The High Time Resolution Universe Pulsar Survey – IV. Discovery and polarimetry of millisecond pulsars

Abstract

We present the discovery of six millisecond pulsars (MSPs) in the High Time Resolution Universe (HTRU) survey for pulsars and fast transients carried out with the Parkes radio telescope. All six are in binary systems with approximately circular orbits and are likely to have white dwarf companions. PSR J1017−7156 has a high flux density and a narrow pulse width, making it ideal for precision timing experiments. PSR J1446−4701 and PSR J1125−5825 are coincident with gamma-ray sources, and folding the high-energy photons with the radio timing ephemeris shows evidence of pulsed gamma-ray emission. PSR J1502−6752 has a spin period of 26.7 ms, and its low period derivative implies that it is a recycled pulsar. The orbital parameters indicate it has a very low mass function, and therefore a companion mass much lower than usually expected for such a mildly recycled pulsar.In addition we present polarization profiles for all 12 MSPs discovered in the HTRU survey to date. Similar to previous observations of MSPs, we find that many have large widths and a wide range of linear and circular polarization fractions. Their polarization profiles can be highly complex, and although the observed position angles often do not obey the rotating vector model, we present several examples of those that do. We speculate that the emission heights of MSPs are a substantial fraction of the light cylinder radius in order to explain broad emission profiles, which then naturally leads to a large number of cases where emission from both poles is observed.

Full text

Page 1

arXiv:1109.4193v2 [astro-ph.SR] 5 Oct 2011
Mon. Not. R. Astron. Soc. 000, 000–000 (0000)Printed 6 October 2011(MN LATEX style file v2.2)
The High Time Resolution Universe Pulsar Survey IV:
Discovery and polarimetry of millisecond pulsars
M. J. Keith1⋆, S. Johnston1, M. Bailes2,3,4, S. D. Bates5, N. D. R. Bhat2,4,
M. Burgay6, S. Burke-Spolaor1, N. D’Amico6, A. Jameson2, M. Kramer7,5,
L. Levin2,1, S. Milia6,8, A. Possenti6, B. W. Stappers5, W. van Straten2,4and
D. Parent9
1Australia Telescope National Facility, CSIRO Astronomy & Space Science, P.O. Box 76, Epping, NSW 1710, Australia
2Swinburne University of Technology, Centre for Astrophysics and Supercomputing Mail H39, PO Box 218, VIC 3122, Australia
3University of California, Berkeley, 601 Campbell Hall 3411, Berkeley, CA 94720, USA
4ARC Centre of Excellence for All-sky Astrophysics (CAASTRO)
5University of Manchester, Jodrell Bank Centre for Astrophysics, Alan Turing Building, Manchester M13 9PL, UK
6INAF-Osservatorio Astronomico di Cagliari, localit` a Poggio dei Pini, strada 54, I-09012 Capoterra, Italy
7Max Planck Institut f¨ ur Radioastronomie, Auf dem H¨ ugel 69, 53121 Bonn, Germany
8Dipartimento di Fisica, Universit` a degli Studi di Cagliari, Cittadella Universitaria, 09042 Monserrato (CA), Italy
9Center for Earth Observing and Space Research, College of Science, George Mason University, Fairfax, VA 22030, USA
ABSTRACT
We present the discovery of six millisecond pulsars (MSPs) in the High Time Reso-
lution Universe (HTRU) survey for pulsars and fast transients carried out with the
Parkes radio telescope. All six are in binary systems with approximately circular or-
bits and are likely to have white dwarf companions. PSR J1017–7156 has a high flux
density and a narrow pulse width, making it ideal for precision timing experiments.
PSRs J1446–4701 and J1125–5825 are coincident with gamma-ray sources, and fold-
ing the high-energy photons with the radio timing ephemeris shows evidence of pulsed
gamma-ray emission. PSR J1502–6752 has a spin period of 26.7 ms, and its low period
derivative implies that it is a recycled pulsar. The orbital parameters indicate it has
a very low mass function, and therefore a companion mass much lower than usually
expected for such a mildly recycled pulsar.
In addition we present polarisation profiles for all 12 MSPs discovered in the
HTRU survey to date. Similar to previous observations of MSPs, we find that many
have large widths and a wide range of linear and circular polarisation fractions. Their
polarisation profiles can be highly complex, and although the observed position angles
often do not obey the rotating vector model, we present several examples of those that
do. We speculate that the emission heights of MSPs are a substantial fraction of the
light cylinder radius in order to explain broad emission profiles, which then naturally
leads to a large number of cases where emission from both poles is observed.
Key words: pulsars: general — pulsars: individual: (PSR J1017–7156, PSR J1337–
6423, PSR J1446–4701, PSR J1502–6752, PSR J1543–5149, PSR J1622–6617)
1INTRODUCTION
Radio pulsars are typically classified in terms of the observed
parameters of pulse period and its derivative, placing them
on the ‘P −˙P diagram’. They are generally thought to be
born with spin periods of a few tens of milliseconds and
⋆Email: mkeith@pulsarastronomy.net
magnetic field strengths of the order 1012G. As the pul-
sar ages, the rotational energy is radiated away and the
rotation rate slows, and therefore the pulsar traces a line
across the P −˙P diagram, until it reaches the ‘death line’
where radio emission ceases. A distinct sub-group of radio
pulsars, the “millisecond pulsars” (MSPs) typically have
spin periods less than ∼ 30 ms, often just a few millisec-
onds. These periods are shorter than the youngest pulsars,
however their small period derivatives indicate spin-down

