MNRAS 527, 882–894 (2024) https://doi.org/10.

1093/mnras/stad3250
Advance Access publication 2023 October 25

A near magnetic-to-kinetic energy equipartition flare from the relativistic

jet in AO 0235 + 164 during 2013–2019
Whee Yeon Cheong ,1,2 Sang-Sung Lee ,1,2 ‹ Sang-Hyun Kim ,1,2 Sincheol Kang ,2
Jae Young Kim ,3,4 Bindu Rani ,2,5,6 Anthony C. S. Readhead,7 Sebastian Kiehlmann ,8,9
Anne Lähteenmäki,10,11 Merja Tornikoski,10 Joni Tammi,10 Venkatessh Ramakrishnan ,12,10
Iván Agudo,13 Antonio Fuentes,13 Efthalia Traianou,4,13 Juan Escudero,13 Clemens Thum,14
Ioannis Myserlis ,14 Carolina Casadio 8,9 and Mark Gurwell 15
1 Astronomy and Space Science Program, University of Science and Technology, 217 Gajeong – ro, Yuseong – gu, Daejeon 34113, Republic of Korea
2 Korea Astronomy and Space Science Institute, 776 Daedeok – daero, Yuseong – gu, Daejeon 34055, Republic of Korea

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

3 Department of Astronomy and Atmospheric Sciences, Kyungpook National University, Daegu 702-701, Republic of Korea
4 Max – Planck – Institut für Radioastronomie, Auf dem Hügel 69, D-53121 Bonn, Germany
5 NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA
6 Department of Physics, American University, Washington, DC 20016, USA
7 Owens Valley Radio Observatory, California Institute of Technology, Pasadena, CA 91125, USA
8 Institute of Astrophysics, Foundation for Research and Technology – Hellas, GR-71110 Heraklion, Greece
9 Department of Physics, University of Crete, GR-70013 Heraklion, Greece
10 Metsähovi Radio Observatory, Aalto University, Metsähovintie 114, FI-02540 Kylmälä, Finland
11 Department of Electronics and Nanoengineering, Aalto University, PO BOX 15500, FI-00076 AALTO, Finland
12 Finnish Centre for Astronomy with ESO (FINCA), University of Turku, Vesilinnantie 5, FI-20014 University of Turku, Finland
13 Instituto de Astrofı́sica de Andalucı́a – CSIC, Glorieta de la Astronomı́a, E-18008 Granada, Spain
14 Institut de Radioastronomie Millimétrique, Avenida Divina Pastora, 7, Local 20, E–18012 Granada, Spain
15 Center for Astrophysics|Harvard & Smithsonian, Cambridge, MA 02138, USA

Accepted 2023 October 13. Received 2023 October 5; in original form 2023 July 20

ABSTRACT
We present the multiwavelength flaring activity of the blazar AO 0235 + 164 during its recent active period from 2013 to 2019.
From a discrete correlation function analysis, we find a significant (> 95 per cent) correlation between radio and gamma-ray
light curves with flares at longer wavelengths following flares at shorter wavelengths. We identify a new jet component in 43 GHz
Very Long Baseline Array data that was ejected from the radio core on MJD 57246+ 26
−30 (2015 August 12), during the peak of the
2015 radio flare. From the analysis of the jet component, we derived a Doppler factor of δ var = 28.5 ± 8.4, a bulk Lorentz factor
of  = 16.8+ +1.07
−3.1 , and an intrinsic viewing angle of θv = 1.42−0.52 degrees. Investigation of the quasi-simultaneous radio data
3.6

revealed a partially absorbed spectrum with the turnover frequency varying in the range of 10−70 GHz and the peak flux density
varying in the range of 0.7−4 Jy. We find the synchrotron self-absorption magnetic field strength to be BSSA = 15.3+ 12.6
−14.0 mG
+10.6
at the peak of the 2015 radio flare, which is comparable to the equipartition magnetic field strength of BEQ = 43.6−10.4 mG
calculated for the same epoch. Additional analysis of the radio emission region in the relativistic jet of AO 0235 + 164 suggests
that it did not significantly deviate from equipartition during its recent flaring activity.
Key words: galaxies: active – BL Lacertae objects: individual: AO 0235 + 164 – galaxies: jets – gamma-rays: galaxies – radio
continuum: galaxies.

astrophysics. We present here a thorough analysis of the multifre-

1 I N T RO D U C T I O N
quency observation of a blazar, AO 0235 + 164, to probe the physical
Powered by accretion onto supermassive black holes, active galactic process(es) responsible for gamma-ray emission in the source.
nuclei (AGN) produce copious amounts of high-energy emission. AO 0235 + 164 is a blazar (a subclass of radio-loud AGN with
Understanding the physical conditions and processes responsible small viewing angles from the jet axis; Angel & Stockman 1980)
for the production of high-energy particles and the emission of at a redshift of z = 0.94 (Cohen et al. 1987). Even among blazars,
gamma-rays is one of the most challenging quests of high-energy AO 0235 + 164 is found to be extremely compact in radio with
no stable extended structure even at parsec-scale resolution (e.g.
Jorstad et al. 2017). On kiloparsec scales, Very Large Array images

at 1.4 GHz (Cooper, Lister & Kochanczyk 2007) and at 5 GHz (Frey
E-mail: [email protected]
et al. 2006) reveal a weak, extended structure north-northwest of
© 2023 The Author(s).
Published by Oxford University Press on behalf of Royal Astronomical Society. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium,
provided the original work is properly cited.
 AO 0235 + 164 883
the compact core. This could be a consequence of the extremely 2.1.2 Metsähovi 37 GHz light curve
small viewing angle found for this source (e.g. see Section 3.4 of
The 14 m telescope at the Aalto University Metsähovi Radio
this paper; Jorstad et al. 2017; Kutkin et al. 2018). A consequence
Observatory is used to monitor multiple AGNs and microquasars at
of such a small viewing angle is the enhancement of the Doppler
a centre frequency of 36.8 GHz. The detection limit of the telescope
factor δ which governs relativistic effects such as Doppler boosting.
at this frequency is on the order of 0.2 Jy under optimal conditions.
A large δ could amplify even small variations in the source into
Data points with a signal-to-noise ratio (S/N) less than 4 are handled
detectable variability. AO 0235 + 164 is known to display variation
as non-detections. The flux density scale is set by observations of
in brightness at various time-scales, from as short as a few hours in
DR 21 with NGC 7027, 3C 274, and 3C 84 being used as secondary
optical (Sagar et al. 2004) to as long as years in radio (Nieppola et al.
calibrators. Details of the data reduction and analysis may be found
2009). Quasi-periodicity was observed in X-rays with a short period
in Teraesranta et al. (1998). We use data obtained from 2013 January
of 17 d (Rani, Wiita & Gupta 2009). Longer periods of a few years
11 (MJD 56303) to 2019 December 30 (MJD 58847). The average
were found in both long-term radio and optical data (e.g. Raiteri
cadence for AO 0235 + 164 during this period is 8.5 d. The rms error
et al. 2001; Tripathi et al. 2021), hinting at the possible existence
during this period is typically nine per cent of the total flux density.
of jet precession. Space-Very Long Baseline Interferometry (Space-

