FAST Observations of FRB 20220912A: Burst Properties and Polarization Characteristics

Figure 1(F) displays the length of each observation, along with the number of detected bursts and event rates. Over the course of 17 observations, we detected a total of 1076 bursts, details of which can be found in Table 1. The event rates of eight observations exceeded 100 hr⁻¹, with the highest event rate of 390 hr⁻¹ during the first observation, which is only surpassed by FRB20201124A's 542 hr⁻¹ (Zhang et al. 2022b), demonstrating that FRB 20220912A is a highly active repeating FRB.

Zoom In Zoom Out Reset image size

Table 1. The Properties of the FRB 20220912A Bursts

Burst ID	MJD a	DM	Peak Flux b	Width c	Peak Frequency d	Bandwidth e	Fluence b	Energy	RM	Linear	Circular
		(pc cm−3)	(mJy)	(ms)	(MHz)	(MHz)	(Jy ms)	(erg)	(rad m−2)	(%)	(%)
B01	59,880.497624400	220.24 ± 1.76	117.4 ± 1.4	4.70	1479.5 ± 5.4	122.4 ± 9.3	0.552 ± 0.007	3.979(49)e+37	−4.3${}_{-1.9}^{+2.1}$	95.0 ± 1.4	5.3 ± 1.0
B02	59,880.497624673	217.82 ± 2.9	115.5 ± 1.4	3.72	1366.2 ± 4.3	231.4 ± 15.0	0.429 ± 0.005	3.097(38)e+37	−0.3${}_{-0.8}^{+0.8}$	98.5 ± 1.8	−16.9 ± 1.3
B03	59,880.497684224	221.87 ± 0.06	242.8 ± 3.0	4.09	1300.0 ± 13.7	735.7 ± 896.2	0.993 ± 0.012	7.161(88)e+37	${1.1}_{-0.2}^{+0.2}$	97.7 ± 1.0	3.3 ± 0.7
B04	59,880.497945389	220.88 ± 3.78	174.6 ± 2.1	3.54	1381.0 ± 6.8	321.4 ± 28.0	0.619 ± 0.008	4.463(55)e+37	${0.5}_{-0.5}^{+0.5}$	98.6 ± 1.4	−6.7 ± 1.0
B05	59,880.497991706	220.91 ± 0.24	32.1 ± 0.4	4.06	1095.7 ± 11.2	251.7 ± 34.6	0.130 ± 0.002	9.398(115)e+36	−3.1${}_{-1.8}^{+1.5}$	94.5 ± 6.7	−16.1 ± 4.9
B06	59,880.498128495	219.64 ± 5.08	95.2 ± 1.2	4.89	1564.0 ± 56.7	338.4 ± 65.8	0.465 ± 0.006	3.355(41)e+37	${2.1}_{-1.2}^{+1.2}$	90.6 ± 1.6	7.2 ± 1.2
B07	59,880.498199601	223.30 ± 0.17	138.2 ± 1.7	8.97	1449.5 ± 26.3	332.1 ± 63.9	1.240 ± 0.015	8.942(109)e+37	${0.1}_{-0.3}^{+0.4}$	97.2 ± 0.9	14.8 ± 0.6
B08	59,880.498605364	222.98 ± 1.36	96.5 ± 1.2	8.65	1399.0 ± 15.2	374.2 ± 61.9	0.834 ± 0.01	6.019(73)e+37	${3.1}_{-0.6}^{+0.6}$	97.8 ± 1.5	−2.7 ± 1.0
B09	59,880.498635029	220.22 ± 0.24	34.5 ± 0.4	0.64	1004.4 ± 22.4	110.3 ± 34.3	0.022 ± 0.0	1.592(19)e+36	⋯	⋯	⋯
B10	59,880.498915641	221.35 ± 0.21	77.7 ± 0.9	8.74	1083.0 ± 5.5	256.5 ± 16.1	0.679 ± 0.008	4.898(59)e+37	${1.9}_{-0.6}^{+0.6}$	98.7 ± 2.2	−3.8 ± 1.6

Notes. The full table is available in ScienceDB, doi:10.57760/sciencedb.08058.

