Polarization Properties of 28 Repeating Fast Radio Burst Sources with CHIME/FRB

FRB 20180916B. A periodic (∼16.35 days) activity window was found for this FRB (CHIME/FRB Collaboration et al. 2020). R. Mckinven et al. (2023a) have reported temporal RM variations that are unrelated to the 16.35 days cycle, showing stochastic behavior between 2019 January and 2021 January, and then a secular increase until early 2022. In our extended data set of FRB 20180916B taken between 2022 January and 2024 May, CHIME recorded 26 baseband events, of which significant polarization measurements can be obtained for 10 events that are bright enough, as listed in Table 3. From these, we report that the RM of FRB 20180916B once again shows a stochastic trend through to mid-2024 (Figure 5). A similar trend was reported by S. Bethapudi et al. (2025) during the preparation of this work.

Zoom In Zoom Out Reset image size

This temporal variation pattern does not support the long-orbit binary models proposed by F. Y. Wang et al. (2022), Z. Y. Zhao et al. (2023), and H.-T. Lan et al. (2024) being the reason for this observed polarization fluctuation, particularly since no clear RM sign change nor turning point is seen during the five years of observations, which argues against a potential magnetic field reversal caused by the orbital motion of the binary star along the line of sight. However, this does not preclude FRB 20180916B from being in a binary system. We note that the expected Galactic RM contribution at the position of FRB 20180916B is quite uncertain, with RM_MW = −94 ± 45 rad m⁻². If the true Galactic contribution is <−59 rad m⁻², then our latest observations would indicate an RM sign change happened in 2022. This may or may not indicate a magnetic field reversal. The ambiguity is due to the fact that we cannot distinguish the fractional RM contribution from the host ISM.

We see no correlation between the RM variation and other burst parameters, namely emission bandwidth, L/I, and DM. Note that the emitting bandwidth shown in Figure 5 represents a by-eye estimation of the portion of the 400–800 MHz CHIME band over which emission is observed, uncorrected for the nonuniform bandpass of the instrument. The emission bandwidth appears to drift to lower frequencies in more recent observations, as was already noted by R. Mckinven et al. (2023b), although correlated changes are not observed in the other parameters. Figure 6 shows L/I versus fluence for the bursts we have detected from FRB 20180916B. We do not see any obvious correlation between the two parameters, with a Spearman correlation coefficient p-value of 0.07. There is perhaps a wider spread of L/I values at the higher fluence end, which is likely due to the fact that it is easier to detect weakly polarized signals in brighter events.

Zoom In Zoom Out Reset image size

FRB 20181119A. Three baseband polarization measurements (RMs of +606.2(1.2), +1343(2), and +480.9(5) rad m⁻²) from FRB 20181119A have already been published by R. Mckinven et al. (2023b). In this paper, we include a fourth event (RM of +355.2(11) rad m⁻²) during the time period 2020 June–2021 June. The RMs vary by hundreds of units during this period. The cable delay for some of these events is quite large (∼19 ns), which introduced some uncertainty in the signs of the RMs, although the positive RM fits showed slightly higher likelihood.

FRB 20190117A. Eight baseband events were recorded between 2019 June and 2023 May. Only one burst with significant RM detection is shown in Table 3, while the remaining bursts have no constraining measurements, likely due to low S/N.

FRB 20190303A. This FRB has been localized to a merging galaxy pair (D. Michilli et al. 2023). Five baseband events were recorded between 2021 June and 2023 September. Only two bursts with significant RM detections are shown in Table 3. Comparing these new measurements to the ones published by R. Mckinven et al. (2023a) (see the left column in Figure 7), FRB 20190303A shows a secular temporal RM trend where the RM values increase by several hundreds of units. Given that the Galactic RM contribution at this position is +21 ± 5 rad m⁻², we can be certain that a sign change (from negative to positive) of the RM took place between 2022 and 2023.