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

VLBI) observations reveal extreme brightness temperatures of up to
1014 K (Frey et al. 2000; Kutkin et al. 2018), implying the presence 2.1.3 POLAMI light curve
of extreme physics within the jet.
The previous flaring activity from 2006 to 2009 was extensively The POLAMI (Polarimetric Monitoring of AGN at Millimetre
investigated during multiple campaigns (e.g. Raiteri et al. 2008; Wavelengths) programme (Thum et al. 2018; Agudo et al. 2018a,
Agudo et al. 2011; Ackermann et al. 2012). After a long period b)3 is a long-term programme to monitor the polarimetric properties
of quiescence, AO 0235 + 164 once again displayed multiple flares (Stokes I, Q, U, and V) of a sample of around 40 bright AGN
from radio to gamma-rays from 2014 to 2018 (e.g. Ciprini 2015; at 86 and 230 GHz frequencies with the IRAM 30 m Telescope
see also Fig. 1). In this paper, we present our analysis of total flux near Granada, Spain. The programme has been running since 2006
measurements in the radio (15–340 GHz) and gamma-ray bands from October, and it currently has a time sampling of ∼2 weeks. The XPOL
2013 to 2019 data collected from multiple monitoring programmes. polarimetric observing setup has been routinely used as described in
We investigate the jet parameters with high-resolution VLBI data Thum et al. (2008) since the start of the programme. The reduction
at 43 GHz obtained with the Very Long Baseline Array (VLBA). and calibration of the POLAMI data presented here are described in
We also estimate the magnetic field strengths within the jet from detail in Agudo et al. (2010, 2014, 2018b).
quasi-simultaneous radio spectra.
We adopt a flat  cold dark matter cosmology of H0 =
2.1.4 ALMA calibrator data base
67.4 km s−1 Mpc−1 , m = 0.315, and  = 0.685 (Planck Collab-
oration VI 2020). The luminosity distance at the redshift of AO The Atacama Large Millimeter Array (ALMA) Calibrator Source
0235 + 164 is DL = 6303.5 Mpc, and the linear scale is 8.12 pc Catalogue4 is a data base containing measurements of various
mas−1 , which was calculated with CosmoCalc (Wright 2006).1 compact radio sources for multiple frequency bands. The primary
purpose of the data base is to assist in selecting bandpass, phase,
and flux calibrators for ALMA observations. AO 0235 + 164 has
2 M U LT I WAV E L E N G T H DATA been observed since mid-2011 at frequencies of 91.5 and 103.5 GHz
(both Band 3), 233 (Band 6), and 343.5 GHz (Band 7). Observations
2.1 Radio and submillimeter data at the two different frequencies in Band 3 are usually simultaneous.
We collect single-dish and compact-array light-curve data for On days where observations were made only at a single frequency
AO 0235+164 from a wide range of observing frequencies (15– in Band 3, this was usually done at 91.5 GHz. Therefore, due to the
340 GHz). We use data obtained from 2013 to 2019. We also collect close proximity of the two frequencies and the higher cadence of
complementary 43 GHz VLBA monitoring data of AO 0235 + 164. measurements at the lower frequency, we choose only the 91.5 GHz
The details of the individual data sets are presented below. data for further evaluation. We use Band 3 data obtained from 2013
June 17 (MJD 56460) to 2019 December 31 (MJD 58848), Band
6 data obtained from 2013 December 21 (MJD 56647) to 2019
2.1.1 OVRO monitoring of Fermi blazars September 16 (MJD 58742), and Band 7 data obtained from 2013
July 1 (MJD 56474) to 2019 December 18 (MJD 58835). The average
AO 0235 + 164 is monitored by the 40 m telescope at the Owens cadence at Band 3 is 11 d with a typical rms flux error of four per cent
Valley Radio Observatory (OVRO) as part of the 15 GHz monitoring of the total flux density. The average cadence at Band 6 is 75 d with
programme (Richards et al. 2011).2 Starting from late 2007, the a typical rms flux error of seven per cent of the total flux density. The
OVRO 15 GHz monitoring programme has regularly monitored the average cadence at Band 7 is 15 d with a typical rms flux error of
flux of ∼1800 AGNs with a typical flux density rms of 4 mJy and nine per cent of the total flux density.
typical systematic errors of 3 per cent. In this paper, we use data
obtained from 2013 January 4 (MJD 56296) to 2019 December 22
(MJD 58839). The average cadence for AO 0235 + 164 during this 2.1.5 SMA calibrator data base
period is 7.5 d. The flux rms error during this period is typically
The Submillimeter Array (SMA) is a connected array of eight 6 m
two per cent of the total flux density.
dishes located on Maunakea, Hawaii. AO 0235 + 164 is included in

1 https://www.astro.ucla.edu/∼wright/CosmoCalc.html 3 http://polami.iaa.es
2 https://sites.astro.caltech.edu/ovroblazars/ 4 https://almascience.eso.org/sc/

MNRAS 527, 882–894 (2024)

884 W. Y. Cheong et al.

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

Figure 1. Multiwavelength light curves of AO 0235 + 164 from 2013 to 2019. From top to bottom, the individual light curves correspond to gamma-rays,
343, 228, 90, 37, and 15 GHz, respectively. Uncertainties correspond to 1σ confidence regions. Additionally, the mean GPR light curves and corresponding 1σ
confidence regions for the radio data are plotted as black solid lines and grey-shaded regions, respectively. In the gamma-ray light curve, values with TS ≥ 10
are marked as black dots while upper-limits for time bins with TS < 10 are plotted as red triangles.

an ongoing monitoring program at the SMA to determine the fluxes of data by fitting elliptical and circular Gaussian model components
compact extragalactic radio sources that can be used as calibrators at to the calibrated visibility data using the MODELFIT procedure in
mm wavelengths (Gurwell et al. 2007). Available potential calibrators DIFMAP (Shepherd, Pearson & Taylor 1994).
are observed for 3 to 5 min, and the measured source signal strength
is calibrated against known standards, typically Solar system objects
2.2 Gamma-ray data
(Titan, Uranus, Neptune, or Callisto). Data from this program are
updated regularly and are available at the SMA website.5 We use We use 100 MeV to 300 GeV gamma-ray data obtained from the
226 GHz data obtained from 2013 January 8 (MJD 56300) to 2019 Fermi Large Area Telescope (LAT; Atwood et al. 2009) between
April 5 (MJD 58578) and 342 GHz data from 2013 February 13 (MJD 2013 January 1 (MJD 56293) and 2019 December 31 (MJD 58848).
56336) to 2018 August 12 (MJD 58342). The average cadences of We perform the data analysis using FERMITOOLS (version 2.2.0) and
the 226 and 342 GHz data are 24 and 223 d, respectively. We note the P8R3 SOURCE V3 instrument response function. Photons in the
that half of the observations at 342 GHz are concentrated between event class 128 (‘source’ class) and within a region of interest of 15
2016 January 2 (MJD 57389) and 2016 February 28 (MJD 57446). deg centred at the position of AO 0235 + 164 were selected for the
The rms flux error is typically seven per cent of the total flux density analysis. We carried out an unbinned likelihood analysis (Abdo et al.
at 226 GHz and ten per cent of the total flux density at 342 GHz. 2009) with a time bin of 7 d. All fourth Fermi Large Area Telescope
catalog (4FGL; Abdollahi et al. 2020) sources within 25 deg of AO
0235 + 164 were included in the initial xml model. We also include
2.1.6 VLBA 43 GHz data Galactic interstellar emission, using the gll iem v07.fits model, and
isotropic diffuse emission, using the iso P8R3 SOURCE V3 v1.txt
AO 0235 + 164 has been monitored by the VLBA at 43 GHz as
model.7 We also left free to vary the normalization of AO 0235+164,
part of the VLBA-BU-BLAZAR programme (Jorstad et al. 2005;
16 variable sources (variability index ≥ 18.48) located within 10
Jorstad & Marscher 2016)6 since 2007 June 14 (MJD 54265). We
deg of AO 0235+164, and 4C +28.07, a gamma-ray bright blazar
use the calibrated VLBI data obtained between 2013 January 15
located within 12 deg of AO 0235+164. Model parameters of all
(MJD 56307) and 2019 October 19 (MJD 58775). We analyse the
other point sources were fixed. Spectral parameters for all sources
were fixed to the 4FGL values. For AO 0235 + 164, we use a
5 http://sma1.sma.hawaii.edu/callist/callist.html
6 https://www.bu.edu/blazars/VLBAproject.html 7 https://fermi.gsfc.nasa.gov/ssc/data/access/lat/BackgroundModels.html

MNRAS 527, 882–894 (2024)

 AO 0235 + 164 885
Table 1. Fit parameters of the DCFs computed τ DCF < 0 indicates that the corresponding light curve is leading the
with respect to the 15 GHz light curve. 15 GHz light curve. We find that all radio light curves from 37 to
343 GHz have τ DCF < 0, indicating that flares at higher frequencies
Frequency∗ ADCF τ DCF occur prior to those at lower frequencies. There is a trend of larger
(GHz) (d) absolute values of τ DCF with increasing frequency of the reference
gamma-ray 0.74 ± 0.04 −48 ± 29 light curve, although this is difficult to track concretely due to the
343 0.87 ± 0.02 −24 ± 10 relatively large uncertainties of τ DCF . We find that the gamma-rays
228 0.89 ± 0.02 −37 ± 15 also lead the radio frequencies, indicating that the gamma-ray region
90 0.90 ± 0.01 −24 ± 12 is located upstream of the radio emission region (see Section 4.1 for
37 0.95 ± 0.01 −17 ± 9 additional analysis).
Notes. Column 1: is the frequency of the com- We evaluate the significance of the correlation by simulating 104
panion light curve with the exception of the light curves using the method of Emmanoulopoulos, McHardy &
gamma-ray data∗. Column 2: is the peak of Papadakis (2013) as implemented in DELCGEN (Connolly 2016). The
the DCF. Column 3: is the time lag. Quoted probability density function (PDF) and the power spectral density