^a Barycentrical arrival time at 1.5 GHz. ^b Calculated within 500 MHz bandwidth. ^c Equivalent width. ^d Obtained with Gaussian fitting. ^e FWHM of Gaussian fitting.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

We calculated the waiting times between bursts for each observation. Similar to FRB 20121102A and FRB 20201124A, FRB 20220912A also exhibits a distinctive bimodal distribution (Figure 2). We utilized two lognormal functions to model the waiting-time distribution, with peaks located around 18 s and 51 ms, respectively.

Zoom In Zoom Out Reset image size

The waiting times of a Poisson process are exponentially distributed. Here, we use an exponential function to fit the waiting-time distribution using the waiting times down to the second valley of the two lognormal distribution (∼0.52 s). We utilize a Kolmogorov–Smirnov (K-S) test to evaluate the goodness of the lognormal and exponential fits, with p-values of 0.969 and 0.963, respectively, indicating that both models effectively describe the distribution. The Poisson process rate obtained from the exponential distribution is 0.041 ± 0.002 s⁻¹ or 147 ± 7 hr⁻¹, which is close to the average observed event rate of 1076/8.67 ∼ 124 hr⁻¹. The right peak of the waiting time represents the activity of the FRB source during the statistical period. The lognormal provides a left peak of the waiting time near 51 ms, which is quite similar to FRB 20201124A (39 ms in Xu et al. 2022 and 51 ms in Zhang et al. 2022b). The left peak of FRB 20121102A is about 3 ms (Li et al. 2021), significantly different from FRB 20220912A here. However, FRB 20121102A appears to have a weak secondary peak around 51 ms (Figure 3 in Li et al. 2021). The characteristic waiting time of 51 ms may signify some fundamental properties of the FRB source emitting the bursts.

Searching for periodicity of FRBs remains an active research area. There have been three non-repeating FRBs discovered to have millisecond-level quasi-periods (Chime/Frb Collaboration et al. 2022). For repeating FRBs, however, no reports of short periods have emerged (Zhang et al. 2018; Li et al. 2021; Niu et al. 2022; Xu et al. 2022). Here, we also conducted a period search for FRB 20220912A using two methods: Lomb–Scargle periodograms (LSPs) and phase folding. The LSP method has been widely applied to nonuniformly sampled time series (Lomb 1976; Scargle 1982), making it suitable for periodic searches in the arrival time series of FRBs. Phase folding is also a simple and foundational method for period searches. Here, we assumed MJD 59,880 as the initial phase and computed the phase of each burst under a preset period, counting the longest continuous phase interval without burst existence, i.e., the void fraction. We iterated over periods ranging from 1 ms to 1000 s and computed the void fraction under these periods. A larger void fraction means a more concentrated burst distribution in the phase space, indicating higher reliability of the corresponding period. This approach is similar to the statistics used in Rajwade et al. (2020), except that we directly calculate the phase of the burst, making it more efficient and accurate. Unfortunately, neither of these two methods yielded a valid periodicity in the burst arrival times.

Only FRB 20121102A and FRB 20180916B have been reported to possess potential periodicities, the former approximately 157 days (Rajwade et al. 2020; Cruces et al. 2021), the latter approximately 16 days (Chime/Frb Collaboration et al. 2020). These periodicities are composed of active and quiescent phases (like square waves), and not all active phases had burst detection. Bursts from FRB 20220912A have been detected in all of our observations, precluding exploration of such active-quiescent periods. However, we can define the event rate variations over time as a "light curve" to search for possible periodicities in activity levels. Given the 54 days duration of FAST observations, we can only search for periodicities up to 27 days. To avoid observational interference on a 1 day period, we commence our search from a 2 day period. Utilizing LSPs, no reliable periodicities were found within the period range of 2 and 27 days.

