255 knot 2, and 3C 266 knot 1—were excluded because their projected nuclear separations equal or exceed the corresponding same-side hotspot distances.

For the fiducial hotspot speed βhs=0.01\beta_{\mathrm{hs}} = 0.01, at least one accepted SPS duration satisfies tSF≤tavailt_{\mathrm{SF}} \le t_{\mathrm{avail}} for 38 of the 41 regions (92.7%). This high fraction is not strongly discriminating because short durations are permitted for nearly every region. The more restrictive all-model condition, in which every accepted SPS duration satisfies tSF≤tavailt_{\mathrm{SF}} \le t_{\mathrm{avail}}, is satisfied by seven regions (17.1%). All seven regions satisfying the all-model timing condition belong to the 12 regions whose colors require a recent onset of star formation. Ten of these 12 regions enter the timing analysis, and seven satisfy tSF,max≤tavailt_{\mathrm{SF,max}} \le t_{\mathrm{avail}} at the fiducial lobe-head advance speed. These are 3C 13 region 2, 3C 210 region 1, 3C 238 region 1, 3C 267 region 1, 3C 326.1 region 1, 3C 356 region 1, and 3C 469.1 region 2. If their UV emission is predominantly stellar, these seven regions constitute the strongest combined SPS and jet-timing candidates for recently initiated, jet-induced star formation.

Six of these seven candidates were evaluated using the localized timing geometry; 3C 210 region 1 uses the host-wide comparison. For localized regions, the timing condition is intrinsically easier to satisfy at smaller projected nuclear distances because tavailt_{\mathrm{avail}} increases as RUVR_{\mathrm{UV}} decreases. The prevalence of near-nuclear regions among the candidates is therefore partly a consequence of the timing geometry and is not independent evidence for jet-induced star formation. In this respect, 3C 356 region 1 and 3C 326.1 region 1 are particularly notable because they satisfy the all-model condition at projected nuclear distances of 2.0 and 5.9 kpc, respectively.

Table 7 shows how the fractions of regions satisfying tSF,min≤tavailt_{\mathrm{SF,min}} \le t_{\mathrm{avail}} or tSF,max≤tavailt_{\mathrm{SF,max}} \le t_{\mathrm{avail}} depend on the assumed lobe-head advance speed. Because tavail∝βhs−1t_{\mathrm{avail}} \propto\beta_{\mathrm{hs}}^{-1}, faster advance speeds leave less time for star formation after the radio structure passes a region, and fewer regions satisfy either criterion. At every tested speed, substantially fewer regions satisfy tSF,max≤tavailt_{\mathrm{SF,max}} \le t_{\mathrm{avail}}, meaning that the available time is sufficient for every accepted SPS duration. All eight host-wide regions have at least

βhs\beta_{\mathrm{hs}}At least one SPS modelAll SPS models
0.00538/41 (92.7%)8/41 (19.5%)
0.01038/41 (92.7%)7/41 (17.1%)
0.02034/41 (82.9%)5/41 (12.2%)
0.05022/41 (53.7%)4/41 (9.8%)
0.10017/41 (41.5%)3/41 (7.3%)

Table 7. Sensitivity of the Jet–Stellar-Population Timing Test

**Note—**The complete table is available in machine-readable form. Geometry codes: L = Dhs−RUVD_{\mathrm{hs}} - R_{\mathrm{UV}}; H = half the full projected radio extent; O = outside the projected host; R = RUV≥DhsR_{\mathrm{UV}} \ge D_{\mathrm{hs}}. None, Some, and All mean that none, some but not all, or all accepted SPS durations satisfy tSF≤tavailt_{\mathrm{SF}} \le t_{\mathrm{avail}}. Not-tested denotes exclusion from the timing comparison; blank SPS entries indicate no accepted stellar-continuum model.

one accepted SPS model with tSF≤tavailt_{\mathrm{SF}} \le t_{\mathrm{avail}} for βhs≤0.02\beta_{\mathrm{hs}} \le0.02, while one satisfies this inequality for every accepted SPS model at the fiducial speed. At βhs=0.05\beta_{\mathrm{hs}} = 0.05 and 0.10, at least one accepted model remains consistent with the available time for four and three of the eight host-wide regions, respectively.

These comparisons establish only whether the temporal ordering required for jet-induced star formation is permitted under the SPS, geometrical, and hotspot-speed assumptions. They do not demonstrate that the jet caused the star formation. Nevertheless, the combination of a required recent onset and consistency with the all-model timing condition identifies a restricted set of seven particularly strong candidates for further investigation.

Radial Trends in Optical-UV Color and Surface Brightness

In Figure 4 we show the UV–optical color, F606W–F140W, as a function of projected distance from the galaxy nucleus. The two F140W nondetections are shown as color upper limits but are not included in the correlation or fit. Among the 46 regions detected in F140W, the UV-emitting regions become redder with increasing projected distance, with Spearman rank correlation coefficient, ρ=0.498\rho= 0.498. We evaluated the significance by randomly permuting the colors among the projected distances 10610^{6} times and recalculating ρ\rho. Only 523 permutations produced ∣ρ∣≥0.498|\rho| \ge0.498, giving a two-sided pp-value of p=5.2×10−4p = 5.2 \times10^{-4}.

Figure 4. F606W–F140W color versus projected distance from the nucleus for the 48 UV-emitting regions. Filled circles show the 46 regions with F140W detections. Open circles with downward arrows show the two F140W nondetections and therefore represent color upper limits. The solid line is the ordinary least-squares fit to the detections, and the shaded region is the corresponding 95% confidence band.

We estimated the uncertainty using 10,000 bootstrap realizations in which the 46 detected regions were resampled with replacement. The resulting 95% bootstrap percentile interval is 0.258≤ρ≤0.6780.258 \le\rho\le0.678. An ordinary least-squares fit gives a color gradient of 0.0675 mag kpc−1^{-1}, with a 95% bootstrap interval of 0.0396–0.1040 mag kpc−1^{-1}. The shaded region in Figure 4 is the pointwise 95% confidence band obtained from these same bootstrap fits. Several physical effects could produce this trend. The bluer colors at smaller radii could reflect younger stellar populations or a stronger contribution from AGN-related UV emission, such as scattered quasar light or nebular continuum, whose importance is expected to decrease as the nuclear radiation field weakens with radius. Alternatively, the redder colors at larger radii could reflect older stellar populations or greater dust extinction in the outer UV-emitting structures.

We next examined the radial behavior of the F606W surface brightness. Given that the apertures were defined on the native F606W images, we calculated μ606\mu_{606} from the native, pre-PSF-matching F606W aperture fluxes and segmentation areas. Figure 5 shows the resulting rest-frame UV surface brightnesses as a function of projected distance from the host-galaxy center for all 48 UV-emitting regions identified in our segmentation analysis (§ 3.2). The F606W surface brightness becomes fainter with increasing radius, with Spearman correlation coefficient ρ=0.433\rho= 0.433. We evaluated the significance by randomly permuting the surface brightnesses among the projected radii 10610^{6} times and recalculating ρ\rho. Of these, 2,277 permutations produced ∣ρ∣≥0.433|\rho| \ge0.433, giving a two-sided pp-value of p=2.28×10−3p = 2.28 \times10^{-3}.3

Figure 5. F606W surface brightness versus projected nuclear radius for all 48 UV regions. Surface brightnesses were calculated from the native, pre-PSF-matching F606W aperture fluxes and native F606W segmentation areas. The solid line is the population fit, and the shading is its pointwise 95% mean-relation confidence band from 2×1052 \times10^{5} bootstrap realizations in which the 48 regions were resampled with replacement. The dashed line is the constant-column scattering reference I606∝Rproj−2I_{606} \propto R_{\mathrm{proj}}^{-2}, with its intercept fitted while fixing Bscat=5B_{\mathrm{scat}} = 5.

The expected radial dependence for scattering follows from geometric dilution of the quasar radiation field. At an observer-frame frequency νo\nu_o in F606W, the corresponding emitted-frame frequency is νe=(1+z)νo\nu_e = (1 + z)\nu_o. For optically thin, single scattering, the observed specific intensity is then

Iνo=Lνe,QSO4π(1+z)3Nssin⁡2θdσνedΩTνeRproj−2,(7)I_{\nu_o} = \frac{L_{\nu_e,\mathrm{QSO}}}{4\pi(1+z)^3}N_s \sin^2\theta\frac{d\sigma_{\nu_e}}{d\Omega}T_{\nu_e}R_{\mathrm{proj}}^{-2}, \tag*{(7)}

where NsN_s is the scatterer column, θ\theta is the scattering angle, dσνe/dΩd\sigma_{\nu_e}/d\Omega is the differential scattering cross section evaluated at the emitted-frame frequency, and TνeT_{\nu_e} is the transmission (the full derivation is given in Appendix § 8). We collect all terms other than the radial dependence into an effective scattering normalization, K606K_{606}, such that

I606=K606Rproj−2,(8)I_{606} = K_{606}R_{\mathrm{proj}}^{-2}, \tag*{(8)}

where

K606=Lνe,QSO4π(1+z)3Nssin⁡2θdσνedΩTνe.(9)K_{606} = \frac{L_{\nu_e,\mathrm{QSO}}}{4\pi(1+z)^3}N_s \sin^2\theta\frac{d\sigma_{\nu_e}}{d\Omega}T_{\nu_e}. \tag*{(9)}

Thus, K606K_{606} contains the quasar luminosity, scatterer column, scattering geometry and phase function, cosmological dimming, and transmission. We adopt this radius-independent scattering normalization as a simple reference model that isolates the radial dependence produced by geometric dilution. This does not require every region to have the same scattering normalization; rather, it assumes that variations in the quantities entering K606K_{606} have no systematic dependence on projected radius and therefore add scatter without changing the mean radial slope. The corresponding AB surface brightness is μ606=−2.5log⁡10I606+C\mu_{606} = -2.5\log_{10} I_{606} + C, where CC is a radius-independent constant set by the AB zero point and intensity units. Combining this relation with Equation (8) gives

μ606(Rproj)=μ606(R0)+5log⁡10(RprojR0).(10)\mu_{606}(R_{\mathrm{proj}}) = \mu_{606}(R_0) + 5\log_{10}\left(\frac{R_{\mathrm{proj}}}{R_0}\right). \tag*{(10)}

We take R0=1R_0 = 1 kpc, so that μ606(R0)\mu_{606}(R_0) corresponds to the intercept AA in the fitted relation below. Thus, the constant-column reference model, with the other scattering terms held fixed, predicts a coefficient Bscat=5B_{\mathrm{scat}} = 5 in a fit of surface brightness against logarithmic radius. An ordinary least-squares fit to the observed regions gives

μ606=A+Blog⁡10(Rprojkpc),B=0.49±0.16.(11)\mu_{606} = A + B\log_{10}\left(\frac{R_{\mathrm{proj}}}{\mathrm{kpc}}\right), \qquad B = 0.49 \pm0.16. \tag*{(11)}

The difference B−Bscat=−4.51±0.16B - B_{\mathrm{scat}} = -4.51 \pm0.16 is 27.8 times the nominal ordinary-least-squares standard error. Resampling the 48 regions with replacement in 2×1052 \times10^5 bootstrap realizations gives a 95% bootstrap percentile interval of 0.18<B<0.840.18 < B < 0.84, which remains far below Bscat=5B_{\mathrm{scat}} = 5. The shaded region in Figure 5 is the 95% confidence band obtained from these same bootstrap fits. The observed radial relation is therefore much shallower than predicted by simple scattering models.

Allowing the effective scattering normalization to vary with radius, the observed relation μ606=A+Blog⁡10(Rproj/kpc)\mu_{606} = A + B\log_{10}(R_{\mathrm{proj}}/\mathrm{kpc}) corresponds to I606∝Rproj−0.4BI_{606} \propto R_{\mathrm{proj}}^{-0.4B}. Because I606=K606Rproj−2I_{606} = K_{606}R_{\mathrm{proj}}^{-2}, the required normalization scales as K606∝Rproj2−0.4BK_{606} \propto R_{\mathrm{proj}}^{2-0.4B}. The measured slope therefore requires K606∝Rproj1.80K_{606} \propto R_{\mathrm{proj}}^{1.80}. If all other scattering quantities are fixed, K606K_{606} is proportional to NsN_s, and the required scatterer column follows the same radial scaling.

The inferred scaling corresponds to an increase in K606K_{606} by a factor of approximately 33 between 2.3 and 16 kpc. Such a strong outward increase is not expected for a smooth, centrally concentrated scattering medium. For example, mass conservation in a steady radial outflow with fixed solid angle gives M˙=Ωflowr2ρv=constant\dot{M} = \Omega_{\mathrm{flow}} r^2 \rho v = \mathrm{constant} ([89 see Section 4.4]), where M˙\dot{M} is the mass-outflow rate, Ωflow\Omega_{\mathrm{flow}} is the solid angle subtended by the flow, rr is the physical distance from the nucleus, ρ\rho is the gas mass density, and vv is the radial flow speed. If the number of scattering particles per unit gas mass is constant with radius, the scatterer number density traces the gas density: ns∝ρ∝(r2v)−1n_{\mathrm{s}} \propto\rho\propto(r^{2}v)^{-1}. Thus, for a constant flow speed, ns∝r−2n_{\mathrm{s}} \propto r^{-2}. A cone with a fixed opening angle widens in proportion to its distance from the nucleus, so the characteristic path length through the scattering material scales as ℓ∝r\ell\propto r. This offsets one power of the density decline: if ns∝r−2n_{\mathrm{s}} \propto r^{-2}, then Ns∼nsℓ∝r−1N_{\mathrm{s}} \sim n_{\mathrm{s}}\ell\propto r^{-1}. The scatterer column therefore still decreases outward. A smooth radial flow would therefore tend to reduce, rather than increase, the effective scattering normalization at larger radii. Clumpy gas accretion through the surrounding halo, gas associated with or stripped from satellite galaxies, and tidal debris or shells produced by galaxy encounters could create localized enhancements in the scatterer column or cloud filling factor within individual galaxies at large radii. However, these localized structures provide no clear physical basis for the large, systematic increase in the effective scattering normalization with radius inferred from the full galaxy sample.

