Deep Synoptic Array Science: Discovery of the Host Galaxy of FRB 20220912A

The DSA-110 ³ is a radio interferometer hosted at the Owens Valley Radio Observatory (OVRO). During the observations presented herein, the array was being commissioned with data from 48 core antennas and 15 outriggers being processed. A full description will be presented in V. Ravi et al. 2023, (in preparation), and we here include essential information to support the observational results for FRB 20220912A.

Essential specifications of the DSA-110 at the time of these observations are presented in Table 1. Each DSA-110 antenna is automatically steerable in elevation only and observes on the meridian. Each antenna is equipped with dual orthogonal linearly polarized receivers and ambient temperature amplifiers that deliver a typical system temperature of 25 K (Weinreb & Shi 2021). A real-time FRB search is enabled through the coherent combination of 48 core antennas located in an east–west line, with a maximum spacing of 400 m. Total-intensity data in 256 coherent, fan-shaped search beams are inspected in real time for FRBs, using a modified version of the heimdall software (Barsdell et al. 2012). Triggers from FRB candidates result in the storage of post-filterbank voltage data from the 48 core antennas and 15 outrigger antennas. The 4-bit voltage data are recorded for 61440 samples at 32.768 μs time resolution in 6144 × 30.518 kHz channels.

Table 1. Specifications of the DSA-110 during Commissioning Observations

Parameter	Value
Dish diameter (m)	4.65
Central frequency (MHz)	1405
Primary beam FWHM (deg)	3.4
Bandwidth (MHz)	187.5
Search SEFD (Jy) a	140
Search-beam channel width (kHz)	244.141
Search-beam time resolution (μs)	262.144
Search-beam width (arcsec) b	134
Number of search beams	256
Search-beam spacing (arcmin)	1
Maximum baseline (m)	2500
Synthesized beam c	354 × 176 @ 834

Notes.

^a System-equivalent flux density (SEFD) at boresight. ^b This is in the east–west direction. The search beams are fan beams formed with an east–west array, and their sizes are only constrained by the primary beam in the north–south direction. ^c Major- and minor-axis diameters at a decl. of +487, assuming natural weighting. The position angle is measured east of north.

Download table as:  ASCIITypeset image

Data from all antennas (core and outriggers) are also continuously correlated in real time in the two linear polarizations only (no cross-polarization products), resulting in the production of visibility data fringe stopped on the meridian at the pointing center. Each FRB candidate is accompanied by the storage of 2 hr of visibility data surrounding the trigger time. These visibility data are used to solve for antenna-based complex gains at the trigger time using a sky model. Any >1 Jy calibrator source from the Very Large Array (VLA) calibrator manual that passes within the primary beam, represents >15% of the flux within the beam, and is classified as "P" or "S" for the VLA C configuration is also used to trigger the storage of 10 min of visibility data fringe stopped at the calibrator location. These visibility data are used to derive antenna-based band-pass calibrations for the beam former and for postprocessing of the voltage data.

Following the report on FRB 20220912A by CHIME/FRB (McKinven & Chime/Frb Collaboration 2022) on 2022 October 15, we began a campaign of near-daily observations of transits of the source with the DSA-110. The array was pointed at a decl. of 487. We report on the period between 2022 October 15 and 2022 October 25, during which we observed seven transits of FRB 20220912A and detected two bursts with the real-time system. During each transit, the source was within the field of view of the search beams for 25.9 minutes. As these observations were undertaken during science commissioning and the precise search completeness was not quantified, we do not quote a derived burst rate from FRB 20220912A.

When not observing FRB 20220912A, the DSA-110 was primarily pointed at a decl. of 716. Daily band-pass calibration was made possible using observations of the source 3C309.1 (J1459+7140), which has a flux density of 7.6 Jy at a wavelength of 20 cm according to the VLA calibrator manual. Approximate polarization calibration was made possible using voltage data obtained on the standard sources 3C48 and 3C286 (Perley & Butler 2013) on 2022 October 20.