Page 2

2
M. J. Keith et al.
ages of > 109years and they are most often found in bi-
naries with white dwarf (WD) or neutron star (NS) com-
panions. The evolution of MSPs must therefore follow a dif-
ferent path to the majority of pulsars. The accepted the-
ory states that MSPs are initially formed in the same way
as other pulsars, however after most are thought to have
spun down to longer periods, they are then ‘recycled’ via
accretion of mass and angular momentum from a compan-
ion star (Alpar et al. 1982). This allows the pulsar to at-
tain very short spin periods (although this is not always the
case), and the accretion process is thought to also reduce
the dipole magnetic field strength and therefore reduce the
observed period derivative (Bisnovatyi-Kogan & Komberg
1974; Taam & van den Heuvel 1986; Cumming et al. 2001).
The majority of MSPs have spin periods less than 10 ms,
and are either solitary or in circular orbits with a WD, or
a very low mass companion (< 0.03M⊙). Longer spin pe-
riod MSPs tend to have larger mass companions. Within
10 < P < 20 ms the companions are always WDs, however
above 20 ms there are a broad range of systems including
WD, NS (Burgay et al. 2003) and even Be star companions
(Johnston et al. 1992). At the time of writing, the pulsar
catalogue1lists 220 pulsars with a spin period2less than
30 ms, of which only 95 are located outside of globular clus-
ters. Although the majority of the observed MSPs are lo-
cated in globular clusters, the fact that many are formed
via exchange interactions, and the effects of the cluster grav-
itational potential means that their evolutionary paths are
different. In this paper we only consider Galactic-field MSPs.
Only 2% of pulsars with periods greater than 30 ms are in
binaries, however there are 63 binary systems in the 95 non-
cluster MSPs.
The pulse profiles of the majority of MSPs share
many characteristics with those of the classically ‘young’
pulsars,i.e.thosewithP
10−15(Thorsett & Stinebring 1990). The profiles of MSPs
are typically broad with polarisation fraction between ∼
1% and ∼ 50% (Xilouris et al. 1998; Stairs et al. 1999;
Ord et al. 2004; Yan et al. 2011). One way to understand
the polarisation of pulsars is through the ‘rotating vector
model’ (RVM; Radhakrishnan & Cooke 1969) in which the
observed polarisation angle is defined by the direction of a
dipolar magnetic field at the point of emission. The model
is defined in terms of the geometric parameters α, the an-
gle between the rotation axis and the magnetic axis, and
β, the angle between the magnetic axis and line of sight
at closest approach. For some pulsars, particularly those
with components separated by ∼ 180◦this model has been
used to measure these geometric parameters with high pre-
cision (e.g. Kramer & Johnston 2008; Keith et al. 2010b).
Although many MSPs have emission over a very wide phase
window, sometimes greater than 180◦, the position angles
(PAs) often have large deviations from the best fit to the
RVM, or the RVM simply does not model the observations
well at all (Ord et al. 2004; Yan et al. 2011).
As well as the open questions intrinsic to the for-
<100msand
˙P>
1http://www.atnf.csiro.au/research/pulsar/psrcat/
2Even though recycling process can produce pulsars with spin
periods longer than 30 ms, we choose this limit to easily distin-
guish MSPs from young pulsars.
mation, evolution and emission of MSPs, they can also
be valuable tools for other physics and astrophysics ex-
periments. Recently many radio MSPs have been discov-
ered to have gamma-ray pulsations (Abdo et al. 2009b),
and indeed searches of Fermi Large Area Telescope (LAT)
gamma-ray sources have led to the discovery of MSPs with
both radio and gamma-ray emission (Ransom et al. 2011;
Cognard et al. 2011; Keith et al. 2011). This opens new pos-
sibilities for studying the broad-band emission phenomenon
of pulsar and comparison between young pulsars and MSPs.
Although gamma-ray pulsations have been detected from
some young pulsars without radio emission (Abdo et al.
2009a; Saz Parkinson et al. 2010), the gamma rays from
most MSPs can only be binned into pulse profiles with the
aid of a timing ephemeris obtained through radio timing.
The high precision time of arrival measurements that can
be obtained from fast spinning MSPs, and their typically
stable rotation rates mean that they can be used as precise
clocks for tests of gravity and as broadband pulsed signals
for probing the interstellar medium. Of particular interest
is the use of an array of MSPs to constrain the solar sys-
tem ephemeris (Champion et al. 2010) and for the detection
of low-frequency gravitational waves (Yardley et al. 2011;
van Haasteren et al. 2011).
Research into the origin of the MSPs, the physics of
their emission mechanism, and their use as tools for other as-
trophysical experiments are all facilitated by expanding the
known sample. To achieve this, the High Time Resolution
Universe (HTRU) survey for pulsars and fast transients be-
gan at the Parkes radio telescope in 2008. The high time and
frequency resolution of the survey makes it much more sen-
sitive to MSPs than previous efforts at Parkes (Keith et al.
2010a). Indeed, we have already announced the discovery of
6 MSPs (Bates et al. 2011; Bailes et al. 2011). In this paper
we present the discovery and timing solutions for a further
6 MSPs, and discuss their origins (Section 2). We also anal-
yse Fermi LAT data associated with PSRs J1125–5825 and
J1406–4701, both of which are coincident with LAT sources
(Section 3). In Section 4 we present high quality polarised
profiles for all 12 MSPs discovered by the HTRU survey to
date and discuss the implications in Section 5.
2DISCOVERY OF SIX MSPS
Here we report on the discovery of six MSPs in the HTRU
survey, namely PSRs J1017–7156, J1337–6423, J1446–4701,
J1502–6752, J1544–5149 and J1622–6617. All six MSPs are
in almost circular orbits, consistent with the standard model
in which the spin-up of the pulsar is associated with Roche
Lobe overflow from a nearby companion that circularised the
orbit. The companions are either white dwarfs or very low
mass companions, such as those in orbit around PSR J2051–
0827 (Stappers et al. 1996). We note that all 12 MSPs dis-
covered by the HTRU survey to date are part of binary sys-
tems, though there does not seem to be any obvious reason
for the HTRU survey to preferentially select binaries.
2.1 Observations and timing
Each of these six pulsars was discovered in 540 s integrations
as part of the mid-latitude portion of the HTRU survey be-

Page 3

HTRU IV: Discovery & polarisation of MSPs
3
ing carried out at the Parkes radio telescope (see Keith et al.
2010a for details). Follow up observations were carried out at
Parkes with the centre beam of the ‘HI-Multibeam’ receiver
and the Berkley-Parkes-Swinburne Recorder (BPSR), or
once the orbital parameters were identified, the 3rd Parkes
Digital Filterbank system (PDFB3). The BPSR observa-
tions had an effective bandwidth of 340 MHz centred at
1352 MHz and the PDFB3 observations had a bandwidth of
256 MHz centred at 1369 MHz. PDFB3 observations were
calibrated for differential gain between the linear feeds by
means of observation of a pulsed noise diode. Each obser-
vation was averaged in time, frequency and converted to
Stokes total intensity, then matched with an analytic ref-
erence profile to produce a single time-of-arrival measure-
ment for each observation. A model of the rotational, astro-
metric and orbital parameters was then fit to these mea-
surements for each pulsar, in accordance with the stan-
dard pulsar timing procedure, using the tempo2 software
(Hobbs et al. 2006). Due to the almost circular nature of the
orbits we have used the ‘ELL1’ binary model as described in
Lange et al. (2001). The best fit timing parameters and ad-
ditional derived and observed properties of these six pulsars
are presented in Tables 1 and 2. Flux densities presented
are averaged over all 1369 MHz observations and are cali-
brated by comparison to the continuum radio source 3C218.
We do not quote a standard error in this value since all of
these pulsars exhibit significant flux density variations due
to interstellar scintillation. Also shown in the table is the oft
used gamma-ray detectability measure˙E1/2/d2(Abdo et al.
2010). MSPs detected by the Fermi satellite typically have
˙E1/2/d2> 1010erg1/2pc−2s−1/2(Abdo et al. 2009b). The
measured frequency derivative can be biased due to motion
transverse to the line of sight (Shklovskii 1970). No proper
motion has yet been measured in these systems, however we
can assume a typical velocity of 100kms−1(Toscano et al.
1999) and the DM derived distance to compute a likely
fraction of the frequency derivative which could be due to
this effect. This is largest for PSR J1017–7156, with this
value ∼ 0.3 ˙ ν, however the other five pulsars have a likely
Shklovskii contribution less than 0.2 ˙ ν.
2.2 Two low mass binary pulsars:
PSR J1017–7156 and PSR J1543–5149
As seen in Tables 1 and 2, PSR J1017–7156 and J1543–5149
have similar properties and fall into the category of ‘low
mass binary pulsars’ (LMBPs; Phinney & Kulkarni 1994;
Edwards & Bailes 2001). Other than the low mass function
(companion mass 0.1 < Mc < 0.5 M⊙), these pulsars are
typified by an orbital period of a few days and a rapid ro-
tation rate. It should be noted however that there are a
number of sources which fall into this category with longer
spin periods (Edwards & Bailes 2001). LMBPs are by far
the most common class of binary MSPs, with some 30 (of
57) such pulsars known.
PSR J1017–7156 has a spin period of 2.34 ms and a DM
of 94.2 cm−3pc and was discovered in an observation centred
∼ 0.07◦from the nominal pulsar position, carried out on
2010-01-02 and detected with a signal-to-noise ratio (S/N)
of 54. The candidate was initially identified by a neural net
system similar to that described in Eatough et al. (2010).
Follow up observations quickly identified the pulsar to be
in a 6.5 day circular orbit around a low mass companion.
The properties of the binary are similar to those seen in
other LMBPs. The high S/N, short pulse period and narrow
profile suggest that this pulsar is a good candidate for high
precision timing. Indeed, we can see that that this pulsar
gives the best RMS residual of any HTRU discovery to date
and it is already included as a target for the Parkes Pulsar
Timing Array (Hobbs et al. 2009).
PSR J1017–7156 has a mean flux density of 0.89 mJy
(the highest of any HTRU MSP), however even though its
position was covered by previous surveys the relatively large
DM ensured that it was not discernible from radio frequency
interference due to excessive dispersion measure smearing.
To demonstrate this, folding the closest pointing in the
(Edwards et al. 2001) survey shows a signal with a S/N of
11 and a pulse width of 1.3 ms which is broader than half
the pulse period. The superior frequency resolution of the
BPSR back-end allowed for an easy detection with a pulse
width of 0.16 ms width and a S/N of 54. This clearly demon-
strates the increased sensitivity of the BPSR back-end for
pulsars with a high DM/P ratio.
PSR J1543–5149 has a pulse period of 2.06 ms, orbital
period of 8 days and was first detected with a S/N of 16 in
an observation carried out on 2010-08-24, offset 0.08◦from
the nominal pulsar position. The minimum companion mass
is 0.22 M⊙, and so this pulsar is likely to be another member
of the LMBP group.
2.3 An intermediate mass binary pulsar:
PSR J1337–6423
Intermediate mass binary pulsars (IMBPs) have heavier
companion masses than the LMBPs and typically have spin
periods of order 10 ms. PSR J1337–6423 is a 9.4 ms pulsar
with a DM of 260 cm−3pc, the second largest DM of any
MSP known to date. It is located 0.04◦from the centre of
the survey beam in which it was discovered with a S/N of
11. The pulsar is in a 4.8 day orbit with an eccentricity of
< 1 × 10−4. These properties, and the minimum compan-
ion mass of 0.74 M⊙ suggest that PSR J1337–6423 is likely
to have a heavy WD companion and fall into the class of
IMBPs.
2.4A very low mass binary pulsar:
PSR J1446–4701
PSR J1446–4701 is a 2.2 ms pulsar in a compact binary
with a 6.6 hour orbital period. The pulsar was discov-
ered in a survey beam centred 0.1◦from the nominal pul-
sar position on 2009-08-20 with a S/N of 12.4. The min-
imum companion mass is just 0.019 M⊙. With the recent
discovery of PSRs J0610–2100, J1731–1845, J2214+3000
and J2241–5236 (Burgay et al. 2006; Bates et al. 2011;
Ransom et al. 2011; Keith et al. 2011), all of which have
very light companions, it is clear that the so-called “Black
Widow” systems (Klu´ zniak et al. 1988; Phinney et al. 1988;
van den Heuvel & van Paradijs 1988) are members of a
larger class of very low mass binary pulsars (VLMBPs;
Freire et al. 2003). These systems all show spin periods less
than ∼ 4 ms, orbital periods less than ∼ 10 hours and a large
˙E. Although many of these systems show eclipses for some