One possible selection bias is incompleteness near our surface-brightness detection limit: we may be missing faint outer UV-emitting regions, which would bias the measured relation toward a shallower slope. However, a robustness test for incompleteness at large radii shows that adding 20 hypothetical undetected regions distributed evenly between 16 and 24 kpc with μ606=26\mu_{606} = 26 mag arcsec−2^{-2} increases the fitted slope only from B=0.49B = 0.49 to B=1.64B = 1.64. Even adding 40 hypothetical regions over the same radial range with μ606=28\mu_{606} = 28 mag arcsec−2^{-2} gives only B=3.17B = 3.17. With 40 such regions, obtaining B≃5B \simeq5 would require μ606≃31\mu_{606} \simeq31 mag arcsec−2^{-2}, more than 6 mag fainter than the faintest detected region. Moreover, the absolute residuals from the fitted relation show no significant radial trend (Spearman ρ=−0.039\rho= -0.039, p=0.79p = 0.79), providing no indication that the scatter among the detected regions changes systematically with radius. These tests indicate that modest faint-end incompleteness can not account for the large difference from the constant-normalization scattering prediction.

A separate possible bias arises from differences in intrinsic hidden-quasar luminosity among the galaxies. A more luminous quasar could make scattered emission detectable at larger radii. Quasar luminosity, however, affects all regions within a given galaxy and therefore acts primarily as a galaxy-wide shift in surface brightness, rather than producing a radial trend within that galaxy. Indeed, 30 of the 48 regions belong to galaxies with multiple regions at different projected radii. Luminosity differences would substantially flatten the combined-sample relation only if the most luminous quasars systematically occurred in the galaxies contributing the largest-radius regions.

One final consideration is that quasar variability, coupled with differences in light-travel time, could in principle flatten the profile if more distant clouds sample an earlier, more luminous phase of AGN activity. As an illustration, consider clouds at nuclear distances of 1 and 11 kpc, both lying in the plane of the sky. Their light-echo delays differ by 10 kpc/c≃3.3×104c \simeq3.3 \times10^{4} yr. The difference in AB surface brightness between two radii is μ606(R)−μ606(R0)=−2.5log⁡10[Iνo(R)/Iνo(R0)]\mu_{606}(R) - \mu_{606}(R_{0}) = -2.5\log_{10}[I_{\nu_{o}}(R)/I_{\nu_{o}}(R_{0})]. Combining this with the fitted relation μ606=A+Blog⁡10(R/R0)\mu_{606} = A + B\log_{10}(R/R_{0}) yields Iνo(R)/Iνo(R0)=(R/R0)−0.4BI_{\nu_{o}}(R)/I_{\nu_{o}}(R_{0}) = (R/R_{0})^{-0.4B}. Holding the scatterer column, scattering geometry, and transmission fixed, the scattering intensity scales as Iνo∝Lνe,QSOR−2I_{\nu_{o}} \propto L_{\nu_{e},\mathrm{QSO}}R^{-2}, where νe=(1+z)νo\nu_{e} = (1+z)\nu_{o} and Lνe,QSOL_{\nu_{e},\mathrm{QSO}} is the nuclear specific luminosity at the epoch sampled by each cloud. Consequently, Lνe,QSO∝IνoR2∝R2−0.4BL_{\nu_{e},\mathrm{QSO}} \propto I_{\nu_{o}}R^{2} \propto R^{2-0.4B}. For B=0.49B = 0.49, the nuclear specific luminosity would therefore have to decrease by a factor of (11/1)2−0.4B≃76(11/1)^{2-0.4B} \simeq76 between the epochs illuminating the outer and inner clouds. Explaining the population trend in this way would invoke ubiquitous dimming of the nuclear UV continuum by similar factors over comparable light-travel timescales. We have no independent evidence for such a common fading trend; stochastic variability without it would primarily add scatter without changing the mean radial slope.

Taken together, the much shallower observed profile provides strong evidence against scattered quasar light as the dominant population-wide origin of the UV emission, under the physically motivated expectation that the effective scattering normalization does not increase strongly with radius. This result does not exclude a sub-dominant scattered component, scattering in individual regions, or more complex scattering models involving systematic variations in scatterer column, extinction, or geometry.

Scattering Energetics of the UV Regions

The radial surface-brightness analysis above tests whether the population of UV regions follows the radial behavior expected if scattered quasar light is the dominant emission mechanism. We next consider the complementary region-by-region question of whether a representative hidden quasar could produce the observed F606W surface brightnesses without requiring an implausibly large mass of scattering material. Specifically, if all of the F606W emission in each UV region were scattered quasar continuum, what ionized-gas or dust mass would be required within the corresponding UV aperture? This calculation tests the feasibility of scattering in individual regions and does not assume that all regions share the same emission mechanism.

We calculate the required line-of-sight scatterer column in each UV region from the optically thin scattering relation derived in Appendix § 8.§ 2. Expressing the observed specific intensity in terms of the F606W surface brightness and normalizing the intrinsic quasar spectrum at 2500 Å gives

Ns=4π(1+z)3k2500ν2500Rproj2ΩasLbolsin⁡2θ[dσ(λrest,piv,θ)/dΩ]T(λrest,piv)×fUV,scat10−0.4(μ606+48.60)(λrest,piv2500 A˚)α(12)N_{\mathrm{s}} = \frac{4\pi(1+z)^{3}k_{2500}\nu_{2500}R_{\mathrm{proj}}^{2}}{\Omega_{\mathrm{as}}L_{\mathrm{bol}}\sin^{2}\theta\left[\mathrm{d}\sigma(\lambda_{\mathrm{rest,piv}},\theta)/\mathrm{d}\Omega\right]T(\lambda_{\mathrm{rest,piv})}} \times f_{\mathrm{UV,scat}}10^{-0.4(\mu_{606}+48.60)}\left(\frac{\lambda_{\mathrm{rest,piv}}}{2500\ \text{Å}}\right)^{\alpha} \tag*{(12)}

where μ606\mu_{606} is the observed F606W surface brightness, fUV,scatf_{\mathrm{UV,scat}} is the fraction of the measured F606W emission attributed to scattered quasar light, Ωas=2.35044×10−11 sr\Omega_{\mathrm{as}} = 2.35044 \times10^{-11}\ \mathrm{sr} is the solid angle of one square arcsecond, LbolL_{\mathrm{bol}} is the intrinsic quasar bolometric luminosity, and k2500=Lbol/(ν2500Lν,2500)k_{2500} = L_{\mathrm{bol}}/(\nu_{2500}L_{\nu,2500}) is the 2500 Å bolometric correction. The emitted-frame wavelength corresponding to the F606W pivot wavelength is λrest,piv=5887 A˚/(1+z)\lambda_{\mathrm{rest,piv}} = 5887\ \text{Å}/(1+z), α\alpha is the intrinsic continuum slope defined by Lν∝ναL_{\nu} \propto\nu^{\alpha}, θ\theta is the scattering angle, and TT is the combined transmission along the quasar–scatterer and scatterer–observer paths.

For the fiducial calculation, we assign all of the measured F606W emission to scattered quasar light, so that fUV,scat=1f_{\mathrm{UV,scat}} = 1, and neglect attenuation along both light paths, so that T=1T = 1. We adopt a representative hidden-quasar luminosity of log⁡(Lbol/erg s−1)=47.28\log(L_{\mathrm{bol}}/\mathrm{erg\ s}^{-1}) = 47.28, a bolometric correction of k2500=2.75k_{2500} = 2.75, and an intrinsic continuum slope of α=−0.44\alpha= -0.44. Equation (12) then becomes

Ns,fidcm−2=8.81×1022(1+z)3(Rprojkpc)210−0.4(μ606+48.60)(13)\frac{N_{\mathrm{s,fid}}}{\mathrm{cm}^{-2}} = 8.81 \times10^{22}(1+z)^{3}\left(\frac{R_{\mathrm{proj}}}{\mathrm{kpc}}\right)^{2}10^{-0.4(\mu_{606}+48.60)} \tag*{(13)}
×(λrest,piv2500 A˚)−0.44{sin⁡2θ[dσ(λrest,piv,θ)/dΩ1 cm2 sr−1]}−1.(14)\times\left(\frac{\lambda_{\mathrm{rest,piv}}}{2500\ \text{Å}}\right)^{-0.44}\left\{\sin^{2}\theta\left[\frac{\mathrm{d}\sigma(\lambda_{\mathrm{rest,piv}},\theta)/\mathrm{d}\Omega}{1\ \mathrm{cm}^{2}\ \mathrm{sr}^{-1}}\right]\right\}^{-1}. \tag*{(14)}

We evaluate Equation 13 separately for electron and dust scattering and convert the inferred columns into ionized-gas and dust masses within the UV apertures. For each model, we adopt the favorable geometry described in § 4.6.1 and 4.6.2, thereby minimizing the column required to account for all of the observed F606W emission. The full derivation and justification of the fiducial assumptions are given in Appendix § 8.§ 2.

Electron Scattering

For electron scattering, the scatterer column is the free-electron column, Ns=NeN_{\mathrm{s}} = N_{\mathrm{e}}. The Thomson differential cross section is

dσTdΩ=3σT16π(1+cos⁡2θ),(15)\frac{\mathrm{d}\sigma_{\mathrm{T}}}{\mathrm{d}\Omega} = \frac{3\sigma_{\mathrm{T}}}{16\pi}(1+\cos^{2}\theta), \tag*{(15)}

where σT\sigma_{\mathrm{T}} is the Thomson cross section. At fixed RprojR_{\mathrm{proj}}, the electron-scattering intensity depends on the scattering angle through the factor

Ee(θ)=sin⁡2θdσTdΩ=3σT16π(1−cos⁡4θ).(16)\mathcal{E}_{\mathrm{e}}(\theta) = \sin^{2}\theta\frac{\mathrm{d}\sigma_{\mathrm{T}}}{\mathrm{d}\Omega} = \frac{3\sigma_{\mathrm{T}}}{16\pi}(1-\cos^{4}\theta). \tag*{(16)}

Unlike dust scattering, Thomson scattering is independent of wavelength. Equation (16) is maximized at θ=90∘\theta= 90^\circ, so we adopt this favorable geometry to obtain the smallest required electron column.

If the scattering material lies along the radio axis, then θ=i\theta= i on the near side and θ=180∘−i\theta= 180^\circ-i on the far side, where ii is the jet inclination to our line of sight. Because Equation (16) is symmetric about θ=90∘\theta= 90^\circ, the corresponding near- and far-side locations have the same Thomson-scattering efficiency. Type 2 3CR radio galaxies are expected to have i≳45∘i \gtrsim45^\circ [9], so geometries reasonably close to the plane of the sky are plausible. For example, i=60∘i = 60^\circ, corresponding to θ=60∘\theta= 60^\circ or 120∘120^\circ, retains

ϵe(60∘)ϵe(90∘)=1−cos⁡4(60∘)=0.9375(17)\frac{\epsilon_e(60^\circ)}{\epsilon_e(90^\circ)} = 1-\cos^4(60^\circ)=0.9375 \tag*{(17)}

of the maximum efficiency. Thus, although θ=90∘\theta=90^\circ provides the most favorable limiting case, moderate departures from this geometry change the required electron column only modestly.

To convert this column into a gas mass, let ρ\rho be the total gas mass density, nen_e the free-electron number density, and mpm_p the proton mass. We define

μe≡ρnemp,(18)\mu_e \equiv\frac{\rho}{n_e m_p}, \tag*{(18)}

so that μe\mu_e is the dimensionless mean gas mass per free electron in units of mpm_p. The ionized-gas mass within an aperture of projected physical area AskyA_{\rm sky} is then

Mion=μempNeAsky.(19)M_{\rm ion}=\mu_e m_p N_e A_{\rm sky}. \tag*{(19)}

To obtain a lower limit on MionM_{\rm ion} at fixed NeN_e, we assume full ionization4. Fully ionized hydrogen supplies one electron per nucleon, while helium supplies two electrons per four nucleons. For the dominant metals, the numbers of protons and neutrons are approximately equal, giving about one electron per two nucleons when fully ionized. Approximating the nucleon mass by mpm_p and adopting the protosolar mass fractions of hydrogen (X=0.7154X=0.7154), helium (Y=0.2703Y=0.2703), and all elements heavier than helium (Z=0.0142Z=0.0142) from Asplund et al. [5] gives

μe≃[X+Y+Z2]−1=1.17.(20)\mu_e \simeq\left[X+\frac{Y+Z}{2}\right]^{-1}=1.17. \tag*{(20)}

The left panel of Figure 6 shows the distribution of the required ionized-gas mass after summing the disjoint UV apertures within each galaxy. For comparison, Nesvadba et al. [88] studied 33 powerful radio galaxies spanning 1.4<z<4.81.4<z<4.8, and estimated jet kinetic powers ranging from a few 104610^{46} to nearly 1048 erg s−110^{48}\ \mathrm{erg\ s^{-1}}. Although their sample extends to higher redshift than our predominantly 1<z<21<z<2 3CR sample, both comprise exceptionally powerful radio galaxies with energetic jets. Their measurements therefore provide a relevant empirical benchmark for the ionized-gas reservoirs associated with this broad class of systems, rather than a redshift- or radio-power-matched control sample.

Figure 6. Scattering requirements for the UV-emitting regions. Left: distribution of the ionized-gas mass required for electron scattering, summed over the disjoint UV apertures in each of the 30 galaxies. The shaded region shows the warm-ionized gas masses measured for powerful high-redshift radio galaxies by Nesvadba et al. [88], assuming ne=500 cm−3n_{\mathrm{e}} = 500\ \mathrm{cm}^{-3}, and the dashed line marks the median required mass in our sample. Right: distribution of the dust mass required within the UV apertures relative to the global SED-based dust mass for the 11 galaxies with available measurements. The dashed line marks the median required fraction, and the dotted line marks 0.5% of the global dust mass. Short vertical marks along the bottom of each panel show the individual galaxies.

