CHARACTERISTICS OF GAMMA-RAY LOUD BLAZARS IN THE VLBA IMAGING AND POLARIMETRY SURVEY

Throughout the rest of this paper, we will use the nonparametric Spearman test (e.g., Press et al. 1986) to look for correlations between the LAT-detected and non-LAT-detected objects. The Spearman correlation coefficient (ρ_s) has a range of 0 < |ρ_s| < 1. A high value of ρ_s indicates a significant correlation. This is a powerful test for statistical correlation, but it does not test an actual physical correlation. In order to be certain our correlations are physically significant, we make sure that a redshift selection effect is not adding a bias to our data.

Imagine a population of sources covering a wide range of redshifts in which the radio and γ-ray emission is not physically correlated, but in which there is a correlation of radio flux, and gamma-ray flux, with redshift. This is actually what one would naively expect because the more distant objects will be fainter. Such a population will show a strong correlation of radio flux with gamma-ray flux via the Spearman test, but this does not indicate a significant physical correlation between these two observables.

However, by investigating the relationship between radio flux density and redshift for the LAT-detected sources (see Figure 1) we can rule out a correlation between the two. We calculated the Spearman ρ_s values for the S_8.5–z relationship for the BL Lac objects and FSRQs separately. For the BL Lac objects, the ρ_s value is 0.0737, with a 79% probability that random sampling would produce this same ρ_s value. For the FSRQs, the ρ_s value is 0.0450, with a 74% probability of getting the same value by random sampling. Therefore, there is no significant correlation between the radio flux density and redshift for our objects.

Zoom In Zoom Out Reset image size

For completeness, we also tested the correlation between γ-ray flux and redshift. The ρ_s values are −0.319 for BL Lac objects and 0.104 for FSRQs, with the probability of getting the same values from random sampling of 21% and 45%, respectively. So, there is no significant correlation between γ-ray flux and redshift, either. See Figure 2 for a plot of γ-ray flux versus redshift.

Zoom In Zoom Out Reset image size

In Figure 3, we plot the LAT flux versus the flux density at 8.5 GHz from the CLASS. The LAT fluxes are broadband fluxes from 100 MeV to 100 GeV. The BL Lac objects have a large range in their flux density and γ-ray flux, but one can see a large fraction of them have relatively low radio flux density (S_8.5< 400 mJy) but high γ-ray flux. The FSRQ objects also have a very large range in both their radio flux density and their γ-ray flux, but the higher flux density FSRQs also tend to have higher γ-ray flux. Overall, there is a lack of strong γ-ray emitters among the weak radio sources, but the weak γ-ray emitters can be bright or faint at radio wavelengths. The median γ-ray flux is 3.5 × 10⁻⁸ photons cm⁻² s⁻¹ for BL Lac objects, 3.7 × 10⁻⁸ photons cm⁻² s⁻¹ for FSRQs, and 3.8 × 10⁻⁸ photons cm⁻² s⁻¹ for the radio galaxies and unclassified objects.

Zoom In Zoom Out Reset image size

The Spearman ρ_s values we found were 0.217 for the BL Lac objects, 0.318 for the FSRQs, and −0.413 for the radio galaxies and unclassified objects. For the BL Lac objects and radio galaxies/unclassified objects, the chances that random samplings would result in the same ρ_s values were greater than 16%, indicating a lack of correlation between radio flux density and γ-ray flux. For the FSRQs, the chance of a random sampling having the same ρ_s was only 1.7%, which indicates that there is a significant correlation between 8.5 GHz flux density and γ-ray flux for the FSRQs. However, when we exclude the brightest 10% of sources in the radio, we find that the remaining FSRQs no longer correlate. The weaker 90% of sources have ρ_s's of 0.135 for BL Lac objects and 0.159 for FSRQs, with the chances of random samples having the same ρ_s values of 18% for BL Lac objects and 33% for FSRQs.

