Detection of Extended Millimeter Emission in the Host Galaxy of 3C 273 and Its Implications for QSO Feedback via High Dynamic Range ALMA Imaging

Object 3C 273 was observed at 93 GHz (Atacama Large Millimeter/submillimeter Array (ALMA) Band 3) as a bandpass and phase calibrator for Cycle 6 (2017.1.00815). The observations were conducted over three scans on 2018 June 3–4 and six scans on 2019 January 16–19. Each scan lasted 5 minutes.

The raw data in ALMA Science Data Model format were downloaded from the ALMA archive and reprocessed using the standard pipeline (version 5.6.1–8). After additional flagging at spectral window edges, imaging was performed using tclean with natural weighting. The resulting peak flux was 9.37 Jy beam⁻¹ with a standard deviation of 1σ = 6.1 mJy beam⁻¹ at emission-free regions and a beam size of 36 × 30. Panel (c) of Figure 1 shows this "standard" image, clearly indicating that the image noise is dominated by side lobes of the bright central source, i.e., that the image noise is limited by DR. Standard ALMA images are expected to reach a DR of ∼100 for lower-frequency bands (ALMA Cycle 8 Proposers Guide, Doc. 8.2., ver. 1.0), which is clearly insufficient to detect any weak emission associated with the host galaxy. We attempted to increase the DR via the following steps.

2.1.1. Self-calibration

2.1.2. Point-source Subtraction

Modeling and subtracting the central point source, which is responsible for the large side lobes, is important to increase the image DR. This was done using the following steps. First, the weights of the visibility of the 2018 and 2019 runs were normalized using the CASA task statwt. We performed tclean separately for each scan, with the savemodel = "modelcolumn" parameter and the clean box set to only the central peak pixel in order to save the central component as a model. The channel width was chosen to be 10 MHz as a reasonable compromise between averaging out phase variations and preserving the small-scale bandpass characteristics of the point source. The resulting point-source model has a frequency resolution of 10 MHz and time resolution of 5 minutes (i.e., the length of one scan). This model was then subtracted from the original visibilities using the task uvsub. Individual scans were then imaged for visual checking. At this step, two out of three scans in the 2018 run were flagged out, as they had features that were apparently 180° symmetric, indicative of the Hermitian nature of interferometric side lobes. These two scans were conducted on the same day and had larger phase variations compared to the other seven scans. After further flagging of spectral window edges and spurious features, the point source–subtracted visibilities were combined with the task concat and imaged together. Panel (d) in Figure 1 shows the point source–subtracted image. The image noise is 1σ = 0.11 mJy beam⁻¹, with a beam size of 36 × 32. This corresponds to DR ∼ 85,000. The theoretical noise of the data, calculated using the effective bandwidth and assuming that the noise is dominated by thermal processes, was 2 μJy beam⁻¹ over an effective bandwidth of 6.5 GHz. Therefore, the image is still DR-limited. An extended component is apparent around the central source at a >5σ level over a linear scale of ∼10'', corresponding to ∼27 kpc.

Zoom In Zoom Out Reset image size

ALMA observations at 233 GHz (Band 6) were obtained in Cycle 7 (2019.1.00634) in order to verify the existence of extended emission detected in Band 3 archival data.

Band 6 observations were conducted with the ACA and 12 m arrays simultaneously, within a time frame of 20 minutes. This ensured that the intrinsic flux variations of the source were negligible and that the ACA and 12 m array could be calibrated together. After standard pipeline calibration, an initial image showed that the peak flux at 233 GHz was 4.04 Jy beam⁻¹ with 1σ = 13 mJy beam⁻¹ (DR ∼ 300), with a beam size of 14 × 12 (panel (e) of Figure 1).

Four iterations of self-calibration were performed, with solution intervals of 10 s and 5 s for the first two iterations solving only for phases, then with a solution interval of 1 s solving for both amplitude and phase, and then lastly with the solnorm = T option to normalize solutions. The final peak flux was 4.67 Jy beam⁻¹ but did not change during the last two iterations. The resulting noise decreased to 1σ =0.2 mJy beam⁻¹.

Subtraction of the central point source was done in the same way as in Band 3. Panel (f) of Figure 1 shows the point source–subtracted image. The image noise is 1σ = 0.12 mJy beam⁻¹. This corresponds to DR ∼ 39,000. The theoretical noise of the data was 7.9 μJy beam⁻¹ over an effective bandwidth of 8.0 GHz. The image is still DR-limited. An extended component can be seen elongated in the northeast–southwest direction at the >2σ level. Two compact components adjacent to the central source are significant at the >20σ level, but these components could arise from insufficient subtraction of the central point source. As the point-spread function at the center with a peak flux of 4.67 Jy beam⁻¹ decreases to the 1σ noise level at a radius of approximately 2'', we disregard the central 4'' from further analyses.

ALMA 343 GHz (Band 7) data were obtained together with Band 6. For Band 7, observations were conducted with the ACA only, with nine antennas. Observations were completed within a time frame of 90 minutes. After standard pipeline calibration, an initial image showed that the peak flux at 343 GHz was 3.13 Jy beam⁻¹ with 1σ = 17 mJy beam⁻¹ (DR ∼ 180), with a beam size of 58 × 34 (panel (g) of Figure 1).