Zoom In Zoom Out Reset image size

FRB 20191106C. This is a relatively active repeater with 11 baseband events between 2019 November and 2022 May. Four events were not recorded properly, hence only seven are presented in Table 2 and Figure 7. There are significant RM variations of several hundred RM units, showing a clear secular trend. There is a low level of linear polarization (red line) in all available waterfalls.

FRB 20200120E. This is the FRB in a globular cluster associated with the galaxy M81 (M. Bhardwaj et al. 2021; F. Kirsten et al. 2022). We have eight events with baseband data recorded between 2022 January and 2023 October. Only three bursts with significant RM detections are shown in Table 3. One of these was previously announced by M. Bhardwaj et al. (2021) but not with complete polarization information. We note that the RMs of this source do not vary a lot (σ(RM) = 2.0 rad m⁻²) throughout the few years of data we have, which is consistent with the expectation from the sparse ISM environment of a globular cluster.

FRB 20200929C. All seven bursts from this repeater have low ∣RM∣ values, although given the relatively high L/I fraction, the detected RM is unlikely to be (purely) instrumental. Given that the Galactic RM contribution at this position is −18 ± 3 rad m⁻², we can say that there are at least two RM sign changes in the corrected, extragalactic RM component between 2020 November and 2022 February (see third column in Figure 7).

FRB 20201124A. 15 events with baseband data were recorded for this repeater between 2021 March and September. Some of these events were described by A. E. Lanman et al. (2022) although polarization information was not included. Only the three bursts with significant RM detections are shown in Table 3. CHIME/FRB has not detected this repeater after 2021 September despite regular operations, but if we combine the CHIME/FRB measurements with those in the literature (G. H. Hilmarsson et al. 2021; P. Kumar et al. 2022; H. Xu et al. 2022), we see that the CHIME/FRB measurements are consistent with those of the other telescopes, showing secular temporal variations where the RMs fluctuate for hundreds of units (see right column of Figure 7). The CHIME/FRB data seem to have a lower L/I than data from the other telescopes, which might be because these other facilities observe at higher frequency than CHIME/FRB and their data suffer less Faraday depolarization. We also note that burst #3 (TNS name 20210526D) shows possibly varying L/I as a function of time during the burst (last panel in Figure 4).

FRB 20201130A. Six baseband events were recorded between 2020 December and 2021 March, although one of them was not saved properly and thus cannot be analyzed. The RM of this source is stable at around 180–190 rad m⁻² within this period. A range of morphologies was observed. All five events shown in Figure 1 have a high degree of linear polarization (red line).

We further study the RM distribution of all 26 repeaters with more than one RM value. In Figure 9, we compare the standard deviation of the RM excess (observed RM_FDF – RM_MW) with the maximum time span of measurements (the time of the newest data point minus the oldest), as well as the averaged RM excess magnitudes. From the left panel, we can see that the majority of these repeaters have at least one year of data, albeit with varying degrees of cadence (refer to Figure 8). RM measurements taken on short time spans are likely probing the same Faraday active media, whereas RM measurements taken a long time apart could give insights into Faraday active media of much larger physical sizes. Intuitively, we would expect RMs taken near in time to show small variations (σ(RM)), whereas larger RM variations could be expected over longer periods of time. This is mostly consistent with what the left panel is showing. However, we do see that a number of FRB repeaters have small RM variations of a few tens of units irrespective of the time span (bottom half of the plot area). These are likely the repeaters associated with stochastic RM variations, and the observed behavior suggests a uniform environment with limited turbulence. On the top half of the plot area, a number of exceptions show much higher levels of RM variations. This division is perhaps more apparent from the right panel of Figure 9, where these repeaters seem to separate into two categories. Those to the right of the dotted line, with a low ratio of σ(RM)/, likely come from stable magnetoionic environments. On the other side (to the left of the dotted line), these sources have a high ratio of σ(RM)/ and likely reside in dynamic RM environments. The cases of FRBs 20190520B, 20121102A, 20181119A, 20190303A, and 20201124A have already been discussed in the literature (D. Michilli et al. 2018; R. Luo et al. 2020; R. Anna-Thomas et al. 2023; R. Mckinven et al. 2023a). This work adds FRB 20191106C to this subgroup of FRBs with high , which are thought to reside in dynamic RM environments, where the extreme circumburst environment contributes to the high-level RM variations. It is also remarkable that the top boundary of the trend in the right panel is very close to a 1–1 relation between σ(RM) and . This is additional evidence that for these repeaters in dynamic RM environments, the majority of the RM variations are likely due to contributions from the local environment (which we expect to vary in time), instead of from the ISM of the host galaxy (which we expect to be stable with time). We could also conclude that the sources located close to the 1–1 relation must have a low fractional contribution from the ISM of the host galaxy to the RM temporal variation, as the majority of the temporal variations are already accounted for by the evolving environment near the FRB source.