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

uncertainties represent 1σ confidence regions (PSD) encapsulate the variability and statistical properties of a
on the parameters. given light curve, and it has been demonstrated that both properties
used for the creation of simulated light curves may significantly
LogParabola function8 (Massaro et al. 2006) to model the spectral affect the estimation of the confidence intervals of a DCF (e.g.
shape of the gamma-ray data. For each 7-d time bin, we ran an initial Emmanoulopoulos, McHardy & Papadakis 2013; Max-Moerbeck
likelihood analysis including all sources. After the initial run, we et al. 2014b). Therefore, it is important that the simulated light curves
removed sources with a test statistic TS < 25 from the xml model accurately represent both the PDF and PSD of the original light curve.
(with the exception of AO 0235 + 164). We then ran a second We model the PDFs of the radio light curves using a lognormal
round of likelihood analysis with the filtered xml model. After the fit distribution (e.g. Goyal 2020; Acciari et al. 2021), whereas we use a
converged, we found the photon count rate (in units of ph cm−2 s−1 ) gamma distribution for the gamma-ray light curve (e.g. Algaba et al.
along with the corresponding uncertainties using pyLikelihood. The 2018a; Kim et al. 2022b). We model the PSDs for all frequencies
full gamma-ray light curve from the start of 2013 to the end of 2019 using a power-law function with an added constant (e.g. Park &
is presented in Fig. 1. In each time bin, we consider AO 0235 + 164 Trippe 2014; Tarnopolski et al. 2020).
to be detected if the TS is ≥10 (e.g. Ackermann et al. 2012). 95 Examples of the calculated DCFs with the corresponding con-
per cent confidence upper limits are plotted for time bins with TS < fidence intervals are presented in Fig. 2. All five light curves,
10. from 37 GHz to gamma-rays, are correlated with the 15 GHz light
curve with the significance of the correlation with the gamma-rays
reaching a level of greater than 2σ . AO 0235 + 164 is known for
3 DATA A N A LY S I S A N D R E S U LT S
significant correlations between radio and gamma-ray flares (e.g.
Max-Moerbeck et al. 2014a). We find that this is the case for the
3.1 Multiwavelength light-curve correlations
period of our study, suggesting that radio and gamma-ray emission
We search for correlations among the various light curves using the are tightly related in this source. The 3σ correlations between the
discrete correlation function (hereafter DCF), which is a method for radio data sets suggest that the radio fluxes of this source arise from
analysing correlations in unevenly sampled data sets. Details on the a common origin and mechanism.
DCF may be found in Edelson & Krolik (1988). To improve the
cadences of the light curves, we integrate data sets taken at similar
frequencies into combined light curves. We do this for the following 3.2 Radio light-curve Gaussian process regression
data sets: (1) we combine the POLAMI 86 GHz data and ALMA Given the variety of multiwavelength observations used in this analy-
91.5 GHz data into the 90 GHz light curve, (2) we combine the sis, our measurements at different wavelengths are not simultaneous.
POLAMI 229 GHz data, the ALMA 233 GHz data, and the SMA To construct quasi-simultaneous radio spectra, we interpolate the
226 GHz data into the 228 GHz light curve, and (3) we combine the data to a common date with Gaussian Process Regression (hereafter
ALMA 343.5 GHz data and the SMA 342 GHz into the 343 GHz GPR; Rasmussen & Williams 2006), which has been used in the past
light curve. The mean frequencies of the combined light curves are to study the variability characteristics of blazars (e.g. Karamanavis
90.4, 228, and 343 GHz, respectively. We evaluate DCFs for all of the et al. 2016). We use SCIKIT-LEARN (Pedregosa et al. 2011) for the GPR,
light curves with respect to the 15 GHz light curve, which has both a using a noisy radial-basis function kernel of the form
high-mean cadence and low-measurement errors. The 37 GHz light   
curve has a comparable mean cadence, but the measurement errors   1 xi − xj 2
are significantly larger. kn xi , xj = σf exp −
2
+ σrms
2
δij (1)
2 l
To evaluate the uncertainties of the measured time lags τ DCF , we
simulate 104 artificial light curves using the Flux Randomization to represent the covariance between two data points at xi and xj . There
and Random Subset Selection methods described in Peterson et al. are a total of three hyperparameters. Parameters l and σ f represent
(1998). For each frequency pair, we fit a normalized distribution to the time-scale and magnitude of the covariation, respectively. The
the array of 104 simulated DCF peaks (ADCF ) and corresponding parameter σ rms corresponds to a global uncertainty due to the
time lags. The fit parameters are presented in Table 1. We note that measurement errors. Together with the Kronecker delta function δ ij ,
2
σ rms is part of the white kernel, σrms δij . The hyperparameters are
optimized separately for the 15, 37, 90, 228, and 343 GHz radio light
8 https://fermi.gsfc.nasa.gov/ssc/data/analysis/scitools/source models. curves using a maximum likelihood analysis. After optimization, the
html#LogParabola mean GPR light curve is obtained for 1-d intervals starting from 2013

MNRAS 527, 882–894 (2024)

886 W. Y. Cheong et al.

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

Figure 2. DCFs of the 37, 90, 228, 343 GHz, and gamma-ray light curves with respect to the 15 GHz light curve. The specific light-curve pair is labelled in the
upper-right corner of each plot. Negative time lags indicate that the paired light curve leads the 15 GHz light curve. The red dashed horizontal lines represent
the boundaries of the 95.45 per cent confidence region. The blue dot-dashed horizontal lines represent the boundaries of the 99.73 per cent confidence region.

January 1 (MJD 56293). We plot the results with the observations

in Fig. 1 using black solid lines and grey-shaded regions to indicate
GPR light curves and their corresponding 1σ confidence regions.

3.3 Radio spectral energy distribution

We investigate spectral evolution in the radio for AO 0235 + 164
during its 2015–2019 flaring period. We limit our evaluation to
periods when the total radio flux is dominated (>90 per cent) by the
radio core as inferred from the 43 GHz VLBA data (see discussion
in Section 4.3). There are two periods that satisfy this criterion – the
flaring activity in 2015 March–November (MJD 57100-57350) and
the flaring activity in 2018 March–August (MJD 58180-58360). We
fit a curved power law of the form (e.g. Massaro et al. 2006; Lee
et al. 2016; Algaba et al. 2018b)
 
 c1 ln ν
ν νc
S (ν) = Sm (2) Figure 3. Evolution of Sm and ν c during the 2015 flare. Quasi-simultaneous
νc
spectrum results are in red squares while GPR spectrum results are in blue
to the single dish and compact array radio light curves for these circles. The vertical dotted lines indicate the boundaries of the different phases
two periods. Sm is the flux at the turnover frequency ν c . c1 is a described in Section 3.3. The time gap between consecutive GPR data points
constant that fits the overall spectral shape. We fit to both the quasi- is 7 d (in order to prevent overlap).
simultaneous spectrum, which was constructed with observations
within 3 d of Metsähovi 37 GHz observations, and the GPR spectrum plotted in red while those from the GPR spectrum are plotted in blue.
constructed from the GPR light curves. In both cases, we remove the The uncertainties correspond to the 1σ confidence regions for the
minimum detected flux from each frequency as a method of exploring parameters which are estimated with a Markov Chain Monte Carlo
the variable radio flux. The evolution of the resultant parameters analysis. We present an example of a fit in Fig. 4. For this example, we
of the spectral analysis for the 2015 flaring activity is plotted in see that spectral shapes (characterized by c1 and ν c ) determined with
Fig. 3. Sm and ν c values from the quasi-simultaneous spectrum are the two different methods are well compatible. The extimated values

MNRAS 527, 882–894 (2024)

 AO 0235 + 164 887

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

Figure 4. Example of a spectral fit for the spectrum during the flare occurring on 2015 September 20 (MJD 57285). The upper row plots the radio spectrum
of the source in black circles with the derived spectral model in red. The lower row shows the joint posterior distribution of the fit parameters. Results for the
quasi-simultaneous spectrum are plotted in the left column while those corresponding to the GPR spectrum are plotted in the right.