From the spatially integrated Balmer-line emission of the detected nebulae, Nesvadba et al. [88] inferred

warm-ionized gas masses of approximately 2×1082\times10^8–5×109 M⊙5\times10^9\ M_\odot, adopting ne=500 cm−3n_e=500\ \mathrm{cm^{-3}}. This warm gas supplies free electrons capable of Thomson scattering, although Balmer emission preferentially traces dense, warm material, whereas scattering depends on the total electron column and may also include diffuse or hotter plasma. Moreover, the SINFONI masses encompass the detected line-emitting nebulae, whereas our estimates include only the much more restricted UV apertures. The comparison is therefore conservative: 14 of our 30 galaxies require more than 5×109 M⊙5\times10^9\ M_\odot within the UV apertures alone. If the scattering medium extends beyond these apertures, its total ionized-gas mass would be larger still. Dominant electron scattering would therefore require ionized-gas reservoirs at or beyond the upper end of those measured in other powerful high-redshift radio galaxies, despite being concentrated within only a spatial subset of each galaxy.

Dust Scattering

Dust is a much more efficient UV scatterer than free electrons. We adopt the optical/UV differential scattering cross sections calculated by Draine [46] for the carbonaceous–silicate grain-size distribution of Weingartner & Draine [128] that reproduces the average Milky Way extinction curve with RV≡AV/E(B−V)=3.1R_V\equiv A_V/E(B-V)=3.1. For each source, we evaluate dσdust/dΩ\mathrm{d}\sigma_{\rm dust}/\mathrm{d}\Omega at rest-frame F606W pivot wavelength and scattering angle by interpolating between neighboring entries in the public numerical tables accompanying Draine [46].

The tabulated differential cross sections are normalized per hydrogen nucleus and already incorporate the assumed abundance and size distribution of dust grains relative to the gas. Accordingly, the scatterer column in Equation 13 is Ns=NHN_s=N_H.

Because dust scattering is anisotropic and forward-peaked, its most favorable geometry at fixed RprojR_{\rm proj} is not θ=90∘\theta=90^\circ. The relevant angular efficiency is

Edust(λ,θ)=sin⁡2θdσdust(λ,θ)dΩ.(21)\mathcal{E}_{\rm dust}(\lambda,\theta)=\sin^2\theta\frac{\mathrm{d}\sigma_{\rm dust}(\lambda,\theta)}{\mathrm{d}\Omega}. \tag*{(21)}

The wavelength dependence of the dust phase function causes the optimum angle to vary slightly across the sample. Maximizing Equation (21) at each source’s rest-frame pivot wavelength gives θmax⁡=35∘\theta_{\max}=35^\circ–41∘41^\circ. For a near-side cloud lying along the radio axis, the scattering angle equals the jet inclination to the line of sight, θ=i\theta=i. In the orientation-based unification model for 3CR sources, Barthel [9] inferred a critical inclination of approximately 45∘45^\circ, with quasars viewed at smaller inclinations and radio galaxies at larger inclinations. We therefore adopt θ=45∘\theta=45^\circ as a favorable boundary case for an on-axis scattering cloud. At this angle, Edust\mathcal{E}_{\mathrm{dust}} retains 96.6–99.3% of its source-specific maximum. The adopted angle is not intended to represent the inclination of every source, and scattering material that does not lie along the radio axis need not have θ=i\theta= i. Adopting θ=60∘\theta= 60^\circ or 90∘90^\circ increases the required NHN_{\mathrm{H}} by median per-region factors of 1.15 or 1.94, respectively.

Equation 13 then gives the Hydrogen-nucleus column NHN_{\mathrm{H}} required by the measured UV surface brightness of each region. To keep the mass conversion consistent with our choice of grain model, we use its tabulated dust mass per Hydrogen nucleus, mdust/H=1.398×10−26 g H−1m_{\mathrm{dust/H}} = 1.398 \times10^{-26}\ \mathrm{g\ H}^{-1}.5 The dust mass within an aperture of projected physical area AskyA_{\mathrm{sky}} is therefore

Mdust=mdust/HNHAsky=(1.398×10−26 g H−1)NHAsky.(22)M_{\mathrm{dust}} = m_{\mathrm{dust/H}}N_{\mathrm{H}}A_{\mathrm{sky}} = \left(1.398 \times10^{-26}\ \mathrm{g\ H}^{-1}\right)N_{\mathrm{H}}A_{\mathrm{sky}}. \tag*{(22)}

Across the rest-frame wavelengths sampled by F606W, the dust differential cross section per Hydrogen nucleus at θ=45∘\theta= 45^\circ is of order 10310^3 times the Thomson differential cross section per electron at θ=90∘\theta= 90^\circ.

As a consistency check on the optically thin solution, we use the inferred Hydrogen-nucleus columns to calculate the corresponding optical depths:

τsca=NHσsca,τext=NHσext.(23)\tau_{\mathrm{sca}} = N_{\mathrm{H}}\sigma_{\mathrm{sca}}, \qquad\tau_{\mathrm{ext}} = N_{\mathrm{H}}\sigma_{\mathrm{ext}}. \tag*{(23)}

Here σsca\sigma_{\mathrm{sca}} and σext\sigma_{\mathrm{ext}} are the total scattering and extinction cross sections per Hydrogen nucleus, respectively, supplied by the grain model at the source’s rest-frame pivot wavelength. The extinction cross section is

σext=σsca+σabs,(24)\sigma_{\mathrm{ext}} = \sigma_{\mathrm{sca}} + \sigma_{\mathrm{abs}}, \tag*{(24)}

where σabs\sigma_{\mathrm{abs}} is the absorption cross section per Hydrogen nucleus. The resulting optical depths have median values of τsca=0.008\tau_{\mathrm{sca}} = 0.008 and τext=0.015\tau_{\mathrm{ext}} = 0.015, with maxima of 0.135 and 0.220, respectively. These small values support the assumptions used in the calculation. The maximum τsca\tau_{\mathrm{sca}} indicates that multiple scattering should remain a minor correction, while the maximum τext\tau_{\mathrm{ext}} corresponds to a minimum transmission of T=e−τext=e−0.22≃0.80T = e^{-\tau_{\mathrm{ext}}} = e^{-0.22} \simeq0.80. Thus, neglecting attenuation is an excellent approximation for a typical region and remains a favorable limiting assumption for the most opaque case. These optical depths are model-dependent consistency checks rather than independent measurements: the Draine cross sections do not impose τ<1\tau< 1, and a larger inferred column could instead have invalidated the optically thin solution. They also do not constrain additional attenuation along the quasar–cloud or cloud–observer paths.

The right panel of Figure 6 compares the inferred dust requirements with the galaxy-wide dust reservoirs. Of the 12 z>1z > 1 3CR radio galaxies with global SED-based dust masses from Podigachoski et al. [98], 11 have detected UV regions and enter this comparison; 3C 68.2 has no such detection. For each of these 11 galaxies, we sum the required dust mass over the disjoint UV apertures and normalize it by the galaxy-wide dust mass. The median required fraction is only 1.2×10−41.2 \times10^{-4} and remains below 0.5% in every case. Even allowing one magnitude of UV attenuation keeps the maximum below 1.2%. Changing to an SMC-bar grain model increases the required Hydrogen-nucleus column by a median factor of 4.9 but still does not strain the available dust reservoir.

Table 8 presents the region-by-region quantities underlying these results, including the projected distance, aperture area, and F606W surface brightness. The electron column and ionized-gas mass are calculated for the favorable θ=90∘\theta= 90^\circ geometry. The hydrogen-nucleus column and dust mass are calculated for the fiducial Milky Way RV=3.1R_V = 3.1 dust model at θ=45∘\theta= 45^\circ.

SourceRegionRprojR_{\mathrm{proj}} (kpc)AskyA_{\mathrm{sky}} (kpc2^{2})μ606\mu_{606} (AB mag arcsec−2^{-2})log⁡Ne\log N_{\mathrm{e}} (cm−2^{-2})log⁡Mion\log M_{\mathrm{ion}} (M⊙M_{\odot})log⁡NH\log N_{\mathrm{H}} (cm−2^{-2})log⁡Mdust\log M_{\mathrm{dust}} (M⊙M_{\odot})
3C 1319.9038.1923.6722.5410.1019.554.96
3C 1320.7078.4622.8320.588.4417.593.31
3C 3610.5227.8421.5720.798.2117.813.08

Table 8. **Note—**All logarithms are base 10. The complete table, including the dust scattering and extinction optical depths, is available in machine-readable form.

UV Strength and Host–Radio Alignment

We next tested whether the luminosity of the UV-emitting regions depends on the orientation of the host galaxy relative to the jet axis. The host-alignment sample contains 26 radio galaxies satisfying b/a<0.8b/a < 0.8, of which 23 contain at least one resolved UV-emitting region and enter this analysis. We derive a power-law spectral slope for each region from its PSF-matched F606W flux density and host-subtracted F140W residual. For regions lacking an F140W detection, we used the median spectral slope of the detected regions in the same galaxy. We normalized each power law to the native, pre-PSF-matching F606W aperture flux density and estimated the rest-frame specific luminosity at 2500 Å. We then summed the luminosities of all detected UV regions within each galaxy, including regions too round to enter the UV–radio alignment tests. For both relations below, we calculated Spearman ρ\rho across the 23 galaxies and obtained the two-sided pp-values from the standard Student-tt approximation to its null distribution, with n−2=21n - 2 = 21 degrees of freedom.

The total UV luminosity decreases with increasing host–radio misalignment (Figure 7a), with Spearman ρ=−0.667\rho= -0.667 and p=5.09×10−4p = 5.09 \times10^{-4}. Because the total rest-frame UV luminosity may scale with the galaxy’s rest-frame optical luminosity, we also normalized LUVL_{\mathrm{UV}} by the rest-frame VV-band host luminosity derived from the F140W photometry using the passive stellar-population KK correction described in § 4.1. The anti-correlation remains for LUV/LhostL_{\mathrm{UV}}/L_{\mathrm{host}} (Figure 7b), with Spearman ρ=−0.561\rho= -0.561 and p=5.33×10−3p = 5.33 \times10^{-3}. The aligned systems therefore exhibit stronger total UV emission not only in absolute terms, but also relative to their rest-frame optical stellar continuum.

Figure 7. Rest-frame 2500 Å specific luminosity of the UV-emitting regions as a function of the misalignment angle between the host-galaxy semimajor axis and the radio axis. Panel (a) shows the summed rest-frame Lν(2500 A˚)L_{\nu}(2500\,\text{Å}) of the resolved UV regions, and panel (b) shows this quantity normalized by the passive-template K-corrected rest-frame VV-band host luminosity. Black points represent individual galaxies. Solid lines are ordinary least-squares fits, and the shaded regions are 95% confidence bands for the fitted mean relations from 10,000 source-bootstrap realizations. The displayed ρ\rho values and two-sided pp-values are from Spearman rank tests, with the pp-values calculated using the Student-tt approximation.

Discussion

The principal result motivating the broader analysis presented in this paper is that the large-scale radio jets of the z>1z > 1 type 2 3CR radio galaxies are preferentially aligned with three distinct components of their host galaxies and surrounding environments: the major axes of their older stellar populations, their clumpy rest-frame UV emission, and the projected axes of close candidate mergers. The alignment between the radio and rest-frame UV emission confirms the well-established UV manifestation of the alignment effect in a statistically complete high-redshift radio-galaxy sample at high significance (McCarthy et al. [80]; Chambers et al. [29]). We further find a strong alignment between the radio axes and the major axes of the smooth rest-frame optical hosts dominated by older stellar populations.

While alignment at rest-frame optical wavelengths has been reported previously (Dunlop & Peacock [49]; Best et al. [12]; Pentericci et al. [95]; Zirm et al. [134]), most of these studies measured the orientation of the total light distribution which contains both the underlying stellar host plus any additional aligned structures. In particular, Best et al. [12] found that the measured (observer-frame) near-IR K-band alignment was strongest in the galaxies for which their SED modeling attributed the largest fraction of the K-band flux to a relatively flat-spectrum component associated with the rest-frame UV alignment effect. After the four galaxies with the strongest contributions from this component were excluded, the remaining K-band alignment was only weakly significant. The clearest previous evidence specific to the older stellar population came from Zirm et al. [134], who analyzed nine 3CR radio galaxies with near-IR HST imaging. They masked neighbors and bright aligned features before measuring host position angles, broadly paralleling our approach (§ 3.1), and found non-random host–radio alignment at 99.9% confidence. Our uniform two-dimensional Sérsic analysis extends that result to the statistically complete z>1z > 1 3CR type 2 parent sample. Of 33 usable hosts, 26 satisfy the b/a<0.8b/a < 0.8 requirement; their host–radio alignment significance, p=9.01×10−5p = 9.01 \times10^{-5}, establishes a population-wide correspondence between the radio axis and the large-scale distribution of older stars. The strength of the UV emission is also related to the host–radio geometry. Galaxies with closer host–radio alignment contain stronger UV emission, both in absolute terms and relative to the rest-frame optical host (Figure 7). The common projected axis is therefore associated not only with the orientation but also with the prominence of the UV component. One possible interpretation is that gas and dust are already distributed along the host or assembly axis. When the radio source follows the same axis, more of this material may lie within the jet-driven cocoon or quasar ionization cone, enhancing jet-induced star formation, nebular emission, or scattering from a hidden quasar. Beyond the trend with UV luminosity, our exploratory comparisons found no robust association between host–radio misalignment and any of the following: the presence of a selected companion, projected radio size, rest-frame 178 MHz radio luminosity, the UV–optical color measured from the combined light of the detected UV regions, or the number of those regions.

A preference for neighboring galaxies to lie near the radio axis has also been reported previously ([95]), although their analysis differed substantially from the galaxy-merger-focused measurement presented here. In a heterogeneous sample of 19 radio galaxies at 1.7<z<3.21.7 < z < 3.2, they defined companions as compact sources with H<23.5H < 23.5 at projected radii of 30–50 kpc. They found a non-random distribution of candidate companion orientations at 99% confidence, with the companions preferentially located along the radio axis, or along the optical axis when the radio source was unresolved. Their result established an anisotropy in the relatively faint, larger-scale neighboring population, but did not specifically test the geometry of close, relatively luminous candidate mergers.

Our analysis extends this earlier environmental result into the inner, close-companion regime explicitly excluded by Pentericci et al. [95]. We require candidate companions to lie at projected separations of 5–25 kpc and within two F140W magnitudes of the host, selecting relatively luminous candidate mergers. Although our merger-axis sample contains only eight systems and spectroscopy is required to confirm the physical association of the candidate companions, the simultaneous alignment of the close-companion axes, smooth stellar hosts, and UV-emitting regions with the radio axes suggests that these orientations may not be independent.