The lack of correlation for BL Lac objects is counter to what has been reported in other studies. Kovalev et al. (2009) find a definite correlation for their MOJAVE sources at 15 GHz. Their radio data were taken closer to the LAT measurements, and the γ-ray blazars are suspected to have higher radio flux densities during or shortly after γ-ray emission (Pushkarev et al. 2010). However, their sample is only about 1/3 the size of ours and it is biased to sources with high radio flux density (>1.5 Jy). Very strong flux–flux correlations are found by Ghirlanda et al. (2010) at 20 GHz. They have a larger sample than ours with 230 LAT sources, about half of which are FSRQs. However, their sample is also mostly comprised of bright sources. More than half of their sources have S₂₀ > 500 mJy, compared to our sample where about 1/3 have S_8.5 > 500 mJy.

To see if we could reproduce the correlation seen by others, we applied the Spearman test to the 36 sources with 8.5 GHz flux densities above 500 mJy. This subsample is roughly the brightest third of our overall sample and contains 11 BL Lac objects and 25 FSRQs. At first, we did not break the subsample into optical types. The result is a ρ_s of 0.390. The chance of a random sampling having the same ρ_s value as our strong source subsample was only 1.9%, indicating that this is a significant correlation. When the subsample was broken up by optical type, the ρ_s's were 0.214 for BL Lac objects and 0.422 for FSRQs. The chances of random samples having the same ρ_s's were 69% and 3.7%, respectively. So we find that FSRQs exhibit a significant correlation between radio flux density and γ-ray flux but BL Lac objects do not, even when limited to the strongest radio sources. Any sample that contains mostly high radio flux density objects will automatically be dominated by FSRQs and will likely show a correlation between radio flux density and γ-ray flux.

The median flux density at 8.5 GHz is 247 mJy for the LAT BL Lac objects and 276 mJy for the non-LAT BL Lac objects. The LAT FSRQs appear to have significantly higher flux densities than the non-LAT FSRQs, with a median of 377 mJy for LAT FSRQs compared to a median of 238 mJy for non-LAT FSRQs. The LAT radio galaxies and unclassified objects also have higher 8.5 GHz flux densities than the non-LAT sources. The median flux density for LAT radio galaxies/unclassified objects is 207 mJy and the median flux density for non-LAT radio galaxies/unclassified objects is 119 mJy. The Kolomogorov–Smirnov (K-S) test probability that the LAT and non-LAT BL Lac objects belong to the same parent population is about 81%. The K-S test probability that the LAT and non-LAT FSRQs are related is only 0.3%. The radio galaxy/unclassified objects K-S test probability is 0.01%.

VIPS sources are classified as point sources (PSs), short jets (SJETs), long jets (LJETs), complex (CPLX), or compact symmetric object candidates (CSO) by the automatic classification scheme of Helmboldt et al. (2007). See Table 2 for the source classifications for both LAT and non-LAT sources in VIPS. From the classifications, it appears that SJETs and PS BL Lac objects are less likely to produce γ-rays. About 75% of the LAT BL Lac objects are classified as LJET, compared to only 45% for the non-LAT BL Lac objects. The major difference between LAT and non-LAT FSRQs is the lack of CSOs in the LAT sample. There is also a lack of CSOs among the LAT radio galaxies and objects without an optical classification.

Polarization is detected from the cores of nearly half (49 of 102) of the LAT sources in our sample. We define the core as the bright compact component at the base of the jet. We detected polarization in the core of 17 of the 40 BL Lac objects, 24 of the 50 FSRQs, and 8 of the 12 radio galaxies and unclassified objects. For the non-LAT sources, core polarization is detected in only 270 of the total 1018 sources. There are 10 of 24 non-LAT BL Lac objects, 158 of 479 FSRQs, and 102 of 515 radio galaxies or unclassified objects detected. The LAT FSRQs and radio galaxies/unclassified objects tend to have polarization in their cores more than their non-LAT counterparts. In contrast, the LAT BL Lac objects do not appear to be any more likely to have polarized cores than the non-LAT BL Lac objects. The distributions of fractional polarization in the cores for LAT and non-LAT FSRQs, and radio galaxies/unclassified objects are shown in Figure 4. The K-S test probabilities that these distributions were drawn from the same parent populations are 84.6% for the BL Lac objects, 39.4% for the FSRQs, and 76.9% for the radio galaxies/unclassified objects.