We detected bursts from FRB 20220912A on 2022 October 18 (hereafter Burst 1) and on 2022 October 25 (hereafter Burst 2), on modified Julian dates (MJDs) 59,870.22278235404 and 59,877.203253573236, respectively. The arrival times correspond to 1530 MHz and the DSA-110 reference position of −118283 longitude, +372334 latitude. Burst 1 was detected with a signal-to-noise ratio (S/N) of 12.2 and an approximate DM of 228 pc cm⁻³, and Burst 2 was detected with an S/N of 8.5 and an approximate DM of 223 pc cm⁻³. Despite the differences between these DMs and the structure-optimized DM reported by CHIME/FRB (McKinven & Chime/Frb Collaboration 2022), we associated these detections with FRB 20220912A based on the correspondance between the CHIME/FRB position and the detection search beam. We attribute these differences to the time-frequency drifting common in repeating FRBs (e.g., Hessels et al. 2019).

Voltage data on each burst were coherently combined in the direction of the detection beams using the 48 core antennas only. Time-frequency data sets in all four Stokes parameters were formed with the native time and frequency resolutions. Incoherent dedispersion at the native time and frequency resolutions was applied. Flattening of the band passes in each received polarization was done on a per-antenna basis using observations of 3C309.1 on the same sidereal days as the burst detections. Voltage data obtained during transits of 3C48 and 3C286 were used for polarization calibration under the ideal feed assumption; coupling between the orthogonal polarizations is detected at only the percent level for DSA-110 (M. Sherman et al. 2023, in preparation). Full-Stokes spectra toward 3C48 and 3C286 were formed using the same beam-forming procedure as applied to the burst data. Spectra on 3C48 were used to verify the equalization of the amplitude gains in the two linear polarizations, and spectra on 3C286 were used to measure the frequency-dependent phase between the two polarizations.

Total-intensity dynamic spectra and temporal profiles are shown in Figure 1, and full-Stokes temporal profiles are shown in Figure 2. Voltage data in one out of 16 sub-bands for Burst 2 were corrupted, resulting in the reduced bandwidth evident in Figure 1. The two bursts exhibit marginally different morphologies and polarization characteristics. Burst 1 exhibits a single component, whereas Burst 2 potentially exhibits two components with visually different total-intensity spectra. For Burst 1, we measure a linear polarization fraction of 64% ±18% and a circular polarization fraction of 41% ± 9%. For Burst 2, we measure a linear polarization fraction of 73% ± 30%, with no significant circular polarization and a different polarization position angle. This is consistent with the report in Zhang et al. (2022) of substantially varying polarization properties between bursts from FRB 20220912A. Although these data have been corrected for an RM of +0.6 rad m⁻² (McKinven & Chime/Frb Collaboration 2022), an independent RM search of each burst revealed measurements consistent with this value. These data do not have sufficiently high S/N for more detailed inferences to be made.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We were able to derive arcsecond-scale localizations for the two bursts detected with DSA-110 from FRB 20220912A. The voltage data obtained for each burst from all core and outrigger antennas were dedispersed with DMs that maximized the S/N, and correlation products in each orthogonal linear polarization were formed and integrated over 8.36 ms centered on each burst arrival time. These correlation products were written into Common Astronomy Software Applications (CASA; THE CASA TEAM et al. 2022) measurement sets (MSs), and all further processing was done using CASA version 5.4.1. Band-pass calibrations were obtained using 10 minute observations of 3C309.1 on the same sidereal days as the burst detections and assuming a flat-spectrum point-source model. So-called "field" MSs were then generated using 10 minutes of visibility data surrounding the arrival times of each burst, fringe stopped at a 1 hr angle of 0 at the midpoint of the observation. Sky models based on the NRAO VLA Sky Survey (NVSS; Condon et al. 1998) were inserted into the field MSs, including all sources brighter than 30 mJy of any size. Extended sources were modeled as elliptical Gaussians. After band-pass calibration, the sky models were used to derive complex gain solutions for each antenna from the field MSs. The band-pass and gain calibration tables were then applied to the burst MSs. Wide-field images of the burst MSs 45 in diameter with 4100 pixels per side were made using the wsclean software (Offringa et al. 2014) to crudely identify the locations of the bursts. The CASA task tclean was then used to make higher-resolution images 5' and 1000 pixels per side centered on the approximate FRB positions. For these images, Briggs weighting with a robust parameter of 0.5 was used to optimize the point-spread function (PSF). All frequencies were weighted equally despite the apparent spectral structure of the bursts (Figure 1); although this results in lower-fidelity deconvolution, the effect is marginal given the small fractional bandwidth. Baselines shorter than 40 m were discarded due to spurious correlated power on short baselines. The localizations were derived by fitting two-dimensional Gaussians in the image plane to deconvolved burst images, derived from 10 image-plane CLEAN iterations.

