Mon. Not. R. Astron. Soc. 405, 1061–1074 (2010) doi:10.1111/j.1365-2966.2010.16497.

The dynamics of pulsar glitches: contrasting phenomenology

with numerical evolutions

T. Sidery,1 A. Passamonti2 and N. Andersson2

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

1 FENS, Sabanci University, Orhanli, 34956 Istanbul, Turkey
2 School of Mathematics, University of Southampton, Southampton SO17 1BJ

Accepted 2010 February 9. Received 2010 January 14; in original form 2009 October 20

ABSTRACT
In this paper, we consider a simple two-fluid model for pulsar glitches. We derive the basic
equations that govern the spin evolution of the system from two-fluid hydrodynamics, ac-
counting for the vortex mediated mutual friction force that determines the glitch rise. This
leads to a simple ‘bulk’ model that can be used to describe the main properties of a glitch event
resulting from vortex unpinning. In order to model the long-term relaxation following the
glitch, our model would require additional assumptions regarding the repinning of vortices, an
issue that we only touch upon briefly. Instead, we focus on comparing the phenomenological
model to results obtained from time-evolutions of the linearized two-fluid equations, i.e. a
‘hydrodynamic’ model for glitches. This allows us to study, for the first time, dynamics that
was ‘averaged’ in the bulk model, i.e. consider the various neutron star oscillation modes
that are excited during a glitch. The hydro-results are of some relevance for efforts to detect
gravitational waves from glitching pulsars, although the conclusions drawn from our rather
simple model are pessimistic as far as the detectability of these events is concerned.
Key words: gravitational waves – methods: numerical –stars: neutron – stars: oscillations –
pulsars: general.

An important question concerns whether differences in internal

1 I N T RO D U C T I O N
structure can be deduced from observations. This is obviously prob-
Neutron stars provide excellent testbeds for extreme physics theory. lematic since most of the data collected for these stars originate from
Since their core density reaches beyond anything we can produce in the star’s surface, atmosphere or magnetosphere. Having said that,
the laboratory, observations of such compact remnants may provide there are phenomena that involve bulk dynamics and which should
unique constraints on models for supranuclear physics (Lattimer depend on the internal composition. Examples of such phenomena
& Prakash 2004). In order to infer useful information from astro- are the radio pulsar glitches (Lyne, Shemar & Smith 2000) and the
physical observations, we need to construct realistic models with quasi-periodic oscillations observed in the X-rays from magnetar
the power to predict the evolution of individual systems (at least to flares (see e.g. Watts & Strohmayer 2007). The long relaxation time
some extent). This is a serious challenge for the modelling com- associated with glitches is seen as indirect evidence for neutron star
munity. Even a moderately reasonable neutron star model should superfluidity (Anderson & Itoh 1975; Ruderman 1976; Alpar et al.
account for the presence of different exotic states of matter. From 1981), and recent considerations of the crust oscillations which are
nuclear physics and Bardeen–Cooper–Schrieffer (BCS) theory, we thought to be the origin of the observed magnetar oscillations sug-
expect the outer neutron star core to consist of superfluid neutrons, gest that these may be affected by the presence of a crust superfluid
superconducting protons, free electrons and muons. Deeper into the as well (Andersson, Glampedakis & Samuelsson 2009b).
core more exotic phases of matter, like superfluid hyperons and/or In this paper, we discuss basic models for pulsar glitches. We
deconfined quarks exhibiting colour-flavour-locked superconduc- focus on the glitch event itself rather than the subsequent long-term
tivity, are likely to be present. Meanwhile, a relatively thin (1 km relaxation. We consider the implications of the standard two-fluid
or so) crust surrounds the fluid core. In the crust, matter changes model from two different points of view. First of all, we derive
from a soup of nucleons in the interior to an elastic nuclear lattice the simple equations that govern the bulk evolution of the system
of heavy iron near the surface. In the inner part of the crust (beyond from two-fluid hydrodynamics, accounting for the vortex mediated
neutron drip), free neutrons are expected to be superfluid. mutual friction force. This leads to a basic model that can be used
to describe a glitch rise resulting from global vortex unpinning, be
it in the crust (Link & Epstein 1996; Melatos, Peralta & Wyithe
 E-mail: [email protected] 2008; Warszawski & Melatos 2008; Melatos & Warszawski 2009)


C 2010 The Authors. Journal compilation 
C 2010 RAS
1062 T. Sidery, A. Passamonti and N. Andersson
or in the core (Link 2003).1 The results demonstrate that the model the Euler equations:
requires additional assumptions regarding the repinning of vortices  
∂
in order to model the long-term evolution. This is as expected (Alpar Ei = + v j ∇j vi + ∇i (μ̃ + φ) = 0, (1)
∂t
et al. 1984b). Secondly, we use time-evolution of the linearized
two-fluid equations as a ‘hydrodynamic’ model for the glitch event. where φ is the gravitational potential and μ̃ = μ/m is the chemical
This allows us to consider dynamics that was ‘averaged’ in the bulk potential divided by the particle mass. In addition, we have the
model. In particular, we consider the various neutron star oscillation continuity equation
modes that are excited during the glitch. In principle, the obtained