of Sm differ by ∼ 5 per cent. Overall, we find that the evolution model fitting of the calibrated 43 GHz VLBI data from the VLBA-
of the spectral parameters (Sm and ν c ) is well interpolated by using BU-BLAZAR programme with DIFMAP. We model the radio core
the GPR light curves. We present Sm with respect to ν c as derived with an elliptical Gaussian (i.e. a 2D elliptical Gaussian brightness
from the GPR spectrum for the 2015 flare in Fig. 5. The evolution distribution) and any other components with circular Gaussians. The
of the flare may be decomposed into three stages. During the onset positions of all components, including the core, were set as free
of the flare, Sm rises from ∼2 to ∼3.6 Jy while ν c is approximately parameters. We evaluate the errors in the fit parameters following
constant at 45 GHz. Then, ν c decreases from ∼44 GHz at MJD the procedures in Fomalont (1999) and Lee et al. (2008). From the
57 230 to ∼25 GHz at MJD 57270. During this period, Sm increases 43 GHz VLBA data, we find the ejection of a new jet component
from ∼3.6 to ∼4.4 Jy, followed by a decrease to ∼3.7 Jy. Finally, Sm (hereafter J5) at a position angle of approximately 57.8 ± 1.7 deg (see
decreases from ∼3.7 to ∼2 Jy as ν c is ∼20 GHz. Fig. 6). We were able to track this component up to approximately
0.4 mas from the core before it became too faint to be reliably
detected. The time evolution of the distance between J5 and the core
3.4 Jet parameters from new superluminal jet components is shown in Fig. 7. Fitting a linear function to the separation (plotted
While a detailed analysis of the VLBI data is beyond the scope of in red), we find a proper motion of μ = 0.230 ± 0.019 mas yr−1 and
this paper, we present a subset of our study on determining the jet a corresponding apparent speed of βapp = 11.84+ 0.96
−0.97 c. The epoch
parameters of AO 0235 + 164 during the 2015 flare (which is required of ejection from the core is estimated to be MJD 57246+ 26
−30 , which
for the following analysis presented in this paper). We conducted corresponds to the peak of the 2015 flare.

MNRAS 527, 882–894 (2024)

888 W. Y. Cheong et al.
δ ≈ 24 and ≈ 14 for the same 2008 flare. Hovatta et al. (2009) also
find δ = 24 and = 12.1 for this source, which were derived from
the variability time-scales of long-term radio monitoring data. Both
Hovatta et al. (2009) and Kutkin et al. (2018) are consistent with the
values found in this paper for the 2015 flare.
We find the presence of a second component (hereafter J6) ejected
close in time to J5. Following the same analysis, we find that J6
has a slower proper motion, likely due to a smaller viewing angle as
compared to J5. The value of δ var is consistent with that of J5 within
uncertainties while the estimated values of agree well between the
two jet components. The results are summarized in Table 2. Weaver
et al. (2022) analyse the same 43 GHz VLBI data during this period
and also find the two jet components J5 and J6 in Table 2 (labelled
respectively as B5 and B6 in their paper). The position angles of

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

the jet components found in Weaver et al. (2022) agree well with
those found in our analysis; on the other hand, the proper motion
of J5 is ∼ 20 per cent slower and that of J6 is ∼ 60 per cent faster
than in Table 2. The variability time-scales in Weaver et al. (2022)
Figure 5. Plot of Sm versus ν c derived from the GPR spectrum (blue circles)
are ∼ 50 per cent lower than found in this paper. Their δ var and
and the quasi-simultaneous spectrum (red squares) during the 2015 flare.
The GPR data start at MJD 57150 and end at MJD 57340. The time gap for J5 are consistent with our results while those for J6 differ by
between consecutive GPR data points is 10 d (in order to prevent overlap). approximately a factor of two.
The quasi-simultaneous data start at MJD 57 180 and end at MJD 57325. Due to J5 having a faster, well-constrained proper motion com-
pared to J6, we consider the jet parameters derived from J5 to be
The light curve of J5 is presented in Fig. 8. Following Jorstad et al. representative of AO 0235 + 164 at the time of the 2015 flare. We
(2017), we estimate the variability Doppler factor from the decaying refer to these values in the following analysis.
light curve of this jet component. We model the light curve as an
exponential decay with the function ln (S(t)/S0 ) = −kvar (t − t0 ), where
S0 is the flux of the component at a certain time t0 and kvar is a decay
constant. The variability time-scale is then found as τ var = |1/kvar |. 3.5 Radio core magnetic field strength
Fitting to the data of J5, we find kvar = 1.32 ± 0.20 yr−1 and τvar = We estimate the magnetic field strength of AO 0235 + 164 assuming
0.76+ 0.14
−0.10 yr. Following Lobanov (2005 ), we calculate the minimum
  S/N 
a synchrotron self-absorption (hereafter SSA) spectrum using the
1−β/2
resolvable size as θmin = 2 bψ ln 2
π
ln S/N−1
, where bψ is the equation (Marscher 1983)
full width at half-maximum (FWHM) at an arbitrary position angle  
BSSA 1 Jy 2 dm 4 νc 5 1+z
ψ, and β is a weighting parameter with β = 0 for uniform weighting = b( α) , (6)
and β = 2 for natural weighting. If the fit FWHM was found to be 10−2 mG Sm 1 mas 1 GHz δ
smaller than θ min , we consider the component to be unresolved. The where b(α) is a dimensionless factor that varies as a function of the
mean FWHM of J5 was found to be <θ vlbi > J5 = 0.234 ± 0.059 mas, optically thin spectral index α (e.g. Marscher 1983; Pushkarev et al.
which was calculated over a total of nine epochs where the component 2019), Sm is the flux density at the turnover frequency ν c , dm is the
was found to be resolved. The variability Doppler factor is found as angular size of the SSA region, δ is the Doppler factor, and z is the
(Jorstad et al. 2017) cosmological redshift
 of the source. The difference in the power of
15.8 dm DL |kvar | the factor 1+δ z in equation (6) from that of Marscher (1983) is due
δvar = = 28.5 ± 8.4, (3) to considering the SSA region to be the radio core as in Lee et al.
1 + z 1 mas 1 Gpc 1 yr−1
(2017).
−1
where we have substituted τvar with |kvar | and estimated the angular We estimate dm from the model-fit analysis of multi-epoch 43 GHz
size assuming a uniform sphere (i.e. 1.8 times θ VLBI for consistency VLBA data as in Section 3.4, but this time focusing on the radio core.
with Section 3.5) rather than the uniform disc assumed in Jorstad For the core, we determine θ min separately for the major and minor
et al. (2017). This difference does not significantly affect the derived axes of the elliptical Gaussian fitted to the radio core. We consider
jet parameters. the core to be unresolved when either of the two are smaller than the
With an independent estimate of δ var and β app , we are now able to corresponding θ min . Following Algaba et al. (2017), the mean size
determine the bulk Lorentz factor  and the intrinsic viewing angle of the fitted elliptical Gaussian is calculated as <θ vlbi > = (2θ major
θ v as (e.g. Boettcher, Harris & Krawczynski 2012) + θ minor )/3, where θ major and θ minor are the FWHMs of the major and
2
βapp + δvar
2
+1 minor axes of the elliptical Gaussian. Then, we have
= = 16.8+ 3.6
−3.1 (4)  
2δvar νc −k
and dm = 1.8 θvlbi  , (7)
νobs
2βapp
tan θv = 2 + δ2 − 1
= 0.025+ 0.019
−0.009 . (5) where k determines the dependence of the core size with frequency.
βapp var An additional factor of 1.8 is used to convert between the FWHM
This corresponds to an intrinsic viewing angle of θv = of a Gaussian function and the total size (Marscher 1977; Hodgson
1.42+ 1.07
−0.52 degrees. We note that the values of δ var and  for J5 are et al. 2017).
significantly smaller than those presented in Jorstad et al. (2017) for We also estimate the magnetic field based on the assumption of
the 2008 flare (δ ≈ 60, ≈ 31). However, Kutkin et al. (2018) report equipartition between the magnetic field and the particle energy

MNRAS 527, 882–894 (2024)

 AO 0235 + 164 889

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

Figure 6. A subset of the 43 GHz VLBA total intensity images of AO 0235 + 164 following the peak of the 2015 flare. All CLEAN maps were restored with
a common beam of 0.36 × 0.15 mas and the beam major axis at a position angle of −10 deg (plotted in the lower-left corner). The contours start at 5 mJy per
beam and incrementally increase by factors of 2. The images from each epoch are shifted in RA and Dec. to avoid overlap. The model-fitted components are
overlaid on top of the CLEAN maps. The radio core (red) is the compact feature located near the intensity peaks. Jet components are plotted in symbols with
different colours used to identify different components. The J5 and J6 components studied in Section 3.4 are plotted in green and purple, respectively.