Energy is one of the basic properties of FRBs, which is a physical quantity that can directly reflect the radiation mechanism of FRBs. The energy function of FRBs is typically modeled as a power-law function, probably with a cutoff at the high end (e.g., Luo et al. 2018; Lu et al. 2020; Luo et al. 2020; Zhang et al. 2021). Li et al. (2021)'s detection of the low-energy outburst of FRB 20121102A reveals the multiple radiation mechanisms that FRBs may possess. Figure 3 displays the energy function of FRB 20220912A, along with its distribution over time. Due to varying observation lengths, the energy function is weighted based on the observation time. Like FRB 20121102A and FRB 20201124A (Aggarwal et al. 2021; Li et al. 2021; Xu et al. 2022; Zhang et al. 2022b; Jahns et al. 2023), the differential energy function of FRB 20220912A cannot be explained using one single function. Two lognormal functions are used for the fit, with the corresponding characteristic energies being 5.29 × 10³⁶ erg and 4.13 × 10³⁷ erg, respectively. The integral energy function also cannot be fit with a single power law. Two sections of power-law fit are utilized, with the power-law exponents being −0.38 ± 0.02 and −2.07 ± 0.07. In Figure 3(C), the energy appears to exhibit an evolutionary characteristic over time, with fewer high-energy bursts observed in the latter observations.

Zoom In Zoom Out Reset image size

After calibrating the time-frequency data, we calculate the fluence of each burst by integrating the flux with respect to time in each frequency channel. Then we average the fluence of all bursts to obtain the synthetic fluence-frequency spectrum of FRB 20220912A. As the different frequency channels of these bursts may be masked as radio frequency interference (RFI) channels, the number of bursts used for averaging fluence varies in frequency channels. We use Poisson counting error as the statistical error and use a simple power-law function to fit the spectral index of the synthetic spectrum:

$$\begin{eqnarray}&&I=A\times {F}^{\alpha }+C\end{eqnarray} \tag{ 2 }$$

In Figure 4 one can see large error bars near 1000, 1200 MHz, etc. due to frequent RFI interference, resulting in fewer effective data points. For the fit, we randomly select 90% of the data points from the available data and perform 1000 fits, resulting in the expectation and error of the FRB 20220912A spectral index being α = − 2.60 ± 0.21. This is the first measurement of the average spectral index of the synthetic spectrum of an FRB source. Such a spectrum can, in principle, be used to test against FRB radiation mechanisms (e.g., curvature radiation; Yang & Zhang 2018).

Zoom In Zoom Out Reset image size

Although the synthetic spectrum of all bursts conforms well to a power-law distribution, it is difficult to describe the relatively narrow bandwidth of individual bursts using a power-law model. Based on the current empirical evidence, a Gaussian function may be a viable alternative to describe the spectra of individual bursts. Prior research has employed a Gaussian model to investigate FRB 20121102A, with the fitting residual demonstrating the potential efficacy of the Gaussian distribution as a model, as illustrated in Figure 5 of Aggarwal et al. (2021). Similarly, Zhou et al. (2022) fitted the spectrum of FRB 20201124A with Gaussian functions and identified a bimodal distribution of central frequencies. Here, we also attempt to fit the spectrum of individual bursts using a Gaussian function:

$$\begin{eqnarray}&&I=A\times \exp \left[-\displaystyle \frac{{\left(F-\mu \right)}^{2}}{2{\sigma }^{2}}\right],\end{eqnarray} \tag{ 3 }$$

where we use the fitted μ as the center frequency ν₀ and FWHM ( $2\sqrt{2\mathrm{ln}2}\sigma$) as the bandwidth Δν. It should be noted that due to the limited bandwidth of FAST, some bursts likely have emission outside the FAST band not fully recorded. This introduces additional uncertainties to the bandwidth fitting results. Nine examples of bandwidth fitting are illustrated in Figure 5.