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

∂ρ
results could be of relevance for efforts to detect gravitational waves + ∇i (ρv i ) = 0 , (2)
∂t
from glitching pulsars. Having said that, our estimates suggest that
the gravitational-wave signals associated with the impulsive events and the Poisson equation
that we consider here are very weak. ∇ 2 φ = 4πGρ. (3)
Note that we use a coordinate basis to represent vector quantities
2 B U L K P RO P E RT I E S throughout this paper. In other words, we distinguish between co-
We want to consider the simplest viable model for large pulsar and contravariant quantities, using the flat space metric gij to re-
glitches. The angular momentum of any superfluid component is late them. That is, we have v i = gij v j . We also make use of the
determined by the density and configuration of vortices threading Einstein summation convention for repeated indices. Finally, we
the fluid. If the vortices are fixed (pinned), there can be no angular have assumed that the internal energy E depends only on the num-
momentum exchange between the superfluid and other components ber density n, i.e. that the fluid is a barotrope. Then,
in the star. Assuming a scenario of catastrophic unpinning, it is dE
straightforward to formulate a simple glitch model. One can simply μ= , (4)
dn
assume that the system has two components, with different moments
of inertia and spin rates. Adding the assumption that vortex pinning and the pressure P of the fluid is defined (in the usual way) by
allows a lag in the rotation between the observed component (e.g. dP = n dμ. (5)
the crust) and the other component (the interior superfluid) and that
the pinning breaks once a critical lag is reached, one arrives at a Viscosity terms have been omitted in anticipation of applying the
phenomenological glitch model. In order to effect the rapid transfer solid-body approximation, which we do now. Assuming uniform
of angular momentum that leads to the observed spin change in the rotation, the fluid velocity can be written as
crust, one would typically assume that the rotation lag relaxes on vi = ij k j x k =  ϕi , (6)
some time-scale. The standard model assumes that this relaxation is
due to the mutual friction acting on the superfluid vortices (Alpar, where we take  = (t),  is the cylindrical distance from the
Langer & Sauls 1984a; Mendell 1991; Andersson, Sidery & Comer rotation (z) axis and ϕ i is a unit vector in the direction of the flow.
2006). By assuming axisymmetry it follows that
While such a phenomenological model is useful, its scope is lim- v j ∇j vi = 0 , (7)
ited. Ultimately, a detailed description will require a hydrodynamics
analysis. In particular, if we want to understand the reason for the v j ∇j (μ̃ + φ) = 0. (8)
sudden vortex unpinning [likely due to an instability, see Andersson,
Comer & Prix (2003); Glampedakis & Andersson (2009) for recent We will now derive the conservation of angular momentum and
ideas] and possible neutron star oscillations excited by the event. energy from the above equations. Contracting (1) with ρv i and
Developing a detailed hydrodynamics model is a severe challenge integrating over a fixed volume V gives
given the many uncertainties in the relevant physics, but we can    
∂E j ∂vj ∂ 1
make some progress by making contact between the general two- = ρv dV = ρv dV = 0.
2
(9)
∂t ∂t ∂t 2
fluid hydrodynamics framework and the phenomenological bulk
dynamics description. As a suitable starting point, we will show This shows that the kinetic energy is conserved. Adding the as-
how the bulk model can be obtained from hydrodynamics. This is sumption that the fluid is in solid-body rotation, we see that
instructive since it provides insight into the validity of the model,  
j
v 2 = j k δk x 2 − x j xk , (10)
and also gives us a better idea of the origin of the different param-
eters (like the global spin-up time). Moreover, we will be able to where x2 = gij xi xj = xj xj . From this, we define the moment of
make direct comparisons with hydrodynamics results. inertia as

 
2.1 Single fluid Ili = ρ δli x 2 − x i xl dV . (11)

It is useful to begin by outlining the analysis of a single fluid body. We can use (10) and (11) to rewrite the change in kinetic energy
In that case, we have a velocity field vi which evolves according to equation (9) as
1 ∂E 1 ∂  i 
In reality, our model is somewhat unrealistic for both crust and core su- = Il i l = 0. (12)
perfluids. In the former case, we have not accounted for the crust elasticity, ∂t 2 ∂t
while in the latter case we are ignoring the expected interaction between neu- Let us now consider the z component of the angular momentum.
tron vortices and proton fluxtubes. These effects may have a decisive impact Assuming that the chemical and gravitational potential only depend
on glitch dynamics. Nevertheless, the present work is state-of-the-art in this on the (spherical) radial position, i.e. assuming slow rotation, then
area, and we expect to add the relevant features to the hydrodynamical model
in future work. ij k x j ∇ k (μ̃ + φ) = 0 for i = z. (13)


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
 The dynamics of pulsar glitches 1063
Contracting Ei with ρ ijk xj and integrating gives of rotating solid bodies. We also assume that the two components
 rotate around the same axis, i.e. we ignore any precessional motion.
∂Ji
= ρij k x j E k dV This means that we can write
∂t  
  
vxi = x  ϕ i and wyx i
= vyi − vxi = y − x  ϕ i . (20)
∂ j
= ρj δi x 2 − xi x j dV
∂t As we are assuming solid-body rotation, x is not a function of
∂  j  position. Moreover, the axial symmetry of the system implies that
= I j = 0 for i = z . (14)
∂t i vxj ∇j vix = 0,