Figure 7. The separation between the newly ejected jet component J5 Figure 8. The decay of the total flux of J5. Fitting an exponential decay
(associated with the 2015 multiwavelength flare) and the radio core. This resulted in a variability time-scale of τvar = 0.76+ 0.14
−0.10 yr.
component was ejected from the 43 GHz core on MJD 57246+ 26
−30 (plotted in
blue) with a proper motion of μ = 0.230 ± 0.019 mas yr−1 (plotted in red).
σ EQ for BSSA and BEQ are evaluated with standard error propagation
densities. Following Kataoka & Stawarz (2005), we have as

BEQ 100 Mpc 2
νc σSSA σS m 2 5σνc 2 4σdm 2 σδ 2
= 0.123 η2 (1 + z)11 = + + + (9)
1 mG DL 5 GHz BSSA Sm νc dm δ
 17 and
2 6
Sm 300 mas −5 
δ , (8) 2 2 2 2
0.1 Jy dm σEQ 2σSm σνc 6σdm 5 σδ
= + + + , (10)
BEQ 7Sm 7νc 7dm 7δ
where η is the ratio of the energy of both the protons and electrons to
the energy of the electrons (η = 1 for a leptonic jet, η = 1836 for a where σSm , σνc , σdm , and σ δ are the uncertainties of Sm , ν c , dm , and
hadronic jet) and DL is the luminosity distance. The errors σ SSA and δ.

MNRAS 527, 882–894 (2024)

890 W. Y. Cheong et al.
Table 2. Jet parameters of AO 0235 + 164 derived from the jet components J5 and J6.

ID To μ PA β app τ var <θ vlbi > δ var  θv

(MJD) (mas yr−1 ) (deg) (c) (yr) (mas) (deg)

J5 57246+ 26
−30 0.230 ± 0.019 57.8 ± 1.7 11.84+ 0.96
−0.97 0.76+ 0.14
−0.10 0.234 ± 0.059 28.5 ± 8.4 16.8+ 3.6
−3.1 1.42+ 1.07
−0.52
J6 56985+ 203
−388 0.073 ± 0.023 50.7 ± 1.4 3.74+ 1.20
−1.18 0.87+ 0.22
−0.22 0.330 ± 0.097 35.0 ± 13.6 17.8+ 6.6
−6.8 0.36+ 0.64
−0.20
Notes. The columns correspond to – Column 1: the ID of the jet component. Coumn 2: the epoch of ejection from the core. Column
3: the proper motion. Coumn 4: the direction of proper motion. Coumn 5: the apparent speed. Column 6: the variability time-scale.
Column 7: the mean size. Column 8: the variability Doppler factor. Column 9: the bulk Lorentz factor. Coulmn 10: the intrinsic
viewing angle. All uncertainties correspond to 1σ confidence regions for the parameters.

Table 3. The magnetic fields of AO 0235 + 164 assuming b(α) = 3.8 and
η = 100.

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

MJD Sm νc dm BSSA BEQ
(d) (Jy) (GHz) (μas) (mG) (mG)

57124 1.85+ 0.14

−0.14 58.4+ 7.22
−5.34 <160 <338 >75.7
57183 2.49+ 0.10
−0.10 44.6+ 2.47
−2.78 <135 <25.2 >91.5
+0.09
57206 3.02−0.09 45.9+ 2.66
−2.92 <128 <15.8 >102
57236 3.99+ 0.09
−0.09 38.5+ 2.37
−2.65 <131 <4.13 >105
+0.07
57288 3.43−0.06 23.1+ 2.63
−3.04 322+ 41
−45 15.3+ 12.6
−14.0 43.6+ 10.6
−10.4
57362 2.30+ 0.18
−0.11 10.3+ 5.38
−5.08 <491 <3.28 >24.1
58188 1.44+ 0.15
−0.14 48.9+ 22.3
−14.6 <194 <497 >58.3
+0.13
58228 1.05−0.11 69.8+ 93.7
−39.3 <103 <422 >97.5
58250 1.25+ 0.11
−0.11
+8.04
65.2−6.86 <128 <515 >83.8
+0.09
58286 1.43−0.09 36.9+ 4.11
−5.27 <257 <381 >44.0
Figure 9. Plot of the calculated BSSA and BEQ with time. BSSA is plotted 58316 1.31+ 0.12
−0.11
+4.56
39.5−6.07 <157 <84.9 >66.6
+0.22
as a red square, and BEQ is plotted as a blue circle. Upper/lower limits are 58356 0.73−0.15 37.6+ 26.3
−21.8 <230 <1020 >40.0
represented by upper/lower triangles respectively. In this plot, we assumed Notes. The columns correspond to – Column 1: the epoch of measurement in
b(α) = 3.8 and η = 100. MJD. Column 2: the peak flux density in units of Jy. Column 3: the turnover
frequency in units of GHz. Column 4: the size of the emission region at
To ensure that the spectra from the multiwavelength data of single- the turnover frequency in units of μas. Column 5: the SSA magnetic field
dish and compact array observations in Section 3.3 represent the strength in units of mG. Column 6: the equipartition magnetic field strength
radio core, we limit the calculation to VLBI observations where the in units of mG. All uncertainties correspond to 1σ confidence regions for
the parameters.
model-fitted core flux is at least nine times greater than the sum of
all other components. We adopt the value of δ = δ var = 28.5 ± 8.4
from Section 3.4 and k = 0.8 ± 0.1 from Kutkin et al. (2018). region in blazars significantly deviates from equipartition. This is
For VLBI observations where the core is unresolved, we use the in contrast with the results of this paper where the single epoch
calculated minimum resolvable size for the major and minor axes to measurement of the magnetic field strengths implies that the SSA
place limits on BSSA and BEQ . The results are presented in Fig. 9 and region of AO 0235 + 164 during the 2015 flare was approximately
summarized in Table 3 along with the values of Sm , ν c , and dm . We in an equipartition state.
note that the uncertainties derived for BSSA and BEQ do not account We briefly consider the effect of some of the assumptions used
for uncertainties due to b(α), η, z, and DL . We find that among the when deriving the magnetic field strengths in Section 3.5. In calcu-
VLBI observations that satisfy the core-dominance criteria, the core lating BSSA , we have taken b(α) = 3.8 (or α = 1, for S(ν) ∝ ν −α ) as a
was resolved in only one epoch (MJD 57288) with values of BSSA ≈ benchmark value. The relatively low cadence of the high-frequency
15.3(b(α)/3.8) mG and BEQ ≈ 43.6(η/100)2/7 mG. In all other epochs, (228, 343 GHz) light curves combined with the highly variable nature
the core was found to be unresolved, leading to upper limits on BSSA of AO 0235 + 164 makes it difficult to concretely track the variation
and lower limits on BEQ . of α. The value of b(α) varies from 3.2 to 3.8 for α of 0.5 to 1.0. The
mean spectral index between 228 and 343 GHz during the decay of
the 2015 flare is ∼0.7, and the change of BSSA due to b(α) is less than
4 DISCUSSION
the uncertainties due to other observables. The values of BEQ may
vary by a factor of ∼0.27 (for a fully leptonic jet) to ∼2.3 (for a fully
4.1 Equipartition of the SSA region
hadronic jet) depending on the composition of the jet. Modelling of
A number of previous papers (Lee et al. 2017, 2020; Algaba et al. the broad-band spectral energy distribution (SED) of AO 0235 + 164
2018b; Kang et al. 2021; Kim et al. 2022a) estimated both BEQ and suggests a moderately hadronic jet (e.g. Ackermann et al. 2012) as
BSSA for various sources using similar methods as those used in was considered in our analysis.
this paper. They find BEQ values in the range of 0.93–1150 mG Another method of cross-checking our magnetic field strength
and BSSA values in the range of 0.01–135 mG. In all previous estimates is to derive the so-called time-lag core shift (Kudryavtseva
cases, the authors found that BEQ  BSSA , implying that the SSA et al. 2011) using the results from Section 3.1 and Section 3.4. The

MNRAS 527, 882–894 (2024)

 AO 0235 + 164 891
time-lag core shift between two frequencies ν h and ν l is defined as 4.2 Shock-in-jet interpretation of the 2015 flare

rν 4.85 × 10−3 μ t As shown in Section 3.3, the evolution of Sm and ν c may be
= decomposed into different stages, starting with a factor of ∼2 rise of
1 pc GHz1/kr (1 + z)2 1 mas yr−1 1 yr
  Sm at roughly constant ν c , followed by an ∼50 per cent reduction of
1/k 1/kr
DL νh r νl ν c with less variation of Sm (remaining within ∼10 per cent of 4 Jy),
× . (11)
1 Mpc νh
1/kr
− νl
1/kr
then finally a decrease of Sm by ∼50 per cent at an approximately
constant ν c . This seems to imply that the flare may be interpreted
This is a modification of the core shift from the form in Lobanov according to the shock-in-jet model (Marscher & Gear 1985) with
(1998) by using the proper motion μ and the observed time lag the three stages each corresponding to the Compton, synchrotron,
t to estimate the angular separation between the radio cores at and adiabatic stage, respectively. We find similar trends for the 2018
two different frequencies. The parameter kr = [(3 + 2α)m + 2n − flare; however, the larger uncertainties complicate the identification
2]/(5 + 2α), where m and n represent the power-law dependence of the three stages. Within the shock-in-jet scenario, we may interpret
of the magnetic field (B(r) ∝ r−m ) and the electron number density the newly ejected jet component J5 (and to a lesser certainty, J6)
(N(r) ∝ r−n ) as a function of distance r from the jet base. α is the found in the 43 GHz VLBA maps (Section 3.4) to be the propagating

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

optically thin spectral index. With an estimate of the core shift, we can relativistic shock that produced the 2015 flare. It should be noted that
define a relation for the distance of the radio core from the jet base for the exact multiwavelength characteristics of radio flares predicted by
rν −1/kr
a given frequency ν: rcore (ν) = sin θv
ν . From our DCF time lags the shock-in-jet scenario depends greatly on the physical parameters
(Section 3.1), a proper motion of μ ≈ 0.23 mas yr−1 and viewing of the jet, such as the energy distribution of the synchrotron-emitting
angle of θ v ≈ 1.42 deg (Section 3.4), we find that all frequency pairs electrons, and the change of the magnetic field strength, particle
considered consistently indicate that the 15 GHz core is located at density, and Doppler factor with distance from the jet base (e.g.
approximately 6 pc from the jet base (see Table 4). The 37–343 GHz Fromm et al. 2011; Fromm, Fuhrmann & Perucho 2015). It may be
cores are found to be in the range of 3–0.4 pc from the jet base. that the Compton stage extends to MJD 57248, with Sm peaking at
Assuming equipartition, the magnetic field strength at 1 pc from 4.4 Jy and a marginal decrease of ν c to 35 GHz. The synchrotron
the jet base may be estimated as (Hirotani 2005; Karamanavis et al. stage then may result in a decrease of Sm to 3.7 Jy and a decrease
2016) of ν c to 25 GHz, followed by the adiabatic stage starting at MJD
0.25 57270. Detailed modelling of the evolution of Sm and ν c , or of the
3rνkr (1 + z)2 ln (γmax /γmin ) multiwavelength radio light curves themselves, may in turn allow for
B1 pc = 14 [mG], (12)
δ 2 φ sin3kr −1 θv constraints on the jet parameters of AO 0235 + 164.
The distance d between the radio and gamma-ray emission
where γ min /γ max are the minimum/maximum Lorentz factors and φ
regions may be estimated as (Pushkarev, Kovalev & Lister 2010)
is the intrinsic jet half-opening angle. The magnetic field strength
at the radio core for a given frequency may also be found from βapp cτDCF
(Karamanavis et al. 2016) d = . (16)
(1 + z) sin θv
−1
Bcore (ν) = B1 pc rcore (ν) . (13) Using τ DCF found from the gamma-ray and the 15 GHz data, we find
Adopting the same jet parameters as in Section 3.5 and taking φ that the gamma-ray emission region is located d ≈ 9.9 ± 6.1 pc
≈ 1 deg from Kutkin et al. (2018) and ln (γ max /γ min ) ≈ 10, we upstream of the 15 GHz radio core. This implies that the gamma-
find that all frequency pairs consistently imply B1pc ≈ 0.2 G and rays were produced near the jet base. Nalewajko, Begelman & Sikora
Bcore (15 GHz) ≈ 33 mG. Evaluating equation (13) at the turnover (2014) use multiple constraints to infer that the 2008 gamma-ray flare
frequency of ν c = 23 GHz for the epoch of MJD 57288, we find was produced at ∼1 pc from the central engine while the 43 GHz
Bcore (23 GHz) ≈ 47 mG. This is consistent with both BEQ and BSSA VLBI core is at ∼6.7 pc. This is consistent with what we find for
found in Section 3.5 for the same epoch. the 2015 flare where we find the gamma-ray production region to
We may define an ‘equipartition size’ by equating BSSA and BEQ . be ∼6.5 pc upstream of the 37 GHz core. The inferred deprojected
From equations (6) and (8), we find distance between the 37 GHz core and the 43 GHz core is <0.5 pc,
which is within the measurement uncertainties. From these distances
 341 and the radio-gamma-ray correlation, we can infer that the 2015 flare
2
dm,eq η2 δ 2 (1 + z)4 100 Mpc
≈ 1.94 may have been caused by a relativistic shock propagating along the
1 mas b(α)7 DL jet. Particle acceleration by this shock near the jet base may have
8 produced the leading gamma-ray flare followed by the subsequent
1 GHz Sm 17
× . (14) radio flares (with lower frequencies peaking later) as the shock
νc 1 Jy
propagates further downstream and interacts with the radio core.
The uncertainties are evaluated using error propagation, It should be noted that the uncertainty in the location of the gamma-
 ray flaring region is large, primarily due to the large uncertainty in
σdm,eq σνc 2 8σSm 2 σδ 2 the gamma-to-radio time lag. Future high-cadence monitoring, as
= + + . (15) well as gamma-ray observations at increased sensitivity (enabling
dm,eq νc 17Sm 17δ
detection within smaller time bins), may help to better constrain the
We evaluate the above equations for our core-dominated epochs and time delay between gamma-ray and radio light curves. In turn, better
compare them to the VLBI-derived SSA region sizes dm,vlbi . The constraints in the time delay will enable better constraints in the
results are presented in Fig. 10. We find that dm,eq varies between gamma-ray emission region.
0.03 and 0.3 mas during our period of interest. We also see that An alternative mechanism that may produce strong gamma-ray
the observed upper limits on dm,vlbi are consistent with the values of flares followed by delayed radio flares is magnetic reconnection
dm,eq . (e.g. Petropoulou, Giannios & Sironi 2016). Emission models based

MNRAS 527, 882–894 (2024)

892 W. Y. Cheong et al.
Table 4. Time-lag core shifts of AO 0235 + 164. In all cases, the lower frequency ν l = 15 GHz. The errors of each
parameter are calculated from the uncertainty in the time lag t. All uncertainties correspond to 1σ confidence regions
for the parameters. Systematic errors arising from the jet and cosmological parameters that are common to the various
frequency pairs are not included in the calculations.

νh r ν rcore (ν l ) rcore (ν h ) B1pc Bcore (ν l ) Bcore (ν h )

(GHz) (pc GHz0.8 ) (pc) (pc) (mG) (mG) (mG)

343 1.16 ± 0.48 5.37 ± 2.22 0.44 ± 0.18 181 ± 70 33.7 ± 0.9 412 ± 10.6
228 1.88 ± 0.78 8.67 ± 3.59 0.98 ± 0.41 284 ± 110 32.7 ± 0.8 288 ± 7.5
90 1.39 ± 0.68 6.45 ± 3.14 1.54 ± 0.75 215 ± 98 33.3 ± 1.0 140 ± 4.3
37 1.44 ± 0.75 6.67 ± 3.50 3.24 ± 1.70 222 ± 109 33.2 ± 1.1 68.5 ± 2.2
Notes. The columns correspond to – Column 1: the paired light-curve frequency in units of GHz. Column 2: the derived
time-lag core shift. Column 3: the distance of the 15 GHz core from the jet base. Column 4: the distance of the ν h core

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

from the jet base. Column 5: the core shift magnetic field strength at 1 pc from the jet base. Column 6: the core shift
magnetic field strength at the 15 GHz core. Column 7: the core shift magnetic field strength at the ν h core.

Figure 10. A plot of the emission region size derived from VLBI observa- Figure 11. Plot of the 7 mm VLBI light curve. The core component is
tions (dm,vlbi ) and the equipartition assumption (dm,eq ). Upper limits on dm,vlbi plotted as red dots while the sum of the remaining components is plotted as
are plotted with triangles. The errors on dm,vlbi represent 1σ confidence blue squares. The total VLBI flux is plotted as black diamonds.
regions. The 90-per cent and 99-per cent confidence regions of dm,eq are
plotted as filled regions of darker and lighter green, respectively. 4.3 Core dominance of the radio flux
AO 0235+164 is compact with minimal apparent structure extending
past a couple mas from the centre. VLBI observations at cm
wavelengths find the flux of AO 0235 + 164 to be core dominated
(e.g. Lister et al. 2016; Kutkin et al. 2018). On the other hand, multi-
epoch mm-VLBI observations are able to resolve the sub-mas jet
on particle-in-cell simulations of magnetic reconnection are able to of this source and find that for certain epochs, the jet components
reproduce the multiwavelength SEDs of both low-frequency and account for a considerable fraction of the total flux density (e.g.
high-frequency peaked blazars (Petropoulou, Giannios & Sironi Piner et al. 2006; Jorstad et al. 2017). We find this to be the case
2016), and of both BL Lacs and flat spectrum radio quasars (Christie during our time period of interest. As shown in Fig. 11, the model
et al. 2019). Gamma-ray flares from magnetic reconnection are fits to the 43 GHz VLBA data suggest that the combined flux for the
expected to vary on time-scales of several hours to days, depending extended structures may account for up to half of the total flux in
on the size of the produced plasmoids. In this paper, we investigate some epochs. However, we note that from the time period of the 2015
the gamma-ray variability of AO 0235 + 164 with 7-d time bins, flare until the jet component is resolved from the core in 2016, the
which lacks the temporal resolution to study flares on such time- total contribution of any external components is only a few per cent.
scales. The radio synchrotron emission from plasmoids created by Therefore, we consider the multiwavelength total flux measurements
magnetic reconnection events is expected to be strongly attenuated by to be good representations of the radio core flux during this period.
SSA and unlikely to be observed directly (Petropoulou, Giannios &
Sironi 2016). The radio emission may be observable once the
5 CONCLUSION
particles escape the reconnection layer (Christie et al. 2019). A
detailed study of the temporal and spectral evolution of radio flares In this paper, we investigated the multiwavelength flaring activity of
following magnetic reconnection may further shed light on whether AO 0235 + 164 during its recent active period from 2013 to 2019.
such processes are able to explain the multiwavelength flares of AO From a DCF time-lag analysis, we find that the radio and gamma-
0235 + 164. ray light curves are correlated with a significance of 2σ or greater,

MNRAS 527, 882–894 (2024)

 AO 0235 + 164 893
which has been previously demonstrated for this source (e.g. Agudo with the Republic of Chile. The Joint ALMA Observatory is operated
et al. 2011; Max-Moerbeck et al. 2014a). We also find that variability by ESO, AUI/NRAO, and NAOJ. The Submillimeter Array is a joint
at lower frequencies is delayed with respect to variability at higher project between the Smithsonian Astrophysical Observatory and the
frequencies, implying that the radio variability of AO 0235 + 164 Academia Sinica Institute of Astronomy and Astrophysics and is
is affected by the opacity of the radio core (which is responsible for funded by the Smithsonian Institution and the Academia Sinica. We
most of the radio flux variability). This is consistent with the results of recognize that Maunakea is a culturally important site for the indige-
Fuhrmann et al. (2014) who performed a cross-correlation analysis of nous Hawaiian people; we are privileged to study the cosmos from
a sample of blazars (including AO 0235 + 164) that are bright in the its summit. This study makes use of VLBA data from the VLBA-
Fermi-LAT band. They also found a trend of increasing lags toward BU Blazar Monitoring Program (BEAM-ME and VLBA-BU-
lower frequencies with source frame time delays on the order of BLAZAR; http://www.bu.edu/blazars/BEAM-ME.html), funded by
days−tens of days. We identify a newly ejected superluminal (β app ≈ NASA through the Fermi Guest Investigator Program. The VLBA
12 c) jet component associated with the 2015 flare. Tracking the time is an instrument of the National Radio Astronomy Observatory. The
evolution of this jet component, we find δ v = 29, γ = 17, and θ v = National Radio Astronomy Observatory is a facility of the National
1.4 deg. We evaluated the SSA spectrum from quasi-simultaneous Science Foundation operated by Associated Universities, Inc.

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

radio spectra constructed with GPR, finding the turnover frequency The Fermi LAT Collaboration acknowledges generous ongoing
ν c to vary between 10 and 70 GHz and the peak flux density Sm to support from a number of agencies and institutes that have supported
vary between 0.7 and 4 Jy. The time-evolution trends in the Sm –ν c both the development and the operation of the LAT as well as
parameter space could be consistent with a shock-in-jet origin of scientific data analysis. These include the National Aeronautics
the flaring activity. We evaluated BSSA and BEQ , finding upper limits and Space Administration and the Department of Energy in the
on BSSA in the range of 4–1020 mG and lower limits on BEQ in the United States, the Commissariat à l’Energie Atomique and the Centre
range of 24–105 mG. A single measurement of BSSA = 15.3+ 12.6
−14.0 mG National de la Recherche Scientifique/Institut National de Physique
+10.6
and BEQ = 43.6−10.4 mG was obtained close to the peak of the 2015 Nucléaire et de Physique des Particules in France, the Agenzia
radio flare when the radio core was resolved in 43 GHz VLBI data. Spaziale Italiana and the Istituto Nazionale di Fisica Nucleare in Italy,
From additional investigation of the time-lag core shifts and the the Ministry of Education, Culture, Sports, Science and Technology
equipartition size, we find that the SSA region of AO 0235 + 164 is (MEXT), High Energy Accelerator Research Organization (KEK)
consistent with the equipartition assumption. We infer that the SSA and Japan Aerospace Exploration Agency (JAXA) in Japan, and the
region was approximately 1–10 pc downstream of the jet base during K. A. Wallenberg Foundation, the Swedish Research Council and
the 2015 flare while the gamma-rays were produced close to the jet the Swedish National Space Board in Sweden.
base. Additional support for science analysis during the operations phase
is gratefully acknowledged from the Istituto Nazionale di Astrofisica
in Italy and the Centre National d’Études Spatiales in France. This
AC K N OW L E D G E M E N T S
work performed in part under DOE Contract DE-AC02-76SF00515.
We thank the anonymous reviewer for valuable comments and This research made use of ASTROPY,9 a community-developed core
suggestions that helped to improve the paper. The manuscript was Python package for Astronomy (Astropy Collaboration 2013, 2018).
improved by the helpful comments of Filippo D’Ammando, Tonia WYC, SSL, SHK and SK were supported by the National Research
Venters, Dave Thompson, and Deirdre Horan. This research has made Foundation of Korea (NRF) grant funded by the Korea government
use of data from the OVRO 40-m monitoring programme (Richards (MIST) (2020R1A2C2009003).
et al. 2011), supported by private funding from the California Insitute
of Technology and the Max Planck Institute for Radio Astron-
omy, and by NASA grants NNX08AW31G, NNX11A043G, and DATA AVA I L A B I L I T Y
NNX14AQ89G and NSF grants AST-0808050 and AST- 1109911. The data underlying this article will be shared on reasonable
This publication makes use of data obtained at Metsähovi Radio request to the corresponding author. The Fermi-LAT photon data are
Observatory, operated by Aalto University in Finland. JYK was available through query at the LAT data server (https://fermi.gsfc.n
supported for this research by the National Research Foundation of asa.gov/cgi-bin/ssc/LAT/LATDataQuery.cgi). The 15 GHz OVRO
Korea (NRF) funded by the Korean government (Ministry of Science data may be available on request to the OVRO 40 m collaboration.
and ICT; grant no. 2022R1C1C1005255). IA acknowledges financial For questions regarding the availability of the 37 GHz Metsähovi
support from the Spanish ‘Ministerio de Ciencia e Innovación’ data, please contact agn-metsahovi at aalto.fi. The ALMA data
(MCINN) through the ‘Center of Excellence Severo Ochoa’ award are available at the ALMA Calibrator Source Catalogue, found
for the Instituto de Astrofı́sica de Andalucı́a-CSIC (SEV-2017-0709). at https://almascience.eso.org/sc/. The SMA data are available at
Acquisition and reduction of the MAPCAT data was supported in http://sma1.sma.hawaii.edu/callist/callist.html. Questions regarding
part by MICINN through grants AYA2016-80889-P and PID2019- the data and data policies should be addressed to Mark Gurwell
107847RB-C44. The POLAMI observations were carried out at (mgurwell at cfa.harvard.edu). The 43 GHz VLBA data from the
the IRAM 30m Telescope. IRAM was supported by INSU,CNRS VLBA-BU-BLAZAR programme are available at https://www.bu.e
(France), MPG (Germany), and IGN (Spain). CC acknowledges du/blazars/VLBAproject.html.
support from the European Research Council (0:funding-sou
rce 3:href="http://dx.doi.org/10.13039/100010663"ERC/0:fundin
g-source) under the HORIZON ERC Grants 2021 programme REFERENCES
under grant agreement no. 101040021. This paper makes use of Abdo A. A. et al., 2009, ApJS, 183, 46
the following ALMA data: ADS/JAO.ALMA#2011.0.00001.CAL. Abdollahi S. et al., 2020, ApJS, 247, 33
ALMA is a partnership of ESO (representing its member states),
NSF (USA), and NINS (Japan), together with NRC (Canada), MOST
and ASIAA (Taiwan), and KASI (Republic of Korea), in cooperation 9 http://www.astropy.org