Zoom In Zoom Out Reset image size

Figure 6 presents the distribution of the central frequency (ν₀) and bandwidth (Δν) of the FRB 20220912A bursts. In order to mitigate the biases introduced by the limitations of observational bandwidth, we excluded bursts with ν₀ falling outside the range of 1–1.5 GHz and bursts with ν₀ within 50 MHz of the bandwidth edges (1 and 1.5 GHz) with bandwidth Δν < 100 MHz as these bursts are associated with high levels of fitting uncertainty. We subsequently analyzed the distribution of ν₀ and Δν for the remaining bursts. The central frequency of FRB 20220912A bursts mainly concentrates on the low-frequency range, and the distribution of high and low frequencies is extremely uneven. We use a single lognormal function to fit the bandwidth distribution. The mode of the lognormal function is located at 181 MHz, indicating that FRB 20220912A's bursts have very narrowband spectra. Furthermore, we did not find significant correlation between central frequency and emission bandwidth according to the sample we collected at the L-band. This is different from the power-law trend found by Kumar et al. (2023) from the bursts detected by the Parkes ultra-wideband lower receiver from the repeater FRB 20180301A, which might imply the actual observed bandwidth matters for this kind of analysis.

Zoom In Zoom Out Reset image size

A more relevant parameter to describe the narrowness of an FRB spectrum is Δν/ν₀. We investigate Δν/ν₀ in more detail. Because many bursts have emission outside of the FAST band (1–1.5 GHz), we limit our analysis to the bursts whose emission completely falls within the FAST band to avoid the bandwidth-selection effect. We select bursts based on two stringent criteria, i.e., the Gaussian fit standard deviation error δ σ is smaller than the standard deviation σ itself, and the FWHM of the burst spectrum is completely within the range of 1.05–1.45 GHz. Figure 7 displays the distribution δ ν/ν₀ of these bursts under these two filtering criteria. We fit the histogram distributions using lognormal functions and obtain μ = − 1.812 ± 0.001 and σ = 0.475 ± 0.002 for the blue histogram and μ = − 1.718 ± 0.002 and σ = 0.217 ± 0.003 for the red histogram, respectively. From the fitting results, we obtain their modes ${e}^{\mu -{\sigma }^{2}}$ located at 0.13 and 0.17, respectively. We also fit the cumulative Δν/ν₀ distribution using a power-law function. The results are suboptimal with significant fitting errors.

Zoom In Zoom Out Reset image size

The DM of each burst was determined using the DM phase software package (Seymour et al. 2019), ¹⁹ which maximizes the coherent power in the pulse across the emission bandwidth. The median value of the DM is 220.70 pc cm⁻³ with a standard deviation of 1.83 pc cm⁻³, which is consistent with CHIME's measurement of ∼220 pc cm⁻³ (McKinven & Chime/Frb Collaboration 2022). Linear fitting indicates the absence of any trend in the time evolution of DM, where the slope is dDM/d t < 8 × 10⁻³ pc cm⁻³ day⁻¹.

Due to the Faraday effect, the polarization plane of the FRB signal undergoes rotation during propagation. We employed the RM synthesis method to fit the RM of the burst using Equation (4),

$$\begin{eqnarray}&&\left(\begin{array}{l}I\\ Q^{\prime} \\ U^{\prime} \\ V\end{array}\right)=\left[\begin{array}{llll}1 & 0 & 0 & 0\\ 0 & \cos 2\theta & \sin 2\theta & 0\\ 0 & -\sin 2\theta & \cos 2\theta & 0\\ 0 & 0 & 0 & 1\end{array}\right]\left(\begin{array}{l}I\\ Q\\ U\\ V\end{array}\right),\end{eqnarray} \tag{ 4 }$$

where θ = RMλ². We selected bursts with RM errors less than 10 rad m⁻² (881 in total) for presentation.

The mean value of the RM is −0.08 rad m⁻², close to 0, indicating that the RM contribution from FRB 20220912A's host galaxy is comparable to that of the Milky Way, which is about −16 rad m⁻² (Hutschenreuter et al. 2022). The low value of RM indicates that the FRB 20220912A may be in a very clean environment. The linear fit also suggests that there is no trend in the evolution of RM with time, with a slope of 0.017 ± 0.018 day⁻¹. Furthermore, several other active repeating FRBs, such as FRB 20121102A (Hilmarsson et al. 2021a), 20201124A (Xu et al. 2022), 20190520B (Anna-Thomas et al. 2023), 20180916B (Mckinven et al. 2023a), and several other bursts (20181030A, 20181119A, 20190117A, 20190208A, 20190303A, 20190417A; Mckinven et al. 2023b) exhibit large RM values and show variations in RM on the timescale of months. In contrast to these repeating FRBs with large RMs, which suggest that all repeaters may have associated synchrotron-emitting persistent radio sources (a supernova remnant, a magnetar wind nebula or a mini-AGN) with a dense and highly magnetized environment, the polarization data of FRB 20220912A suggest that a large and varying RM is not the necessary condition to make active repeaters.