Four iterations of self-calibration were performed, with solution intervals of 60 s and 5 s for the first two iterations solving only for phases, then with a solution interval of 1 s solving for both amplitude and phase, and then lastly with the solnorm = T option. The final peak flux was 3.21 Jy beam⁻¹ with 1σ = 1.5 mJy beam⁻¹, with a beam size of 58 × 34.

Point-source subtraction was done in the same way as in Band 3. Panel (h) of Figure 1 shows the point source–subtracted image. The image noise is 1σ = 1.3 mJy beam⁻¹, with a beam size of 60 × 34. This corresponds to DR ∼ 2500. The theoretical noise of the data was 81 μJy beam⁻¹ over an effective bandwidth of 8.0 GHz. The image is still DR-limited.

Residual emission can be seen in the center, but no significant extended emission was found.

The most prominent structure in the ALMA bands after the subtraction of the central source is the synchrotron jet emitted from the central source in the southwest direction. The jet itself is orders of magnitude weaker compared to the central peak and invisible at higher-frequency ALMA bands when only the standard imaging procedure is applied, due to the low DR. We measured its total flux to compare with the centimeter flux obtained with the Very Large Array (VLA). The region where the flux was measured is shown in panel (b) of Figure 1. The region is selected to include the bulk of the millimeter and centimeter jet, but the boundary of the jet in the vicinity of the nuclear source cannot be determined unambiguously. A millimeter knot can be seen at the boundary of the jet region in panels (d) and (f) of Figure 1, but its flux is ∼2 mJy, which accounts for less than 1% of the total jet flux; thus, its effect is negligible for the discussion presented here.

Table 1 lists the VLA and ALMA flux, and the corresponding spectral energy distribution (SED) is shown in Figure 2. The jet SED can be well fit by a simple power law of the form F_ν ∝ ν^α , where α = −0.85 ± 0.03. The reduced χ² of the fit was 0.007.

Zoom In Zoom Out Reset image size

The high DR attained by the ALMA image reveals a diffuse, extended component around the central source. The emission is annotated in panels (d) and (f) of Figure 1. The emission extends to the northeast, opposite from the jet, showing that this component is unlikely to be the root of the jet.

We measure the flux of the extended component within an ellipse with major axis 13'', minor axis 6'', and a position angle of 0° centered on the central source, which is a reasonable compromise between capturing the bulk of the extended component and avoiding the root of the jet extending to the southwest. This region is also consistent with the area where extended [O iii] is detected in Husemann et al. (2019b). For reasons explained in Section 2.2, we subtract the flux in the central 4''. Table 1 lists the measured flux of the ALMA bands. Flux are listed separately for the lower and upper sidebands, as well as for the aggregated band in Bands 3 and 6. For Band 6, the flux becomes larger by approximately 20% if the image is smoothed to a common angular resolution with the Band 3 image. For the following analyses, we use the average flux of 5.8 mJy over Bands 3 and 6. For Band 7, the error in the flux is large, and we do not detect any significant diffuse emission at this frequency.

Extended centimeter emission, or the radio lobe, has been detected in the vicinity of the central source with the VLA (Punsly & Kharb 2016; Perley & Meisenheimer 2017). Here we coin the term "radio lobe" specifically for the centimeter detection to make the distinction from the extended millimeter component. Punsly & Kharb (2016) discussed the radio lobe in detail and concluded that the lobes are relaxed synchrotron cooling plasma associated with the jet and counterjet. In order to estimate the extended centimeter flux of the region where we measure the extended millimeter flux, we assume that the spatial extrapolation of the radio lobe flux to the central region gives its upper limit. This is a reasonable assumption, as radio lobes are bipolar structures that peak at the lobes, created by their interaction with the intergalactic medium. We measured the average flux density in the VLA 1.4 GHz (L band) and 4.9 GHz (C band) images for regions where the radio lobe was strongest, in the northeast direction from the nucleus (counterjet side; shown in panel (b) of Figure 1). Then, we scaled the flux assuming that the same flux density applies for the central 13'' × 6''. The flux limit estimates for the VLA bands are listed in Table 1.

From the SED of the extended millimeter component shown in Figure 2, it is clear that in the millimeter range, the SED is flat or can allow for only a very shallow (α = −0.03 ± 0.07) spectral index. The centimeter flux is significantly stronger than in the millimeter. We spectrally extrapolate the radio lobe flux to the millimeter assuming α = −0.85, the same value as the jet. This is justified because the nuclear source is the common origin for the relativistic electrons that give rise to the radio lobe and the jet, and the electron number distribution is likely to be similar as well. The assumed value is also in line with the synchrotron spectral index of the KINGFISHER sample of nearby galaxies (−0.97 ± 0.16; Tabatabaei et al. 2017). The expected flux is ≤1 mJy at 93 GHz and ≤0.5 mJy at 233 GHz. Thus, synchrotron emission originating from the nuclear source or the jet accounts for less than 17% of the extended 93 GHz flux and less than 9% of the extended 233 GHz flux, indicating that it is not the dominant mechanism responsible for extended emission in the millimeter range.