Zoom In Zoom Out Reset image size

In general, practically all FRB repeaters demonstrate temporal RM variations that have larger amplitudes than those found for Galactic pulsars, which are typically no more than a few rad m⁻² (H. M. Wahl et al. 2022). There are certain pulsar systems that show a comparably high magnitude of RM variations (see right panel of Figure 9). Examples include pulsars near the Galactic center close to Sagittarius A* (e.g., PSRs J1745−2900, J1746−2849, J1746−2850, J1746−2856, and J1745−2912; G. Desvignes et al. 2018; F. Abbate et al. 2023), “windy”²⁵ pulsar binaries (e.g., PSRs J2051−0827 and B1259−63; T. W. Connors et al. 2002; S. Q. Wang et al. 2023), and a redback²⁶ system (PSR B1744−24A; D. Li et al. 2023). The FRB repeaters from evolving RM environments appear to agree quite well with the pulsars from (eclipsing) binary systems, but not so well with the solitary ones near the Galactic center. Given the similarities, it might be suggestive to say that coronal winds from a binary stellar companion are a potential explanation of the circumburst environments of these FRB systems with a high ratio of σ(RM)/, and that these repeating FRBs probably did not come from the Galactic center of their host galaxies, assuming similar host ISM properties to our Milky Way. This is indeed the case for the three high-σ(RM)/ sources that have localizations from very long baseline interferometry, namely FRB 20190520B at 5 kpc from the center of its host galaxy (C. H. Niu et al. 2022), FRB 20180916B with a 4.7 kpc offset (B. Marcote et al. 2020), and FRB 20220912A with a 0.8 kpc offset (D. M. Hewitt et al. 2024). In addition, in the cases of the Galactic center pulsars, large variations in DM were also reported (F. Abbate et al. 2023), which is not the case for the FRB repeaters.

Beyond stochastic variations, some sources show deterministic trends that are not captured in Figure 9 and that can cause an overestimation of stochastic variation. To investigate this, we have defined four simple empirical models for the RM evolution: (1) a constant RM (one parameter), (2) a linear trend (one parameter), (3) a model with constant RM episodes connected by a linear trend (four parameters, motivated by FRB 20180916B), and (4) a periodic model (four parameters, motivated by FRB 20190520B). We have used least-squares fitting to fit all four models to the 19 sources with more than two RM measurements and selected the best-fit model using the Akaike information criterion (see Section 2.4 in K. Burnham & D. Anderson 2002), modified for small sample sizes.

In this framework, 12 FRBs are best fit by a constant RM. FRBs 20180301A and 20191106C are best fit by a linear trend with slopes of 0.24 and −2.0 rad m⁻² day⁻¹ over ≥418 and ≥413 days, respectively. FRBs 20121102A, 20180916B, and 20190303A are best fit by a model with constant episodes connected by a linear trend with slopes of −90, 0.19, and 0.72 rad m⁻² day⁻¹ over 378, 323, and ≥1275 days, respectively. FRB 20190520B is best fit by a periodic model with a period of 350 days (note that this does not imply it is in an orbit) and FRB 20201124A shows too much temporal variation to produce a meaningful fit using these four simple models.