Page 4

4
M. J. Keith et al.
Table 1. Observed parameters from timing of three previously unknown MSPs. Values in parenthesis are the nominal 1-σ tempo2
uncertainties in the last digits.
ParameterJ1017−7156J1337−6423 J1446−4701
Right ascension, α (J2000) ..............
Declination, δ (J2000) ...................
l (◦) .....................................
b (◦).....................................
Pulse frequency, ν (s−1) .................
Frequency derivative, ˙ ν (s−2) ............
Epoch of model (MJD) ..................
Dispersion measure, DM (cm−3pc) ......
10:17:51.32832(5)
−71:56:41.64003(12)
291.56
−12.55
427.62190510627(9)
−4.75(6)×10−16
55329.1
94.2256
13:37:31.928(18)
−64:23:04.88(3)
307.89
−1.96
106.11873502(3)
−2.2(8)×10−15
55234.7
260.32
14:46:35.71432(13)
−47:01:26.7616(15)
322.50
11.43
455.64401644177(16)
−2.10(4)×10−15
55647.8
55.8340(11)
Binary model............................
Orbital period, Pb(d) ...................
Projected semi-major axis, asini (lt-s)...
Epoch of ascending node, Tasc (MJD)....
ecosω ...................................
esinω ...................................
Inferred eccentricity, e ...................
Minimum companion mass, mc,min(M⊙)
ELL1
6.5118988121(19)
4.83004527(11)
55329.10065316(5)
−7.174(5)×10−5
1.2268(5)×10−4
1.4212(4)×10−4
0.19
ELL1
4.78533407(9)
13.086499(11)
55234.770356(6)
1.82(13)×10−5
8.4(14)×10−6
1.97(13) × 10−5
0.78
ELL1
0.2776660759(19)
0.0640128(7)
55647.8044387(7)
1.3(18)×10−5
1.1(21)×10−5
< 6 × 10−5
0.019
Fit time span (MJD) ....................
RMS of residuals (µs) ...................
Reduced χ2..............................
55343.2—55681.2
0.8
1.3
55460.0—55694.3
37.5
1.7
55358.6—55734.5
2.5
0.8
Mean flux density, S1400 (mJy) . .........
Pulse width at 50% of peak, W50 (◦).....
Pulse width at 10% of peak, W10 (◦).....
Spin down energy loss rate,˙E (ergs−1) ..
Characteristic age, tc (years).............
Dipole magnetic field strength, Bsurf(G)
DM derived distance, d (kpc)∗...........
˙E1/2/d2(×1010erg1/2pc−2s−1/2).......
0.89
10
20
8.0 × 1033
1.4 × 1010
7.8 × 107
3.0
1.0
0.32
23
52
9.2 × 1033
7.6 × 108
1.4 × 109
5.1
0.5
0.37
13
48
3.8 × 1034
3.4 × 109
1.5 × 108
1.5
9.0
of the orbit, as with PSR J0610–2100 and J2241-5236, PSR
J1446–4701 does not show any sign of eclipse at 1369 MHz.
2.5Two intermediate-spin-period, low-mass
binary pulsars: PSR J1502–6752 and
PSR J1622–6617
Although LMBPs typically have spin periods of a few ms,
there are a handful of LMBPs that have spin periods
greater than 10 ms. Examples are PSRs J1745–0952 and
J1841+0130, which have spin periods of 19 and 29 ms respec-
tively (Edwards & Bailes 2001; Lorimer et al. 2006), but are
otherwise similar to the LMBPs.
PSR J1622–6617, with pulse period 23.6 ms, perhaps
fits into the category of intermediate period LMBPs. It was
first observed on 2010-01-03, in a survey beam 0.07◦from
the nominal pulsar position and with a S/N of 12. We note
that the observed spin and orbital parameters are similar
to the known intermediate spin period LMBP J1841+0130
(Lorimer et al. 2006).
A more puzzling situation presents itself with the dis-
covery of PSR J1502–6752. Discovered with a S/N of 15.8
in an observation taken on 2010-06-11 (offset 0.08◦from the
nominal pulsar position) its spin period is 26.7 ms and the
mass function implies a minimum companion mass of just
0.02 M⊙, well below any known LMBP. Indeed, PSR J1502–
6752 and J1622–6617 have the smallest mass functions of
any of the binary ‘intermediate period’ MSPs (selected by
P > 10 ms and˙P < 10−18). The true companion mass does,
of course, depend on the inclination angle, and it is possible
that we are observing these LMBPs close to face on. For
the companion mass of PSR J1502–6752 to be greater than
0.1 M⊙, the inclination angle must be less than 13◦, which
has a probability of 0.025 of being drawn from a uniform dis-
tribution of three-dimensional orientations. It is also worth
noting that, as mentioned by Freire et al. (2003), there is a
distinct gap in the mass functions of the lowest mass binary
MSPs between the VLMBPs, all of which have minimum
masses less than 0.03 M⊙ and the LMBPs of which all have
masses greater than 0.09 M⊙. This suggests that there is a
fundamental difference in the formation mechanism of the
two systems. The gap suggests that the LMBPs with the
lowest mass functions must be somewhat inclined to the
line of sight, and therefore we feel that the anomalously low
mass function of PSR J1502–6752 is unlikely to be due to
an inclination effect.
To better understand the unique nature of J1502–6752,
we can consider where it lies amongst the known population
of MSPs. By plotting the binary period and minimum com-
panion mass of the known binary MSPs, as done in Figure
1, we can identify at least three main populations of binary
MSPs. The VLMBPs appear in the lower left corner of the
diagram, with masses lower than 0.03M⊙, and binary pe-
riods which are typically shorter than a day. The LMBPs