The best-fit position for Burst 1 was (R.A. J2000, decl. J2000) = (23:09:04.90, +48:42:25.4), and the best-fit position for Burst 2 was (R.A. J2000, decl. J2000) = (23:09:04.88, +48:42:25.6). ⁴ Dirty and clean images of Burst 1, together with the PSF, are shown in Figure 3. In the wide-field wsclean images, the image S/N values for the bursts were 11 and 7, found after 10 image-plane CLEAN iterations and measured far from the sources. As the image of Burst 2 is of somewhat worse quality than that of Burst 1, we adopt the position of Burst 1 as the best-fit DSA-110 position of FRB 20220912A.

Zoom In Zoom Out Reset image size

We verified the astrometric accuracy of our results in four ways. First, we compared images of the primary-beam full-width half-maximum (FWHM) areas made with the entire 2.1 s voltage data sets, and with the field MSs, and ensured that bright sources were detected at consistent positions. We then checked the efficacy of the gain calibrations derived from the field MSs by checking the positions of bright (>50 mJy) compact (<30'') NVSS sources within the primary-beam FWHM in images of these MSs against NVSS catalog positions. The results are shown in the top panel of Figure 4. Some offsets in derived positions due to the differing synthesized-beam shapes of the DSA-110 and NVSS are expected for resolved sources, and so this check is best used to search for systematic errors in the gain calibrations rather than quantify the localization accuracy. Instead, we checked the absolute astrometry by forming MSs on and imaging two compact bright sources from the Radio Fundamental Catalog (rfc_2022c), J2258+4937 and J2325+4806. These sources are close in decl. to FRB 20220912A and thus transit through similar portions of the DSA-110 primary beam, and are close in R.A. and thus could be calibrated using gains from the field MSs. The derived position offsets were 07 and 04 respectively from the rfc_2022c positions on 2022 October 18, and 05 and 06 respectively on 2022 October 25. We also conducted observations of the bright pulsar J1935+1616 over three transits between 2022 October 22 and 2022 October 25, and performed a localization analysis on 11 automatically detected single pulses in exactly the same way as for the FRB 20220912A burst detections. The resulting offsets from the true position of PSR J1935+1616 (Manchester et al. 2005), corrected for proper motion, are shown in the bottom panel of Figure 4. The S/Ns of these pulses ranged between 10 and 21.

Zoom In Zoom Out Reset image size

Given these tests, we find no evidence for systematic offsets or effects of non-Gaussian noise. We therefore quote the localization as measured along with a conservative theoretical estimate of the statistical uncertainty for a source with an image S/N of 10. Conservatively, the standard error in each coordinate is given by half the synthesized-beam FWHM (Reid et al. 1988). Each error is scaled by 2.14 (e.g., Condon et al. 1998) to derive diameters of the 90% confidence error ellipse. We quote an approximate 90% confidence error ellipse with diameters of 4'' in R.A. and 2'' in decl., as shown in Figure 4.

We obtained an optical image of PSO J347+48 in the "Ellis-R" filter of the Echellette Spectrograph and Imager on the Keck II Telescope (Keck II/ESI; Sheinis et al. 2002). Four 300 s exposures were obtained on 2022 October 24 at an air mass of 1.45 in conditions of 08 seeing. The data were bias subtracted, flat fielded, combined, and registered using a custom astropy-based pipeline (Astropy Collaboration et al. 2022). An astrometric solution was obtained using the ten nearest unsaturated Gaia Data Release 3 (Gaia Collaboration et al. 2022) stars to the center of PSO J347+48. The image is shown in Figure 5, and an approximate calibration against Gaia data implies a limiting magnitude of 26. There is no evidence for any object besides PSO J347+48 within the DSA-110 localization region of FRB 20220912A.