There is also a morphological degeneracy between a single elongated host galaxy and two strongly overlapping galaxies in a late-stage merger. Our requirement that candidate companions have 5<Rproj<255 < R_{\rm proj} < 25 kpc excludes systems with centers separated by less than 5 kpc, where overlapping stellar envelopes can cause two galaxies to be modeled as one elongated host. Such systems would not enter our companion-selected merger-axis sample, although their projected separation axes could already contribute to the host–radio alignment. The measured merger–radio alignment may therefore be conservative, since identifying additional radio-aligned systems of this type would tend to strengthen the signal.

A Common Axis for the Radio Source, Host Galaxy, and Galaxy Mergers

While the clumpy rest-frame UV emission could, in principle, reflect a local and rapid response to the jet, the smooth rest-frame optical light traces the galaxy-wide distribution of older stars. This stellar structure can not represent the same rapid response to the current radio episode. Star formation associated with earlier episodes of jet activity may have contributed to the older stellar population, but the host–radio alignment could instead indicate that both orientations trace a longer-lived structure established during galaxy assembly.

Stellar Rotation and the Jet Axis

The physical origin of the host–radio alignment is not immediately obvious. The launch direction of a radio jet is expected to be linked to the angular-momentum axis of the innermost accretion flow and/or the spin axis of the black hole. These axes determine the preferred outflow direction in accretion-disk-launched and black-hole-spin-powered models, respectively [17, 18]. In a rotationally supported oblate galaxy, the stellar angular-momentum axis is perpendicular to the disk plane. If the nuclear gas disk is coplanar with the stellar disk, this simple geometry predicts that the jet should be preferentially aligned with the host minor axis rather than its major axis. Such a minor-axis preference has been observed among some lower-redshift early-type galaxies [10]. Among the z<0.5z < 0.5 radio AGN studied by Zheng et al. [132], this alignment along the host minor axis is stronger at lower radio luminosities and in more elongated hosts. They suggest that coherent and secular accretion between galaxy mergers could align black hole spins with the angular momentum of their host galaxies and thereby favor minor-axis jets in oblate hosts. If the larger-scale ISM is also distributed in a disk, the minor axis provides a lower-column-density path, while jets propagating through the disk plane are expected to interact more strongly with the surrounding gas [85].

The major-axis alignment found here is therefore opposite to this simple disk-based expectation. This distinction is important because the hosts in our 3CR sample span a broad range of Sérsic indices, including many low values consistent with disk-like surface-brightness profiles. However, a low Sérsic index alone does not establish that the host galaxy’s stellar component is a rotationally supported oblate disk [105]. If some of these hosts are genuine stellar disks, the observed alignment would instead suggest that the inner accretion flow and/or black-hole spin is preferentially misaligned with the stellar angular-momentum vector, with its projected direction lying closer to the stellar disk plane. A strongly misaligned or polar nuclear gas disk could produce such a geometry. Merger-driven torques, externally accreted gas, and instabilities within the inflow can generate this decoupling between galactic and nuclear scales [68, 117].

Dust absorption can trace the distribution of cold gas. In nearby radio galaxies, the radio axis has been found to be perpendicular in projection to compact dust features [38] or to dust lanes misaligned with the host major axis [123, 129]. If the dust traces a rotating disk, this projected radio–dust geometry is consistent with a jet directed along the gas angular-momentum axis.

A different possibility is that some hosts are triaxial or prolate spheroids rather than oblate stellar disks. In a general triaxial system, the projected photometric axes do not uniquely determine the intrinsic stellar angular-momentum direction, and substantial photometric–kinematic misalignments are possible [72]. One particular configuration that could produce the observed geometry is prolate rotation. In a prolate rotator, the stars rotate about the intrinsic long axis, so their angular-momentum vector lies approximately along that axis. Except for unfavorable viewing angles, the intrinsic long axis projects near the observed photometric major axis; the projected stellar angular-momentum direction can therefore lie closer to the host major axis than to its minor axis. This is seen in the nearby radio galaxy NGC 4261, where the stellar rotation axis is substantially misaligned with the radio axis [37]. In a sample of nearby radio galaxies, Heckman et al. [66] found little evidence for a general alignment between the radio and stellar rotation axes. If the gas feeding the black hole rotates about the same long axis as the stars and retains this orientation as it flows inward, a disk-launched jet could follow the host major axis in projection. Accretion that consistently supplies angular momentum along this axis could also bring the black-hole spin into alignment with it [45], giving a spin-powered jet the same orientation.

Deep JWST/NIRSpec integral-field spectroscopy could test for short- or long-axis rotation by using Doppler shifts in rest-frame optical absorption features to map the motion of the older stars that contribute to the radio–host alignment observed in this study [66, 112, 58]. NIRSpec has resolved stellar rotation in a post-starburst AGN host at z=3.06z = 3.06 [40] and in the central ∼2\sim2 kpc of the nearby radio galaxy 3C 293 [39]. For ordered stellar rotation, the line-of-sight velocity gradient is nearly perpendicular to the projected rotation axis. If the jet follows the host major axis, a gradient along that axis would favor short-axis rotation, with the projected rotation axis nearly perpendicular to the jet. A gradient along the host minor axis would be consistent with long-axis rotation, with the rotation axis close to the jet in projection.

Merger History and the Large-Scale Environment

Major galaxy mergers provide one possible route to long-axis stellar rotation, although the outcome depends strongly on the merger configuration. Idealized simulations show that a major polar merger can produce a prolate remnant in which long-axis rotation is inherited primarily from the progenitor disk that was initially perpendicular to the merger orbital plane [120]. Cosmological simulations likewise find that the emergence of prolate shape and rotation is frequently associated with a significant merger, and that the major axis of the remnant often follows the direction from which the progenitors approached during the final stages of coalescence [51].

The high incidence of major mergers among the z>1z > 1 3CR radio galaxies makes prolate rotation an attractive possible connection among the host, merger, and radio alignments. A preferred merger or infall direction could establish the long axis of a prolate stellar system, while angular momentum retained from the progenitors could produce rotation around that axis. If the inner accretion flow and black-hole spin subsequently became aligned with the same long axis, the jet could emerge along the host major axis and remain aligned with the preferred merger direction.

However, these simulations constrain the stellar remnant and do not demonstrate that either the nuclear gas or black-hole spin shares its long-axis orientation. Nor does the projected companion axis reveal either the polar progenitor configuration modeled by Tsatsi et al. [120] or the orbital angular momentum, which also depends on the relative velocity.

Figure 8 presents six multicomponent systems whose rest-frame optical morphologies illustrate geometries relevant to the merger–radio alignment. Among the systems entering the quantitative subsample (Table 5), 3C 194 and 3C 267 each contain two selected companion candidates, producing suggestive three-component configurations. Although 3C 255 and 3C 256 do not satisfy all of the companion-selection criteria, 3C 255 shows a possible three-component configuration broadly along the radio axis, while the central components and lower-surface-brightness extensions of 3C 256 share approximately the same orientation.

SourceNcompN_{\mathrm{comp}}PAmergerPA_{\mathrm{merger}}PAradioPA_{\mathrm{radio}}Δθmerger\Delta\theta_{\mathrm{merger}}
(deg)(deg)(deg)
3C 1942171.5159.312.1
3C 21012.5164.118.5
3C 250142.954.511.6
3C 267252.078.826.7
3C 305.1237.910.427.6
3C 469.11137.3171.233.9
3C 4701140.938.077.1
4C 13.6612.07.75.7

Table 5. Candidate-Merger Axes and Radio Alignments

Selected multicomponent systems with rest-frame optical morphologies relevant to the merger–radio alignment

Figure 8. Selected multicomponent systems with rest-frame optical morphologies relevant to the merger–radio alignment. The panels show HST/WFC3 F140W imaging with radio contours overlaid; white circles mark projected radii of 25 kpc from the host galaxies. 3C 194, 3C 210, and 3C 267 enter the quantitative merger-axis analysis. 3C 255 and 3C 256 are included as illustrative morphological examples but do not enter that subsample. For 3C 356, the close and wide views show the two principal galaxies and place them within the large-scale radio structure. The images are displayed with logarithmic intensity stretches spanning 24.7–31.5 AB mag per pixel. All images are rotated north up and east to the left.

3C 356 provides a wider-separation comparison: its two massive galaxies are separated approximately along the radio axis and are both associated with compact radio emission, although the identity of the galaxy powering the large-scale source has historically been ambiguous. Simpson & Rawlings [111] interpreted 3C 356 as a late-stage merger between two galaxy groups in which an interpenetrating collision shock-heated the intragroup gas and funneled material into the northern source, which they identified as powering the large-scale jet.

The close-companion systems considered here have generally not yet coalesced and therefore can not be the fully merged prolate remnants described by these simulations, although previous close passages may already have disturbed their structures. Nor can the current early-stage interactions have produced the binary-SMBH coalescence proposed by Breiding et al. [22] to enable the present radio jet. An earlier major merger could instead have supplied the binary SMBH, with the observed companions tracing a subsequent episode of assembly. Connecting their geometry to the radio axis would then require successive galaxy mergers to follow a longer-lived preferred direction.

Anisotropic assembly along the surrounding cosmic web provides one possible origin for this longer-lived preferred direction. Using SDSS DR7, Zhang et al. [131] found that the major axes of red central galaxies preferentially follow cosmic-web filaments, with the alignment becoming stronger toward higher luminosities and stellar masses. More recently, Jung et al. [70] found a similar alignment among massive radio galaxies at z<0.6z < 0.6, particularly for the most massive systems located closest to the filament axes. These results are consistent with massive galaxies and their halos assembling through repeated accretion and mergers directed preferentially along filaments. A long-lived filament could therefore establish the host major axis and provide a preferred direction for subsequent companions, without requiring the currently observed companion to have produced the existing stellar structure. Extending this coherence to the radio axis would still require the connection between galaxy assembly and the central engine discussed above. Wide-field imaging and spectroscopy around the 3CR galaxies could determine whether their host, companion, and radio axes trace a common filamentary environment.

If a cosmic web filament supplies gas along this common assembly axis, radio lobes expanding in this direction could be more tightly confined, increasing the magnetic field energy density and synchrotron luminosity (Hardcastle [62]). Since the 3CR survey is flux-limited, this effect could favor the inclusion of sources whose jets point along the assembly axis (Eales [50]).

One way to examine this possibility is to repeat the alignment measurements presented here at lower radio luminosities for a suitable subset of radio galaxies at 1≲z≲21 \lesssim z \lesssim2 in the LOFAR Two-metre Sky Survey Deep Fields (Best et al. [15]). The comparison would require rest-frame optical imaging from HST or JWST comparable to our HST data and high-angular-resolution radio maps that resolve both terminal hot spots. Low-frequency radio maps sensitive to diffuse emission would also allow one to measure the fraction of the integrated flux density contributed by the lobe plasma. This would allow one to test how the host–radio and companion–radio angle distributions change with the flux limit for samples matched in redshift, host galaxy properties, larger-scale environmental properties where possible, and radio morphology. The lobe flux measurements would also test whether the alignment is stronger among sources for which the lobes contribute a larger fraction of the flux used for selection.

The merger geometry is also important when interpreting the aligned UV emission. Galaxy interactions can trigger star formation through tidal compression, gas inflows, and shocks. Because the merger axes in this sample are themselves preferentially aligned with the radio axes, star-forming knots produced in tidal bridges or debris could appear radio-aligned even if the jet did not trigger them. The radio–UV alignment alone is therefore insufficient to establish positive jet feedback. Merger- and jet-triggered star formation need not, however, be mutually exclusive: the merger may supply and organize the cold gas, while the jet or its overpressured cocoon compresses particular clouds and determines where within that gas the most recent stars form.

Jet-Triggered Star Formation

The combined SPS and jet-timing analysis identifies seven regions as the strongest candidates for jet-triggered star formation (§ 4.4). For each, all accepted stellar-continuum models require a recent onset of star formation and satisfy tSF,max≲tavailt_{\mathrm{SF,max}} \lesssim t_{\mathrm{avail}} at the fiducial lobe-head advance speed of 0.01c0.01c.

This temporal consistency is necessary but not sufficient evidence for jet triggering. The comparison depends on the assumed lobe-head speed and projected geometry and establishes that the inferred sequence of events is possible, but not that the radio structure physically interacted with the star-forming gas.

A direct spatial relationship with specific radio features would strengthen the causal interpretation. If the UV emission is stellar, regions located at jet bends, hot spots, lobe edges, or independently identified shock fronts would be more difficult to explain through merger-triggered star formation alone. A UV structure that follows a bend in the radio jet would favor a local jet–gas interaction over illumination, particularly if the structure extends beyond the ionizing radiation cone from the hidden quasar. Exact coincidence with the narrow radio-jet spine is not required, because the jet-driven bow shock and overpressured cocoon can affect gas over a substantially wider area ([59, 85]). A spatial sequence in which star formation began earlier in inner regions than in regions farther outward, as expected in the simplest picture of an advancing radio source, would provide stronger evidence still. Figure 9 presents the F606W images of 3C 124, 3C 210, 3C 266, and 3C 324 with the radio contours overlaid. These galaxies were selected for their suggestive rest-frame UV morphologies rather than by the combined SPS and jet-timing criteria: their emission is resolved into multiple discrete knots or narrow, irregular structures containing several surface-brightness peaks, with the ensemble of components distributed approximately along the radio axis. Among the four galaxies, only 3C 210 satisfies the strict all-model timing condition.

Examples of particularly clumpy, radio-aligned rest-frame UV morphologies

Figure 9. Examples of particularly clumpy, radio-aligned rest-frame UV morphologies. Shown are the HST/WFC3 UVIS F606W images of 3C 124, 3C 210, 3C 266, and 3C 324, with the radio emission overlaid as green contours. Yellow ellipses show the intensity-weighted second-moment regions used to characterize the UV morphology, and the white circles mark projected radii of 25 kpc from the host galaxies. All panels are oriented with north up and east to the left; the white arrows show 3″ angular scales. The images are displayed with logarithmic intensity stretches spanning 27.4–32.4 AB mag per pixel.