MNRAS 527, 882–894 (2024)

894 W. Y. Cheong et al.
Acciari V. A. et al., 2021, MNRAS, 504, 1427 Kutkin A. M. et al., 2018, MNRAS, 475, 4994
Ackermann M. et al., 2012, ApJ, 751, 159 Lee S.-S., Lobanov A. P., Krichbaum T. P., Witzel A., Zensus A., Bremer M.,
Agudo I., Thum C., Wiesemeyer H., Krichbaum T. P., 2010, ApJS, 189, 1 Greve A., Grewing M., 2008, AJ, 136, 159
Agudo I. et al., 2011, ApJ, 735, L10 Lee S.-S. et al., 2016, ApJS, 227, 8
Agudo I., Thum C., Gómez J. L., Wiesemeyer H., 2014, A&A, 566, A59 Lee J. W., Lee S.-S., Hodgson J. A., Kim D.-W., Algaba J.-C., Kang S., Kang
Agudo I., Thum C., Ramakrishnan V., Molina S. N., Casadio C., Gómez J. J., Kim S. S., 2017, ApJ, 841, 119
L., 2018a, MNRAS, 473, 1850 Lee J. W. et al., 2020, ApJ, 902, 104
Agudo I. et al., 2018b, MNRAS, 474, 1427 Lister M. L. et al., 2016, AJ, 152, 12
Algaba J. C., Nakamura M., Asada K., Lee S. S., 2017, ApJ, 834, 65 Lobanov A. P., 1998, A&A, 330, 79
Algaba J.-C. et al., 2018a, ApJ, 852, 30 Lobanov A. P., 2005, preprint (arXiv:astro-ph/0503225)
Algaba J.-C. et al., 2018b, ApJ, 859, 128 Marscher A. P., 1977, ApJ, 216, 244
Angel J. R. P., Stockman H. S., 1980, ARA&A, 18, 321 Marscher A. P., 1983, ApJ, 264, 296
Astropy Collaboration 2013, A&A, 558, A33 Marscher A. P., Gear W. K., 1985, ApJ, 298, 114
Astropy Collaboration 2018, AJ, 156, 123 Massaro E., Tramacere A., Perri M., Giommi P., Tosti G., 2006, A&A, 448,
Atwood W. B. et al., 2009, ApJ, 697, 1071 861
Boettcher M., Harris D. E., Krawczynski H., 2012, Relativistic Jets from Max-Moerbeck W. et al., 2014a, MNRAS, 445, 428

Downloaded from https://academic.oup.com/mnras/article/527/1/882/7330181 by guest on 12 May 2025

Active Galactic Nuclei. Wiley, Berlin Max-Moerbeck W., Richards J. L., Hovatta T., Pavlidou V., Pearson T. J.,
Christie I. M., Petropoulou M., Sironi L., Giannios D., 2019, MNRAS, 482, Readhead A. C. S., 2014b, MNRAS, 445, 437
65 Nalewajko K., Begelman M. C., Sikora M., 2014, ApJ, 789, 161
Ciprini S., 2015, Astron. Telegram, 7975, 1 Nieppola E., Hovatta T., Tornikoski M., Valtaoja E., Aller M. F., Aller H. D.,
Cohen R. D., Smith H. E., Junkkarinen V. T., Burbidge E. M., 1987, ApJ, 2009, AJ, 137, 5022
318, 577 Park J.-H., Trippe S., 2014, ApJ, 785, 76
Connolly S. D., 2016, Astrophysics Source Code Library, record Pedregosa F. et al., 2011, J. Mach. Learn. Res., 12, 2825
ascl:1602.012 Peterson B. M., Wanders I., Horne K., Collier S., Alexander T., Kaspi S.,
Cooper N. J., Lister M. L., Kochanczyk M. D., 2007, ApJS, 171, 376 Maoz D., 1998, PASP, 110, 660
Edelson R. A., Krolik J. H., 1988, ApJ, 333, 646 Petropoulou M., Giannios D., Sironi L., 2016, MNRAS, 462, 3325
Emmanoulopoulos D., McHardy I. M., Papadakis I. E., 2013, MNRAS, 433, Piner B. G., Bhattarai D., Edwards P. G., Jones D. L., 2006, ApJ, 640, 196
907 Planck Collaboration VI 2020, A&A, 641, A6
Fomalont E. B., 1999, in Taylor G. B., Carilli C. L., Perley R. A., eds, ASP Pushkarev A. B., Kovalev Y. Y., Lister M. L., 2010, ApJ, 722, L7
Conf. Ser. Vol. 180, Synthesis Imaging in Radio Astronomy II. Astron. Pushkarev A. B., Butuzova M. S., Kovalev Y. Y., Hovatta T., 2019, MNRAS,
Soc. Pac., San Francisco, p. 301 482, 2336
Frey S. et al., 2000, PASJ, 52, 975 Raiteri C. M. et al., 2001, A&A, 377, 396
Frey S. et al., 2006, PASJ, 58, 217 Raiteri C. M. et al., 2008, A&A, 480, 339
Fromm C. M. et al., 2011, A&A, 531, A95 Rani B., Wiita P. J., Gupta A. C., 2009, ApJ, 696, 2170
Fromm C. M., Fuhrmann L., Perucho M., 2015, A&A, 580, A94 Rasmussen C. E., Williams C. K. I., 2006, in Gaussian Processes for Machine
Fuhrmann L. et al., 2014, MNRAS, 441, 1899 Learning MIT Press, Cambridge, MA, USA,
Goyal A., 2020, MNRAS, 494, 3432 Richards J. L. et al., 2011, ApJS, 194, 29
Gurwell M. A., Peck A. B., Hostler S. R., Darrah M. R., Katz C. A., 2007, Sagar R., Stalin C. S., Gopal-Krishna, Wiita P. J., 2004, MNRAS, 348, 176
in Baker A. J., Glenn J., Harris A. I., Mangum J. G., Yun M. S., eds, Shepherd M. C., Pearson T. J., Taylor G. B., 1994, BAAS, 26, 987
ASP Conf. Ser. Vol. 375, From Z-Machines to ALMA: (Sub)Millimeter Tarnopolski M., Żywucka N., Marchenko V., Pascual-Granado J., 2020, ApJS,
Spectroscopy of Galaxies. Astron. Soc. Pac., San Francisco, p. 234 250, 1
Hirotani K., 2005, ApJ, 619, 73 Teraesranta H. et al., 1998, A&AS, 132, 305
Hodgson J. A. et al., 2017, A&A, 597, A80 Thum C., Wiesemeyer H., Paubert G., Navarro S., Morris D., 2008, PASP,
Hovatta T., Valtaoja E., Tornikoski M., Lähteenmäki A., 2009, A&A, 494, 120, 777
527 Thum C., Agudo I., Molina S. N., Casadio C., Gómez J. L., Morris D.,
Jorstad S., Marscher A., 2016, Galaxies, 4, 47 Ramakrishnan V., Sievers A., 2018, MNRAS, 473, 2506
Jorstad S. G. et al., 2005, AJ, 130, 1418 Tripathi A., Gupta A. C., Aller M. F., Wiita P. J., Bambi C., Aller H., Gu M.,
Jorstad S. G. et al., 2017, ApJ, 846, 98 2021, MNRAS, 501, 5997
Kang S. et al., 2021, A&A, 651, A74 Weaver Z. R. et al., 2022, ApJS, 260, 12
Karamanavis V. et al., 2016, A&A, 590, A48 Wright E. L., 2006, PASP, 118, 1711
Kataoka J., Stawarz Ł., 2005, ApJ, 622, 797
Kim S.-H. et al., 2022a, MNRAS, 510, 815
Kim D.-W. et al., 2022b, ApJ, 925, 64
Kudryavtseva N. A., Gabuzda D. C., Aller M. F., Aller H. D., 2011, MNRAS,
415, 1631 This paper has been typeset from a TEX/LATEX file prepared by the author.

© 2023 The Author(s).

Published by Oxford University Press on behalf of Royal Astronomical Society. This is an Open Access article distributed under the terms of the Creative Commons Attribution License
(https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

MNRAS 527, 882–894 (2024)