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

(21)
We see that, for cylindrical polar coordinates with  = (0, 0, )
i

and Izz = I , we get the standard results vxj ∇j (φ + μ̃x ) = 0. (22)

1 We continue to follow the method used in the single fluid case,
E = I 2 and J z = I . (15)
2 contract Eix with ρx vix and integrate over a fixed volume to find the
Both these quantities are conserved. global change in energy of each constituent. This leads to

∂Ex
= ρx vxi Eix dV
2.2 Two-fluid model ∂t
 y

We now consider a two constituent stellar model. Our particular 1 ∂ 2 ∂vj
= (ρx − 2α) vx + 4αvx j
dV = 0. (23)
interest concerns two effects, the entrainment and the mutual friction 2 ∂t ∂t
between the components. In order to simplify the initial analysis, We obtain the final result for the total change in energy by adding
we first ignore the mutual friction. the expressions for the two components,
Our formulation for multifluid hydrodynamics in Newtonian
∂E ∂En ∂Ep
gravity derives from the work by Prix (2004), see also Andersson = +
& Comer (2006). The analysis is based on a variational principle, ∂t ∂t ∂t
 
where the action is minimized by varying the fluid flow lines. Key in 1 ∂  2 
= ρn vn + ρp vp2 − 4αwnp
2
dV = 0 . (24)
this analysis is the allowance that the internal energy may depend on 2 ∂t
np p
the velocity difference between the two constituents, wi = vin −vi , This result shows how the entrainment affects the conserved energy
such that (cf. Carter & Chamel 2004 for a detailed discussion) and agrees
∂E ∂E ∂E perfectly with the modified ‘kinetic energy’ used by Mendell (1991).
dE = dnn + dnp + dwnp2
∂nn ∂np 2
∂wnp In the particular case of (aligned) solid-body rotation, with xi
aligned with the z-axis, we have
= μn dnn + μp dnp + αdwnp
2
. (16) 
∂E ∂ 1
Here, the constituent indices ‘n’ and ‘p’ denote the neutron and = In n [n + εn (p − n )]
∂t ∂t 2
‘proton’ (incorporating also electrons and muons in the usual way) 
components, respectively. An important consequence is that the 1
+ Ip p [p + εp (n − p )] = 0 . (25)
conjugate momentum density for each constituent is modified, so 2
that Here, we have defined the constituent moment of inertia as
  y    
pix = ρx vix + εx vi − vix , (17) j j
Ix i = ρx δi x 2 − xi x j dV , (26)
where x and y is either ‘n’ or ‘p’, with x = y. In this relation,
the entrainment is represented by the parameter ε x . This is a non- and Ix = Ix zz . Similarly, we can calculate the total change in angular
dissipative effect, in the neutron star case due to the strong interac- momentum. To do this, we note that
tion between neutrons and protons, which leads to the momentum ij k x j vxl ∇ k vlx = ij k x j x k 2x − ij k x j kx xl lx = 0 for i = z,
no longer being aligned with the individual components velocity.
(27)
We refer the reader to Prix, Comer & Andersson (2004), Carter,
Chamel & Haensel (2005), Gusakov & Haensel (2005), Chamel & as xi
is parallel to the z-axis. Contracting Eix j
with ρ x  ijk x and
Carter (2006) for discussions of the role of entrainment in neutron integrating over the volume V we arrive at

star dynamics. Note that ∂Jix
= ρx ij k x j E k dV
εx ρx = εy ρy = 2α. (18) ∂t
 
We now have a set of Euler equations for each constituent. These ∂vxk ∂vyk
= ρx ij k x (1 − εx )
j
+ εx dV
equations can be written as ∂t ∂t
 
∂  x yx  yx =0 for i = z. (28)
Eix = + vxj ∇j vi + εx wi + ∇i (φ + μ̃x ) + εx wj ∇i vxj
∂t This can be rewritten as
= 0. (19) ∂Jix ∂  j x  y 
= Ix i j + εx j − xj =0 for i = z,
In addition, we have one continuity equation for each component ∂t ∂t
and the Poisson equation for φ is now sourced by the total mass den- (29)
sity ρ = ρ n + ρ p . Following the analysis in the previous section, we from which it follows that the total change in angular momentum
want to derive the equations that represent the global conservation is given by
of energy and angular momentum. In doing this, we will, for sim- ∂Jz ∂
plicity, assume that the velocity fields of the constituents are those = (In n + Ip p ) = 0. (30)
∂t ∂t


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
1064 T. Sidery, A. Passamonti and N. Andersson
Hence, the total angular momentum is conserved. This is obviously A can be calculated from the standard magnetic dipole model
not surprising. The non-trivial result concerns how the entrainment (Shapiro & Teukolsky 1983). Modifying the result so that the torque
affects the evolution of the individual components. This will be acts on the proton fluid rather than the whole star, we find
important later.
Bp2 R 6 sin2 θ
A= , (41)
6c3 Ip
3 A S I M P L E S P I N - D OW N M O D E L
where Bp is the strength of the magnetic dipole with axis at an angle

