The FRB 121102 host is atypical among nearby FRBs

We search for host galaxy candidates of nearby fast radio bursts (FRBs), FRB 180729.J1316+55, FRB 171020, FRB 171213, FRB 180810.J1159+83, and FRB 180814.J0422+73 (the second repeating FRB source). We compare these candidates with that of the first repeating FRB source, FRB 121102, in absolute magnitudes and the expected host dispersion measure $\rm DM_{host}$. We find that the hosts of nearby FRBs are not similar to that of FRB 121102. They tend to have either a smaller $\rm DM_{host}$ or be fainter. Based on the host galaxy star formation rate (SFR), radius, and offsets, we also estimate the contribution of $\rm DM_{host}$ from the host of long gamma ray bursts (LGRBs) and superluminous supernovae (SLSNe), the proposed progenitor systems of FRB 121102. The FRB 121102 host is consistent with those of LGRBs and SLSNe, whereas the nearby FRB host candidates either have a smaller $\rm DM_{host}$ or are fainter than the hosts of LGRBs and SLSNe. In order to avoid the uncertainty in estimating $\rm DM_{host}$ due to the line-of-sight effect, we propose a galaxy-group-based method to estimate the electron density in the inter-galactic regions, and hence, $\rm DM_{IGM}$. The result strengthens our conclusion that the host galaxies of nearby FRBs, at least for FRB 180729.J1316+55, FRB 171020, and FRB180814.J0422+73, are not similar to that of FRB 121102 or the typical hosts of LGRBs and SLSNe. We conclude that the host galaxy of FRB 121102 is atypical and LGRBs and SLSNe are likely not the progenitor systems of at least most nearby FRB sources.


Introduction
Fast radio bursts (FRBs) are bright objects in radio, with durations of a few milliseconds (Lorimer et al. 2007; Thornton et al. 2013;Spitler et al. 2014;Petroff et al. 2016, see Lorimer 2018 for a review). The values of their dispersion measure (DM), an indicator of the electron column density along the line of sight, are much larger than the predicted values from the Milky Way galaxy, so they are expected to be of an extragalactic origin.
The origin of FRBs is highly debated. It is known that at least some FRB sources produce repeating bursts (Spitler et al. 2016;CHIME/FRB Collaboration et al. 2019a). These FRBs are usually explained within the "intrinsic" models that invoke young pulsars (Connor et al. 2016;Cordes & Wasserman 2016;Katz 2016), or magnetars (Beloborodov 2017; Kashiyama & Murase 2017;Metzger et al. 2017Metzger et al. , 2019, with the ultimate energy coming either from the spindown power or the magnetic power of a neutron star. Alternatively, some "extrinsic" models invoking the kinetic energy of an external source (e.g., the socalled "cosmic comb" model, Zhang 2017Zhang , 2018b or the gravitational energy of an external object (e.g., asteroids hitting neutron stars; Geng & Huang 2015;Dai et al. 2016) have also been discussed in the literature. It is possible that not all FRB sources repeat (Palaniswamy et al. 2018;Caleb et al. 2019). If this is the case, there might be FRBs produced from catastrophic events, such as compact star mergers (Piro 2012;Kashiyama et al. 2013;Totani 2013;Liu et al. 2016;Wang et al. 2016b;Zhang 2016Zhang , 2019Dai 2019) and collapse of supramassive neutron stars to black holes (Falcke & Rezzolla 2014;Zhang 2014).
The extragalactic origin of FRBs is confirmed by the precise localization of the first repeating FRB 121102 (Spitler et al. 2014(Spitler et al. , 2016Marcote et al. 2017) and the identification of its host galaxy Chatterjee et al. 2017;Kokubo et al. 2017;Tendulkar et al. 2017). The host galaxy of FRB 121102 is an irregular, low-metallicity dwarf galaxy. FRB 121102 resides in the bright star-forming region in the galaxy. The properties of the host and the subgalactic localization of the source are similar to those of long gamma ray bursts (LGRBs) and superluminous supernova (SLSNe), some of which have been suggested to leave behind rapidly spinning magnetars. As a result, young magnetars born from massive star core collapse events that produced LGRBs or SLSNe are regarded as the leading candidates to power FRBs, and it has been expected that the host galaxy of FRB 121102 should be typical for FRB sources Nicholl et al. 2017).
The search for host galaxies of other FRBs has been carried out. FRB 150418 was proposed to be associated with a fading radio transient, which is located in an elliptical galaxy (Keane et al. 2016). However, the association is not secure because the radio counterpart is a radio persistent source with significant variability (Akiyama & Johnson 2016;Vedantham et al. 2016;Williams & Berger 2016;Johnston et al. 2017). Mahony et al. (2018) searched for the host galaxy of FRB 171020 with a small DM (which means it is nearby) and found a host candidate ESO 601-G036. It is a lowmetallicity Sc galaxy at redshift z=0.00867, which is similar to that of FRB 121102. However, the chance coincidence probability is quite large, and the allowable host DM of FRB 171020 is on the lower end of FRB 121102.
So it is unclear whether FRBs in general (both repeating and nonrepeating ones) have host galaxies and subgalactic environments similar to those of FRB 121102. 6 We intend to investigate this problem by searching for host galaxy candidates of nearby FRBs (those with small DMs) in this paper. We define our nearby FRB sample in Section 2. To prepare for the host DM estimation of the candidates, we propose a galaxy-group-based method to estimate DM IGM in Section 3. We then search for the nearby FRB host candidates, and compare them with the host of FRB 121102 in Section 4. We also estimate the host DM values of LGRB and SLSNe host galaxies, and compare them with those of our host candidates as well as FRB 121102 in Section 5. We draw the conclusion that the FRB 121102 host is atypical and rare. The results are summarized in Section 6 with some discussion. The following cosmological parameters have been adopted: (Dunkley et al. 2009).