The linear polarization that is measured can be subject to overestimation when noise is present. As a result, we employ the frequency-averaged total linear polarization that has been de-biased (Everett & Weisberg 2001),

$$\begin{eqnarray}&&{L}_{\mathrm{de}{\rm{-}}\mathrm{bias}}=\left\{\begin{array}{l}{\sigma }_{I}\sqrt{{\left(\displaystyle \frac{{L}_{i}}{{\sigma }_{I}}\right)}^{2}-1}\quad \displaystyle \frac{{L}_{i}}{{\sigma }_{I}}\gt 1.57\\ 0\quad \displaystyle \frac{{L}_{i}}{{\sigma }_{I}}\leqslant 1.57\end{array}\right.,\end{eqnarray} \tag{ 5 }$$

where σ_I represents the off-pulse standard deviation of Stokes I, while L_i is the frequency-averaged linear polarization that has been measured for time sample i. The degree of linear and circular polarization are calculated with

$$\begin{eqnarray}&&L=\displaystyle \frac{{\sum }_{i}{L}_{\mathrm{de}{\rm{-}}\mathrm{bias},i}}{{\sum }_{i}{I}_{i}}\quad \mathrm{and}\quad V=\displaystyle \frac{{\sum }_{i}{V}_{i}}{{\sum }_{i}{I}_{i}},\end{eqnarray} \tag{ 6 }$$

where V_i is defined similarly to L_i . The uncertainties on the linear polarization fraction and circular polarization fraction are calculated as

$$\begin{eqnarray}&&\sigma =\displaystyle \frac{{\sigma }_{I}}{I}\sqrt{N+N\left({\rho }^{2}/{I}^{2}\right)},\end{eqnarray} \tag{ 7 }$$

where N is the number of time samples of the burst and ρ is ∑_i L_de-bias,i or ∑_i V_i , depending on whether we are calculating the linear or circular polarization fraction.

Most of the bursts from FRB 20220912A exhibit almost 100% linear polarization, with a noticeable fraction of bursts exhibiting significant circular polarization. Figure 8 displays the two bursts with the highest circular polarization degrees, which are −78.0% ± 10.8% and −69.4% ± 1.4%, respectively.

Zoom In Zoom Out Reset image size

Additionally, the circular polarization dynamic spectra of FRB 20220912A show various morphologies. We display 16 bursts in Figure 9. It can be seen that some bursts that cannot be distinguished in Stokes I appear as multiple bursts in Stokes V, such as bursts 1, 2, 7, 11, and 16. The bursts that show sign changes in circular polarization also have sub-pulse structures, so that the sign change may be caused by sub-pulses with different circular polarization modes at different times. The mechanism responsible for circular polarization is still under extensive discussion and no definitive conclusion has been reached (Qu & Zhang 2023). Curvature radiation by emitting bunches can be circularly polarized if the line of sight (LOS) is not confined in a beaming angle (Wang et al. 2022c). A sign change of circular polarization can be seen if the opening angle of the bunch is not much larger than 1/γ, where γ is the Lorentz factor of the bunch (Wang et al. 2022b). The curvature radiation mechanism predicts an average circular polarization fraction smaller than 55% when a sign change of circular polarization occurs. Another intrinsic radiation mechanism is the coherent inverse Compton scattering through charged bunches (Zhang 2022b; Qu et al. 2023). The scattered waves can be circularly polarized by adding up linearly polarized waves with different phases and polarization angles (Qu & Zhang 2023). When the LOS sweeps across the bunch's central axis, the sign of circular polarization can change, and the maximum circular polarization fraction can be larger than that of curvature radiation. The observation of FRB 20220912A could be understood within either scenario.