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

We are not yet in a position where we can model glitches. To do this, θ to the rotation axis, R is the radius of the star and c is the speed
even in the most basic fashion, we need to account for the coupling of light. As I p  I n ≈ I (typically), we can use
due to mutual friction. However, before we discuss that problem,
it is interesting to consider how the conservation equations that Ip Ip 2MR 2
Ip ≈ I≈ , (42)
we have obtained can be used to model the rotational evolution of In In 5
the system. The main purpose of doing this is to understand how where M is the mass of the star. Using typical parameters B =
the entrainment enters the problem. As we will see, the result can 1012 G, R = 106 cm, I p /I n = 0.05, 0 = 2π/0.1 s−1 and M =
be quite surprising. 1.4 M and setting θ = π/2, we find τ ≈ 7 × 104 yr.
Equation (29) shows that the angular momentum of each con- Let us now consider the evolution of the, unseen, neutron com-
stituent is conserved. In a neutron star, we would expect the protons ponent. From equation (35), we have
to be locked to the magnetic field and the crust. Hence, they should
be spun down due to a magnetic torque J˙ em εĨ
i . We can include this n − 0n = − (p − 0 ), (43)
torque in the equations by breaking the conservation of the proton 1 − εĨ
angular momentum, so that where 0n = n (t = 0). Substituting equation (39) into this, we get
p
J˙i = −J˙em
i (31) εĨ 0 t
n ≈ 0n + . (44)
(from now on we will often represent time derivatives by dots). To 1 − εĨ τ
be specific, we assume that the magnetic torque is related to the This relation allows us to estimate how long it takes for a rotational
angular velocity of the protons by lag to develop between the two constituents. Assuming that the
J˙em = AIp 3p . (32) two components rotate together at time t = 0, then 0n = 0 . The
rotational lag then evolves according to
This is in accord with the standard magnetic dipole model. We will
also assume that the constituent moments of inertia are constant. εĨ 0 t 0 t
 = n − p = +
Noting that I n εn = I p εp and defining ε = εp , equation (29) leads to 1 − εĨ τ τ
 
˙ n + Ip ε(
˙p−
˙ n ) = 0, εĨ 0 t 1 0 t
In  (33) = 1+ = (45)
1 − εĨ τ 1 − εĨ τ
˙ p + Ip ε(
Ip  ˙n−
˙ p ) = −AIp 3p . (34) or
 1 t
Defining Ĩ = Ip /In equation (33) gives ≈ . (46)
p 1 − εĨ τ
˙ n = − εĨ 
 ˙ p. (35) It has been argued (Lyne et al. 2000) that a rotational lag of /p
1 − εĨ ≈ 10−4 is needed in order to ‘explain’ Vela sized glitches. From
Substituting this into equation (34), we get equation (46) we see
  t
˙p ≡ 1− ε
Ī  ˙ p = −A3p .
 (36) ≈ 10−4 → t ∼ 7 yr. (47)
1 − εĨ τ
This is a separable equation so we can integrate to get Hence, this simple model is consistent with large glitches occurring
 p  once in a few years in a typical young pulsar.
dp A t Finally, let us consider the entrainment coupling in more detail.
= − dt, (37)
0 3p Ī t0 In general, the evolution of the rotation of the proton and neutron
fluids is given by (39) and (44), respectively. If we focus on a
where 0 is p at time t0 . Setting t0 = 0, we find the solution sufficiently short evolutionary time-scale, then we can assume that
 −1/2 J˙em is approximately constant. From equations (33)–(35), we find
2A20 t
p = 0 1 + . (38)  −1
Ī
˙ p = − 1 1 − εIn J˙em . (48)
As we expect the evolution to be slow, the second term in the bracket Ip In − εIp
is small and we can expand to get This is an interesting, and perhaps surprising result. It appears that,
 
A20 t even though a spin-down torque acts on the protons, their rotational
p ≈ 0 1 − . (39) velocity may increase. This happens when
Ī
From this, we can find the characteristic evolution time-scale τ for In In
<ε< . (49)
the crust (the protons). Ignoring entrainment, we have In + Ip Ip
1 Would this happen for realistic parameter values? Setting ε = 0.05
τ= . (40)
A20 and I p /I n = 0.1, typical values in the outer neutron star core (Chamel


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
 The dynamics of pulsar glitches 1065
2006), we find that the crust should spin down. However, if we repinning should determine the long-term relaxation of the glitch,
consider the neutron fluid, we find the condition for spin up is i.e. the spin evolution on timescales longer than tens of seconds.
In In order to model this phase, one would likely need to account for
0<ε< . (50) vortex creep. Eventually, the system will reach a state where the
In + Ip
rotational lag increases, and the pulsar may glitch again.
For the expected values of ε and I p /I n , we see that the neutron fluid We focus on the glitch event itself, i.e. the short-term evolution
spins up. This is somewhat counterintuitive, but does not violate following global vortex unpinning. During this phase, one would
any fundamental principles. In fact, it is easy to see how this effect expect the main dynamics to be determined by the mutual friction