The resolved, clumpy F606W morphologies are qualitatively suggestive of recent star formation, most clearly in 3C 324. Its chain of UV-emitting knots is reminiscent of the “beads-on-a-string” complex in SDSS J1531+3414, where the individual beads are young stellar clusters associated with a reservoir of multiphase gas ([92]). Similar structures also clearly extend across the host galaxy and its companion in 3C 356 (Figure 8). Although this morphological resemblance does not by itself establish a stellar origin, it provides additional support for identifying at least some of the clumpy UV components as candidate regions of recent star formation. However, even if the UV knots trace star formation, their linear arrangement alone does not establish jet triggering. Elmegreen & Scalo [52] identified galaxies with chains of large clumps in HST images of the Tadpole field and interpreted the clumps as star-forming regions in nearly edge-on disks. In simulations without a jet, gravitational fragmentation of a gas-rich disk produces star-forming clumps that appear as a chain when the disk is viewed edge-on ([20]).

Even if confirmed, jet-triggered star formation in these regions would demonstrate a localized positive response rather than a net increase in the galaxy-wide star-formation rate. Simulations show that jets can compress some clouds while turbulence and gas dispersal suppress star formation elsewhere ([85, 78]). The present UV measurements therefore can not determine the net effect of the jet on star formation throughout the host galaxy.

The effect of the radio source on star formation may change as the jets propagate through and eventually beyond the host ISM. While the radio source remains confined, its bow shock and overpressured cocoon can compress dense gas throughout the host; after breakout, the interaction may instead become increasingly dominated by turbulence, heating, and gas removal. Falkendal et al. [54] found no significant relation between global star-formation rate and radio size, but noted that their robust detections all occurred in sources smaller than approximately 3″, suggesting enhanced star formation before the radio source breaks out of the host ISM. We likewise found no significant relation between projected radio size and either the luminosity or number of the segmented UV regions. This provides no evidence for such evolution in our sample, although projected radio size is an imperfect age indicator and the UV emission is not necessarily stellar.

Scattered Quasar Light

Our finding that the surface brightness of the UV-emitting regions depends only weakly on projected radius argues against a simple scattering model as their dominant population-wide origin. Geometric dilution of the incident quasar radiation predicts a substantially steeper decline unless the properties of the scattering material change systematically with radius. Explaining the observed profile entirely through scattering would therefore require the effective scattering efficiency to increase strongly outward, which is not a natural expectation for a smooth medium or steady radial outflow.

The large ionized-gas masses required by electron scattering provide a second constraint. Even assuming a favorable scattering geometry and no attenuation, many of our radio galaxies require ionized-gas reservoirs exceeding those typically inferred for other high-redshift radio-galaxy samples. Previous studies of extended UV emission in high-redshift radio galaxies have found similarly large mass requirements and have generally favored dust over electron scattering (e.g., [30, 31, 41]). Our results do not rule out a contribution from electron scattering in individual regions, but make an unlikely explanation for the UV emission across the sample.

Dust scattering is energetically much more feasible. Spectropolarimetric studies have demonstrated that dust-scattered quasar light can dominate the rest-frame UV continuum of individual high-redshift radio galaxies and that the scattering medium can be highly clumped (e.g., [124, 113]). In our study, the required dust mass within the UV apertures is a small fraction of the global dust reservoir for every galaxy with an SED-based dust-mass estimate. The principal requirement is therefore not an unusually large dust mass, but the presence of illuminated dust-bearing structures at projected distances reaching tens of kpc from their host galaxy centers. The high incidence of mergers in the sample provides a natural means of distributing dusty gas over these scales. Observations of nearby interacting systems directly show that tidal stripping can displace dust into extended tidal structures [127]. Nebular line and continuum emission may contribute to the measured F606W and F140W fluxes. In the HST snapshot sample examined by Hilbert et al. [67], published line-flux measurements were available for only two of 12 radio galaxies: Mg II in 3C 356 and [O III] in 3C 297, which contributed 1.6% and 0.03%, respectively, of the integrated broadband flux. These results show that those particular lines do not dominate the corresponding images, but they do not constrain other lines or the nebular continuum across the sample. Diffuse F140W line emission could also contribute to the fitted host component (§ 3.1).

Nebular Line and Continuum Emission

The nebular continuum can be constrained from spatially matched spectroscopy because HβH\beta and the nebular continuum are produced by the same recombining gas. After correction for reddening, the measured HβH\beta flux can normalize the expected free-bound, free-free, two-photon, and high-order Balmer emission across the HST bandpass. Using this method, Tadhunter et al. [116] inferred contributions of approximately 3–40% immediately below the Balmer break in nuclear spectra of 2-Jy radio galaxies. In 3C 368, which is included in our sample, spectroscopic analyses estimated that nebular continuum contributes approximately 60–80% of the UV continuum in its brightest aligned components [44, 115].

The ionization required to produce the nebular continuum and recombination lines can be supplied by young stars, radiation from the hidden quasar, or jet-driven shocks. Young stars and the quasar photoionize the gas, whereas shocks driven by the jet, bow shock, or expanding cocoon can heat and collisionally ionize it [16]. As the gas recombines, it produces nebular continuum and emission lines regardless of the original ionization source, with different implications for the alignment effect. Emission powered by young stars would still trace recent star formation, but would bias ages inferred by treating the broadband flux as pure stellar continuum. Quasar radiation escaping through an ionization cone could instead produce aligned emission without a mechanical interaction with the jet. Shock-powered emission would remain causally associated with the radio source, but the F606W flux would not directly measure recent star formation. These mechanisms may also coexist within the same galaxy.

Because the relevant lines vary with redshift and spatially matched recombination-line measurements are not presently available, we can not apply a reliable nebular correction. This uncertainty does not affect the principal geometric result that the F606W emission is aligned with the radio source, but it does affect its physical interpretation and the stellar ages inferred from its broadband color. We therefore treat the SPS and jet-timing results as conditional on the F606W emission being predominantly stellar. The observations required to quantify the nebular contribution and distinguish its ionization source are discussed in the following section.

Distinguishing the UV Emission Mechanisms

The stellar-population analysis presented in this work is limited by a single rest-frame UV–optical color. Although F606W–F140W spans the wavelength interval containing the Balmer and 4000 A˚4000\,\AA breaks, neither filter measures the continuum immediately around these features. Consequently, an older stellar population and a younger, dust-reddened population can produce the same color. Nor can this color establish that the F606W emission is stellar. Additional HST or JWST bands redward of F140W would constrain the less dust-sensitive rest-frame optical-to-near-infrared spectral shape, restricting the allowed combinations of stellar age and extinction and permitting more flexible star-formation histories than the constant-SFR models used here. Rest-frame UV spectroscopy could reveal photospheric absorption and stellar-wind features from massive young stars. Spectroscopy covering the Balmer and 4000 A˚4000\,\AA breaks and the Balmer absorption lines would help constrain the ages of the stellar populations.

Rest-frame UV imaging polarimetry provides the most direct spatial test for scattered quasar light. Scattering of light from a hidden quasar predicts polarization vectors oriented perpendicular to the projected direction from the nucleus, producing a centrosymmetric pattern. For an extended scattering region aligned with the radio source, the integrated electric-vector position angle should therefore lie broadly perpendicular to the radio axis [43]. Spectropolarimetry provides a stronger spectral test: broad permitted lines detected in polarized flux directly reveal the hidden quasar spectrum.

Rest-frame UV spectropolarimetry has shown that dust-scattered nuclear light dominates the UV continuum in some powerful radio galaxies [118, 124, 113].

HST/ACS narrow-band imaging of Hα\alpha+[N II] and [O III] in 19 3CR radio galaxies at z<0.3z < 0.3 revealed kiloparsec-scale line-emitting structures preferentially aligned with the radio axes [119, 8]. These maps locate ionized gas and variations in its excitation, but do not uniquely identify star-forming regions because AGN photoionization and jet-driven shocks can also produce strong line emission.

JWST/NIRSpec integral-field maps of six radio galaxies at z>3.5z > 3.5 show that regions of enhanced [O III] line width tend to coincide with radio structures [102]. Similar observations of our strongest candidates could test whether the UV emission coincides with disturbed ionized gas. Hβ\beta, [O III] λ\lambda5007, Hα\alpha, and [N II] are accessible over the sample redshift range, permitting spatially and spectrally resolved excitation and kinematic maps. The Balmer decrement would provide a model-dependent constraint on reddening, while the corrected Hβ\beta luminosity would constrain the nebular contribution to F606W. Line ratios, widths, velocity fields, and radio morphology would help distinguish H II-region-like emission from AGN photoionization and jet-driven shocks. H II-region-like excitation co-spatial with disturbed gas at a radio interaction site would substantially strengthen the case for jet-induced star formation, particularly if rest-frame UV spectroscopy detected young-star features and polarimetry limited scattered quasar light.

Establishing jet-triggered star formation requires both evidence for young stars and a physical connection between those stars and the radio source. The observations of 4C 41.17 provide a well-studied example of this combination. Keck/LRIS spectropolarimetry found low polarization and stellar-wind features from massive young stars. Spectroscopy covering the Balmer and 4000 A˚4000\,\AA breaks and the Balmer absorption lines would help constrain the ages of the stellar populations. larization, photospheric absorption and stellar-wind features from massive young stars, together with strongly disturbed emission-line profiles [42]. HST imaging registered against VLA maps placed bright UV knots and Lyα emission near a bright radio knot [121]. Building on these observations, Bicknell et al. [16] modeled the interaction as a jet-driven bow shock and found that the compressed gas could become gravitationally unstable within 2–6 Myr, well within the estimated age of the radio source. These stellar, spatial, kinematic, and timescale constraints jointly support a jet-triggered origin.

In the nearby radio galaxy Coma A (3C 277.3), the southern jet bends beside a bright emission-line cloud. Chandra observations support a jet–cloud interaction at this site [130]. VLT/MUSE line ratios identify star-forming knots along filaments at the edges of the radio lobes, where the surrounding line emission is consistent with a contribution from shocks [25]. At candidate jet–gas interaction sites in our sample, spectroscopy and polarimetry could help distinguish stellar emission, nebular emission, and scattered quasar light.

High-resolution ALMA imaging could test whether molecular gas is concentrated or kinematically disturbed at candidate jet–ISM interaction sites. An intermediate-JJ transition such as CO(5–4) would trace more highly excited gas, but AGN heating and jet-driven shocks can also enhance its emission [93, 36]. Pairing it with a lower-JJ transition would constrain the excitation correction and better trace the bulk molecular-gas reservoir, reducing the uncertainty in MmolM_{\mathrm{mol}} [26, 35]. The inferred mass would nevertheless remain sensitive to the choice of CO-to-H2_2 conversion factor [19]. With an independent star-formation tracer, the depletion time, tdep=Mmol/SFRt_{\mathrm{dep}} = M_{\mathrm{mol}}/\mathrm{SFR}, could then be compared on and off the radio structure. A shorter depletion time co-spatial with disturbed gas would be consistent with locally enhanced star-formation efficiency.

JWST/MIRI would provide a complementary test for dust-obscured star formation missed by F606W. Galaxy-integrated Spitzer and Herschel SED modeling inferred luminous star-forming components in a substantial subset of z>1z > 1 3CR hosts but could not locate that activity relative to the radio source [99, 98]. For the 12 Herschel-selected radio galaxies studied by Podigachoski et al. [98], the rest-frame mid-infrared continuum was attributed largely to the AGN torus, and PAH features were not prominent in the four objects with Spitzer/IRS spectra.

Multi-band MIRI imaging could search for spatially resolved warm-dust continuum and excess emission in filters containing redshifted PAH features, identifying candidate regions of obscured star formation. PAH luminosities could help constrain SFRs even in galaxies hosting an AGN [108]. MIRI/MRS could also map rotational H2_2 and ionic emission lines accessible at each source’s redshift. PAH-to-H2_2 ratios, ionic-line ratios, and gas kinematics could help assess the contributions of star formation, shocks, and AGN photoionization to the emission line in the z>1z > 1 3CR radio galaxies studied here (e.g., [96]). Registration with the NIRSpec, molecular-gas, and radio maps would then show whether regions consistent with obscured star formation coincide with candidate jet impact sites.

Summary & Concluding Remarks

In this study, we have combined HST/WFC3 imaging in F140W and F606W with high-spatial-resolution radio imaging for the statistically complete sample of type 2 3CR radio galaxies at z>1z > 1. We subsequently used these data to examine the relationships among radio-source properties, host-galaxy structure in the rest-frame optical, rest-frame UV emission, and nearby candidate mergers. Our principal conclusions are as follows.

  1. The radio axes are preferentially aligned with the rest-frame UV emission, the major axes of the smooth stellar hosts, and the projected axes of close candidate mergers. The close-companion result is based on a small sample and requires spectroscopic confirmation, but together these alignments indicate a common projected axis connecting the radio source to the structure and assembly of the host galaxy.

  1. Galaxies with closer host–radio alignment contain stronger total UV emission, both in absolute terms and relative to the rest-frame optical host. The common geometry is therefore related not only to the orientation but also to the prominence of the aligned UV component.

  1. The UV regions become redder and fainter in surface brightness with increasing distance from the nucleus. However, the surface brightness declines much more slowly than predicted by a simple centrally illuminated scattering model. The large ionized-gas masses required for electron scattering further argue against it as the dominant population-wide mechanism, while the much smaller dust requirements leave dust-scattered quasar light viable in individual regions.

  1. The stellar-population and jet-propagation timescales identify a restricted subset of UV regions as the strongest candidates for recently initiated, jet-triggered star formation. Their timing is consistent with jet triggering but does not demonstrate causation, and the interpretation remains conditional on the UV emission being predominantly stellar. Merger-triggered star formation, dust scattering, and line and nebular emission remain plausible contributions to the rest-frame UV emission observed in our sample of radio galaxies.

These results favor a composite origin for the alignment effect. The common projected geometry may trace a long-lived axis of galaxy assembly. Repeated mergers or anisotropic infall along a cosmic-web filament could establish the host major axis, potentially imprinting this direction on a triaxial or prolate stellar remnant, while delivering subsequent companions and gas along the same axis. Because the observed companions have generally not yet coalesced, their positions may trace the continuation of this assembly history rather than the origin of the host’s existing structure. If the nuclear accretion flow or black-hole spin also became aligned with this assembly axis, the radio jet could emerge along the same direction, although the present data do not demonstrate such coupling. The jet could then shock or compress gas distributed along this axis, mergers could trigger star formation independently of the jet, and radiation from the hidden quasar could produce scattered and nebular emission. Galaxy assembly may therefore establish the common geometry within which several mechanisms contribute to the alignment effect.