Zoom In Zoom Out Reset image size

Our polarization results are different than those reported by Hovatta et al. (2010) with their MOJAVE data at 15 GHz. They found that the median fractional polarization for the cores of LAT-detected sources was higher than for non-LAT sources. Their median values for their 2008–2009 observations are 2.5% for LAT-detected sources and 1.86% for non-LAT sources. Our overall median values are 3.5% for LAT sources and 4.4% for non-LAT sources. However, our result is not of high significance as the standard deviations are 2.5% for LAT sources and 3.4% for non-LAT sources. It should also be noted that our VIPS data are not contemporaneous with LAT observations. It is possible that polarization increases with γ-ray emission. In fact, the MOJAVE data indicate that the polarization levels were stronger during the LAT detections than in previous years (2002–2008), but only by a factor of 1.2.

We have obtained the brightness temperatures from automatic modelfits to source components as described in Helmboldt et al. (2007). For sources with two or fewer components we have further refined our modelfitting procedure by fitting to the visibility data directly (Taylor et al. 2007). Only image plane modelfitting was carried out in Helmboldt et al. (2007) due to the tendency for the automatic visibility modelfitting to go awry for complicated sources.

There is strong evidence that the brightness temperature for the cores of the LAT-detected sources may be higher than the γ-ray quiet population (see the FSRQ and radio galaxy/unclassified distributions in Figure 5). The BL Lac objects are the exception to this trend. The median core brightness temperature for BL Lac objects detected by LAT is about 2.6 × 10¹⁰ K, while the median for remaining non-LAT BL Lac objects is 2.7 × 10¹⁰ K. The K-S test probability for these two distributions coming from the same parent population is 93.6%. The median core brightness temperature for FSRQ objects in the LAT sample is 8.2 × 10¹⁰ K, while the non-LAT FSRQ median is only about 2.5 × 10¹⁰ K. The K-S test probability that the two FSRQ distributions come from the same parent population is very low (0.001%). The median core brightness temperature for LAT radio galaxies and unclassified objects is 3.9 × 10¹⁰ K, and for non-LAT objects the median is 1.0 × 10¹⁰ K. The K-S test probability that these distributions are drawn from the same parent sample is only 4.4%.

Zoom In Zoom Out Reset image size

Interestingly, γ-ray flux does not appear to have any relation to core brightness temperature. We applied the Spearman test to the γ-ray versus core brightness temperature distributions and found correlation coefficients of 0.093 for BL Lac objects, 0.111 for FSRQs, and −0.181 for radio galaxies and unclassified objects.

We have measured a mean opening half-angle by the following procedure (Taylor et al. 2007): We measure the separation of each jet component from the core along the jet axis (taken to be a linear fit to the component positions) and the distance of each component from the jet axis, i.e., x' and y' positions in a rotated coordinate system with the jet axis along the x'-axis. For each component, we measure its extent from its center along a line perpendicular to the jet axis using the parameters of its elliptical fit, and then deconvolve this using the extent of the Gaussian restoring beam along the same line. The opening half-angle measured from each component is then taken to be

$$$ \psi = {\rm arctan}[(|y^{\prime }|+dr)/|x^{\prime }|], $$$

where dr is the deconvolved Gaussian size perpendicular to the jet axis. After measuring this for each jet component, we average them to get a single value. This is only done for sources with more than two total components, (i.e., at least two jet components). Unfortunately, we can only make quantitative estimates for 30 of the LAT sources; 16 BL Lac, 9 FSRQ, and 5 radio galaxy or unclassified objects. The distributions of opening angles are shown in Figure 6. Although this is based on very small statistics there is evidence that the LAT source jets have unusually large opening angles. This is especially true for the FSRQ detected by LAT. The MOJAVE (Pushkarev et al. 2009) and TANAMI (Ojha et al. 2010) groups both reported larger opening angles for γ-ray loud blazars. Taylor et al. (2007) also found evidence for larger opening angles in EGRET sources.