Sample Selection
We use the DM values of FRBs to select nearby FRBs. We decompose the total observed DM into four terms: where DM MW is the contribution from the Milky Way disk, which is estimated using the NE2001 (Cordes & Lazio 2002) or YMW16 (Yao et al. 2017) models constructed with the observed pulsar DM data; DM halo is the contribution from Milky Way halo, which is estimated to be 30 pc cm −3 in Dolag et al. (2015) or 50-80 pc cm −3 in Prochaska & Zheng (2019) from simulations-to be conservative, we used 30 pc cm −3 for our estimation; and DM IGM and DM host are the contributions from the intergalactic medium (IGM) and from the host galaxy, respectively. The latter also includes the contribution from the FRB local environment. We would like to use DM IGM to constrain the distance, and investigate DM host in this paper. We select the FRBs with the excess DM,  Table 1.

IGM DM
In order to investigate the allowable redshift range, and estimate the host DM, a relation between redshift and DM IGM , i.e., DM IGM =fz, is usually applied (Ioka 2003;Inoue 2004;Deng & Zhang 2014;Zhang 2018a), with f in the range of ∼850-1200 pc cm −3 . However, the relation between DM IGM and redshift suffers from large uncertainties. Cosmological simulations reveal that the line-of-sight fluctuations dominate the DM IGM uncertainties. The difference resulting from different lines of sight can be substantial (McQuinn 2014; Jaroszynski 2019; Pol et al. 2019).
In order to eliminate the line-of-sight uncertainty, here we propose to directly use the observed galaxy group information to estimate the cosmic density field, and hence DM IGM , along the lines of sight of FRBs in our sample.
With the observed galaxy groups, Wang et al. (2009Wang et al. ( , 2016a) developed a halo-domain method to reconstruct the cosmic density field. However, Wang et al. (2016a) only covered the SDSS DR7 region, which contains only one of our object, FRB 180729.J1316+55. We therefore use a nearly all-sky galaxygroup catalog in Lim et al. (2017). To the first-order estimate, we adopt the empirical Navarro-Frenk-White (NFW) dark matter density profile (Navarro et al. 1997) to reconstruct the cosmic density field.
There are four galaxy group catalogs in Lim et al. (2017), which are produced with the galaxy catalogs from the 2MASS redshift survey (2MRS), 6dF Galaxy Survey (6dFGS), Sloan Digital Sky Survey (SDSS), and 2dF Galaxy Redshift Survey (2dFGRS). Among them, 2MRS has 91% sky coverage, nearly all except the galactic plane. Most of our FRBs are only covered by 2MRS so we use the 2MRS catalog here. However, redshift is not a good indicator of distance for nearby galaxies. We thus update the distance of the galaxies with the nearby galaxy catalog of Karachentsev et al. (2013), and then propagate the updated distances to the corresponding groups.  6 During the review process of this paper, the host galaxies of two more FRBs, FRB 180924 (Bannister et al. 2019) andFRB 190523 (Ravi et al. 2019) were reported, which are different from the host of FRB 121102. This is consistent with the conclusion of our paper. 