We obtained an optical spectrum of PSO J347+48 using the Low Resolution Imaging Spectrometer on the Keck I Telescope (Keck I/LRIS; Oke et al. 1995). A single 300 s exposure was obtained on 2022 October 19 in conditions of 09 seeing at an air mass of 1.15, using a 1'' slit at a position angle of 2547. Light was split between the blue and red arms using the D560 dichroic and dispersed using the 400/3400 grism on the blue arm and the 400/8500 grating on the red arm. The line FWHM was approximately 7 Å. The data were processed using standard techniques with the lpipe software (Perley 2019); flux calibration was carried out using observations of the standard star BD+28 4211. A spectrum was extracted from the central square arcsecond of the galaxy; this is shown in Figure 6.

Zoom In Zoom Out Reset image size

Several emission lines are evident in the spectrum of PSO J347+48, at a redshift of 0.0771 ± 0.0001. We measure an Hα line flux of (1.23 ± 0.05) × 10⁻¹⁵ erg s⁻¹ cm⁻². The line ratios $\mathrm{log}([{\rm{O}}\,\mathrm{III}]/{\rm{H}}\beta )=0.00\pm 0.05$, $\mathrm{log}([N\mathrm{II}]/{\rm{H}}\alpha )\,=-0.04\pm 0.01$, $\mathrm{log}([{\rm{S}}\mathrm{II}]/{\rm{H}}\alpha )=-0.43\pm 0.02$, and $\mathrm{log}([OI]/{\rm{H}}\alpha )=-1.2\pm 0.1$ imply a composite Baldwin–Phillips–Terlevich (BPT) classification (Kauffmann et al. 2003; Kewley et al. 2006). A two-dimensional Sérsic profile fit to the morphology of PSO J347+48 in the Keck II/ESI image, combined with the redshift measurement, indicates an effective radius of 2.2 ± 0.1 kpc. This is consistent with the z ∼ 0 size−mass relation for late-type galaxies (Shen et al. 2003). The Hα line flux implies a star formation rate in the central square arcsecond of the galaxy of approximately 0.1 M_⊙ yr⁻¹, assuming no active galactic nucleus contribution and using a conversion factor consistent with Heintz et al. (2020). We note that the Keck I/LRIS two-dimensional spectrum shows evidence for off-nuclear Hα emission. Assuming a disklike galaxy, the inclination derived from a half-light isophote fit is 51°.

We performed a fit to the optical spectrum of PSO J347+48 using the latest iteration of the penalized pixel fitting code pPXF (Cappellari 2022) to jointly fit the stellar continuum and nebular emission. We adopted the default MILES stellar templates and ran the fit using recommended procedures and parameters. The fit results are shown in Figure 6. An internal dust extinction corresponding to A_V = 0.5 was found, and the fit results were also used for the redshift, line-flux, and line-ratio measurements given above.

We then fit archival photometry of PSO J347+48 from the Pan-STARRS (Magnier et al. 2020) and ALLWISE (Cutri et al. 2021) source catalogs using the Prospector stellar population synthesis modeling code (Johnson et al. 2021). Although the galaxy is marginally detected in 2MASS images, it is not cataloged, and we thus did not consider photometry from 2MASS data. ALLWISE detections in bands w1 and w2 were used, together with all five Pan-STARRS bands. We ran Prospector using recommended techniques and standard priors with a "delay-tau" parametric star formation history and sampled from the posterior using emcee (Foreman-Mackey et al. 2013). We fixed the internal dust attenuation to the pPXF result and included a model for dust reradiation in the likelihood. Results for the galaxy stellar mass and metallicity are given in Table 2. The star formation history was poorly constrained by the fit. The spectral energy distribution of PSO J347+48 is shown in Figure 7, together with the results from the Prospector analysis.

Zoom In Zoom Out Reset image size