Zoom In Zoom Out Reset image size

We found 10 LAT-detected objects with opening angles greater than 30°. The large opening angle objects are BL Lac objects and FSRQs, five of each. See Figure 7 for contour maps of these 10 sources.

Zoom In Zoom Out Reset image size

We define the core fraction as the ratio of flux density in the core compared to the total flux density at 5 GHz. A high core fraction can be an indication of a high Doppler factor. Taylor et al. (2007) reported that EGRET sources tended to have greater core fractions than sources not detected by EGRET. With the LAT sources, we find that this is not the case. The median core fractions are 0.868 for LAT BL Lac objects and 0.896 for non-LAT BL Lac objects, 0.939 for LAT FSRQs and 0.929 for non-LAT FSRQs, and 0.889 for LAT radio galaxies/unclassified objects and 0.886 for non-LAT radio galaxies/unclassified objects. The K-S test probabilities that the distributions were drawn from the same parent distribution are 26.4% for BL Lac objects, 16.1% for FSRQs, and 23.5% for radio galaxies/unclassified objects, respectively.

We can also compare the amount of bending in the jet on the parsec scale for the LAT sources compared to non-LAT sources. Unfortunately, as with the jet opening angle measurements, we are limited to only 30 LAT sources and the results are inconclusive.

We compared the distribution of jet length in LAT sources with non-LAT sources. There does not appear to be an appreciable difference between the two populations of FSRQ an radio galaxies/unclassified objects. The K-S test results are 63.3% for the FSRQ objects, and 64.6% for the radio galaxy and unclassified objects. Interestingly, the BL Lac objects in the LAT sample appear to have longer jets than the non-LAT BL Lac objects. The distributions for the BL Lac objects are shown in Figure 8. The K-S test probability that the two are drawn from the same distribution is 2.1%. The FSRQ and AGNs/other objects have essentially the same distributions.

Zoom In Zoom Out Reset image size

Unlike the core brightness temperatures, the jet brightness temperatures (formally, the brightness temperature of the brightest jet component) for the LAT and non-LAT distributions look fairly similar. The K-S test probabilities for the BL Lac object, FSRQ, and radio galaxy/other pairs of distributions are 34.2%, 92.1%, and 83.5%, respectively.

J09576+5522. With a core brightness temperature of only 9.7 × 10⁸ K, this FSRQ has a γ-ray flux of (106.5 ± 7.4) × 10⁻⁹ photons cm⁻² s⁻¹, nearly three times the median for FSRQs. Also known as [HB89] 0954+556, it has a redshift of 0.896 (Abazajian et al. 2004) and a 8.5 GHz flux density is 1.5 Jy. This FSRQ has been studied in the UV (Pian et al. 2005), radio, X-ray (Tavecchio et al. 2007), and was detected in the γ-rays by EGRET (Hartman et al. 1999). Tavecchio et al. (2007) noted that X-ray flux was only found at the terminal portion of the jet. The automated program initially reported the core brightness temperature for this object to be 1.2 × 10⁸ K. We suspected program did not generate an accurate model for this source. When we did the model fitting by hand in DIFMAP, we calculated a core brightness temperature of 9.7 × 10⁸ K. This is the lowest core temperature of all objects in our LAT sample. This object is featured in Figure 7 because its jet opening angle is about 60°.

J08163+5739. The automated program found that this BL Lac object had a core brightness temperature of only 4.16 × 10⁷ K. Again, when we did the model fitting by hand we calculated a core brightness temperature of 1.6 × 10⁹ K. Even with this higher core brightness temperature, this object is still the coolest BL Lac object in our LAT sample. This object has a γ-ray flux of (15.2 ± 0.3) × 10⁻⁹ photons cm⁻² s⁻¹. It is a compact and weak radio source with a radio flux density of 92 mJy at 8.5 GHz. There is no measured redshift for this object. It is also known as SBS 0812+578 and 1RXS J081624.6+573910. It has a small jet extending to the east. This object is featured in Figure 7 because its jet opening angle is approximately 52°.