7 www.frbcat.org Note that 2MRS is not complete for z>0.033 (Tully 2015). The DM IGM with z>0.033 is only a lower limit of DM. For each of the galaxy groups, Lim et al. (2017) estimated their dark matter halo masses, log M h (M e h −1 ). For each dark matter halo, we estimate the dark matter density profile as where r is the distance from the center. R s =r 200 /c 200 is a scale radius. r 200 is the radius where the average density of the halo is 200 times of cosmic critical density, ρ c =3H 2 /8πG; c 200 is the concentration of the halo, depending the halo mass and redshift, and we use - . However, the observational group catalog is flux limited. Wang et al. (2009Wang et al. ( , 2016a  3 for h=0.724) of the universe, where Ω m is the normalized mass density. We thus limit our galaxy groups to those with halo masses larger than  -M 10 h 12 1 , and consider ρ=0.2ρ m as the background density in the intergalactic space in addition to the NFW density profile for the groups.
We convert the dark matter mass density ρ to baryon mass density by the ratio between Ω b and Ω dm , the normalized baryon and dark matter mass densities. If the baryon in the IGM traces dark matter and is composed of totally ionized hydrogen and helium, then the free electron number density n e can be related to the dark matter density ρ by because their positional uncertainties are large, it is still possible that their lines of sight indeed pass through galaxy groups or even the center of the groups. In such cases, their DM IGM values are boosted a lot, even higher than 100 pc cm −3 . The line of sight of FRB 180814.J0422+73, the second repeating FRB, goes through many galaxy groups within z=0.02. Its DM IGM is larger than the value from the DM IGM =855z relation even if the 2MRS catalog is incomplete. Its DM IGM reaches DM exc around z=0.01, indicating that its host is likely extremely nearby.
For comparison, We have also examined FRB 121102. However, FRB 121102 is too close to the Galactic plane, with a galactic latitude −0°.2. This region is avoided by most galaxy group catalogs. So, we are unable to constrain its DM IGM .
To compare with other cosmological results, we calculate DM IGM for different redshifts and all sky, with 360 bins in R. A., and 180 bins in decl. The distribution of the DM IGM as a function of redshift z is plotted in the lowest right panel of Figure 1. The black thick curve indicates the median value for each redshift, and the gray curve presents its mean value. The orange and yellow regions show the 68% and 90% confidence levels, respectively. The black dashed curve is again the DM IGM =855z relation. It turns out that the median and mean values bracket the DM IGM =855z relation with z<0.033, and follows nearly the same shape. It indicates that our result is generally consistent with previous rough estimation by Zhang (2018a), and our 2MRS galaxy group sample is generally complete at z<0.033. However, our results flatten when reaching redshift 0.04 due to the incompleteness of 2MRS at higher redshifts. Thus, our estimation should be considered as the lower limit for z>0.033.
Even without knowing the true redshift, our analysis gives a relation between DM IGM and z for individual FRBs with certain uncertainties. With such a preparation, we can then estimate the values of the host DM, i.e., DM host =DM exc −DM IGM , of each FRB for different redshifts. For z>0.033, our derived DM host can be regarded as the upper limits. These derived values can be then compared with that of FRB 121102 (see Section 4).