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

can result from an exchange of angular momentum in a coupled force. Hence, we do not have to consider either the electromagnetic
system. Unfortunately, since we do not observe the neutron compo- torque or the vortex creep. These are, of course, important in a
nent directly, the result does not help us constrain the entrainment complete glitch model (cf. Larson & Link 2002, for an interesting
parameter. It is just an example of the drastic effect that entrainment discussion of the relaxation phase); but since they require dynamics
can have on the evolution of a system. on rather different timescales, it is natural to (at least initially)
consider the different phases separately. Our main aim is to compare
a simple ‘global’ glitch model to a numerical evolution that takes
4 MODELLING A GLITCH
into account the actual hydrodynamics.
The two-component model that we have described cannot be used
to model glitch events unless we add more physics to it. Key to
4.1 The evolution equations
the problem is the motion of the superfluid vortices. There
are two aspects to this problem; we need to account for the friction To model a glitch event, we need to account for the mutual friction
that arises due to the presence of the vortices and the associated force. As discussed by Andersson et al. (2006), this means that we
dissipative coupling between the two fluids in the model (Alpar consider the evolution equations
 
et al. 1984a; Mendell 1991; Andersson et al. 2006). We should also ∂  x yx  yx
consider the interaction between the vortices and the nuclei in the Eix = + vxj ∇j vi + εx wi + ∇i (φ + μ̃x ) + εx wj ∇i vxj
∂t
neutron star crust, the potential pinning of the vortices to the lattice
ρn ρn
B |ωn | wi − B ij k ωnj wyx
yx
(Donati & Pizzochero 2006; Avogadro et al. 2008) and the extent to = k
, (51)
ρx ρx
which the vortices exhibit creep in the presence of a rotational lag
(Anderson & Itoh 1975; Alpar et al. 1984b; Alpar, Cheng & Pines where B and B are the mutual friction parameters (Andersson
1989; Alpar et al. 1993; Link, Epstein & Baym 1993). et al. 2006). We have assumed that the vortex array remains straight
In the following, we will focus on the role of the mutual friction and that the rotational lag remains orthogonal to the rotation axis.
and the glitch event itself. The vortex pinning will be dealt with This is the simplest assumption. If we were to relax it, we would
in a very simplistic fashion. We will simply assume that the vor- have to consider possible precession of the system and the mutual
tices are either perfectly pinned or completely unpinned. In such friction force would be more complicated. In the case that we are
a model, a glitch would proceed as follows. Assume that the su- considering, we have
perfluid vortices form a uniform, straight, array aligned with the  p 
ωin = 2ni + εn 2i − 2ni . (52)
rotation axis [i.e. we are not considering potential vortex tangles
and turbulence (Andersson, Sidery & Comer 2007), which would Generalizing the prescription from the previous section, we can
make the problem quite a lot harder since the model would have to find the energy equation for each constituent. This leads to

contain local information (Peralta et al. 2005)]. In this case, vortex ∂Ex
= ρx vxj Ejx dV
pinning simply fixes the number of vortices per unit area. This in ∂t
turn fixes the neutron fluids angular momentum, so the superfluid 

ρn ρn
ρx vxj B |ωn | wj − B j kl ωnk wyx
yx
component rotates at a constant rate. If we assume that the charged = l
dV . (53)
ρx ρx
fluid is locked to the crust via magnetic effects, then the vortices
yx
will be rotating with the charged fluid component. As the crust spins The second term in the integral will vanish as vix is parallel to wi .
j
down due to the electromagnetic torque, a velocity difference will Written in terms of the moments of inertia Ix i (cf. equation 26), the
build up between the two constituents. This will lead to an increas- total change in energy is given by
ing Magnus force acting on the vortices. Eventually, when some 
∂E ∂ 1
critical lag, c , is reached, this force will be strong enough to = In n [n + εn (p − n )]
∂t ∂t 2
overcome the nuclear pinning and the vortices are suddenly free to 
move. At this point the vortex mutual friction becomes relevant and 1
+ Ip p [p + εp (n − p )]
serves to transfer angular momentum between the two components. 2
This becomes the mechanism by which the two components couple = −B|ωn |(p − n )2 In < 0 . (54)
and the lag decays. The crust spins up leading to the observed glitch
That is, the mutual friction leads to a loss of kinetic energy (as
jump. If the system relaxes completely, the end state should be such
expected). The equilibrium (minimum energy) state is reached when
that the two components rotate at the same rate. The glitch event
the two fluids are rotating together (p = n ).
itself is relatively sudden. The best resolved event to date is the
We can also calculate the global change in angular momentum.
so-called Christmas glitch in the Vela pulsar, where the glitch rise
Focusing on the z-component of the angular momentum we find
time was shorter than a few tens of seconds (Dodson, McCulloch & 
Lewis 2002). In other words, the angular momentum is transferred ∂Jix
= ρx ij k x j E k dV
to the crust in less than a few hundred rotation periods. On a longer ∂t

time-scale, one would expect the vortices to repin. After all, in the  
relaxed state the Magnus force is absent (or at least very small). The = ij k x j Bρn |ωn | wyx
k
− B ρn  klm ωln wmyx dV . (55)