Page 5

HTRU IV: Discovery & polarisation of MSPs
5
Table 2. Observed parameters from timing of a further three previously unknown MSPs. Values in parenthesis are the nominal 1-σ
tempo2 uncertainties in the last digits.
ParameterJ1502−6752J1543−5149 J1622−6617
Right ascension, α (J2000) ..............
Declination, δ (J2000) ...................
l (◦) .....................................
b (◦).....................................
Pulse frequency, ν (s−1) .................
Frequency derivative, ˙ ν (s−2) ............
Epoch of model (MJD) ..................
Dispersion measure, DM (cm−3pc) ......
15:02:18.610(4)
−67:52:16.78(2)
314.80
−8.07
37.3909719910(4)
−4.0(3)×10−16
55421.2
151.75
15:43:44.1498(10)
−51:49:54.681(7)
327.92
2.48
486.154232083(12)
−3.8(2)×10−15
55522
50.92
16:22:03.6669(6)
−66:17:16.978(4)
321.98
−11.56
42.33082901485(8)
−1.14(4)×10−16
55253.1
87.94
Binary model............................
Orbital period, Pb(d) ...................
Projected semi-major axis, asini (lt-s)...
Epoch of ascending node, Tasc (MJD)....
ecosω ...................................
esinω ...................................
Inferred eccentricity, e ...................
Minimum companion mass, mc,min(M⊙)
ELL1
2.4844570(5)
0.31756(3)
55421.21202(4)
−5.3(125)×10−5
−3.9(144)×10−5
< 2 × 10−4
0.022
ELL1
8.06077304(9)
6.480281(5)
54929.067833(7)
2.02(11)×10−5
6.3(12)×10−6
2.2(1)×10−5
0.22
ELL1
1.640635183(20)
0.979380(6)
55253.087284(3)
4.3(117)×10−6
6.7(127)×10−6
< 2 × 10−5
0.092
Fit time span (MJD) ....................
RMS of residuals (µs) ...................
Reduced χ2..............................
55360.4—55757.5
67.9
0.7
55432.3—55758.5
13.5
3.0
55256.9—55733.5
22.2
1.3
Mean flux density, S1400 (mJy) . .........
Pulse width at 50% of peak, W50 (◦).....
Pulse width at 10% of peak, W10 (◦).....
Spin down energy loss rate,˙E (ergs−1) ..
Characteristic age, tc (years).............
Dipole magnetic field strength, Bsurf(G)
DM derived distance, d (kpc)∗...........
˙E1/2/d2(×1010erg1/2pc−2s−1/2).......
0.68
40
–
5.9 × 1032
1.5 × 109
2.8 × 109
4.2
0.2
0.70
38
–
6.1 × 1034
2.4 × 109
1.7 × 108
2.4
4.2
0.52
50
9.8
1.9 × 1032
5.9 × 109
1.2 × 109
2.2
0.3
0.010.11
Minimum Companion Mass (solar mass)
0.1
1
10
100
1000
Binary Period (days)
VLMBPsLMBPsIMBPs
PSR J1502–6752
Figure 1.
nary period for all known binary pulsars with spin period below
100 ms. Pulsars with discoveries presented in this paper are cir-
cled. We have split the binary MSP population into three cate-
gories as shown, the different symbols mark VLMBP (triangle),
LMBP (square) and IMBPs (circle) as described in the text.
Figure showing minimum companion mass and bi-
contain the majority of the pulsars, and occupy a broad re-
gion in the centre of the plot, showing a weak trend between
companion mass and binary period. A third population, the
IMBPs, have much more massive companions than LMBPs
with similar binary periods. There is some overlap between
the LMBPs and IMBPs at binary periods greater than a few
days, however it is clear that there is a distinct population
for short period binaries. We choose to split the LMBPs and
IMBPs at by a minimum companion mass cut-off of 0.5M⊙.
As described above, PSRs J1446–4701 and J1502–6752 fall
into the VLMBP range, PSRs J1017–7156, J1543-5149 and
J1622–6617 are LMBPs and PSR J1337-6423 is amongst
those classified as IMBPs.
Taking the same sample of pulsars, we now plot mini-
mum companion mass against spin period, resulting in Fig-
ure 2. In this plane the IMBPs and LMBPs are much harder
to distinguish, but we retain the classification from Figure
1. There are two clear outliers, the first is PSR J1903+0327,
a known anomaly, which possesses a main sequence com-
panion, and is probably the descendant of a triple sys-
tem (Champion et al. 2008; Freire et al. 2011). The second
clear outlier is PSR J1502–6752, which sits alone in a large
space in the lower right corner of the figure. We must there-
fore consider if PSR J1502–6752 formed through the same
channel as the VLMBPs, or if it underwent some unique
process.
VLMBPs probably form when mass transfer begins be-

Page 6

6
M. J. Keith et al.
0.0010.01 0.1
Spin Period (s)
0.01
0.1
1
Minimum Companion Mass (solar mass)
PSR J1502–6752
PSR J1903+0327
Figure 2. Figure showing spin period and minimum companion
mass for all known binary pulsars with spin period below 100 ms.
Pulsars with discoveries presented in this paper are circled. The
different symbols mark VLMBP (triangle), LMBP (square) and
IMBPs (circle) as described in the text.
fore a large degenerate core has formed as a result of the
proximity of the neutron star to the companion. This is a
stable process, and leads to a long period of spin-up and
a low-mass companion that is often prone to ablation and
given the right geometry, can eclipse the pulsar. How does
PSR J1502−6752 fit into this scenario? If we exclude the
face-on orbit argument our only hint comes from the or-
bital period of the system, which at 2.5d is much greater
than the other VLMBPs. Mass transfer in wider orbits hap-
pens later in the donor star’s life, leaving less time for mass
transfer. What is unclear is why similar orbital period bina-
ries have more evolved (and heavier) companions. So while
the spin period of PSR J1502−6752 is nicely explained by
the larger orbital period, the mass of the companion is not.
More speculative scenarios might involve the post-supernova
collision of the neutron star with a still main sequence com-
panion due to a favourably-oriented asymmetric kick (Bailes
1989; Tauris et al. 1999) and the subsequent disruption of
the host star and mild spin-up of the pulsar. The formation
mechanism of VLMBPs is not yet well understood, and the
discovery of PSR J1502–6752 undoubtedly complicates the
picture further, although the uniqueness of this system may
indeed suggest a process that is rare and improbable. The
discovery of further VLMBPs would be greatly beneficial in
determining if PSR J1502−6752 really is unique, or if the
spin period distribution of the VLMBPs is in fact wider than
currently understood.
3GAMMA-RAY ASSOCIATIONS
Upon discovery of PSR J1446–4701 we determined that the
gamma-ray source 2FGL J1446.8–4701 lies only 2.7′from
the nominal radio position, well within the 95% confidence
radius of r95 = 7.2′(Abdo et al. 2011). From the radio tim-
ing, we measure the pulsar’s spin-down energy loss rate˙E =
4 × 1034ergs−1, and˙E1/2/d2= 9 × 1010erg1/2pc−2s−1/2,
assuming a DM-derived distance of 1.5 kpc. This value is
similar to that of other MSPs for which pulsed gamma rays
have been detected (Abdo et al. 2010).
Since the discovery of the HTRU MSP J1125–5825
Pulse Phase
00.20.40.60.81 1.21.4 1.61.82
Counts/bin
10
20
30
40
> 0.77 GeV
Figure 3. Result of folding the LAT gamma-ray photons with
the radio ephemeris of PSR J1446–4701 (histogram), overlaid
with the 1369 MHz radio profile (curve). The profiles are aligned
assuming a DM of 55.83 cm−3pc, and an arbitrary scaling for
the radio flux density.
(Bates et al. 2011), the gamma-ray source 2FGL J1125.0–
5821 has been discovered, lying only 6′away, also within
the gamma-ray error radius, r95=9′(Abdo et al. 2011). The
energy loss rate of the pulsar is˙E = 7.9×1034and˙E1/2/d2=
4×1010erg1/2pc−2s−1/2, assuming a DM-derived distance
of 2.6 kpc (Bates et al. 2011).
We searched for gamma-ray pulsations from both of
these pulsars by phase binning Fermi LAT photons (arriving
between MJD 54683 and 55774) using the radio ephemeris.
We used “Pass-6 diffuse” class events (highest probability of
being gamma-ray photons) and excluded events with zenith
angles > 100◦to reject atmospheric gamma rays from the
Earth’s limb. We used the bin-independent H-test statistic
(de Jager & B¨ usching 2010) to search over two parameters:
the maximum angular separation from the pulsar position
(0.1◦< r < 2◦), and the minimum photon energy cutoff
(100 < Ecutoff < 1000 MeV), with a maximum energy fixed
at 50 GeV. This truncates the point-spread function at low
energies and decreases the number of background events.
The best signal for PSR J1446–4701 (r = 0.8◦, Ecutoff > 770
MeV) is shown in Figure 3 with a post-trials significance of
5.4-σ, while the best signal for PSR J1125–5825 (r = 0.4◦,
Ecutoff > 1000 MeV) is presented in Figure 4 with a post-
trials significance of 4.9-σ.
The alignment with the radio profile is accurate to
about 0.01 in pulse phase as the DM is well constrained.
The gamma-ray profile of PSR J1446–4701 is fairly broad
and has its peak at a phase of 0.5 relative to the narrow radio
peak. Although much lower in S/N, the radio and gamma-
ray profile shape and alignment are somewhat similar to
PSR J0437–4715 (Abdo et al. 2009b). In PSR J1125–5825
the gamma-ray peak trails the radio peak by a phase of 0.6,
which is large, but not uncommon for millisecond pulsars
with complex radio profiles. The gamma-ray profile may be
similar to PSR J2124–3358, which also shows a very com-
plex radio profile, or perhaps similar to the double peaked
profiles such as PSR J0030+0451 but with an absent leading
peak (Abdo et al. 2009b).