Host Galaxy Candidates
We search for host galaxy candidates using R.A., decl. of each FRB and its 99% errors. For FRB 171020 and FRB 171213, the localization probability images provided by Shannon et al. (2018) are employed. Since our FRBs are expected to be nearby, we first explore the Galaxy List for the Advanced Detector Era (GLADE) catalog (Dálya et al. 2018). It is a nearby galaxy catalog aiming at providing host galaxy candidates to gravitational wave events. It combines the galaxies in the Gravitational Wave Galaxy Catalog  In order to be more complete, we also explore the extended sources in the Pan-STARRS catalog 11 Flewelling et al. 2016). Following the menu of Pan-STARRS, we select objects in the StackObjectThin database table, exclude spurious sources by requiring ndetections >1, and select Pan-STARRS galaxies by requiring mag PSF -mag kron >0.05. We find and delete duplicate objects whose coordinates are off by 1″. We then assign redshifts from SDSS, 2MPZ, and NED to Pan-STARRS sources, allowing a coordinate offset by 3″.
We use R.A., decl., and their 99% errors to select host galaxy candidates. For FRB 171020 and FRB 171213, we use the localization probability images provided by Shannon et al. (2018) to select the galaxy candidates. The candidates with redshifts, spectroscopic or photometric, less than 0.15 are presented in Table 2. The numbers after the FRB names give the number of galaxies with redshift less than 0.15, the number of galaxies with redshifts, and the total number of galaxies  Laskar et al. 2011), where A g is the Galactic extinction in g band, D L is luminosity distance, and z is the redshift, λ is the observational effective wavelength, 4900 Å for PanSTARRS g band, λ 0 is the rest frame effective wavelength, 4300 Å for B band, and β is the index of the assumed power-law spectrum,   Table 2. For those without redshifts, we estimate their DM host and M B by assuming redshifts z=0-0.1, following the same method in the last paragraph, and then plot them as solid curves. To be clear, we only plot the brightest 50 candidates without redshifts for each FRB. There are many more galaxies fainter than what we presented. Candidates with photometric redshifts are presented as open stars, with dashed curves indicating different redshifts also.

FRB 180729.J1316+55
There are 695 extended sources within the positional region of FRB 180729.J1316+55. Among them, 59 have spectroscopic redshifts from SDSS. There are seven galaxies with spectroscopic redshifts less than 0.15. Two of them, SDSS J131613.66+553741.5 and 2MASS 13170558+5529488 have relatively large values of DM host . 12 The first one, SDSS J131613.66+553741.5, is a faint source within a big disk galaxy SDSS J131613.95+553749.5. It is likely a star-forming region in the galaxy. Its expected DM host is 35 pc cm −3 and 43 pc cm −3 for the NE2001 and YMW16 models, respectively. The second one, SDSS 131705.58 +552948.8, is an edge-on disk galaxy with a significant bulge. SED fitting gives stellar mass M * =3×10 10 M ☉ , SFR=0.007 M ☉ yr −1 (Chang et al. 2015). Its expected DM host are 20 pc cm −3 and 28 pc cm −3 for the NE2001 and YMW16 models, respectively. These two sources both have a smaller DM host than FRB 121102. Other host galaxy candidates have even smaller DM host than FRB 121102.

FRB 171020
There are 4974 extended sources within the error box of FRB 171020. Among them, 31 have redshift information. Four of them have spectroscopic redshifts smaller than 0.15, and 8 of them have photometric redshifts smaller than 0.15. The one with the lowest redshift, ESO 601-G 036, is the galaxy candidate proposed by Mahony et al. (2018). It has DM host =41 pc cm −3 and 54 pc cm −3 for NE2001 and YMW16 models, respectively. For most possible redshifts, the derived DM host is much smaller than that of FRB 121102. Only if the host galaxy is intrinsically very faint (so they are much closer) could its DM host reach the lower limit of FRB 121102 DM host . In this case, the host galaxy candidate should have an absolute magnitude similar to or larger (fainter) than that of FRB 121102. As shown in Figure 2, galaxies without LGRB and SLSNe host galaxies are denoted as gray and blue symbols. Filled dots are for the MW template, and diamonds are for the spherical electron density profile. redshift information may achieve DM host =40-60 pc cm −3 if they are extremely nearby.

FRB 171213 and FRB 180810.J1159+83
Both FRB 171213 and FRB 180810.J1159+83 have DM exc ∼90 pc cm −3 . It is possible to find a host galaxy candidate similar to that of FRB 121102. Also, they are out of the redshift range for our galaxy-group-based method for the z-DM IGM relation. We thus do not explore them in detail.