The highest-priority follow-up is JWST/NIRSpec integral-field spectroscopy of the strongest candidates and close companions. Spatially resolved Balmer-decrement, excitation, and kinematic maps could constrain the nebular contribution to F606W and help distinguish H II-region-like emission from AGN photoionization and jet-driven shocks. Mapping the emission lines that fall within F140W would allow us to subtract their contribution from the HST images and refit the host position angles, testing whether diffuse line emission contributed to the host–radio alignment found in this study. Deep NIRSpec observations of stellar absorption lines could test for short- or long-axis rotation. A projected stellar rotation axis aligned with both the host major axis and the radio axis would be consistent with a scenario in which filamentary accretion and galaxy assembly help set a common axis for stellar rotation and jet launching. JWST/MIRI imaging and spectroscopy could search for dust-obscured star formation at candidate jet impact sites and assess contributions from shocks and AGN photoionization.

Spatially resolved optical spectroscopy could search for redshifted stellar photospheric absorption and wind features in the candidate UV regions using HST/STIS or VLT/MUSE adaptive optics. HST/ACS F606W imaging polarimetry could test for the polarization pattern expected from scattered quasar light, and ALMA CO line maps could test for disturbed molecular gas along the radio structures. Together, these observations would test whether recent star formation and disturbed molecular gas coincide with specific radio features, establish which candidate companions are physically associated, and clarify how the alignment effect connects powerful radio activity with host-galaxy assembly.

We thank Barbara Balmaverde, Peter D. Barthel, Alessandro Capetti, F. Duccio Macchetto, William B. Sparks, Carlo Stanghellini, and Grant R. Tremblay for their contributions as co-investigators on HST program GO-16281. This research is based on observations made with the NASA/ESA Hubble Space Telescope obtained from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5–26555. These observations are associated with programs GO-16281 and SNAP-13023. Some of the data presented in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute. The specific observations analyzed can be accessed via DOI. This work was supported by the HST-GO grant No. 16281.018-A.

PSF homogenization

Our WFC3/UVIS F606W images have a natively sharper PSF than our WFC3/IR F140W images. The WFC3 Instrument Handbook states that the core FWHMs are approximately 0.067 and 0.141 arcsec for F606W and F140W, respectively [79]. Therefore, before performing aperture photometry, we convolved the sharper F606W images to the broader F140W PSF.

To construct representative PSFs, we followed the method described by Breiding et al. [22] for one radio galaxy from each HST observing program used in this study: 3C 222 for the GO-16281 program and 3C 322 for the SNAP-13023 program. We used TinyTim [74] to construct a synthetic PSF for every calibrated flt exposure in WFC3/IR F140W and every calibrated flc exposure in WFC3/UVIS F606W. For each representative galaxy, we selected a common sky position within the UVIS/F606W emission region used for aperture photometry and transformed that position to the appropriate detector coordinates in every exposure. We assumed a flat spectrum in flux density, fν∝ν0f_\nu\propto\nu^0, and used the telescope focus predicted at the time of each exposure by the STScI HST focus model.6 Because the native WFC3/IR pixels do not fully sample the F140W PSF core, we mapped each F140W model onto the distorted detector grid using the TinyTim tiny3 stage with SUB=5. This produced a PSF sampled five times more finely than the native detector pixels. We placed each model at the precise detector position of the selected sky location and then integrated it onto the native detector pixels. After this integration, we applied the standard WFC3/IR detector-response kernel that TinyTim normally includes in its native-sampled PSFs. The F606W PSFs were generated using the normal TinyTim sampling. The exposure-level PSFs were then combined with AstroDrizzle, using the same drizzle parameters and output pixel scales as the corresponding science images.

We used the create_matching_kernel routine in photutils.psf.matching to derive a normalized kernel that transforms the representative F606W PSF into the F140W PSF. The F140W science images, host-galaxy models, and PSFs were mapped from their 0.060 arcsec pixel−1^{-1} grids onto the F606W grids (0.030 arcsec pixel−1^{-1} for GO-16281 targets and 0.04 arcsec pixel−1^{-1} for SNAP-13023 targets) using bilinear reproject_interp from reproject [101]. Pixel areas were calculated with proj_plane_pixel_area in Astropy [6] to preserve integrated flux. The reprojected F140W PSF was used as the target, and the resulting kernel was applied to F606W using scipy.signal.fftconvolve [125].

Because the observing setup was common within each program, we used the kernel derived from 3C 222 for the remaining GO-16281 radio galaxies and the kernel derived from 3C 322 for the remaining SNAP-13023 radio galaxies. The validity of this program-level approach is tested in Section A.1.

Validation of the PSF homogenization

We tested whether the matched F606W and F140W PSFs place the same fraction of their total light inside a given aperture. After normalizing each PSF to a total flux of one, we measured the enclosed-flux fraction in each filter as a function of aperture radius. We then calculated R(r)=F606(<r)/F140(<r)R(r) = F_{606}(<r)/F_{140}(<r), which is shown in the left panels of Figure 10. Before matching, R(r)R(r) quantifies the difference between the original PSFs; after matching, it measures how well the convolved F606W PSF reproduces the F140W PSF. Perfect matching corresponds to R=1R = 1.

Figure 10. PSF-matching validation for the representative GO field, 3C 222 (top), and SNAP field, 3C 322 (bottom). The left panels show the ratio R(r)R(r) of the normalized F606W and F140W enclosed-flux fractions. The black and blue curves show the ratios before matching for the TinyTim PSFs and empirical STScI encircled energies, respectively. The green curves show the ratios after matching F606W to F140W. The right panels show the equivalent PSF-induced F606W–F140W color offset for each representative field (green) and after transferring the unchanged program kernel to the independent 3C 238 and 3C 257 PSF pairs (purple). The dash-dotted gray line and green circle in each row mark the representative-field science aperture; the dotted purple line and diamond mark the independent-field science aperture. The curves are displayed over 0.15–1.5 arcsec, while the smallest science-aperture equivalent radius is 0.17 arcsec. Perfect matching corresponds to R=1R = 1 in the left panels and ΔPSF=0\Delta_{\mathrm{PSF}} = 0 mag in the right panels.

Because the F606W PSF is sharper, it places a larger fraction of its light inside a given aperture than the F140W PSF. Without PSF matching, the measured F606W flux would therefore be biased high relative to F140W, producing systematically bluer colors and potentially biasing the stellar-population modeling toward younger ages. To express the remaining post-matching error in the same units as our measured colors, we converted the flux ratio to the equivalent PSF-induced color offset, ΔPSF(F606W−F140W)=−2.5log⁡10R\Delta_{\mathrm{PSF}}(F606W - F140W) = -2.5\log_{10} R. This quantity is shown in the right panels of Figure 10; perfect matching corresponds to ΔPSF=0\Delta_{\mathrm{PSF}} = 0 mag.

Figure 10 presents the GO result for 3C 222 in the top row and the SNAP result for 3C 322 in the bottom row. In the left panels, the black curves show the ratios calculated from the original TinyTim PSFs, while the blue curves show independent ratios derived from the empirical STScI encircled-energy measurements [79]. At the equivalent circular radii of the representative-field science apertures used for aperture photometry (as described in § 3.3), the corresponding synthetic-PSF and STScI enclosed-energy color offsets differ by approximately 0.015 mag for both programs. This agreement supports our PSF construction and shows that the synthetic PSFs reproduce the relative F606W and F140W aperture losses. The green curves show R(r)R(r) after matching F606W to F140W, and the right panels show these same post-matching ratios converted to ΔPSF(F606W−F140W)\Delta_{\mathrm{PSF}}(F606W-F140W). The green circles in the right panels mark the residuals at the equivalent circular radii of the representative-field science apertures.

To test whether a single kernel could be transferred within each program, we independently constructed a second PSF pair for 3C 238 (GO) and 3C 257 (SNAP), using the same procedure, and applied the corresponding 3C 222 and 3C 322 kernels without refitting them. These independent-field results are shown by the purple curves in the right panels, with purple diamonds marking the equivalent circular radii of their science apertures. At the marked radii, the absolute post-matching color residuals are 0.0025 mag for 3C 222, 0.0008 mag for 3C 238, 0.0003 mag for 3C 322, and 0.0020 mag for 3C 257. These values are point-source diagnostics evaluated from the circular-aperture curves at the equivalent circular radii of the science apertures; they are not corrections applied to the measured photometry. Taken together, the tests show that the residuals at the marked radii are only a few millimagnitudes and should not materially affect the stellar-population synthesis results presented in this study. The figure is displayed over 0.15–1.5 arcsec. The smallest science-aperture equivalent radius is 0.17 arcsec, so structure below this radius is diagnostic and is not used as a photometric correction.

Scattering Formalism and Energetics Calculations

Scattering Intensity and Geometry

The expected radial dependence for scattering follows from geometric dilution of the quasar radiation field. At the emitted-frame frequency νe=(1+z)νo\nu_e=(1+z)\nu_o, where νo\nu_o is the observed-frame frequency, a quasar with intrinsic spectral luminosity Lνe,QSOL_{\nu_e,\mathrm{QSO}} produces an incident spectral flux at a cloud a physical distance rr from the nucleus of

Fνe=Lνe,QSO4πr2.(25)F_{\nu_e}=\frac{L_{\nu_e,\mathrm{QSO}}}{4\pi r^2}. \tag*{(25)}

For an optically thin, single-scattering cloud with scatterer number density nsn_s, a path length dℓ\mathrm{d}\ell contributes an emitted specific intensity dIνe=Fνens(dσνe/dΩ)dℓ\mathrm{d}I_{\nu_e}=F_{\nu_e}n_s(\mathrm{d}\sigma_{\nu_e}/\mathrm{d}\Omega)\mathrm{d}\ell toward the observer. Integrating through the cloud gives Iνe=FνeNs(dσνe/dΩ)I_{\nu_e}=F_{\nu_e}N_s(\mathrm{d}\sigma_{\nu_e}/\mathrm{d}\Omega), where Ns=∫nsdℓN_s=\int n_s\mathrm{d}\ell is the line-of-sight scatterer column. Specific intensity per unit frequency is cosmologically dimmed according to Iνo=Iνe/(1+z)3I_{\nu_o}=I_{\nu_e}/(1+z)^3. Using the geometric relation Rproj=rsin⁡θR_{\mathrm{proj}}=r\sin\theta shown in Figure 1 then gives

Iνo=Lνe,QSO4π(1+z)3[Nssin⁡2θdσνedΩTνe]Rproj−2.(26)I_{\nu_o}=\frac{L_{\nu_e,\mathrm{QSO}}}{4\pi(1+z)^3}\left[N_s\sin^2\theta\frac{\mathrm{d}\sigma_{\nu_e}}{\mathrm{d}\Omega}T_{\nu_e}\right]R_{\mathrm{proj}}^{-2}. \tag*{(26)}

Here IνoI_{\nu_o} is the observed specific intensity per unit solid angle. Extinction along the incident quasar–cloud path and the outgoing cloud–observer path is represented by the transmission

Tνe=e−τνe=10−0.4Aνe,(27)T_{\nu_e}=e^{-\tau_{\nu_e}}=10^{-0.4A_{\nu_e}}, \tag*{(27)}

where τνe\tau_{\nu_e} is the total extinction optical depth and Aνe=1.086τνeA_{\nu_e}=1.086\tau_{\nu_e} is the corresponding extinction in magnitudes. The total extinction is

Aνe=Aνe,QSO→cloud+Aνe,cloud→obs.(28)A_{\nu_e}=A_{\nu_e,\mathrm{QSO}\to\mathrm{cloud}}+A_{\nu_e,\mathrm{cloud}\to\mathrm{obs}}. \tag*{(28)}

Region-by-Region Scattering Energetics

We begin with the general relation before specifying the intrinsic quasar spectrum or scattering model. Solving the scattering-intensity relation above for the line-of-sight scatterer column gives

Ns=4π(1+z)3I606Rproj2Lν(λrest,piv)sin⁡2θ[dσ(λrest,piv,θ)/dΩ]T(λrest,piv).(29)N_s=\frac{4\pi(1+z)^3I_{606}R_{\mathrm{proj}}^2}{L_\nu(\lambda_{\mathrm{rest,piv}})\sin^2\theta\left[\mathrm{d}\sigma(\lambda_{\mathrm{rest,piv}},\theta)/\mathrm{d}\Omega\right]T(\lambda_{\mathrm{rest,piv}})}. \tag*{(29)}

Here I606I_{606} is the observed F606W flux-density intensity and Lν(λrest,piv)L_\nu(\lambda_{\mathrm{rest,piv}}) is the intrinsic quasar specific luminosity at the emitted-frame wavelength sampled by F606W. The remaining terms are the projected distance RprojR_{\mathrm{proj}}, scattering angle θ\theta, differential scattering cross section dσ/dΩ\mathrm{d}\sigma/\mathrm{d}\Omega, and combined transmission TT along the quasar–scatterer and scatterer–observer paths.

We calculate I606I_{606} from the native, pre-PSF-matching F606W aperture flux density divided by the native F606W segmentation-aperture area. Equivalently, the intensity is related to the AB surface brightness μ606\mu_{606} by

I606=10−0.4(μ606+48.60)Ωas,(30)I_{606}=\frac{10^{-0.4(\mu_{606}+48.60)}}{\Omega_{\mathrm{as}}}, \tag*{(30)}

where μ606\mu_{606} is expressed in mag arcsec−2^{-2} and Ωas=2.35044×10−11 sr\Omega_{\mathrm{as}}=2.35044\times10^{-11}\ \mathrm{sr} is the solid angle of one square arcsecond.