Page 7

HTRU IV: Discovery & polarisation of MSPs
7
Table 3. Pulse period (P), DM and derived energy loss rate (˙E) obtained from radio timing of the 12 MSPs discovered in the HTRU
survey to date. Note that˙E is not corrected for the Shklovskii effect, which may be large in some pulsars. Also provided is the rotation
measure (RM), pulse width at 50% and 10% of the peak flux density (W50 and W10), as derived from the 1369 MHz polarised pulse
profiles described in this work. References for discovery and full parameters are: [a] this work, [b] Bates et al. (2011) and [c] Bailes et al.
(2011).
PSRP (ms)DM (cm−3pc)
˙E (×1033ergs−1)RM (radm−2)W50 (◦)W10 (◦) Ref.
J1017–7156
J1125–5825
J1337–6423
J1446–4701
J1502–5143
J1543–5149
J1622–6617
J1708–3506
J1719–1438
J1731–1845
J1801–3212
J1811–2404
2.34
3.10
9.42
2.19
26.7
2.06
23.6
4.51
5.79
2.34
7.45
2.66
94.2
124.8
260.3
55.8
151.8
50.9
88.0
146.8
36.8
106.6
176.7
60.6
8.0
79
9.2
38
0.59
61
0.19
9.9
1.5
76
0.25
28
−78(3)
−7(3)
−135(17)
−14(3)
−225(2)
0(25)
70(30)
−15(5)
16(4)
19(1)
226(4)
23(3)
11
32
23
13
40
38
9.8
60
20
11
25
14
24
200
52
48
230
200
50
139
63
28
120
280
[a]
[b]
[a]
[a]
[a]
[a]
[a]
[b]
[c]
[b]
[b]
[b]
Pulse Phase
00.2 0.40.6 0.811.21.41.61.82
Counts/bin
10
20
30
> 1.0 GeV
Figure 4. Result of folding the LAT gamma-ray photons with
the radio ephemeris of PSR J1125–5825 (histogram), overlaid
with the 1369 MHz radio profile (curve). The profiles are aligned
assuming a DM of 124.79 cm−3pc, and an arbitrary scaling for
the radio flux density.
4POLARISATION PROFILES
We now consider the polarisation properties of all 12 MSPs
discovered in the HTRU survey to date (Bates et al. 2011;
Bailes et al. 2011), the basic parameters of which are given
in Table 3. To form high S/N profiles we summed all
the observations used for timing of each pulsar with 64,
256 and 1024 MHz wide bands centred at 732, 1369 and
3100 MHz respectively. The signals were passed through a
poly-phase filterbank and folded on-line using the Parkes
Digital Filterbank System (PDFB3 and PDFB4). In the case
of PSR J1017–7156 the ATNF-Parkes-Swinburne Recorder
(APSR) was used to obtain coherently dedisperesd profiles
at 732 and 1369 MHz. Observations at 732 and 3100 MHz
were performed with the “10-50” receiver. Observations at
1369 MHz were performed with the centre beam of the “HI-
Multibeam” receiver. Before each observation a noise diode
coupled to the receptors in the feed is observed to calibrate
for differential gain and phase between the feeds. To cor-
rect for cross-coupling of the receptors in the Multibeam
receiver, we used a model of the Jones matrix for the re-
ceiver computed by observation of the bright pulsar PSR
J0437–4715 over the entire range of hour angles visible, using
the ‘measurement equation modelling’ technique described
in van Straten (2004). The majority of timing is carried
out at 1369 MHz, therefore profiles at the other frequencies
are only included if the S/N ratio was high enough. Note
that we follow the ‘RVM sign convention’ for the PA (see
Everett & Weisberg 2001). We measure the Faraday rota-
tion observed towards each pulsar by fitting PA variations
across the 256 MHz band centred at 1369 MHz, using the
algorithm described in Noutsos et al. (2008). The best fit
values of rotation measure (RM) are provided in Table 3.
We now present Figures 5-16, showing profiles of the
12 MSPs in total intensity, linear and circular polarisation.
When multi-frequency data are present we include profiles at
each of the frequencies, with an arbitrary phase alignment.
Additionally we show the polarisation PAs, corrected with
the RM from Table 3, and therefore absolute PAs as emit-
ted at the pulsar, under the assumption that the RM used
is correct. The figures are marked with a horizontal bar in-
dicating the amount of uncorrected dispersion smearing due
to the finite channel width of the DFBs.
4.1Description of profiles
PSR J1017–7156
J1017–7156 exhibits a single emission region with width at
half the peak value of 10◦at 1369 MHz, and a slight broad-
ening at low frequencies. In general, the frequency evolution
of the profile is small, typical for MSPs (Kramer et al. 1999),
although there is some evolution noticeable on the leading
edge between 732 and 1369 MHz. The linear polarisation
fraction is about 30% at all three frequencies, however the
circular polarisation goes from 20% at 732 MHz to 30% at
1369 MHz to 45% at 3100 MHz. At 732 MHz the circular po-
larisation is predominantly positive, however by 1369 MHz
it is entirely negative in sign. Combined with the PAs (as-
suming a rotation measure of −76 radm−2), this strongly
suggests that the emission at 732 MHz is dominated by a
polarisation mode orthogonal to that which is dominant at
1369 and 3100 MHz. The polarisation PAs also show con-
siderable variation with frequency, especially towards the
(Figure 5). The profile of PSR

Page 8

8
M. J. Keith et al.
trailing edge of the profile. At 732 MHz the PAs show a
downwards ‘S-curve‘, however at 1369 MHz the PAs exhibit
a ‘U’ shape, and at 3100 MHz, the PAs show an upwards
slope. We discuss the atypical properties of the polarised
emission observed from this pulsar in Section 5.1.
PSR J1125–5825
file of PSR J1125–5825 exhibits three distinct components.
Though the S/N is much lower, the profile shape is simi-
lar at other frequencies, with some dispersion broadening
apparent at low frequencies. At 1369 MHz the brightest of
the three main components appears to be composed of at
least three overlapping sub-components. At the higher fre-
quency it appears that this component is starting to break
up into three or four narrower peaks.. In total, emission ex-
tends over ∼ 200◦of the profile. The brightest component
exhibits varying linear and circular polarisation, with an or-
thogonal mode jump aligned with the centre of the peak.
There is also a change in the handedness of circular polari-
sation associated with the jump. The other two components
show no circular polarisation, one is 100% linearly polarised
and the other is completely unpolarised. The spectral index
of the components at phase −45◦is flatter than that of the
component at −80◦.
(Figure 6). At 1369 MHz, the pro-
PSR J1337–6423
J1337–6423 extends over ∼ 90◦of pulse phase, with a half-
width of 23◦. This profile is ∼ 5% polarised in both linear
and circular, with the polarised intensity roughly following
the shape of the total intensity. The PA shows a typical
‘S-shaped’ swing.
(Figure 7). The emission of PSR
PSR J1446–4701
J1446–4701 is fairly simple, showing a single peak with half
width ∼ 13◦lead by a small shoulder. The profile has ∼ 20%
linear and circular polarisation, with the polarised intensity
roughly following the shape of the total intensity. The po-
larisation PAs are flat over the entire profile. The on the
trailing edge of the peak (phases 0-20◦), the ellipticity of
the polarisation increases since the circular polarisation re-
mains constant whilst the linear polarisation drops sharply.
(Figure 8). The profile of PSR
PSR J1502–6752
J1502–6752 covers nearly 300◦of pulse phase. The main
component has a polarisation fraction of ∼ 40% linearly and
∼ 30% circular. The remainder of the profile does not have
sufficient S/N to determine the polarisation fraction. The
observed PAs appear to change linearly with pulse phase.
(Figure 9). The profile of PSR
PSR J1543–5149
J1543–5149 is composed of a single, broad, peak with small
shoulders on the leading and trailing edges. There is less
than 5% polarisation fraction for both linear and circular.
(Figure 10). The profile of PSR
PSR J1622–6617
J1622–6617 is composed of a narrow peak with a broad
shoulder on the trailing edge. The polarisation fraction is
small throughout, however the linear and circular polarisa-
tion track each other well, with a ∼ 10% polarisation frac-
tion in each.
(Figure 11). The profile of PSR
Figure 5. Polarisation profiles of PSR J1017–7156 at (a) 732, (b)
1369 and (c) 3100 MHz, showing phases within ±40◦of the pulse
peak. The solid black line shows total intensity, the thinner solid
line shows linear polarisation and the dotted line shows circular
polarisation. The inset figures show the profile over the full 360◦of
pulse phase. The data at 732 and 1369 MHz have been coherently
dedispersed and so do not exhibit any DM smearing.