J07464+2549. With a redshift of 2.979 (Abazajian et al. 2004), this FSRQ has a γ-ray flux (42.3 ± 7.9) × 10⁻⁹ photons cm⁻² s⁻¹. Its core brightness temperature is 5.07 × 10¹¹ K, which is well above the median for the LAT-detected FSRQs. It is also known as B2 0743+25 and RX J0746.4+2549. It is a relatively compact and bright radio source with a flux density of 731 mJy at 8.5 GHz (see Figure 9). It has a very small jet extending to the north. In 2005, the Burst Alert Telescope on board the Swift satellite detected a flare of 15–195 keV emission from this object (Sambruna et al. 2006).

Zoom In Zoom Out Reset image size

J08053+6144. With a redshift of 3.033 (Sowards-Emmerd et al. 2005), this FSRQ has the highest redshift among the LAT-detected objects. It has a γ-ray flux 36.0 ± 8.9 × 10⁻⁹ photons cm⁻² s⁻¹. Its core brightness temperature is 1.30 × 10¹¹ K, which is slightly higher than the median for LAT-detected FSRQs. This object is fairly bright in the radio with a flux density of 725 mJy at 8.5 GHz. It has a short jet extending to the southeast (see Figure 9).

Our sample contains six sources identified as radio galaxies, although in some cases there is controversy in this identification as described below.

J09235+4125. This object is also known as B3 0920+416 and GB6 J0923+4125. It has a γ-ray flux of (40.6 ± 10.0) × 10⁻⁹ photons cm⁻² s⁻¹ and a redshift of 0.028 (Falco et al. 1998). Its peak radio flux density is 120 mJy at 8.5 GHz. It has a short jet extending to the east (see Figure 9). The NASA/IPAC Extragalactic Database (NED) classifies this object as an FSRQ. However, this is probably an error. Falco et al. (1998) call it a "late-type galaxy."

J12030+6031. This object is also known as SBS 1200+608 and RX J1203.0+6031. It has a γ-ray flux of (44.8 ± 0.3) × 10⁻⁹ photons cm⁻² s⁻¹ and a redshift of 0.065 (Falco et al. 1998). Its peak radio flux density is 105 mJy at 8.5 GHz. It has a short jet extending to the south (see Figure 9). Véron-Cetty & Véron (2006) classify this object as a low ionization nuclear emission region (LINER) galaxy, sometimes referred to as a Seyfert 3.

J13307+5202. This radio galaxy is also known as CGRaBS J1330+5202. It has an upper limit to its γ-ray flux of 41.5 × 10⁻⁹ photons cm⁻² s⁻¹ and a redshift of 0.688 (Healey et al. 2008). It has a peak radio flux density of 127 mJy at 8.5 GHz and a jet extending to the southwest (see Figure 9). Sowards-Emmerd et al. (2005) list it as an unknown optical type.

J16071+1551. This object is also known as [HB89] 1604+159. It is an EGRET-detected source known as 3EG J1605+1553. It has a LAT γ-ray flux of (33.9 ± 7.7) × 10⁻⁹ photons cm⁻² s⁻¹ and a redshift of 0.496 (Adelman-McCarthy et al. 2008). It has a peak radio flux density of 214 mJy at 8.5 GHz and a jet extending to the east (see Figure 9). Véron-Cetty & Véron (2006) classify this object as a BL Lac object.

J16475+4950. This object is also known as SBS 1646+499 and 1RXS J164735.4+495001. It has a γ-ray flux of 33.3 ± 8.5 × 10⁻⁹ photons cm⁻² s⁻¹ and a redshift of 0.047 (Falco et al. 1998). It has a peak radio flux density of 131 mJy at 8.5 GHz. It has a very short jet extending to the southeast (see Figure 9). The spectrum in Marcha et al. (1996) indicates that this object is a Seyfert 1 blazar.