Zoom In Zoom Out Reset image size

Circular polarization in some bursts also appears to vary with frequency (Figure 10). An oscillation of polarization parameters as a function of wavelength may be an indication of Faraday conversion (Xu et al. 2022; Qu & Zhang 2023). However, the polarization degree of FRB 20220912A oscillates without an obvious oscillation frequency like FRB 20201124A (Xu et al. 2022). Furthermore, the conditions for significant Faraday conversion are usually quite stringent (Qu & Zhang 2023), requiring special magnetic field reversal environments, e.g., in a binary star system (Wang et al. 2022a) or when the source is surrounded by a supernova remnant (Yang et al. 2023). Such a required environment is not consistent with the very clean environment inferred from the small and non-variable RM of FRB 20220912A.

Zoom In Zoom Out Reset image size

By utilizing the isotropic energies derived from each burst, it is possible to impose limitations on the total energy allocation of the intrinsic FRB source, which can be employed to restrict the various models of FRB sources. When deducing the total source energy based on the energy of each burst, various factors must be considered, including radio radiation efficiency η_r , beaming factor f_b , and observation duty cycle ζ (Zhang 2022a).

We use Equation (1) for energy calculation with the isotropic emission assumption. However, coherent radiation from FRB bursts generally has a small solid angle δΩ. Therefore, the true burst energy should be δΩ/4π of the isotropic burst energy. Additionally, bursts may exist in directions that are not observable. Assuming the global emission beam to be ΔΩ, we can get the global beaming factor f_b = ΔΩ/4π.

Due to the limitation of the observation duty cycle, we cannot observe all bursts emitted by the FRB source during the observation period. For example, our observations of FRB 20220912A were conducted over 17 days with a total observation time of only 8.67 hr, which does not imply that there was no activity at other hours. Therefore, to estimate the total source energy during the observation period, we can use duty-cycle scaling to compensate for the unobserved periods of FRB radiation energy. If the total radio energy of the observed bursts is E_b , the total source energy should be ${E}_{b}\times {f}_{b}\,\times {\eta }_{r}^{-1}\times {\zeta }^{-1}$.

Table 2 presents our estimates of the total source energy for FRB 20121102A (Li et al. 2021), FRB 20190520B (Niu et al. 2022), FRB 20201124A (Xu et al. 2022; Zhang et al. 2022b), and FRB 20220912A (this paper) using FAST observations. In our calculations, we assumed typical values for η_r and f_b of 10⁻⁴ and 0.1, respectively. It should be noted that the burst energies reported in the literature for FRB 20121102A and FRB 20190520B were calculated using a center frequency of 1.25 GHz, rather than the bandwidth Δν. Therefore, we have divided the total energies of these two FRBs by 2.5 to allow for comparison with the other FRBs (see footnote 18 for more discussion). FRB 20220912A emitted an energy of 3.49 × 10⁴⁵ erg during the 17 day period, which exceeds 2% of the total dipolar magnetic energy (E_B ∼ 1.7 × 10⁴⁷ erg) of a magnetar. This implies that the dipolar magnetic energy of the magnetar would be completely depleted in only ∼850 days if the radio efficiency is indeed as low as 10⁻⁴. Certain magnetar models (e.g., low-efficiency models invoking relativistic shocks) would suffer from an energy budget problem.

Table 2. Energy Budget of Four Repeating FRBs

FRB Name	Duty Cycle a	Radio E b	Averaged E c	Source E d
	ζ	(erg)	(erg hr−1)	(${\eta }_{r,-4}^{-1}{f}_{b,-1}$ erg) e
20121102A	0.053	1.36 × 1041	2.29 × 1039	2.59 × 1045
20190520B	0.070	4.39 × 1039	2.37 × 1038	6.26 × 1043
20201124A f	0.063	1.65 × 1041	2.01 × 1039	2.60 × 1045
20201124A g	0.042	6.42 × 1040	1.60 × 1040	1.54 × 1045
20220912A	0.021	7.42 × 1040	8.55 × 1039	3.49 × 1045

Notes.