We normalize the intrinsic quasar spectrum at 2500 Å because our empirical bolometric correction relates the bolometric luminosity directly to νLν\nu L_\nu at this wavelength. The corresponding frequency is ν2500=c/(2500 A˚)=1.20×1015 Hz\nu_{2500}=c/(2500\ \text{\AA})=1.20\times10^{15}\ \mathrm{Hz}. We define the dimensionless bolometric correction as

k2500≡Lbolν2500Lν,2500,(31)k_{2500} \equiv\frac{L_{\mathrm{bol}}}{\nu_{2500}L_{\nu,2500}}, \tag*{(31)}

where LbolL_{\mathrm{bol}} is the total radiative luminosity and Lν,2500L_{\nu,2500} is the intrinsic specific luminosity per unit frequency at 2500 Å. We adopt k2500=2.75k_{2500} = 2.75, the mean 2500 Å correction recovered by Krawczyk et al. [73] through mid-IR-to-UV SED fitting of approximately 120,000 type 1 quasars at 0.06<z<5.50.06 < z < 5.5. This value is also broadly consistent with the UV bolometric corrections predicted for powerful 1<z<21 < z < 2 radio-loud quasars from SED modeling of the 3CRR sample (Azadi et al. [7], the appropriate counterparts to the type 2 radio galaxies studied here).

We adopt log⁡(Lbol/erg s−1)=47.28\log(L_{\mathrm{bol}}/\mathrm{erg\ s}^{-1}) = 47.28, the median of the 29 z>1z > 1 3CR quasars studied by Breiding et al. [22]. Assigning this representative type 1 quasar luminosity to the type 2 radio galaxies follows the orientation-based unification picture (i.e., Lawrence & Elvis [75]; Antonucci & Miller [4]; Barthel [9]; Antonucci [3]).

For simplicity, we represent each F606W measurement monochromatically at the filter pivot wavelength, λpiv=\lambda_{\mathrm{piv}} = 5887 Å.7 For a source at redshift zz, the corresponding emitted-frame wavelength is

Geometry used in the scattering calculation

Figure 11. Geometry used in the scattering calculation. The scattering cloud lies at a physical distance rr from the quasar and at a projected sky-plane distance RprojR_{\mathrm{proj}}. The scattering angle θ\theta is measured between the incident photon direction and the outgoing direction toward the observer, such that Rproj=rsin⁡θR_{\mathrm{proj}} = r \sin\theta. The diagram is schematic and not to scale.

λrest,piv=λpiv1+z.(32)\lambda_{\mathrm{rest,piv}} = \frac{\lambda_{\mathrm{piv}}}{1+z}. \tag*{(32)}

Because F606W samples different rest-frame wavelengths across the sample, λrest,piv\lambda_{\mathrm{rest,piv}} ranges from approximately 1695 to 2831 Å. We approximate the intrinsic quasar continuum as Lν∝ναL_\nu\propto\nu^\alpha, with α=−0.44\alpha= -0.44 (Vanden Berk et al. 2001). The intrinsic specific luminosity at the emitted-frame wavelength sampled by F606W is therefore

Lν(λrest,piv)=Lbolk2500ν2500(2500 A˚λrest,piv)α.(33)L_\nu(\lambda_{\mathrm{rest,piv}}) = \frac{L_{\mathrm{bol}}}{k_{2500}\nu_{2500}}\left(\frac{2500\ \text{Å}}{\lambda_{\mathrm{rest,piv}}}\right)^\alpha. \tag*{(33)}

If a fraction fUV,scatf_{\mathrm{UV,scat}} of the measured F606W emission is attributed to scattered quasar light, the intensity produced by scattering is I606,scat=fUV,scatI606I_{606,\mathrm{scat}} = f_{\mathrm{UV,scat}} I_{606}. Substituting this relation together with Equations (30) and (33) into the general scatterer-column relation gives

Ns=4π(1+z)3k2500ν2500Rproj2ΩasLbolsin⁡2θ [dσ(λrest,piv,θ)/dΩ] T(λrest,piv)×fUV,scat10−0.4(μ606+48.60)(λrest,piv2500 A˚)α.(34)\begin{aligned} N_{\mathrm{s}} ={}& \frac{4\pi(1+z)^3 k_{2500}\nu_{2500}R_{\mathrm{proj}}^2}{\Omega_{\mathrm{as}}L_{\mathrm{bol}}\sin^2\theta\,[d\sigma(\lambda_{\mathrm{rest,piv}},\theta)/d\Omega]\,T(\lambda_{\mathrm{rest,piv}})} \\ &\times f_{\mathrm{UV,scat}}10^{-0.4(\mu_{606}+48.60)} \left(\frac{\lambda_{\mathrm{rest,piv}}}{2500\ \text{Å}}\right)^\alpha. \tag*{(34)} \end{aligned}

This is the general relation presented in Equation (12) in the main text. For the fiducial calculation, we assign all of the measured F606W emission to scattered quasar light, so that fUV,scat=1f_{\mathrm{UV,scat}} = 1, and neglect attenuation along both light paths, so that T=1T = 1. Adopting k2500=2.75k_{2500} = 2.75, log⁡(Lbol/erg s−1)=47.28\log(L_{\mathrm{bol}}/\mathrm{erg\ s}^{-1}) = 47.28, λpiv=5887\lambda_{\mathrm{piv}} = 5887 Å, and α=−0.44\alpha= -0.44 then reduces Equation (34) to the fiducial relation presented in Equation 13 in the main text and used in the region-by-region calculations.

The measured inputs to this calculation are μ606\mu_{606}, RprojR_{\mathrm{proj}}, and zz. The remaining quantities specify the scattering model and geometry. We evaluate the fiducial relation separately for electron and dust scattering and then convert the inferred columns into ionized-gas and dust masses within the UV apertures. For electron scattering, Ns,fid=NeN_{\mathrm{s,fid}} = N_{\mathrm{e}} and the Thomson differential cross section is evaluated at the scattering angle. For dust scattering, Ns,fid=NHN_{\mathrm{s,fid}} = N_{\mathrm{H}} and the wavelength-dependent differential cross section per H nucleus is evaluated at λrest,piv\lambda_{\mathrm{rest,piv}} and the scattering angle.

For a given UV region and a fixed scattering model, geometry, and quasar continuum slope, the required scatterer column scales with the assumed scattered fraction, attenuation, bolometric correction, and hidden-quasar luminosity as

Ns=Ns,fidfUV,scat100.4AUV(k25002.75)(1047.28 erg s−1Lbol).(35)N_{\mathrm{s}} = N_{\mathrm{s,fid}} f_{\mathrm{UV,scat}} 10^{0.4A_{\mathrm{UV}}} \left(\frac{k_{2500}}{2.75}\right) \left(\frac{10^{47.28}\ \mathrm{erg\ s}^{-1}}{L_{\mathrm{bol}}}\right). \tag*{(35)}

where AUVA_{\mathrm{UV}} is the combined effective attenuation along the quasar–scatterer and scatterer–observer paths. The inferred ionized-gas and dust masses scale in the same way because, at fixed aperture area and scattering model, each mass is directly proportional to the scatterer column. A stellar or nebular contribution therefore lowers the required column and mass, whereas attenuation or a less luminous hidden quasar raises them.