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
1066 T. Sidery, A. Passamonti and N. Andersson
Noting that This rearranges to give
yx yx
j klm
ij k x  ωln wmyx = xxj ωin wj −x j
ωjn wi = 0 for i = z, (56) In
V= W0 + 0 ≈ 0 , (67)
I
and
yx yx yx
which should hold since W0  1. We then arrive at the final result
ij k x j wyx
k
= ij k x j  klm l xm = i x j xj − j x j xi , (57)  
−1
0 t/τ Ip
we can write the change in angular momentum in terms of the W ≈ 0 e + − εn (et/τ − 1) ≈ W0 e−t/τ ,
W0 I
constituent moments of inertia to get

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

(68)
∂Jix  y  j
= B |ωn | j − xj In i . (58) where
∂t (1 − ε̄)Ip
The total angular momentum is given by τ≈ . (69)
2BI 0
p
∂Ji ∂Jin ∂Ji It is easy to show that the observed component (the protons) evolves
= + = 0. (59)
∂t ∂t ∂t according to
Hence, the angular momentum is conserved. This is as expected In
since no external torques have been accounted for. It is worth noting p ≈ 0 + W0 (1 − e−t/τ ). (70)
I
that the B coefficient does not feature in the final equations.

4.2 An explicit solution 4.3 Matching observational data

We will now solve the global evolution equations for the system. The model we have described is obviously quite simplistic. Most
To do this, it is useful to rewrite them in terms of the rotational lag importantly, the assumption of constant parameters (which allowed
and a quantity directly related to the total angular momentum. The us to carry out the integration over the body of the star in the
conservation of angular momentum makes the latter variable trivial first place) is quite unrealistic. Having said that, it would not be
to deal with, while the lag is a key (more or less directly observable) surprising if the final model were to retain some of the bulk dynamics
quantity. of the more complex system. Of course, the various quantities in
From (30), (59) and assuming that the moment of inertia of e.g. (69) must be take to represent ‘body averages’ in some sense.
each constituent remains constant, we define V from the angular Moreover, the model is only relevant on the relatively short time-
momentum such that scale of the glitch jump itself. In order to describe the subsequent
˙ n + Ip 
˙ p = I V̇ = 0. long-term evolution, we would need to include both the magnetic
In  (60)
spin-down torque and the repinning of the vortex lines to the crust.
Then, it is obvious that V remains constant during the evolution. The latter could possibly be accounted for by ‘switching off’ the
From (30) and (58), we next find that the rate of change of the lag mutual friction ‘gradually’. That is, one could simply take B to
W = n − p is given by be time-dependent, reflecting the amount of vortex pinning or the
  nature of the vortex creep. This idea has been considered by Sidery
(1 − ε̄)Ẇ = ˙n− ˙ p = −2(n − εn W )B 1 + In W, (61) (2008), and we think it would be interesting to develop it further.
Ip Despite these caveats, it is interesting to consider how obser-
where we have defined ε̄ = εn + εp . From (60), we can rewrite V as vations may constrain the various parameters. Let us consider the
In Ip In Ip Ip scenario where the observed component represents a small fraction
V= n + p = n + (n − W) = n − W. of the total moment of inertia. This would be the case for a typical
I I I I I
neutron star crust coupled to a large superfluid reservoir in the core,
(62)
when we may have I p /I ∼ 10−2 . Then, the observed glitch jump
Substituting this into (61) gives would be
 

Ip I p − 0 In W0 W0
(1 − ε̄)Ẇ = −2B V + − εn W W. (63) ≈ ≈ . (71)
I Ip 0 I 0 0
This is the stage at which the change of variables helps us. Because V That is, W0 would correspond (more or less directly) to the observed
is constant, equation (63) is separable. Straightforward integration, glitch size. At the same time, the available constraint on the glitch
assuming that the glitch occurs at time t = 0 and defining W0 = rise time can be compared to the spin-up time of the model. Let us,
W(0), leads to for simplicity, impose the constraint that the glitch happens in less
 
−1 than 100 rotations. Then, we need
V t/τ Ip
W=V e + − εn (et/τ − 1) . (64) (1 − ε̄)Ip
W0 I τ 0 ≈ < 102 . (72)
2BI
The spin-up time τ is given by
We can rewrite the entrainment factor in terms of the effective proton
(1 − ε̄)Ip mass in the usual way (Prix, Comer & Andersson 2002). Then,
τ= . (65)
2BI V
m∗p m∗p
For practical purposes, it is better to express V in terms of the initial εp = 1 − −→ 1 − ε̄ ≈ , (73)
conditions. Defining the initial rotation of the protons as 0 , we mp mp
easily find V at time t = 0. From (60), it follows that and we need
 