J17240+4004. This object is also known as B2 1722+40 and RX J1724.0+4004. It has a γ-ray flux of (47.1 ± 9.0) × 10⁻⁹ photons cm⁻² s⁻¹ and a redshift of 1.049 (Vermeulen et al. 1996). This is the brightest radio galaxy in our sample with a peak flux density of 323 mJy at 8.5 GHz. It has a long jet extending to the northwest (see Figure 9). Véron-Cetty & Véron (2006) tentatively classify this object as a BL Lac object. However, the spectrum in Vermeulen et al. (1996) does not support this classification. The spectrum has only narrow lines, so it cannot be an FSRQ, and the equivalent widths are greater than 5 Å, so it cannot be a BL Lac object.

Almost two-thirds of the BL Lac objects in VIPS are detected by LAT, and the only significant difference between the LAT and non-LAT populations is in their jet lengths. We have shown that the LAT BL Lac objects have longer jets than the non-LAT BL Lac objects (see Section 4.6). There are no differences in the fraction of polarized BL Lac objects or the distributions of the polarization. The LAT and non-LAT BL Lac objects have nearly identical core brightness temperature distributions. Why are the non-LAT BL Lac objects quiet in the γ-ray? It is possible that the non-LAT BL Lac objects are fundamentally different from the LAT BL Lac objects, but it must be in some subtle way. It seems more likely that all the BL Lac objects are emitting in the γ-ray, but some are emitting below the threshold for the LAT to detect.

Initially, we suspected that the non-LAT BL Lac objects might have been too far away to detect their γ-rays. The more distant BL Lac objects definitely have lower γ-ray flux, but it is clear that the non-LAT BL Lac objects are not at higher redshifts than the LAT BL Lac objects. While we only have redshift measurements for 16 of the LAT BL Lac objects and 15 of the non-LAT BL Lac objects, that is still enough to rule out redshift as the cause of LAT non-detection. The maximum measured redshift for the LAT BL Lac objects is 0.56. Nearly half (11/24) of the non-LAT BL Lac objects have redshifts lower than 0.56. Also, there is no strong correlation between redshift and γ-ray flux for the BL Lac objects (see Section 3.1).

The prevalence of LJET BL Lac objects is an interesting clue to this puzzle. The K-S test result of 2.1% for the jet length distributions (see Section 4.6) is only of modest statistical significance, but the fact that a larger percentage of the LAT BL Lac objects are classified as LJET than the non-LAT BL Lac objects (see Table 2) supports it. One would expect the jets to appear shorter if they are pointed more directly at us. However, if the γ-ray loud BL Lac objects have higher Doppler factors than their γ-ray quiet counterparts, that would indicate that the LAT BL Lac objects have more powerful jets than the non-LAT BL Lac objects. A more powerful jet will appear longer even when viewed at a small angle.

Because we find that the LAT BL Lac objects have longer jets than non-LAT BL Lac objects, the orientation angle for LAT BL Lac objects cannot be smaller than non-LAT BL Lac objects. The non-LAT BL Lac objects could have slightly lower jet speeds. This can explain why the LAT BL Lac objects have longer jets and why they are γ-ray loud. Unfortunately, this theory is not supported by recent work by Lister et al. (2009b), where they find that two non-LAT BL Lac objects have higher maximum jet material velocities than LAT BL Lac objects. However, they only had 10 BL Lac objects in their LAT-detected sample, so this is still an open question.

BL Lac objects are known to have high variability in their flux densities. The non-LAT BL Lac objects could currently be in low γ-ray luminosity states, but may enter states of enhanced γ-ray activity at a later time. It is possible that the jet material speed is related to this flux variability. Based on the number of BL Lac objects in our sample, the duty cycle would need to be about 2/3 over the course of one year.