Indeed, the standard deviation of the residual RM after removing the best-fit models goes down and all sources show stochastic changes σ(RM) ≲ 80 rad m⁻², except for FRBs 20121102A, 20181119A, and 20190520B, which still show σ(RM) rad m⁻². Interestingly, the five best-fit linear trends normalized by the median RM of the respective sources all fall within 0.1%–0.3% day⁻¹. The time evolution could be further quantified using correlation or structure functions, but we defer this investigation to future work.

In this work, we report the change of signs in RMs for two FRBs, namely FRBs 20200929C and 20190303A. Together with one other previously published case of FRB 20190520B (R. Anna-Thomas et al. 2023), there are now three examples (see red star symbols in Figure 9) in this subgroup. A possible fourth source is FRB 20180916B (see Figure 5), whose low Galactic latitude of 37 implies high uncertainty in its Milky Way RM contributions (RM_MW = −94 ± 45 rad m⁻²), which may or may not indicate that the RM variations we see in FRB 20180916B fluctuate around zero. We note that these sources have the highest ratios of σ(RM)/, which are all very close to 1:1 (yellow line), occupying the further left region in the right panel of Figure 9.

This subsample of FRB repeaters with RM sign changes is a stark contrast to the Galactic pulsar population, where out of the over 3000 known pulsars, an RM sign change has only been reported in two systems with windy binaries, namely the black widow binary PSR J2051−0827 (S. Q. Wang et al. 2023) and the binary system PSR B1259−63, which has a Be-star companion (T. W. Connors et al. 2002; S. Johnston et al. 2005). In both pulsar systems, the RM sign change is attributed to the change in magnetic field strength along the line of sight due to binary orbital motions, and indeed the RM temporal variations of these two pulsar systems show orbital modulations that match the orbital periods. The fact that RM sign changes have not been seen more commonly in other windy pulsar binaries might be mostly due to the detection limit of the telescope. D. Li et al. (2023) have shown that RM changes of thousands of rad m⁻² can take place rapidly in these binary systems, and so depolarization could occur in these pulsar observations where an integration time of tens of seconds was typically used. In contrast, this is less of a problem for the much brighter FRB signals where the RM can be measured from each individual burst. Nonetheless, the four repeating FRBs with RM sign changes do not (yet) show any orbital modulations in their RM variations that resemble those observed in the windy pulsars. Further monitoring of the RM variations of these four repeaters would be insightful in understanding any potential relationship between them and the pulsar systems.

Recently, A. Pandhi et al. (2024) compared the polarization properties of 118 nonrepeating FRBs (89 polarized and 29 unpolarized) to 13 repeating sources. They find that the linear polarization fractions and rotation measure distributions of repeaters and nonrepeaters are consistent with being drawn from the same population; however, they present only marginal evidence for repeaters originating from more magnetized environments. In this section, we update their analysis comparing repeaters and nonrepeaters by adding the 75 bursts from 28 repeating FRBs presented in this work. For brevity, we do not detail the full methodology and statistical techniques involved; instead, we refer the reader to A. Pandhi et al. (2024) for this information.

In the top panel of Figure 10, we show the cumulative distribution function (CDF) of the linear polarization fraction for repeaters in blue and nonrepeaters in red. The CDF for nonrepeating sources is computed using the Kaplan–Meier method (E. L. Kaplan & P. Meier 1958), which also accounts for upper limits on L/I for unpolarized bursts, and the red shaded region is the 95% confidence interval. The blue shaded region shows the full range of L/I measurements across bursts for each repeating source. The middle panel of Figure 10 shows the CDFs for the RM with the Milky Way foreground subtracted (∣RM_EG∣) for repeaters and nonrepeaters, respectively. Finally, the bottom panel of Figure 10 depicts the CDFs for the observer-frame lower limit on the average line-of-sight amplitude of magnetic field strength (∣β∣), where and DM_EG is the extragalactic component in the DM.