References

  1. [1]Aird, J., Coil, A. L., Georgakakis, A., et al. 2015, MNRAS, 451, 1892, doi: 10.1093/mnras/stv1062DOI
  2. [2]Alberts, S., Pope, A., Brodwin, M., et al. 2016, ApJ, 825, 72, doi: 10.3847/0004-637X/825/1/72DOI
  3. [3]Antonucci, R. 1993, ARA&A, 31, 473, doi: 10.1146/annurev.aa.31.090193.002353DOI
  4. [4]Antonucci, R. R. J., & Miller, J. S. 1985, ApJ, 297, 621, doi: 10.1086/163559DOI
  5. [5]Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481, doi: 10.1146/annurev.astro.46.060407.145222DOI
  6. [6]Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167, doi: 10.3847/1538-4357/ac7c74DOI
  7. [7]Azadi, M., Wilkes, B., Kuraszkiewicz, J., Ashby, M., & Willner, S. P. 2022, arXiv e-prints, arXiv:2204.12697. https://arxiv.org/abs/2204.12697
  8. [8]Baldi, R. D., Rodríguez Zaurín, J., Chiaberge, M., et al. 2019, ApJ, 870, 53, doi: 10.3847/1538-4357/aaf002DOI
  9. [9]Barthel, P. D. 1989, ApJ, 336, 606, doi: 10.1086/167038DOI
  10. [10]Battye, R. A., & Browne, I. W. A. 2009, MNRAS, 399, 1888, doi: 10.1111/j.1365-2966.2009.15429.xDOI
  11. [11]Belfiore, F., Westfall, K. B., Schaefer, A., et al. 2019, AJ, 158, 160, doi: 10.3847/1538-3881/ab3e4eDOI
  12. [12]Best, P. N., Longair, M. S., & Roettgering, H. J. A. 1998, MNRAS, 295, 549, doi: 10.1046/j.1365-8711.1998.01245.xDOI
  13. [13]Best, P. N., Longair, M. S., & Röttgering, H. J. A. 1996, MNRAS, 280, L9, doi: 10.1093/mnras/280.1.L9DOI
  14. [14]Best, P. N., Röttgering, H. J. A., & Longair, M. S. 2000, MNRAS, 311, 23, doi: 10.1046/j.1365-8711.2000.03028.xDOI
  15. [15]Best, P. N., Kondapally, R., Williams, W. L., et al. 2023, MNRAS, 523, 1729, doi: 10.1093/mnras/stad1308DOI
  16. [16]Bicknell, G. V., Sutherland, R. S., van Breugel, W. J. M., et al. 2000, ApJ, 540, 678, doi: 10.1086/309343DOI
  17. [17]Blandford, R. D., & Payne, D. G. 1982, MNRAS, 199, 883, doi: 10.1093/mnras/199.4.883DOI
  18. [18]Blandford, R. D., & Znajek, R. L. 1977, MNRAS, 179, 433, doi: 10.1093/mnras/179.3.433DOI
  19. [19]Bolatto, A. D., Wolfire, M., & Leroy, A. K. 2013, ARA&A, 51, 207, doi: 10.1146/annurev-astro-082812-140944DOI
  20. [20]Bournaud, F., Elmegreen, B. G., & Elmegreen, D. M. 2007, ApJ, 670, 237, doi: 10.1086/522077DOI
  21. [21]Bradley, L., Sipócz, B., Robitaille, T., et al. 2022, astropy/photutils: 1.5.0, 1.5.0, Zenodo, doi: 10.5281/zenodo.6825092
  22. [22]Breiding, P., Chiaberge, M., Lambirdes, E., et al. 2024, ApJ, 963, 91, doi: 10.3847/1538-4357/ad19dbDOI
  23. [23]Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000, doi: 10.1046/j.1365-8711.2003.06897.xDOI
  24. [24]Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682, doi: 10.1086/308692DOI
  25. [25]Capetti, A., Balmaverde, B., Tadhunter, C., et al. 2022, A&A, 657, A114, doi: 10.1051/0004-6361/202141965DOI
  26. [26]Carilli, C. L., & Walter, F. 2013, ARA&A, 51, 105, doi: 10.1146/annurev-astro-082812-140953DOI
  27. [27]Carnall, A. C., McLure, R. J., Dunlop, J. S., & Davé, R. 2018, MNRAS, 480, 4379, doi: 10.1093/mnras/sty2169DOI
  28. [28]Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392DOI
  29. [29]Chambers, K. C., Miley, G. K., & van Breugel, W. 1987, Nature, 329, 604, doi: 10.1038/329604a0DOI
  30. [30]Cimatti, A., de Serego Alighieri, S., Field, G. B., & Fosbury, R. A. E. 1994, ApJ, 422, 562, doi: 10.1086/173749DOI
  31. [31]Cimatti, A., Dey, A., van Breugel, W., Antonucci, R., & Spinrad, H. 1996, ApJ, 465, 145, doi: 10.1086/177408DOI
  32. [32]Conroy, C., & Gunn, J. E. 2010, ApJ, 712, 833, doi: 10.1088/0004-637X/712/2/833DOI
  33. [33]Conroy, C., Gunn, J. E., & White, M. 2009, ApJ, 699, 486, doi: 10.1088/0004-637X/699/1/486DOI
  34. [34]Croft, S., van Breugel, W., de Vries, W., et al. 2006, ApJ, 647, 1040, doi: 10.1086/505526DOI
  35. [35]Daddi, E., Dannerbauer, H., Liu, D., et al. 2015, A&A, 577, A46, doi: 10.1051/0004-6361/201425043DOI
  36. [36]Dasyra, K. M., Combes, F., Oosterloo, T., et al. 2016, A&A, 595, L7, doi: 10.1051/0004-6361/201629689DOI
  37. [37]Davies, R. L., & Birkinshaw, M. 1988, ApJS, 68, 409, doi: 10.1086/191294DOI
  38. [38]de Koff, S., Best, P., Baum, S. A., et al. 2000, ApJS, 129, 33, doi: 10.1086/313402DOI
  39. [39]de Mellos, M. S. Z., Riffel, R. A., Souza-Oliveira, G. L., et al. 2026, ApJ, 997, 40, doi: 10.3847/1538-4357/ae1f19DOI
  40. [40]D’Eugenio, F., Pérez-González, P. G., Maiolino, R., et al. 2024, Nature Astronomy, 8, 1443, doi: 10.1038/s41550-024-02345-1
  41. [41]Dey, A., Cimatti, A., van Breugel, W., Antonucci, R., & Spinrad, H. 1996, ApJ, 465, 157, doi: 10.1086/177409DOI
  42. [42]Dey, A., van Breugel, W., Vacca, W. D., & Antonucci, R. 1997, ApJ, 490, 698, doi: 10.1086/304911DOI
  43. [43]di Serego Alighieri, S., Fosbury, R. A. E., Quinn, P. J., & Tadhunter, C. N. 1989, Nature, 341, 307, doi: 10.1038/341307a0DOI
  44. [44]Dickson, R., Tadhunter, C., Shaw, M., Clark, N., & Morganti, R. 1995, MNRAS, 273, L29, doi: 10.1093/mnras/273.1.L29
  45. [45]Dotti, M., Colpi, M., Pallini, S., Perego, A., & Volonteri, M. 2013, ApJ, 762, 68, doi: 10.1088/0004-637X/762/2/68DOI
  46. [46]Draine, B. T. 2003, ApJ, 598, 1017, doi: 10.1086/379118DOI
  47. [47]Dressel, L. 2022, in WFC3 Instrument Handbook for Cycle 30 v. 14, Vol. 14, 14
  48. [48]Duggal, C., O’Dea, C. P., Baum, S. A., et al. 2024, ApJ, 965, 17, doi: 10.3847/1538-4357/ad2513DOI
  49. [49]Dunlop, J. S., & Peacock, J. A. 1993, MNRAS, 263, 936, doi: 10.1093/mnras/263.4.936DOI
  50. [50]Eales, S. A. 1992, ApJ, 397, 49, doi: 10.1086/171765DOI
  51. [51]Ebrová, I., & Łokas, E. L. 2017, ApJ, 850, 144, doi: 10.3847/1538-4357/aa96ffDOI
  52. [52]Elmegreen, B. G., & Scalo, J. 2004, ARA&A, 42, 211, doi: 10.1146/annurev.astro.41.011802.094859DOI
  53. [53]Fabian, A. C. 2012, ARA&A, 50, 455, doi: 10.1146/annurev-astro-081811-125521DOI
  54. [54]Falkendal, T., De Breuck, C., Lehnert, M. D., et al. 2019, A&A, 621, A27, doi: 10.1051/0004-6361/201732485DOI
  55. [55]Fanaroff, B. L., & Riley, J. M. 1974, MNRAS, 167, 31P, doi: 10.1093/mnras/167.1.31PDOI
  56. [56]Feain, I. J., Papadopoulos, P. P., Ekers, R. D., & Middelberg, E. 2007, ApJ, 662, 872, doi: 10.1086/518027DOI
  57. [57]Ferrarese, L., & Merritt, D. 2000, ApJL, 539, L9, doi: 10.1086/312838DOI
  58. [58]Forrest, B., Muzzin, A., Marchesini, D., et al. 2026, Nature Astronomy, 10, 1232, doi: 10.1038/s41550-026-02855-0DOI
  59. [59]Gaibler, V., Khochfar, S., Krause, M., & Silk, J. 2012, MNRAS, 425, 438, doi: 10.1111/j.1365-2966.2012.21479.xDOI
  60. [60]Gebhardt, K., Bender, R., Bower, G., et al. 2000, ApJL, 539, L13, doi: 10.1086/312840DOI
  61. [61]Gonzaga, S., Hack, W., Fruchter, A., & Mack, J. 2012, The DrizzlePac Handbook
  62. [62]Hardcastle, M. J. 2018, MNRAS, 475, 2768, doi: 10.1093/mnras/stx3358DOI
  63. [63]Harwood, J. J., Hardcastle, M. J., Morganti, R., et al. 2017, MNRAS, 469, 639, doi: 10.1093/mnras/stx820DOI
  64. [64]Heckman, T. M., & Best, P. N. 2014, ARA&A, 52, 589, doi: 10.1146/annurev-astro-081913-035722DOI
  65. [66]Heckman, T. M., Illingworth, G. D., Miley, G. K., & van Breugel, W. J. M. 1985, ApJ, 299, 41, doi: 10.1086/163681DOI
  66. [67]Hilbert, B., Chiaberge, M., Kotyla, J. P., et al. 2016, ApJS, 225, 12, doi: 10.3847/0067-0049/225/1/12DOI
  67. [68]Hopkins, P. F., Hernquist, L., Hayward, C. C., & Narayanan, D. 2012, MNRAS, 425, 1121, doi: 10.1111/j.1365-2966.2012.21449.xDOI
  68. [69]Johnson, B., Foreman-Mackey, D., Sick, J., et al. 2026, dfm/python-fsps: v0.4.8, v0.4.8, Zenodo, doi: 10.5281/zenodo.591505
  69. [70]Jung, S. L., Whittam, I. H., Jarvis, M. J., et al. 2025, MNRAS, 539, 2362, doi: 10.1093/mnras/staf613DOI
  70. [71]Kormendy, J., & Ho, L. C. 2013, ARA&A, 51, 511, doi: 10.1146/annurev-astro-082708-101811DOI
  71. [72]Krajnović, D., Emsellem, E., Cappellari, M., et al. 2011, MNRAS, 414, 2923, doi: 10.1111/j.1365-2966.2011.18560.xDOI
  72. [73]Krawczyk, C. M., Richards, G. T., Mehta, S. S., et al. 2013, ApJS, 206, 4, doi: 10.1088/0067-0049/206/1/4DOI
  73. [74]Krist, J. E., Hook, R. N., & Stoehr, F. 2011, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 8127, Optical Modeling and Performance Predictions V, ed. M. A. Kahan, 81270J, doi: 10.1117/12.892762
  74. [75]Lawrence, A., & Elvis, M. 1982, ApJ, 256, 410, doi: 10.1086/159918DOI
  75. [76]Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415, doi: 10.1146/annurev-astro-081811-125615DOI
  76. [77]Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285, doi: 10.1086/300353DOI
  77. [78]Mandal, A., Mukherjee, D., Federrath, C., et al. 2021, MNRAS, 508, 4738, doi: 10.1093/mnras/stab2822DOI
  78. [79]Marinelli, M., & Green, J. 2025, in WFC3 Instrument Handbook for Cycle 34 v. 18, Vol. 18
  79. [80]McCarthy, P. J., van Breugel, W., Spinrad, H., & Djorgovski, S. 1987, ApJL, 321, L29, doi: 10.1086/185000DOI
  80. [81]Miley, G. 1980, ARA&A, 18, 165, doi: 10.1146/annurev.aa.18.090180.001121DOI
  81. [82]Miley, G., & De Breuck, C. 2008, A&A Rev, 15, 67, doi: 10.1007/s00159-007-0008-zDOI
  82. [83]Molnár, D. C., Sargent, M. T., Elbaz, D., Papadopoulos, P. P., & Silk, J. 2017, MNRAS, 467, 586, doi: 10.1093/mnras/stx089DOI
  83. [84]Mould, J. R., Ridgewell, A., Gallagher, III, J. S., et al. 2000, ApJ, 536, 266, doi: 10.1086/308927DOI
  84. [85]Mukherjee, D., Bicknell, G. V., Wagner, A. Y., Sutherland, R. S., & Silk, J. 2018, MNRAS, 479, 5544, doi: 10.1093/mnras/sty1776DOI
  85. [86]Nayakshin, S., & Zubovas, K. 2012, MNRAS, 427, 372, doi: 10.1111/j.1365-2966.2012.21950.xDOI
  86. [87]Nesvadba, N. P. H., Boulanger, F., Lehnert, M. D., Guillard, P., & Salome, P. 2011, A&A, 536, L5, doi: 10.1051/0004-6361/201118018DOI
  87. [88]Nesvadba, N. P. H., De Breuck, C., Lehnert, M. D., Best, P. N., & Collet, C. 2017, A&A, 599, A123, doi: 10.1051/0004-6361/201528040DOI
  88. [89]Obied, G., Zakamska, N. L., Wylezalek, D., & Liu, G. 2016, MNRAS, 456, 2861, doi: 10.1093/mnras/stv2850DOI
  89. [90]O’Dea, C. P., Daly, R. A., Kharb, P., Freeman, K. A., & Baum, S. A. 2009, A&A, 494, 471, doi: 10.1051/0004-6361:200809416DOI
  90. [91]Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713, doi: 10.1086/160817DOI
  91. [92]Omoruyi, O., Tremblay, G. R., Combes, F., et al. 2024, ApJ, 963, 1, doi: 10.3847/1538-4357/ad1101DOI
  92. [93]Papadopoulos, P. P., Kovacs, A., Evans, A. S., & Barthel, P. 2008, A&A, 491, 483, doi: 10.1051/0004-6361:200810513DOI
  93. [94]Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2010, AJ, 139, 2097, doi: 10.1088/0004-6256/139/6/2097DOI
  94. [95]Pentericci, L., McCarthy, P. J., Röttgering, H. J. A., et al. 2001, ApJS, 135, 63, doi: 10.1086/321781DOI
  95. [96]Pereira-Santaella, M., Álvarez-Márquez, J., García-Bernete, I., et al. 2022, A&A, 665, L11, doi: 10.1051/0004-6361/202244725
  96. [97]Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2016, A&A, 594, A13, doi: 10.1051/0004-6361/201525830
  97. [98]Podigachoski, P., Rocca-Volmerange, B., Barthel, P., Drouart, G., & Fioc, M. 2016, MNRAS, 462, 4183, doi: 10.1093/mnras/stw1946
  98. [99]Podigachoski, P., Barthel, P. D., Haas, M., et al. 2015, A&A, 575, A80, doi: 10.1051/0004-6361/201425137
  99. [100]Rees, M. J. 1989, MNRAS, 239, 1P, doi: 10.1093/mnras/239.1.1PDOI
  100. [101]Robitaille, T., Deil, C., & Ginsburg, A. 2020, reproject: Python-based astronomical image reprojection, Astrophysics Source Code Library, record ascl:2011.023. http://ascl.net/2011.023
  101. [102]Roy, N., Heckman, T., Henry, A., Wang, W., & Kukreti, P. 2026, ApJ, 1002, 57, doi: 10.3847/1538-4357/ae5b68DOI
  102. [103]Scheuer, P. A. G. 1995, MNRAS, 277, 331, doi: 10.1093/mnras/277.1.331DOI
  103. [104]Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103, doi: 10.1088/0004-637X/737/2/103DOI
  104. [105]Schulze, F., Remus, R.-S., Dolag, K., et al. 2018, MNRAS, 480, 4636, doi: 10.1093/mnras/sty2090DOI
  105. [106]Sérsic, J. L. 1963, Boletin de la Asociacion Argentina de Astronomia La Plata Argentina, 6, 41
  106. [107]Sersic, J. L. 1968, Atlas de Galaxias Australes
  107. [108]Shipley, H. V., Papovich, C., Rieke, G. H., Brown, M. J. I., & Moustakas, J. 2016, ApJ, 818, 60, doi: 10.3847/0004-637X/818/1/60DOI
  108. [109]Silk, J. 2013, ApJ, 772, 112, doi: 10.1088/0004-637X/772/2/112DOI
  109. [110]Silk, J., & Norman, C. 2009, ApJ, 700, 262, doi: 10.1088/0004-637X/700/1/262DOI
  110. [111]Simpson, C., & Rawlings, S. 2002, MNRAS, 334, 511, doi: 10.1046/j.1365-8711.2002.05453.xDOI
  111. [112]Smith, E. P., Heckman, T. M., & Illingworth, G. D. 1990, ApJ, 356, 399, doi: 10.1086/168848DOI
  112. [113]Solórzano-Iñárrea, C., Best, P. N., Röttgering, H. J. A., & Cimatti, A. 2004, MNRAS, 351, 997, doi: 10.1111/j.1365-2966.2004.07842.x
  113. [115]Stockton, A., Ridgway, S. E., & Kellogg, M. 1996, AJ, 112, 902, doi: 10.1086/118063DOI
  114. [116]Tadhunter, C., Dickson, R., Morganti, R., et al. 2002, MNRAS, 330, 977, doi: 10.1046/j.1365-8711.2002.05153.xDOI
  115. [117]Talbot, R. Y., Sijacki, D., & Bourne, M. A. 2024, MNRAS, 528, 5432, doi: 10.1093/mnras/stae392DOI
  116. [118]Tran, H. D., Cohen, M. H., Ogle, P. M., Goodrich, R. W., & di Serego Alighieri, S. 1998, ApJ, 500, 660, doi: 10.1086/305752DOI
  117. [119]Tremblay, G. R., Chiaberge, M., Sparks, W. B., et al. 2009, ApJS, 183, 278, doi: 10.1088/0067-0049/183/2/278DOI
  118. [120]Tsatsi, A., Lyubenova, M., van de Ven, G., et al. 2017, A&A, 606, A62, doi: 10.1051/0004-6361/201630218DOI
  119. [121]van Breugel, W. J. M., Stanford, S. A., Spinrad, H., Stern, D., & Graham, J. R. 1998, ApJ, 502, 614, doi: 10.1086/305925DOI
  120. [123]Verdoes Kleijn, G. A., & de Zeeuw, P. T. 2005, A&A, 435, 43, doi: 10.1051/0004-6361:20042271DOI
  121. [124]Vernet, J., Fosbury, R. A. E., Villar-Martin, M., et al. 2001, A&A, 366, 7, doi: 10.1051/0004-6361:2000076
  122. [125]Virtanen, P., Gommers, R., Oliphant, T. E., et al. 2020, Nature Methods, 17, 261, doi: 10.1038/s41592-019-0686-2DOI
  123. [126]Wagner, A. Y., Bicknell, G. V., & Umemura, M. 2012, ApJ, 757, 136, doi: 10.1088/0004-637X/757/2/136DOI
  124. [127]Walter, F., Sandstrom, K., Aniano, G., et al. 2011, ApJL, 726, L11, doi: 10.1088/2041-8205/726/1/L11DOI
  125. [128]Weingartner, J. C., & Draine, B. T. 2001, ApJ, 548, 296, doi: 10.1086/318651DOI
  126. [129]Weller, E. J., & van Dokkum, P. 2026, ApJ, 999, 33, doi: 10.3847/1538-4357/ae3ddaDOI
  127. [130]Worrall, D. M., Birkinshaw, M., & Young, A. J. 2016, MNRAS, 458, 174, doi: 10.1093/mnras/stw277DOI
  128. [131]Zhang, Y., Yang, X., Wang, H., et al. 2013, ApJ, 779, 160, doi: 10.1088/0004-637X/779/2/160DOI
  129. [132]Zheng, K., Zhang, Y., & Röttgering, H. 2024, A&A, 686, A169, doi: 10.1051/0004-6361/202347096DOI
  130. [133]Zinn, P.-C., Middelberg, E., Norris, R. P., & Dettmar, R.-J. 2013, ApJ, 774, 66, doi: 10.1088/0004-637X/774/1/66DOI
  131. [134]Zirm, A. W., Dickinson, M., & Dey, A. 2003, ApJ, 585, 90, doi: 10.1086/346021DOI
  132. [135]Zubovas, K., Nayakshin, S., King, A., & Wilkinson, M. 2013, MNRAS, 433, 3079, doi: 10.1093/mnras/stt952DOI

Paper details

vlnc:W_ccqvfaub
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents