The University of Manchester Research
Spin-down evolution and radio disappearance of the
magnetar PSR J1622-4950
DOI:
10.3847/1538-4357/aa73de
Document Version
Final published version
Link to publication record in Manchester Research Explorer
Citation for published version (APA):
Scholz, P., Camilo, F., Sarkissian, J., Reynolds, J. E., Levin, L., Bailes, M., Burgay, M., Johnston, S., Kramer, M.,
& Possenti, A. (2017). Spin-down evolution and radio disappearance of the magnetar PSR J1622-4950. The
Astrophysical Journal, 841(2). https://doi.org/10.3847/1538-4357/aa73de
Published in:
The Astrophysical Journal
Citing this paper
Please note that where the full-text provided on Manchester Research Explorer is the Author Accepted Manuscript
or Proof version this may differ from the final Published version. If citing, it is advised that you check and use the
publisher's definitive version.
General rights
Copyright and moral rights for the publications made accessible in the Research Explorer are retained by the
authors and/or other copyright owners and it is a condition of accessing publications that users recognise and
abide by the legal requirements associated with these rights.
Takedown policy
If you believe that this document breaches copyright please refer to the University of Manchester’s Takedown
Procedures [http://man.ac.uk/04Y6Bo] or contact uml.scholarlycommunications@manchester.ac.uk providing
relevant details, so we can investigate your claim.
Download date:10. Aug. 2022
Spin-down Evolution and Radio Disappearance of the Magnetar PSR J1622–4950
P. Scholz
1,2
, F. Camilo
3
, J. Sarkissian
4
, J. E. Reynolds
5
, L. Levin
6
, M. Bailes
7,8
, M. Burgay
9
,
S. Johnston
5
, M. Kramer
6,10
, and A. Possenti
9
1
National Research Council of Canada, Herzberg Astronomy and Astrophysics, Dominion Radio Astrophysical Observatory, P. O. Box 248,
Penticton, BC V2A 6J9, Canada; paul.scholz@nrc-cnrc.gc.ca
2
Department of Physics and McGill Space Institute, Rutherford Physics Building, McGill University, 3600 University Street, Montreal, Quebec, H3A 2T8, Canada
3
SKA South Africa, Pinelands, 7405, South Africa
4
CSIRO Parkes Observatory, Parkes, NSW 2870, Australia
5
CSIRO Astronomy and Space Science, Australia Telescope National Facility, Epping, NSW 1710, Australia
6
Jodrell Bank Centre for Astrophysics, School of Physics and Astronomy, The University of Manchester, Manchester M13 9PL, UK
7
Centre for Astrophysics and Supercomputing, Swinburne University of Technology, Mail H30, P. O. Box 218, Hawthorn, VIC 3122, Australia
8
ARC Centre of Excellence for All-Sky Astronomy (CAASTRO), School of Physics, The University of Sydney, NSW 2006, Australia
9
INAF—Osservatorio Astronomico di Cagliari, Via della Scienza 5, I-09047 Selargius (CA), Italy
10
Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, D-53121 Bonn, Germany
Received 2016 December 21; revised 2017 May 13; accepted 2017 May 15; published 2017 June 5
Abstract
We report on 2.4 yr of radio timing measurements of the magnetar PSRJ1622−4950 using the Parkes
Observatory, between 2011 November and 2014 March. During this period the torque on the neutron star (inferred
from the rotational frequency derivative) varied greatly, though much less erratically than during the 2 yr following
its discovery in 2009. During the last year of our measurements the frequency derivative decreased in magnitude
monotonically by 20%, to a value of −1.3×10
−13
s
−2
, a factor of 8 smaller than when it was discovered. The flux
density continued to vary greatly during our monitoring through 2014 March, reaching a relatively steady low level
after late 2012. The pulse profile varied secularly on a similar timescale as the flux density and torque. A relatively
rapid transition in all three properties was evident in early 2013. After PSRJ1622−4950 was detected in all of our
87 observations up to 2014 March, we did not detect the magnetar in our resumed monitoring starting in 2015
January and have not detected it in any of the 30 observations conducted through 2016 September.
Key words: pulsars: general – pulsars: individual (PSR J1622–4950) – stars: magnetars – stars: neutron
1. Introduction
Magnetars are a class of neutron stars with extremely high
magnetic fields (B ∼ 10
13−15
G) and long spin periods (2–12 s).
Their high-energy emission is powered via the decay of their
magnetic fields, rather than through rotation. This is revealed
through large outbursts and X-ray luminosities that exceed the
available rotational spin-down luminosity (for reviews see
Woods & Thompson 2006;Mereghetti2008). During outburst,
magnetars can increase their X-ray fluxes by orders of magnitude
and then fade on a timescale of months to years.
Most magnetars have been discovered and monitored in
X-rays. The best characterized are those that have been
monitored for more than 15 years with the Rossi X-Ray Timing
Explorer (Dib & Kaspi 2014), now continued by Swift (e.g.,
Archibald et al. 2013, 2015). This may be a biased sample, as
only five of the 23 known magnetars (Olausen & Kaspi 2014)
11
are persistently bright enough to be monitored in this way (Dib
& Kaspi 2014). In order to expand our understanding of
magnetars it is desirable to perform detailed, long-term
monitoring of the rotational and radiative behavior of more
objects.
Radio emission has been detected from only four of the
known magnetars, but it is often quite bright (e.g., Camilo et al.
2007b; Shannon & Johnston 2013). We can therefore expand
the sample of well-characterized magnetars by performing
long-term monitoring using radio telescopes. The study of
radio emission from magnetars also provides a new
electromagnetic window into the behavior of these most
magnetic objects known.
XTEJ1810−197 was the first magnetar to be detected in
radio (Camilo et al. 2006), followed shortly thereafter by
1E1547.0−5408 (Camilo et al. 2007b). They were found to
have highly variable radio flux densities and pulse profiles,
unlike ordinary pulsars. Very unusual compared to standard
radio pulsars, their radio spectra are generally flat (e.g., Camilo
et al. 2007c). Both are transient radio sources: radio emission
from XTEJ1810−197 turned off in 2008 (Camilo et al. 2016)
and 1E1547.0−5408 was detected intermittently following its
2009 outburst (Burgay et al. 2009; Camilo et al. 2009). The
third radio magnetar, PSRJ1622−4950, is the subject of this
work. A fourth was more recently discovered 2″ away from the
Galactic center and is the closest known pulsar to SgrA
*
(Eatough et al. 2013; Shannon & Johnston 2013).
PSRJ1622−4950 was discovered with the CSIRO Parkes
telescope as a radio pulsar with a period of P=4.3 s and a
dispersion measure of DM=820 pc cm
−3
(Levin et al. 2010).
To date it remains the only magnetar to have been detected in
radio without prior knowledge of a corresponding X-ray
source. Like other radio magnetars it has a flat spectrum, nearly
100% linear polarization, and highly variable flux density and
pulse profiles. Its rotational behavior following discovery was
characterized by Parkes observations between 2009 April and
2011 February (Levin et al. 2012). Long-term phase-connected
timing solutions were not possible due to the rapidly evolving
spin-frequency derivative,
n
˙
, and an insufficient observing
cadence. From short-term timing solutions,
n
∣
˙
∣
was found to
have decreased by a factor of 2 in the 2 yr following discovery.
The Astrophysical Journal, 841:126 (8pp), 2017 June 1 https://doi.org/10.3847/1538-4357/aa73de
© 2017. The American Astronomical Society. All rights reserved.
11
See the online Magnetar Catalog at http://www.physics.mcgill.ca/~pulsar/
magnetar/main.html.
1
PSRJ1622−4950 was i den tified as an X-ray source using
archival and dedicat ed Ch andr a and XMM-Newton observa-
tions. Its X-ray flux decreased by a f actor of ∼50 between
2007 June and 2011 February, presum abl y followi ng a
predisco ver y outburst. X-ray pulsatio ns have not been
detected, implying a 70% limit on the pulsed fraction
(Anders on et al. 2012).
Here we present the analysis and results of an additional
2.4 yr of Parkes observations of PSRJ1622−4950. We
describe our data set in Section 2. In Sections 3.1 and 3.2 we
show the pulse profile and flux density evolution of the source.
In Section 3.3 we present a timing analysis and the resulting
phase-connected timing solutions. We discuss our results in
Section 4 and conclude in Section 5.
2. Observations
We observed PSRJ1622−4950 with the Parkes telescope on
a regular basis between 2011 November and 2014 March.
These observations were typically conducted on the same days
as those when we monitored the magnetar 1E1547.0−5408,
which was largely observed at frequencies near 3 GHz because
severe scattering renders its pulse hard to detect at 1.4 GHz
(Camilo et al. 2007b). We made a total of 87 observations on
81 days, 90% of them at 3 GHz using the 10–50 cm receiver,
and the remainder at 1.4 GHz using the center beam of the
20 cm multibeam receiver (Staveley-Smith et al. 1996), once
every 10 days on average. Integration times were typically 5 or
10 minutes per observation.
A total of 69 observations through 2013 September were
conducted with the analog filterbank system (AFB; see, e.g.,
Manchester et al. 2001), used to sample a bandwidth of
864 MHz centered on 3078 MHz or a 288 MHz band
centered on 1374 MHz. In each case the individual channel
widths were 3 MHz. In 2013 April we began using PDFB3, a
digital filterbank (DFB )
, centered on 3100 MHz to sample
512 2 MHz-wide channels. In all cases we recorded total-
intensity (polarization-summed) search-mode data using
1ms samples.
Each of the data sets was subsequently dedispersed and
folded using a known ephemeris (Levin et al. 2010). Each
folded observation was inspected for frequency channels and
subintegrations that were highly contaminated by radio
frequency interference (RFI). The contaminated channels and
subintegrations were then masked in all subsequent analysis.
PSRJ1622−4950 was detected in every observation we
made during 2011–2014. We resumed observations on 2015
January 11 but have not detected the pulsar in any of 30 epochs
through 2016 September 16. These observations, largely at
3 GHz using the PDFB4 DFB, lasted for 15 minutes on
average.
In addition to the new observations conducted between 2011
and 2014, we use the flux densities and pulse times-of-arrival
(TOAs) reported in Levin et al. (2012) from Parkes observa-
tions between 2009 and 2011. We also utilize 26 archival
observations made between the data set presented in Levin
et al. (2012) and the beginning of our campaign. These
observations included 15 observations at 1.4 GHz and three
observations at 3.1 GHz using PDFB3/4, and eight 1.4 GHz
observations with the CASPER-Parkes-Swinburne Recorder.
3. Analysis and Results
3.1. Pulse Profile Variations
Similar to what was observed between 2009 and 2011 by
Levin et al. (2012), the pulse profile of PSRJ1622−4950 in
our observations is made up of multiple components that vary
in relative amplitude and separation over time. Figure 1 shows
the profiles for all of our 2011–2014 observations. Long-period
pulsars observed with the AFB system display artifacts caused
by a high-pass filter with a ≈0.9 s time constant. We used the
prescription given by Manchester et al. (2001) to correct for
this effect in the profiles presented in Figure 1.
In late 2011, the profiles were clearly composed of two
peaks, with the second fainter than the first. In late 2012, the
pulsar became significantly fainter (see Section 3.2) and was
more affected by RFI. Often the profile could only be resolved
as a broad single peak. This persisted until late 2013, when the
flux density increased slightly and the pulse profile narrowed
(Figure 1).
To quantify the narrowing of the pulse profile, we fit a two-
Gaussian model to the profiles. The model fit to the profiles is
PA A
A
,,, exp
2
exp
2
,1
i
i
i 1
1
2
1
2
2
2
2
2
2
fms
fm
s
fm
s
=
--
+
--
()
()
()
()
where A
i
are the amplitudes, μ
i
are the peak phases, and σ
i
are
the widths of the Gaussian components. The FWHM of the
components is
2
2ln2
i
s
. The results of these fits are shown in
Figure 2 and clear evolution is evident. The leading component
remained relatively constant in width and the trailing Gaussian
component became narrower and closer in phase to the first
component as a function of time. This change was occurring on
a similar timescale as the radio flux density decrease (see
Section 3.2).
3.2. Flux Density Evolution
Our filterbank data were not flux calibrated. Nevertheless,
we could extract useful flux density measurements by
computing the area under each profile and scaling it to a
Jansky scale using the system equivalent flux density (SEFD)
at the location of the pulsar. First we set the off-pulse level to
zero and normalized the summed pulse profile counts by the
off-pulse rms. We then scaled the profile into units of flux
density using the off-pulse rms from the radiometer equation
(Dewey et al. 1985):
nt f
SEFD
,2
p int
b
D
()
where β is a loss factor due to the digitization of the signal (1.5
for the AFB, 1.1 for the DFB), n
p
=2 is the number of
polarizations summed, t
int
is the integration time per phase bin,
and Δf is the bandwidth. We determined the SEFD by
analyzing with PSRCHIVE (Hotan et al. 2004) full-Stokes
calibrated observations made with PDFB3. The SEFD at
3.1 GHz was 61 Jy (based on an observation done on 2011
December 12), while the SEFD at 1.4 GHz was 69 Jy (based on
an observation done on 2010 November 3), both measured with
about a 5% precision. Flux densities measured in this way are
2
The Astrophysical Journal, 841:126 (8pp), 2017 June 1 Scholz et al.
shown in Figure 3. In the absence of residual RFI and other
profile artifacts, we estimate that the absolute fractional
uncertainty for each measurement was ≈25%. However, some
profiles were contaminated by residual RFI (see Figure 1).To
address this, we made the measurements using two independent
tools: one in which the off-pulse regions are chosen arbitrarily
by a user and one in which the off-pulse regions are determined
automatically by growing the off-pulse region until the
variance of the off-pulse data changes by more than 10%.
These two tools yield different off-pulse baseline estimations,
normalizations, and flux density values. While on occasion the
two measures differed by up to ≈50%, in most cases they
agreed more closely and these discrepancies do not affect the
trends visible in Figure 3.
Following the PSRJ1622−4950 discovery, Levin et al.
(2010) realized that prediscovery observations existed for the
years 1998–2006 in the form of archival search-mode data for
two nearby pulsars: PSRJ1623−4949 (11′ away) and
PSRJ1622−4944 (7′ away). When pointing at the latter,
PSRJ1622−4950 was near the half-power point of the Parkes
1.4 GHz primary beam (FWHM = 14
4), with a reduction in
sensitivity by a factor of 1.8. When pointing at the former, the
sensitivity was reduced by a factor of 6.3 assuming a Gaussian
beam (which may not be appropriate so far off boresight).
Astonishingly, Levin et al. (2010) recovered many bright
detections of PSR J1622−4950 and estimated flux densities,
even that far off axis. We have reanalyzed those data (52
individual AFB observations) in the same manner as for our
new data set in order to place both sets of detections on the
same flux density scale. We made 14 detections in prediscovery
data, the same as Levin et al. (2010).
In Figure 3 we also include the flux densities corresponding to
the data presented in Levin et al. (2010). However, Levin et al.
(2010) used
TSEFD Gain 24 K 0.735 K Jy 33
sys
1
== =
-
Jy,
which is a factor of 2.1 less than our measured value of 69 Jy at
1.4 GHz. In addition, they used a loss factor of β=1.0, while
β=1.1 for the DFB data and β= 1.5 for the AFB data. Thus we
multiply the flux density values presented in Levin et al. (2010) by
2.3 for the DFB data (obtained during 2009–2011) and 3.1 for the
AFB data (prior to 2007). Our SEFD was measured at the position
ofPSRJ1622−4950, and the correction factor for the pre-2007
data (during which the telescope was pointed several arcminutes
away fromPSRJ1622−49 50) is therefore uncertain. Nevertheless,
Figure 1. Radio pulse profiles of PSRJ1622−4950. Black (gray) profiles correspond to 3 GHz (1.4 GHz) observations. The full pulse period of 4.3 s is displayed,
with 64 phase bins, and profiles are arbitrarily aligned. We list the MJD and integration time (in minutes) of each observation, along with calendar dates in select
instances. Those observations that used a DFB are denoted by a “D.” All other profiles, obtained with an AFB, have been corrected to account for instrumental
artifacts (see Section 3.1). Some profiles remain somewhat contaminated by RFI.
3
The Astrophysical Journal, 841:126 (8pp), 2017 June 1 Scholz et al.
we judge that our SEFD is a closer approximation to the true value
thanthecold-skyvalueassumedinLevinetal.(2010).
All flux density measurements are summarized in Figure 3.
Many of the prediscovery values are much larger, as well as
more variable, than the more recent ones. Another notable point
is that we have numerous detections in 2012–2014 with flux
densities below the (corrected) 3.8 mJy detection limit of the
off-axis observations (Levin et al. 2010 used a limit of
1.2 mJy). Therefore, it is quite possible that some nondetections
for the 1998–2006 period (Levin et al. 2010) simply reflect a
lack of sensitivity and that the pulsar would have been detected
had it been observed on-axis. Those nondetections hence do
not necessarily imply a turnoff in radio emission. By contrast,
our consistent nondetections starting in 2015 (Section 2) reflect
a different state compared to those noted in 2009–2014. For the
first time in the study of PSRJ1622−4950, we can entertain
the possibility that the radio emission effectively turned off or
at least transitioned to a significantly fainter state.
3.3. Phase-coherent Timing
In principle, the timing of radio magnetars presents particular
challenges owing to the varying pulse profiles. In practice, for
PSRJ1622−4950 this did not present substantial difficulties
for the post-2011 data used in this paper.
To account for coarse changes in pulse profiles, we used
three separate templates for TOA extraction. For observations
prior to MJD56250, we used the profile observed on
MJD55924 as the template. For AFB observations from
MJD56250 onward, the MJD56284 profile was used. Finally,
the MJD56685 profile was used to extract TOAs from all DFB
observations (see Figure 1). All TOAs were obtained with the
PRESTO (Ransom et al. 2002 ) tool get_TOAs.py.
In order to quantify the effect of the evolving profiles on the
accuracy of the TOAs, we also extracted TOAs using templates
built from multi-Gaussian fits to the profiles observed on
MJD55924 (for the AFB data) and MJD56669 (for the DFB
data). We then measured the difference between the corresp-
onding original TOAs and the Gaussian-template TOAs. The
standard deviation of these differences ranged over 20–40 ms
for AFB TOAs and was 30 ms for the DFB TOAs, i.e., about
1% of the pulse period. We added these standard deviations in
quadrature to our nominal TOA uncertainties to account for the
error introduced in timing the pulsar with a restricted set of
templates in the face of varying pulse pro files. One observation,
on MJD 56545, was too faint to provide a reliable TOA.
The TOAs were fit to a timing model describing the pulsar
rotation where the pulse phase as a function of time is
described by a Taylor series expansion. Initially, only the spin
frequency ν=1/P was fit for, to a set of four TOAs extracted
from each observation. A frequency derivative
n
˙
was estimated
from those measurements and was used as a starting point for
the iterative process of long-term phase connection using
TEMPO2 (Hobbs et al. 2006). For the final fits we extracted one
TOA per observation in order to improve parameter precision.
Using simple timing models with only ν,
n
˙
, and
¨
n
,itis
possible to phase-connect the data set in two separate date
ranges. These solutions are shown in Table 1. In Figures 4(a)
and (b), the
n
˙
evolution and the phase residuals of these two
solutions are shown in red and blue. In order to probe the
evolution of
n
˙
in more detail, we also fit short-term overlapping
timing models using only ν and
n
˙
. Each short-term model was
fit over a minimum of five observations spanning a minimum
of 61 days and a maximum of 100 days. The resulting values of
n
˙
are shown in Figure 4(a), where the horizontal bars represent
the time span of the fits.
Formal pulsar TOA uncertainties obtained from cross-
correlating observed profiles with templates are often some-
what underestimated. It is therefore standard practice to
increase the TOA errors by a scaling error factor (EFAC) that
yields a reduced χ
2
≡1, ensuring more realistic parameter
uncertainties. We determined that for our data set EFAC=1.3
by considering short-term timing solutions where the effects of
timing noise were negligible.
To probe the timing evolution between the end of the data set
from Levin et al. (2012) and the beginning of our 2011–2014
campaign, we also extracted TOAs from 26 archival observa-
tions (Section 2) using PSRCHIVE’s pat utility. We fit timing
solutions with ν and
n
˙
to these data in two time spans where
phase connection was possible. These frequency-derivative
measurements are shown as red crosses in Figure 5.
4. Discussion
PSRJ1622−4950 is the only magnetar whose rotation has
been studied exclusively at radio wavelengths. Much of what
we know about its radiative behavior also relies on radio
observations, but note that we know that its X-ray flux
decreased by a factor of 50 between mid-2007 and early 2011,
with an exponential timescale of 1 yr, following a presumed
earlier outburst (Anderson et al. 2012). The high-cadence
Parkes monitoring observations that we have presented here
along with previously published radio results (Levin
et al. 2010, 2012) allow us to consider the evolution of
PSRJ1622−4950 over many years and to place it in the
context of other magnetars.
Figure 2. Gaussian fits to the pulse profiles of PSRJ1622−4950. Top panel:
two example profiles are shown with their two-component Gaussian fits shown.
Middle panel: the FWHM of the two Gaussian components for each profile.
Bottom panel: the separation between the peak phases of the two Gaussian
components for each pro file. The gray bar marks the period between MJD
56300 and MJD 56400 in order to illustrate the correlated behavior between the
flux density (Figure 3), spin-down (Figure 4), and pulse profile evolution at
that time.
4
The Astrophysical Journal, 841:126 (8pp), 2017 June 1 Scholz et al.