Synchrotron emission from galaxies can originate from cosmic rays accelerated in supernova remnants, i.e., star formation within the disk of the host galaxy. If the extended millimeter emission is synchrotron emission tracing massive star formation, we can convert the millimeter flux to the star formation rate (SFR). Assuming a synchrotron spectral index of −0.85, the observed millimeter flux of 5.8 mJy corresponds to an SFR of ∼3500 M_⊙ yr⁻¹ (Condon 1992). This is an order of magnitude higher than estimates in other radio-loud quasar hosts (Zakamska et al. 2016; Hardcastle et al. 2019). The stellar mass of the host elliptical galaxy of 3C 273 is estimated to be M_* ∼ 10^11.1 M_⊙ with an average stellar population of 2.5 Gyr (Zhang et al. 2019). For nearby star-forming galaxies on the main sequence, the expected SFR at this stellar mass is only a few × 10 M_⊙ yr⁻¹ (Peng et al. 2010). Thus, if the observed millimeter flux is due to synchrotron emission from massive star formation, the host of 3C 273 must be an extreme starburst, but such a bursty phase contradicts with the relatively old stellar population. In addition, the molecular gas mass in 3C 273 estimated from ¹²CO (J = 1 − 0) emission is only ∼10⁹ M_⊙ (Husemann et al. 2019b), which is comparable to local spirals that have SFRs of only a few M_⊙ yr⁻¹. We therefore conclude that synchrotron originating from massive star formation in the host galaxy is not the source of extended millimeter emission in 3C 273.

An alternative mechanism of radio emission is thermal bremsstrahlung. Thermal bremsstrahlung is weak and normally swamped by synchrotron at the centimeter and dust at the submillimeter wavelengths. Thermal bremsstrahlung normally has a shallow spectral slope of α ∼ −0.1, which is consistent with our nearly constant flux observed in the ALMA bands. Assuming thermal bremsstrahlung, we can convert the observed millimeter flux of 5.8 mJy to the production rate of ionizing photons using the approximation by Rubin (1968),

$$\begin{eqnarray}&&\begin{array}{l}\left[\displaystyle \frac{Q({H}^{0})}{{{\rm{s}}}^{-1}}\right]\geqslant 6.3\times {10}^{25}\\ \times \ \ {\left(\displaystyle \frac{{T}_{e}}{{10}^{4}\ {\rm{K}}}\right)}^{-0.45}{\left(\displaystyle \frac{\nu }{\mathrm{GHz}}\right)}^{0.1}\left(\displaystyle \frac{{L}_{\nu }}{\mathrm{erg}\,{{\rm{s}}}^{-1}{\mathrm{Hz}}^{-1}}\right),\end{array}\end{eqnarray} \tag{ 1 }$$

which assumes equilibrium between ionization and recombination, and the equality holds only if the system is ionization-bound. For an electron temperature of T_e ∼ 10⁴ K, we have Q(H⁰) > 4.0 × 10⁵⁶ s⁻¹. Here we have omitted the equality, since the radio jet protruding from the host galaxy indicates that the system is at least partially density-bound (i.e., some ionizing photons escape from the host galaxy). We denote this lower-limit value as $Q{({H}^{0})}_{\min }=4.0\times {10}^{56}\ {{\rm{s}}}^{-1}$, which indicates the actual number of ionizing photons that were used to create the nebula. There are two candidate sources of ionizing photons: massive star formation in the host galaxy or direct ionization of the host ISM by the central source.

If the extended emission is due to massive star formation, the ionizing photon rate above corresponds to SFR >3000 M_⊙ yr⁻¹ (Condon 1992; Murphy et al. 2011; Bendo et al. 2016; Michiyama et al. 2020). Following the discussion presented in Section 4.2, this SFR is unreasonably high, making thermal bremsstrahlung from massive star formation an unlikely origin of the extended millimeter emission. This is consistent with results by Husemann et al. (2019b) that the optical emission line ratios of the host galaxy are inconsistent with star formation.

Alternatively, the central source can directly ionize atomic gas in the host galaxy, giving rise to the thermal bremsstrahlung (Baskin & Laor 2021). Discussion in the previous sections likely rules out other origins, leaving this "QSO feedback" scenario as the plausible explanation of the extended millimeter flux discovered in 3C 273. In this case, the thermal emission can be regarded as the radio counterpart of EELRs observed through optical forbidden lines.

Martel et al. (2003) observed 3C 273 with the Hubble Space Telescope (HST) coronagraph and found that the outer halo of the host galaxy is lopsided in the northeast direction out to approximately 10'' (1'' = 2.7 kpc). The EELR detected in [O iii] by Husemann et al. (2019b) shows that the [O iii] emission typically extends to 7''. Our detection of the extended millimeter component extends to radii of ∼7'', similar to the EELR, while extending to the northeast direction. The distribution of ionized gas detected in thermal bremsstrahlung is thus consistent with that detected in [O iii].