1 In Ip Ip m∗p Ip
V = (In n + Ip p ) = V + W0 + 0 . (66) τ 0 ≈ < 102 , (74)
I I I I mp 2BI


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
 The dynamics of pulsar glitches 1067
or, for the suggested moment of inertia ratio, easy to show that the available energy is
m∗p E1 ≈
1
B > 5 × 10−5 . (75) 2
I . (77)
mp
As discussed by e.g. Andersson & Comer (2001), this estimate
This constraint is not very severe. In particular, the canonical value suggests that pulsar glitches may be of interest for future generations
B ∼ 10−4 (Alpar et al. 1984a; Mendell 1991; Andersson et al. 2006) of gravitational-wave astronomers. However, if we consider the two-
lies within the required range. However, if we use the constraint of component model, we get a rather different picture. In this case, for

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

a spin-up of the order of a day suggested by a partially resolved constant I x , the conservation of angular momentum in the glitch
glitch in the Crab pulsar (Lyne, Smith & Pritchard 1992), then the leads to
mutual friction parameter would be constrained to being weaker
Ip
than B ∼ 10−10 . This result would not accord well with our current n = − p . (78)
mutual friction models, likely illustrating our level of ignorance In
about the relevant physics. That is, the superfluid (neutrons) spin down as the crust (protons)
In principle, we would now want to consider the case when the spin up. Estimating the available energy, we find that
superfluid component represents only the free neutrons in the crust. 1
That is, when we have I n /I ≈ 10−2 (Link, Epstein & Lattimer 1999). E2 ≈ Ip ()2 . (79)
2
However, the model that we have developed does not immediately
Here,  = p and we have assumed that I p  I n . For typical
apply to this situation. This is obvious since we have assumed that
parameters, I p /I ≈ 0.1 and / ≈ 10−6 , we see that
two-fluid hydrodynamics applies throughout the system. In order to
address the crust superfluid problem, we would have to add a compo- E2 ∼ 5 × 10−8 E1 . (80)
nent representing the single fluid core, and ensure that it is coupled
In other words, in the two-component model the energy available for
to the two-fluid region in a suitable way. In particular, this core com-
radiation is much smaller than in the starquake case. Even though
ponent would affect (59). This generalization is complicated by the
it is not clear how the estimate will change if we account for the
fact that we would have to add appropriate boundary conditions at
energy radiated as heat, changes in internal and potential energy,
the interfaces. As our main aim is to compare the averaged model to
etc., it is clear that the result is rather pessimistic. If the estimate is
the detailed hydrodynamics, we will leave consideration of models
taken seriously, and glitches really represent a transfer of angular
with distinct superfluid regions for future work.
momentum as in the two-fluid model, then the gravitational-wave
signal from a pulsar glitch is unlikely to be detected by any future
generation of detectors. Of course, our level of understanding of
4.4 Energetics
this problem is still rudimentary. We need to improve our models
In the next section, we will consider the hydrodynamics associated considerably if we are to make more reliable estimates. The simu-
with a (core) glitch event. A key motivation for this discussion is the lations that we will now discuss provide an important step in this
need to understand the actual details of how a macroscopic glitch direction.
is triggered (presumably through a large-scale instability) and how
the system evolves once vortices become unpinned. By modelling
5 H Y D RO DY N A M I C S M O D E L
the required hydrodynamics, we hope to understand the nature of
glitches better. We should, for example, be able to establish to what The ‘bulk’ model that we have discussed so far is able to describe
extent the simple ‘bulk model’ we have discussed represents the some key properties of glitches. However, the model has obvious
behaviour of a true two-fluid system. We can also address other restrictions. In particular, it does not provide any information what-
interesting questions, concerning, for example, the modes of oscil- soever about the actual hydrodynamics of the event. By focusing on
lation that are excited in a glitch. This is an interesting question to solid-body motion, we are obviously considering only the averaged
ask because, first of all, there may be additional variability in the dynamics. In principle, one would expect the result to be relevant
glitch event and, secondly, the fluid oscillations may be associated when the dynamics is much slower than, say, the speed of sound
with gravitational-wave emission. It is obviously relevant to try to in the fluid. However, it is clear that we need to move beyond the
understand the nature of this gravitational-wave signal and estimate averaged model if we want to understand issues like the trigger
its amplitude. One should probably not expect glitching pulsar to mechanism for glitches, consider potential gravitational-wave sig-
be supreme gravitational-wave sources, but these estimates are nev- nals, etc. It is thus natural to consider the hydrodynamical aspects
ertheless interesting. Most importantly, since glitches are common of the glitch problem. To do this, we have extended the numer-
in young pulsars (and magnetars), it may be ‘reasonable’ to assume ical code that was recently developed by Passamonti, Haskell &
that the corresponding level of energy is associated with regular Andersson (2009) (see Peralta et al. 2005, 2006; Peralta & Melatos
dramatic events in a neutron star’s life. 2009 for a parallel effort). Within the two-fluid framework, we
Let us therefore consider the energetics of the problem. In past evolve perturbations of rotating, superfluid Newtonian stars in time.
studies, it has been common to estimate the energy available for The new version of the code includes the effects of mutual friction
radiation based on a single component model. In that case, the total and the perturbed gravitational potential (i.e. we are not working
kinetic energy and angular momentum are (obviously) given by in the Cowling approximation). We initiate the time-evolution with
suitable conditions that mimic a pre-glitch configuration, where
1 2
E= I and J = I . (76) the two fluids rotate uniformly with different velocities. Assum-
2 ing that the vortex pinning that is required to reach this state is
Assume that a glitch of size  results from a change in the moment instantaneously broken, we can evolve the system. This allows us
of inertia I. This would represent a ‘starquake’ in an elastic star. to determine the mutual friction damping, extract the associated
Then, assuming that the total angular momentum is conserved, it is gravitational signal and infer the oscillation modes that are excited