FRB 180814.J0422+73
FRB 180814.J0422+73 is the second repeating FRB (CHIME/FRB Collaboration et al. 2019a). There are only 50 extended sources found in Pan-STARRS, and 1 in GLADE in the error box. This is due to the smaller positional uncertainty compared with other objects. The brightest galaxy is the one found in GLADE, with a g-band kron magnitude 17.5 mag, and a 2MASS photometric redshift 0.078. The second brightest one is a point source with another fainter point source 1 7 away. It is quite likely spurious, so we do not show it in the plot. Other galaxies are more than one order of magnitude fainter than these two.
There are many galaxy groups near the line of sight of FRB 180814.J0422+73 for z<0.02. The host galaxy should have to be very nearby, if they have a DM host similar to that of FRB 121102. They should then be intrinsically very faint. As shown in Figure 2, the host of FRB 180814.J0422+73 has to be much fainter than −14 magnitude, more than 3 orders of magnitude fainter than that of FRB 121102, if a DM host similar to that of FRB 121102 is assumed. The DM host of FRB 180814.J0422 +73 must be very small (<7 pc cm −3 ), if its host is as bright as FRB 121102. In this case, the galaxy with redshift, 2MASS J042221.4+734710.2, is not the host, if its photometric redshift is correct.
Even if we use the empirical z−DM IGM relation, the conclusion is similar. If 2MASS J042221.4+734710.2 is the host galaxy of FRB 180814.J0422+73, the estimated DM IGM =67 pc cm −3 (Deng & Zhang 2014), indicating DM host ∼5 pc cm −3 DM. If 2MASS J042221.4+734710.2 is not the host, the host galaxy should be at least three orders of magnitude fainter than the host of FRB 121102, that is, M r >−14 mag. For comparison, LMC and SMC have absolute magnitudes −18.36 and −16.82, respectively. In this case, the FRB would be quite local although still extragalactic. Its isotropic energy would have to be two or three orders of magnitude smaller than typical FRBs, e.g., FRB 121102.
In general, we conclude that the host galaxies of nearby FRBs typically have small DM host values, or are intrinsically faint, much fainter than the hosts of LGRBs and SLSNe (Metzger et al. 2017). This is in contrast to the conclusion drawn from the FRB 121102 measurement (Kokubo et al. 2017;Tendulkar et al. 2017) and the statistical analysis of Yang et al. (2017). Future observations of more localized FRBs will test whether a small DM host is typical for nearby FRBs only or for most FRBs in general.

DM Contribution from the Host Galaxies of LGRBs and SLSNe
Due to the similarity of the FRB 121102 host with LGRB/ SLSNe hosts, FRBs are highly believed to be powered by magnetars born during LGRBs and SLSNe Kashiyama & Murase 2017;Metzger et al. 2017;Nicholl et al. 2017). We want to further explore whether the host galaxies of nearby FRBs are similar to those of LGRBs and SLSNe.
Host galaxies of LGRBs and SLSNe are generally starforming dwarf galaxies (Sahu et al. 1997;Bloom et al. 1998Bloom et al. , 2002Chary et al. 2002;Christensen et al. 2004;Savaglio et al. 2009;Krühler et al. 2015). If the galaxy electron density is known, the host galaxy DM contribution can be estimated based on the scale length r e , and the offset of the transient from the center of the galaxy r off (Bloom et al. 2002;Fruchter et al. 2006;Blanchard et al. 2016;. The free electrons in the interstellar medium are generally ionized by the death of massive stars. They are thus likely correlated to the star formation rate (SFR) and Hα emission (Reynolds 1977;Luo et al. 2018). We can thus estimate their electron density n e based on their Hα emission lines, or SFR. Since resolved optical emission is not always available, we test two possible distributions, i.e., the spherical Gaussian distribution and Milky Way-like distribution.
We obtained SFR, r e , r off and absolute magnitude of SLSNe from Lunnan et al. (2015), Schulze et al. (2018), andPerley et al. (2016), and those of LGRBs from . We then estimate the DM host from LGRBs/SLSNe-like host galaxies as follows.

Spherical Gaussian Distribution
LGRB and SLSN hosts are dwarf star-forming galaxies, which resemble SMC in many aspects. Following the treatment of SMC by Yao et al. (2017), we assume that the electron density follows where r e is the scale length of the galaxy. Since LGRBs and SLSNe both highly trace massive stars, it is reasonable to assume that they are in the disk plane. If the the host is face on, for a specific offset r off , one has According to Reynolds (1977), Hα surface density is a tracer of EM, i.e., where F Hα is in units of erg s −1 cm −2 , r e and r off are in units of arcseconds, r e,pc is in units of parsecs, and DM is in units of cm −3 pc.

Milky Way-like Distribution
We also consider a Milky Way-like electron density distribution as the template of a disk galaxy (Yao et al. 2017 Figure 2, for SLSNe and LGRBs, respectively.