The LAT FSRQs seem to be substantially different from the non-LAT FSRQs. Only about 9% (50/529) of the FSRQs in VIPS are γ-ray loud. This indicates that the LAT FSRQs are special in some way. Recall from Section 3.2 that γ-ray flux correlates with radio flux density for FSRQs, especially for high flux density sources. Also, in general the LAT FSRQs have significantly higher core brightness temperatures, with the notable exception of J09576+5522, and they appear to have larger opening angles. So, brighter FSRQs are more likely to produce γ-rays. We showed in Section 3.1 that γ-ray flux does not correlate with redshift, so we do not expect the LAT FSRQs to have lower redshift than their non-LAT counterparts. We explored the possibility that LAT FSRQs were at low redshift, but found that they cover a range similar to the non-LAT FSRQs.

The γ-ray flux from FSRQs can be explained with Doppler boosting, but requires a more dramatic difference than with the BL Lac objects. Lister et al. (2009b) reported that LAT-detected FSRQs had higher median jet speeds than LAT-detected BL Lac objects by more than a factor of two. It is also possible that there are two subclasses for FSRQs: γ-ray loud and γ-ray quiet. The difference in γ-ray detection rates for FSRQs could also be due to variations in the γ-ray emission, however, the duty cycle required (about 10%) is much more extreme than for the BL Lac objects.

A larger fraction of the LAT FSRQs have polarization in their cores compared to the non-LAT FSRQs. We detected polarization in the cores of 48% of the LAT FSRQs compared to 33% of the non-LAT FSRQs, a difference that is significant at about the 1σ level. This points to an intrinsic difference in γ-ray quiet versus γ-ray loud FSRQs. The higher polarization could indicate more organized magnetic fields, possibly pointing to a greater spin of the supermassive black hole (Meier 2001). A higher black hole spin would also lead to higher material velocities in the jets (Meier 1999).

There is a definite lack of CSOs among the LAT-detected objects. This is not surprising from a Doppler boosting point of view. For an object to appear symmetric, we need to see it close to "edge on" so we can see both jets. With the jets pointed nearly perpendicular to our line of sight, we would not expect to see much Doppler boosting in the radiation. However, Stawarz et al. (2008) suggested that CSOs should account for a "relatively numerous class of extragalactic sources" for Fermi due to ultrarelativistic electrons in the lobes of these compact objects. We have yet to detect any γ-rays from a CSO.

The LAT radio galaxies and unclassified objects are a puzzle. They do not appear to differ significantly from their non-LAT counterparts in any appreciable way. Their opening angles appear to be very similar to the non-LAT objects and they are not particularly bright in the radio. The LAT radio galaxies/unclassified objects may have higher core temperatures than the non-LAT objects. The one difference that stands out is that we detected polarization in the cores of 67% of the LAT radio galaxies/unclassified objects compared to only 20% for their non-LAT counterparts. Unfortunately, our current sample is too small to make any definite claims about these objects. There is also controversy about the optical types of several of the objects we label as radio galaxies.

The major differences between LAT and non-LAT sources are confined to the cores of the sources. The LAT BL Lac objects have stronger jets, suggesting higher ejection velocities from the cores. The LAT FSRQs have higher core temperatures and more frequently exhibit polarized cores. The LAT radio galaxies/unclassified objects are also more polarized than their non-LAT counterparts. We suspect that the γ-rays are being produced very near to the cores of these objects. To put a limit on the size of the γ-ray emitting region, we measured the sizes of the cores for several objects. Using all of the objects for which we have measured redshifts, the median upper bound on the core size is 4.54 pc with a standard deviation of 10.98 pc. However, a better estimate for the upper limit on the size of the emitting region should come from measurements of the objects nearest to us. Limiting ourselves to objects with z ⩽ 0.1, we have four BL Lac objects and three radio galaxies (a LINER, a Seyfert 1, and an object that may actually be an FSRQ). Using this subset of sources, we find a median upper bound on the core size of 0.58 pc with a standard deviation of 0.34 pc. Therefore, the γ-ray emission region must be within about 0.9 pc of the central engine.