Page 9

HTRU IV: Discovery & polarisation of MSPs
9
Figure 6. Polarisation profiles of PSR J1125–5825 at (a) 732,
(b) 1369 and (c) 3100 MHz.
Figure 7. Polarisation profile of PSR J1337–6423 at 1369 MHz.
Figure 8. Polarisation profile of PSR J1446–4701 at 1369 MHz.
Figure 9. Polarisation profile of PSR J1502–6752 at 1369 MHz.

Page 10

10
M. J. Keith et al.
Figure 10. Polarisation profile of PSR J1543–5149 at 1369 MHz.
Figure 11. Polarisation profile of PSR J1622–6617 at 1369 MHz.
PSR J1708–3506
J1708–3506 appears to be formed of two main components,
with the trailing component much broader than the lead-
ing. The broad component appears to have a steeper spec-
tral index, dominating the profile at 732 MHz and appearing
only as a small shoulder at 3100 MHz, causing a relatively
large profile evolution with frequency, as compared to most
MSPs. The level of polarisation is less than 10% through-
out, with the intensity of linear and circular following each
other. An orthogonal mode jump is apparent ∼ 5◦after the
peak of the profile. There appears to be a corresponding
change in the handedness of circular polarisation associated
with the jump, however this later reverses again without a
corresponding change in PA.
(Figure 12). The profile of PSR
PSR J1719–1438
J1719–1438 is relatively narrow, and shows a high shoul-
der on the leading edge and a low shoulder on the trailing
edge. The linear polarisation fraction is ∼ 16% without de-
tectable circular polarisation. The PA swing is steep with
the steepest gradient leading the peak of the profile by 6◦.
(Figure 13). The profile of PSR
Figure 12. Polarisation profiles of PSR J1708–3506 at (a) 732,
(b) 1369 and (c) 3100 MHz.

Page 11

HTRU IV: Discovery & polarisation of MSPs
11
Figure 13. Polarisation profile of PSR J1719–1438 at 1369 MHz.
Figure 14. Polarisation profile of PSR J1731–1845 at 1369 MHz.
PSR J1731–1845
J1731–1845 is composed of two main components, separated
by 155◦. The brightest component is polarised with a linear
fraction of 50% and circular fraction of 20%. The second
component is much less polarised, with only a hint of linear
polarisation.
(Figure 14). The profile of PSR
PSR J1801–3212
J1801–3212 consists of a bright leading component with a
shoulder. The emission is elliptically polarised with a linear
and circular polarisation fraction of ∼ 15% each. The peak
of the polarised intensity trails the peak of the total intensity
by ∼ 18◦. The PA does not appear to follow the typical ‘S-
shaped’ swing predicted by the RVM. The emission changes
hand of circular polarisation at a phase corresponding to the
peak of the total intensity profile. Just prior to this phase,
the PA of the linear polarisation jumps by ∼ 50◦. The lim-
ited S/N makes it hard to discern, but there may also be a
true 90◦jump in the PA between the polarisation associated
with the main peak and that of the shoulder, occurring at a
phase of ∼ 40◦. There may also be a hint that handedness
(Figure 15). The profile of PSR
Figure 15. Polarisation profile of PSR J1801–3212 at 1369 MHz.
of the circular polarisation may also reverse around, or just
preceding this phase.
PSR J1811–2404
discern at least four independent components with emis-
sion spread over 250◦of pulse phase. The emission has a
mean linear polarisation fraction of 16% and a circular po-
larisation fraction of 12%. Of the two brightest components,
the leading component has a flatter spectral index and a
higher polarisation fraction. The trailing component domi-
nates the profile at 732 MHz, and the two components are
equal at 3100 MHz. The profile appears to be almost com-
pletely unpolarised at 732 MHz, even though the leading
component has a considerable polarisation fraction at 1369
and 3100 MHz. The inner components are also polarised and
form a bridge of emission between the two main components.
(Figure 16). At 1369 MHz we can
5DISCUSSION OF POLARISATION AND
SPECULATION ON GEOMETRY
We can use the RVM to model the observed polarisation PA
Ψ in terms of geometrical parameters α and β by
tan(Ψ + Ψ0) =
sinα sin(φ − φ0)
sinζ cosα − cosζ sinαcos(φ − φ0),(1)
where ζ = α + β, and Ψ0 and φ0 are the fiducial phase and
PA, nominally corresponding to the magnetic axis crossing
the line of sight. The relation between φ0 and the observed
emission is however affected by aberration and retardation
effects in the pulsar magnetosphere. If the emission height,
rem, is small compared to the light cylinder radius, RLC,
then φ0 arrives later with respect to the corresponding total
intensity emission, with a shift given by
∆φ = 4rem/RLC
(2)
(Blaskiewicz et al. 1991; Hibschman & Arons 2001; Dyks
2008).
The magnetospheres of MSPs are very compact, a pul-
sar with a spin period of 5 ms has a light cylinder radius
of only 240 km. Conventionally it is assumed that the radio
emission occurs from a radius which is only a small fraction

Page 12

12
M. J. Keith et al.
Figure 16. Polarisation profiles of PSR J1811–2404 at (a) 732,
(b) 1369 and (c) 3100 MHz.
(few percent) of the light cylinder although in young pulsars
higher emission heights are common (Johnston & Weisberg
2006). In MSPs, low emission heights cause a problem be-
cause an emission height of 10% of the light cylinder yields
a pulse width of only ∼60◦unless the rotation and mag-
netic axes are close to alignment. Observationally it has
been clear for over a decade that MSP widths are much
larger than this on average (Xilouris et al. 1998; Yan et al.
2011). Indeed, the observed large widths preclude any emis-
sion height less than the light cylinder in the conventional
picture, unless the majority of MSPs are aligned rotators.
Conversely, for a number of MSPs where there has been an
attempt to constrain the geometry, it has been found that
the inferred opening angle is much smaller than that ex-
pected by extrapolation from the main population of radio
pulsars (Kramer et al. 1998).
In principle, the polarisation properties could help shed
light on this conundrum, especially as large profile widths
are a boon to RVM fitting. Unfortunately, the PA variations
in MSPs can be extremely complex and are often far from
that expected in the RVM. Furthermore, the lack of obvious
inflexion point in the PA variations and the lack of identi-
fiable core/cone components makes it difficult to estimate
emission heights via the aberration/retardation methods of
Blaskiewicz et al. (1991). We note in passing however that
MSPs appear to be no better or worse than normal pul-
sars in this regard (Everett & Weisberg 2001). Indeed the
geometry of the wide profile pulsars B0950+08, B1822−09
and B1929+10 to take but three examples remains unclear
(Everett & Weisberg 2001). In spite of these inherent diffi-
culties we have attempted RVM fits to several of the MSPs
in our sample in an attempt to get a handle on geometry
and emission locations.
5.1PSR J1017–7156
Neither the drastic change of polarisation with frequency,
nor the ‘U’ shaped PAs observed in this pulsar can be sim-
ply explained within the framework of the RVM. We also
observe a change in the handedness of circular polarisation,
and a ∼ 90◦change in the absolute PA of the linear po-
larisation between the profile at 732 MHz and 1369 MHz.
We have considered the possibility that this effect is caused
by an error in calibration, however multiple observations of
this pulsar and the lack of similar effects in other pulsars
largely rules this out. This is, therefore, a strong suggestion
that the profile is composed of two competing elliptically
polarised orthogonal modes, with different spectral indices
(Smits et al. 2006). This superposition of modes can also ex-
plain the rapid sweeps in PA observed at the trailing edge
of the pulse at 732 and 1369 MHz. The strong depolarisa-
tion of the profile at the trailing edge suggests that there is
a change in the dominant polarisation mode at this phase,
and the PA should jump by 90◦. Broadening of the pulse
in the interstellar medium can smooth these discrete jumps
somewhat, causing the appearance of a rapid swing in PA
(Noutsos et al. 2009; Karastergiou 2009).
5.2 PSR J1125–5825
The ‘main pulse’ of this pulsar extends over 100◦of lon-
gitude and has a well behaved swing of polarisation once