Comparison between the Candidates and LGRB/SLSNe Hosts
For both spherical Gaussian distribution and MW-like distribution, the hosts of LGRBs and SLSNe contribute ∼100 cm −3 pc to DM. Only <10% of LGRBs have DM host <30 cm −3 pc. None of the SLSNe have a DM host <30 cm −3 pc. They are plotted in Figure 2 for comparison. Blue and gray colors are for SLSNe and LGRBs respectively. Dots and diamonds are for spherical and MW-like distributions, respectively. FRB 121102 has an estimated DM host in the range of 55-225 pc cm −3 . This is consistent with the estimated value of LGRBs and SLSNe. Also, the absolute magnitude of its host galaxy, −17, is consistent with the values for the LGRB and SLSN host samples.
Other FRB host candidates, on the other hand, are not consistent with the LGRB and SLSN host samples. The host galaxy candidates for FRB 180729.J1316+55, FRB 171020, FRB 171213, and FRB 180814.J0422+73 all have a smaller DM host than the lower limit of FRB 121102, 55 pc cm −3 . The host galaxy candidates of FRB 180814.J0422+73 do not overlap with either LGRBs or SLSNe at all. The host candidates of FRB 180729.J1316+55 overlap with 5 of the 37 LGRB hosts, but no SLSN host. It is located in the faint, low DM host corner of the LGRB/SLSN host distribution. FRB 171020, FRB 171213, and FRB 180810.J1159+83 pass through the DM host range of LGRB and SLSN hosts, so the possibility that their hosts are LGRB/SLSN-like is not ruled out. However, All of them are located within the very low end of the LGRB/SLSN DM host distribution. So, collectively, the probability that all the nearby FRB hosts are consistent with the LGRB/SLSN hosts is extremely low.

FRB 180924 and FRB 190523
During the review of this paper, two FRBs were located to their host galaxies. FRB 180924 was in a massive passive galaxy z=0.3214 (Bannister et al. 2019), and FRB 190523 was in a massive galaxy at z=0.66 (Ravi et al. 2019). Both host galaxies are unlike that of FRB 121102, supporting our conclusion that the host of FRB 121102 is atypical. On the other hand, those two host galaxies are also brighter than most of our host candidates.
In order to apply our method to these FRBs, we extend the galaxy group catalog to higher redshifts with Wen et al. (2018), which covers all of the sky except for the Galactic plane, extends to redshift 0.4, and has a median redshift 0.24. 479 galaxy groups within it are excluded because they are duplicated with the 2MRS galaxy group. The electron density and cumulative DM IGM of FRB 180924 and FRB 190523 are also presented in Figure 1. The derived DM IGM of FRB 180924 at redshift z=0.3214 is 121 pc cm −3 , and that of FRB 180924 at redshift z=0.66 is 339 pc cm −3 . However, the total halo mass range of the galaxy group sample in Wen et al. (2018) is [7×10 13 , 2×10 15 ] M ☉ , much larger than the mass threshold 10 12 M ☉ we applied. As a result, many galaxy groups are likely missed. Furthermore, our galaxy group catalogs do not extend to redshift z>0.4, so we are unable to constrain the 0.4-0.66 range for FRB 190523. As a result, the DM IGM obtained with our method should be considered as very loose lower limits for these two FRBs. By subtracting the DM MW  pc cm −3 , respectively. These two FRBs are also presented in Figure 2. The very loose DM host upper limits are also plotted. One can see that most host candidates of nearby FRBs are also much fainter than these two hosts. As these two hosts also differ from that of FRB 121102 in terms of SFR and offset between the FRB and the host, one can draw the conclusion that the FRB hosts are very diverse among bursts.

Conclusion and Discussion
We have searched the host galaxy candidates of nearby FRBs whose DM exc is below 100 pc cm −3 . Due to the selection criteria, their DM host are expected to be smaller than 100 pc cm −3 . The following conclusions can be drawn: 1. Not all FRBs reside in environments similar to FRB 121102. The existence of FRBs with DM host less than the lower limit of FRB 121102 DM host reveals that not all FRBs are located in environments similar to FRB 121102. The fact that the hosts of the recently localized FRB 180924 and FRB 190523 are also different from that of FRB 121102 and the host candidates studied in this paper strengthens our conclusion and suggests that FRB hosts are very diverse.