Zoom In Zoom Out Reset image size

Applying the Kolmogorov–Smirnov (KS; N. Smirnov 1948; A. Kolmogorov 1956) and Anderson–Darling (AD; T. Anderson & D. Darling 1954) tests to each pair of CDFs presented in Figure 10, we find no evidence for a dichotomy between repeating and nonrepeating FRBs in terms of their L/I using Kendall’s τ test (M. G. Kendall 1938) or their ∣RM_EG∣ using the KS and AD tests. We find marginal evidence for a dichotomy in the repeater and nonrepeater ∣β∣ distributions (with p values of 0.033 and 0.015 with the KS and AD tests, respectively), consistent with the results from A. Pandhi et al. (2024), who used a smaller sample of repeating FRBs.

To test whether the DMs and RMs in our FRBs originate from the same type of media along the line of sight, we analyze the correlation between the RM with the Milky Way foreground subtracted (∣RM_EG∣) and the foreground-subtracted DM (DM_EG). A plot of ∣RM_EG∣ versus DM_EG for all CHIME/FRB observed repeaters and nonrepeaters with polarimetric measurements to date, including the 28 repeating sources in this paper, is presented in Figure 11. Mirroring the analysis by A. Pandhi et al. (2024), we use Spearman’s rank correlation coefficient test (C. Spearman 1904) to gauge the extent of a possible correlation between and in our expanded repeater data set. We find a Spearman rank p-value of 0.804 (i.e., no statistical significance) when evaluating the correlation across all polarized repeating FRBs detected with CHIME/FRB. We note that in the literature, A. Pandhi et al. (2024) also find no correlation for repeaters but they did find a marginal correlation between DM_EG and ∣RM_EG∣ in their nonrepeating FRB samples. Meanwhile, M. B. Sherman et al. (2023) point to a correlation in RM_host versus DM_host for both repeaters and nonrepeaters in their sample of localized FRBs where redshift information is available and so the host contributions to RM and DM are estimated separately.

Zoom In Zoom Out Reset image size

One possible explanation for the lack of a correlation in repeating FRBs, and for the marginal dichotomy in ∣β∣, is that the media contributing the bulk of the RM may be distinct from the media primarily contributing to the DM for repeating FRBs. As suggested by A. Pandhi et al. (2024), repeating FRBs might be embedded in more strongly magnetized local environments (which contribute significantly more to the total RM than to the total DM) than nonrepeating FRBs. In addition, the lack of redshift information (and hence a lack of precise knowledge of the host and local environment) for the FRBs in the samples used in this study may also contribute to the lack of observed correlation.

We note that in this analysis, we have restricted the comparison to CHIME/FRB-detected bursts only in order to limit bias due to the instrumental properties of different telescopes. This means that our sample does not include some of the repeaters with the highest σ(RM)/ ratios (e.g., FRBs 20190520B and 20121102A, and only very few bursts from FRB 20201124A). Intrachannel depolarization could be a reason why CHIME/FRB is less likely to detect high-σ(RM)/ sources than telescopes that operate at higher observing frequencies (e.g., FAST and ASKAP). Another possible reason why CHIME/FRB might not have detected these sources is scattering, although based on the scattering measurements from A. P. Curtin et al. (2024), we do not observe any correlation between σ(RM)/ and the scattering timescale among the events presented in this paper. As suggested in Section 4.1, if indeed there are two categories of magnetoionic environments and if the high-σ(RM)/ repeaters are primarily the ones from more dynamic environments, the actual ∣β∣ and DM–RM dichotomy between nonrepeaters and repeaters might have been more significant.