Page 13

HTRU IV: Discovery & polarisation of MSPs
13
Figure 17. Polarisation profile of PSR J1125–5825 at 1369 MHz,
showing the observed PAs overplayed with the model obtained
from fitting the RVM. The dashed line is separated by 90◦from
the model PA to indicate orthogonal modes.
the orthogonal jump near the centre of the profile is taken
into account. The presence of a further polarised component
some 130◦away from the main pulse is an added complica-
tion; it is not possible to fit the RVM to all the data. How-
ever, it is possible to obtain a good RVM solution using only
the main pulse data. In this case, α = 128◦, β = −21◦with
the inflexion point coincident with the negative peak of the
circular polarisation. The RVM correctly predicts the swing
of the polarisation at the distant component but the values
of PA are ∼30◦offset from the observed values. The data
and model are shown in Figure 17.
We note also the similarity between this profile
and that of the MSPJ1012+5307 (Xilouris et al. 1998;
Stairs et al. 1999) and also with the normal pulsar B1055–
52 (Weltevrede & Wright 2009). In the former case the au-
thors were unsure whether the pulsar is an aligned or an
orthogonal rotator whereas the latter appears to be clearly
orthogonal. For PSR J1125–5825 the ambiguity remains due
to the problematical nature of the RVM fit. If the pulsar is
an aligned rotator this could explain the large width and
the complex RVM can arise if different components emit at
different heights. On the other hand, if the pulsar is an or-
thogonal rotator the likely implication is that the main and
interpulse emission arise from significantly different heights
or indeed from the outer magnetosphere, resulting in a non-
standard RVM. With more time, and therefore greater S/N,
the gamma-ray profile of PSR J1125–5825 may be able to
provide further constraints on the geometry, and clues to
the location of the radio emission regions.
5.3PSR 1502–6748
This pulsar appears to be a very wide triple profile with the
central peak dominating, and two weak outriders located
some 80◦away. The PA swing is smooth, the RVM fit is not
very constraining on α and β but does locate the inflexion
point some 40◦distant from the main peak. This yields a
height of 220 km, about 20% of the light cylinder radius,
using Equation 2. Although this seems plausible it is difficult
to reconcile this relatively low height with the large pulse
width unless α is small.
5.4PSR J1708–3506
The emission from this pulsar seems to conform to standard
behaviour. It is a blended triple, with the central component
having a steep spectral index and located equidistant from
the flatter spectrum conal components. The PA swing is rel-
atively smooth apart from an orthogonal jump. The RVM
fit is reasonably well constrained with α between 60◦and
90◦and β of 20◦. The inflexion point occurs some 40◦after
the location of the central component. If we interpret the
offset as due to aberration, this yields a height of ∼40 km,
which, assuming an emission region filled to the edge of the
open field line region, would imply a profile width of ∼80◦.
The observed profile is however much broader than this,
placing doubt on the RVM-derived geometry, or otherwise
suggesting a more complex model of the emission region is
required. The large frequency-dependant evolution of the
profile may suggest that the profile is composed of compo-
nents that are originating in different parts of the magneto-
sphere and therefore confusing our picture.
5.5 PSR J1719–1438
This pulsar exhibits an almost classical S-shaped swing of
linear polarisation across the pulse. The profile is again ex-
tremely wide, and this width coupled with the PA swing
implies a small value of both α and β. The inflexion point
of the traverse is located some 6◦prior to the pulse peak.
The geometry of this star is thus rather uncertain and can-
not shed much light as to the origins and evolution of its
planetary mass companion (Bailes et al. 2011).
5.6PSR J1811–2404
Polarised emission extends over nearly 300◦of longitude in
four distinct components. By inserting an orthogonal mode
jump between the dominant component and the weaker
components we can obtain a remarkably good and strongly
constrained RVM fit with α=89.7, β=21 and the location of
the inflexion point 145◦after the main pulse. The data and
model are shown in Figure 18. The pulsar therefore seems to
be an orthogonal rotator where we see emission from both
poles. The question remains as to which components belong
to which pole, and which (if any) are centrally located. There
are two possibilities.
The first option is that the core is located at phase
90◦and −90◦. The main pulse would then consist only of
the core component with no conal outriders. The interpulse
consists of the core plus two trailing conal components. The
spectral index evolution does not fit this picture very well,
and the extreme width of the interpulse is also difficult to
reconcile. An alternative form of this picture would assign
the two central components to a different emission height, or
from an emission region located in the outer magnetosphere.
In the second possibility, the core is located at phase
135◦and −45◦. In this case, the main pulse consists only of
a leading edge cone whereas the interpulse is a mostly sym-
metric triple structure. Under this picture the offset between

Page 14

14
M. J. Keith et al.
Figure 18. Polarisation profile of PSR J1811–2404 at 1369 MHz,
showing the observed PAs overlaid with the model obtained from
fitting the RVM. The dashed line is separated by 90◦from the
model PA to indicate orthogonal modes.
the inflexion point and the core location yields a height of
54 km (c.f. the light cylinder radius of 124 km) and this
height is consistent with a pulse width of some 120◦, about
that of the interpulse emission.
5.7Summary
There now exists a substantial body of high quality polari-
sation measurements of MSPs built up over the last decade.
Key points to re-iterate from these observations are (a)
many MSPs have large widths especially when observed with
high dynamic range (Yan et al. 2011), (b) as in normal pul-
sars a wide range of linear and circular polarisation fractions
are observed (c) the polarisation profiles can be highly com-
plex and the PA swing often does not obey RVM, (d) a
non-negligible fraction are amenable to RVM fitting.
It seems difficult to reconcile the observations with a
low emission height assuming an isotropic distribution of
cos(α). However, it may be that α is biased towards small
values in MSPs, linked to their formation process. This has
to be investigated further but the presence of bridges of
emission between the various components might support this
idea. However, it is also not a given from the data that the
entire phenomenological view of the radio emission has to
be overturned in MSPs by invoking e.g. caustic emission
or extreme magnetosphere effects (Dyks et al. 2010). Con-
sider a range of allowable emission heights up to ∼50% of
the light cylinder radius. For isotropic values of cos(α) one
would then expect minimum pulse widths of ∼50◦and high
fraction (30%) of interpulses. This is more or less consis-
tent with the data where it could be argued that at least 3
of our sample and 7 of the Yan et al. (2011) sample show
evidence of interpulse emission. In these interpulse pulsars
there is no a priori reason why either the flux density of
the emission from the two poles should be similar (e.g the
extreme case of PSR J1603−7202) nor that the emission
heights should be similar which would lead to highly com-
plex PA traverses. The bridge emission seen in e.g. PSRs
1939+2134 and 1045−4509 is not evidently part of the in-
terpulse picture but the wide beams mean that naturally
there will be blending between the main and interpulse pro-
files, so we do not see this as a major issue. For the pulsars
which clearly do not show interpulse emission, some have
relatively narrow profiles (PSRs 1017−7156, 1022+1001,
1446−4701, 1600−3053, 1713+0747) whereas some are wide
(PSRs 1543−5149, 1708−3506, 1732−5049) perhaps indicat-
ing lower values of α. In summary, we believe that emission
heights in MSPs are high, up to 50% of the light cylinder.
This leads to a substantial fraction of MSPs with interpulses,
further complicating the identification of components in the
profile and the PA swing.
Although some MSPs have polarisation consistent with
the RVM, it is still not clear that the derived geometry can
be used to determine a consistent model for the emission lo-
cation, and it is possible that the accretion process itself has
modified the magnetosphere to a point where a simple dipo-
lar model is no longer applicable (Xilouris et al. 1998). Over-
all, however, we find that the polarised emission of MSPs is
not more complex than that observed in many slow pul-
sars. There has been success in untangling these complex
PA variations in slow pulsars through studies of polarised
intensity on a pulse-by-pulse basis (e.g. Backer & Rankin
1980; Gil & Lyne 1995; Karastergiou et al. 2002). Future
high-sensitivity observations may well be able to repeat this
success in MSPs, giving us greater confidence in geometric
interpretations of the RVM and a fuller picture MSP emis-
sion. High significance gamma-ray profiles of MSPs could
also provide additional constraints on geometry, and test
theories of radio emission from the outer magnetosphere.
6CONCLUSION
The High Time Resolution Universe survey for pulsars and
fast transients has discovered 12 MSPs to date, six of which
are announced in this work. These six MSPs (indeed, all 12)
are in binaries. Two, PSRs J1017-7156 and J1543-5149, with
spin periods of ∼ 2 ms, binary periods of about 1 week and
companions with mc,min ∼ 0.2 M⊙ are easily classified as
LMBPs, the most common category of MSPs. PSR J1337–
6423, with its larger companion mass (mc,min = 0.8 M⊙)
and longer spin period (9.4 ms) better fits into the class
of IMBPs, likely with a heavy WD companion. In con-
trast, PSR J1446–4701 has a much lower mass companion
(mc,min = 0.02), falling easily into the distinct category of
VLMBPs, typified by the ‘black widow’ pulsars. The high
rotational energy loss rate,˙E ∼ 4 × 1034, and inferred dis-
tance of only 1.5 kpc suggests that this is a good candidate
for detection as a gamma-ray pulsar. Indeed, we find a tanta-
lising indication of gamma-ray pulsations by using the radio
ephemeris to fold gamma-ray photons detected by the Fermi
LAT.
The final two pulsars, PSRs J1502–6752 and J1622–
6617, both have spin periods of ∼ 25 ms, typically considered
intermediate spin periods for recycled pulsars. PSR J1622–
6617 can be classified as an intermediate period LMBP, sim-
ilar to other known systems such as PSR J1841+0130. There
are however no similar systems to PSR J1502–6752 in the
literature. The mass function implies mc,min = 0.02, typical
of the VLMBPs, however the formation of these systems is
thought to spin the pulsar up to very short periods. Whilst