^a The observation duty cycle, e.g., for FRB 20220912A in this paper, the duty cycle is 8.67 hr out of 17 days. ^b Sum of the observed isotropic radio energies of all bursts. ^c The total radio energy divided by observation time, e.g., for FRB 20220912A in this paper, the averaged energy is 7.42 × 10⁴⁰ erg/8.67 hr. ^d The total source energy. ^e The source energy calculation uses η_r = 10⁻⁴ and f_b = 0.1. ^f FAST observation of FRB 20201124A in 2021.04 by Xu et al. (2022). ^g FAST observation of FRB 20201124A in 2021.09 by Zhang et al. (2022b).

Download table as:  ASCIITypeset image

FRB 20201124A was the first repeating FRB discovered with circular polarization (Hilmarsson et al. 2021b). Prior to this, only some non-repeating bursts were found to have significant circular polarization (Cho et al. 2020; Day et al. 2020; Feng et al. 2022a). Recently, FRB 20121102A and FRB 20190520B were detected by FAST with very few bursts exhibiting circular polarization (Feng et al. 2022b). Here, we present that FRB 20220912A has a large number of bursts exhibiting circular polarization, suggesting that this may be a common feature of repeating FRBs.

All the above-mentioned four repeating FRBs have a large number of detected bursts by FAST. We counted the proportion of bursts exhibiting significant circular polarization (circular degree ≥10%) for each of these four repeating FRBs. FRB 20121102A exhibited 12 out of 1652 detected bursts with circular polarization (Li et al. 2021; Feng et al. 2022b). Similarly, FRB 20190520B displayed circular polarization in 3 out of 75 detected bursts (Feng et al. 2022b; Niu et al. 2022). FRB 20201124 showed circular polarization in 302 out of 1863 detected bursts (Xu et al. 2022), while FRB 20220912A exhibited circular polarization in 303 out of 1076 detected bursts (this paper).

Figure 11 shows a tentative relationship between the fraction of bursts with circular polarization degree >10% and the absolute value of each FRB's RM. It appears that there is a negative correlation. The larger the RM value, the lower the fraction of bursts with circular polarization. If such a correlation is physical, it may be related to the intrinsic radiation mechanism and complicated environments of repeating FRB sources. First, the larger fraction of circular polarization in the clean-environment FRB 20121102A suggests that the origin of circular polarization may be more related to intrinsic radiation mechanisms than environment effects. According to Qu & Zhang (2023), significant circular polarization can be made from magnetospheric radiation mechanisms such as coherent curvature radiation or inverse Compton scattering by bunches. Relativistic shock models invoking synchrotron maser radiation mechanism, on the other hand, mostly emit bursts with nearly 100% linear polarization (Metzger et al. 2019; Plotnikov & Sironi 2019; Qu & Zhang 2023). The detection of circular polarization from all four sources favors the magnetospheric origin of the bursts. Second, in case of both curvature and inverse Compton scattering processes, given a random viewing angle, the fraction of bursts that have high circular polarization degree is high if the bunch shape is point-like. In order to reduce the fraction of circularly polarized bursts, the cross section of the bunches should be large so that most emitting leptons are viewed on beam (e.g., Wang et al. 2022b). The correlation seen in Figure 11 would then require that the FRB engine residing in a more magnetized environment (e.g., a younger magnetar) should be able to generate bunches with larger cross sections. Such a scenario may predict that the bursts from a more magnetized environment are systematically brighter, which is not observed. An alternative, possibly a more likely, scenario is that the circular polarization fraction is modified by environments. Qu & Zhang (2023) showed that polarization-mode-selected synchrotron absorption tends to convert circular polarization to linear polarization. The observed trend in Figure 11 then suggests that there is more significant synchrotron absorption in more magnetized environments. More repeater data are needed to confirm whether the circular polarization fraction and RM correlation is physical.

Zoom In Zoom Out Reset image size