Page 15

HTRU IV: Discovery & polarisation of MSPs
15
it is possible that the low mass function is an inclination
effect, it is certainly striking that there are no MSP systems
with 0.026 < mc,min < 0.09, with a large number of MSPs
having mc,min ∼ 0.1 or mc,min ∼ 0.02. Therefore we con-
clude that the companion of PSR J1502–6752 is likely to
have formed through the same channel as the companions
for the VLMBPs, and so that formation mechanism does not
require, or does not guarantee, a short spin period MSP.
We have also undertaken polarimetric observations of
all 12 of the MSPs discovered by the HTRU survey to date.
The calibrated profiles show the wide variety of profiles typ-
ical of MSPs, with pulse widths ranging from 24◦to 280◦
(measured at 10% of peak flux). The profiles generally con-
sist of multiple components, often showing broad shoulders
running off either the leading or trailing edge of a narrow
central feature. PSRs J1125–5825, J1731–1855 and J1811–
2404 show the dominant components separated by ∼ 180◦,
and may either be classified as interpulse pulsars or ex-
tremely wide cones. For a number of the MSPs, we find
that the observed swing in PA can be fit by the RVM, how-
ever there are clearly cases such as PSR J1017–7156 where
effects such as smoothed orthogonal mode jumps make this
impossible. We believe that emission heights in MSPs are a
substantial fraction of the light cylinder, leading to a large
fraction of MSPs showing emission from both poles.
7ACKNOWLEDGEMENTS
The Parkes Observatory is part of the Australia Telescope
which is funded by the Commonwealth of Australia for op-
eration as a National Facility managed by CSIRO.
The Fermi LAT Collaboration acknowledges generous
ongoing support from a number of agencies and institutes
that have supported both the development and the oper-
ation of the LAT as well as scientific data analysis. These
include the National Aeronautics and Space Administration
and the Department of Energy in the United States, the
Commissariat ` a l’Energie Atomique and the Centre National
de la Recherche Scientifique / Institut National de Physique
Nucl´ eaire et de Physique des Particules in France, the Agen-
zia Spaziale Italiana and the Istituto Nazionale di Fisica Nu-
cleare in Italy, the Ministry of Education, Culture, Sports,
Science and Technology (MEXT), High Energy Accelerator
Research Organization (KEK) and Japan Aerospace Explo-
ration Agency (JAXA) in Japan, and the K. A. Wallenberg
Foundation, the Swedish Research Council and the Swedish
National Space Board in Sweden. 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’´Etudes Spatiales in France.
The authors would like to thank P. Ray for help with
the gamma-ray analysis section.
REFERENCES
Abdo
(arXiv:1108.1435)
Abdo A. A. et al., 2009a, Science, 325, 840
Abdo A. A. et al., 2009b, Science, 325, 848
Abdo A. A. et al., 2010, ApJS, 187, 460
A. A. etal.,2011,ApJS,submitted
Alpar M. A., Cheng A. F., Ruderman M. A., Shaham J.,
1982, Nature, 300, 728
Backer D. C., Rankin J. M., 1980, ApJS, 42, 143
Bailes M., 1989, ApJ, 342, 917
Bailes M. et al., 2011, Science, in press (arXiv:1108.5201)
BatesS.D.etal.,2011,
(arXiv:1101.4778)
Bisnovatyi-Kogan G. S., Komberg B. V., 1974, Sov. As-
tron., 18, 217
Blaskiewicz M., Cordes J. M., Wasserman I., 1991, ApJ,
370, 643
Burgay M. et al., 2003, Nature, 426, 531
Burgay M. et al., 2006, MNRAS, 368, 283
Champion D. J. et al., 2010, ApJ, 720, L201
Champion D. J. et al., 2008, Science, 320, 1309
Cognard I. et al., 2011, ApJ, 732, 47
Cumming A., Zweibel E., Bildsten L., 2001, ApJ, 557, 958
de Jager O. C., B¨ usching I., 2010, A&A, 517, L9
Dyks J., 2008, MNRAS, 391, 859
Dyks J., Wright G. A. E., Demorest P., 2010, MNRAS, 405,
509
Eatough R. P., Molkenthin N., Kramer M., Noutsos A.,
Keith M. J., Stappers B. W., Lyne A. G., 2010, MNRAS,
407, 2443
Edwards R. T., Bailes M., 2001, ApJ, 547, L37
Edwards R. T., Bailes M., van Straten W., Britton M. C.,
2001, MNRAS, 326, 358
Everett J. E., Weisberg J. M., 2001, ApJ, 553, 341
Freire P. C., Camilo F., Kramer M., Lorimer D. R., Lyne
A. G., Manchester R. N., D’Amico N., 2003, MNRAS,
340, 1359
Freire P. C. C. et al., 2011, MNRAS, 412, 2763
Gil J. A., Lyne A. G., 1995, MNRAS, 276, L55
Hibschman J. A., Arons J., 2001, ApJ, 546, 382
Hobbs G. B. et al., 2009, PASA, 26, 103
Hobbs G. B., Edwards R. T., Manchester R. N., 2006, MN-
RAS, 369, 655
Johnston S., Manchester R. N., Lyne A. G., Bailes M.,
Kaspi V. M., Qiao G., D’Amico N., 1992, ApJ, 387, L37
Johnston S., Weisberg J. M., 2006, MNRAS, 368, 1856
Karastergiou A., 2009, MNRAS, 392, L60
Karastergiou A., Kramer M., Johnston S., Lyne A. G.,
Bhat N. D. R., Gupta Y., 2002, A&A, 391, 247
Keith M. J. et al., 2010a, MNRAS, 409, 619
Keith M. J. et al., 2011, MNRAS, 414, 1292
Keith M. J., Johnston S., Weltevrede P., Kramer M., 2010b,
MNRAS, 402, 745
Klu´ zniak W., Ruderman M., Shaham J., Tavani M., 1988,
Nature, 334, 225
Kramer M., Johnston S., 2008, MNRAS, 390, 87
Kramer M., Lange C., Lorimer D. R., Backer D. C., Xilouris
K. M., Jessner A., Wielebinski R., 1999, ApJ, 526, 957
Kramer M., Xilouris K. M., Lorimer D. R., Doroshenko
O., Jessner A., Wielebinski R., Wolszczan A., Camilo F.,
1998, ApJ, 501, 270
Lange C., Camilo F., Wex N., Kramer M., Backer D., Lyne
A., Doroshenko O., 2001, MNRAS, 326, 274
Lorimer D. R. et al., 2006, MNRAS, 372, 777
Noutsos A., Johnston S., Kramer M., Karastergiou A.,
2008, MNRAS, 386, 1881
Noutsos A., Karastergiou A., Kramer M., Johnston S.,
Stappers B. W., 2009, MNRAS, 396, 1559
MNRAS,inpress