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
1068 T. Sidery, A. Passamonti and N. Andersson
during a ‘glitch’. In particular, we can test the analytical formula functional (16):
for the global spin-up time-scale τ (equation 69).  
E = E ρn , ρp , wnp
2
, (81)
The main motivation for our perturbative treatment is to consider
stellar models where the relative velocity lag between neutrons and where we have replaced the number densities nx with the mass
protons is very small as a deviation from stationary equilibrium densities ρ x (assuming for simplicity that the neutron and proton
configurations. These background models, which are such that the masses are equal, mp = mn ). When the relative velocity between the
two fluids corotate, are in β-equilibrium and co-exist throughout two fluids is small, equation (81) can be expanded in a Taylor series
the star’s volume, can be constructed by extending the standard (Prix et al. 2002; Passamonti et al. 2009). Then we have

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

self-consistent field method of Hachisu (1986). The details of this  4
method, and its application to superfluid stars can be found in E = E0 (ρn , ρp ) + α0 (ρn , ρp )wnp
2
+ O wnp , (82)
Yoshida & Eriguchi (2004) and Passamonti et al. (2009). In our where the entrainment function, α 0 , and the bulk EoS, E0 , can
current models, the crust is neglected (for simplicity). Our aim is to be independently specified on the corotating background, where
continue to add key physics to the model, and we plan to consider i
wpn = 0. In this paper, we consider two sets of polytropic super-
crustal effects in future work. fluid equations of state already used by Passamonti et al. (2009).
Non-corotating configurations can be determined using the per- Despite their simplicity, these models are very useful for investigat-
turbative approach developed by Yoshida & Eriguchi (2004), where ing the effects of the different physical parameters on the oscillation
the deviations from corotation are numerically computed by means dynamics. In particular, one of these two classes of EoS describes
of a variation of the self-consistent field method. We extend this models with composition gradients that couple the co- and counter-
numerical approach to study different superfluid equations of state moving degrees of freedom.
and generate rotating stellar sequences with constant mass. These The first EoS is given by the following expression:
non-corotating deviations are then implemented in the hydrodynam-
K 2Kσ
ical code as initial conditions and evolved in time with a system of E0 = ρ2 − ρn ρp
perturbation equations. This system is formed by the linearized ver- 1 − (1 + σ ) xp n 1 − (1 + σ ) xp
sions of the two-fluid mass conservation equations, the two Euler- K[1 + σ − (1 + 2σ ) xp ] 2
+ ρp , (83)
type equations (51) and the Poisson equation for the gravitational xp [1 − (1 + σ ) xp ]
potential. Technical details on the construction of the initial data
and the time domain numerical code will be provided elsewhere where K and the proton fraction xp are taken to be constant. As
(Passamonti & Andersson, in preparation). Here, we report only discussed by Prix et al. (2002), the parameter σ is related to the
results that can be directly compared with the analytic expressions ‘symmetry energy’ of the EoS. For this model, we have constructed
from the previous sections. a sequence of corotating axisymmetric configurations which do not
In a two-fluid model, the relative motion between protons and have composition gradients. These non-stratified stars correspond
neutrons can be approximately decomposed in co- and counter- to the A models used by Passamonti et al. (2009). Table 1 gives
moving components. With an appropriate choice of perturbation the main quantities of some rotating models that we consider in
variables, we can study the effects of these two degrees of freedom this paper. Details for more rapidly rotating models can be found in
on the oscillation spectrum and the factors that generate their cou- Passamonti et al. (2009).
pling (Andersson, Glampedakis & Haskell 2009a; Passamonti et al. The second analytical EoS, that describes stratified superfluid
2009). stars, can be written (Andersson, Comer & Langlois 2002; Prix &
The perturbation equations must be completed with an equa- Rieutord 2002; Passamonti et al. 2009) as
γ
tion of state (EoS), which can be described by the energy E0 = kn ρnγn + kp ρp p . (84)

Table 1. This table provides the main parameters of the corotating background models for both the A and C sequences.
√
Model Rp /Req / Gρ0 /K T/|W| × 102 5 )
Ip /(ρ0 Req 5 )
In /(ρ0 Req 3 )
M/(ρ0 Req

A0 1.000 00 0.000 00 0.000 00 0.000 00 0.033 28 0.299 51 1.2732

A1 0.997 92 0.059 13 0.081 56 0.058 02 0.033 19 0.298 68 1.2701
A2 0.983 33 0.166 75 0.229 99 0.384 82 0.032 53 0.292 78 1.2479
A3 0.950 00 0.287 99 0.396 27 1.169 18 0.031 02 0.279 15 1.1967
A4 0.933 33 0.330 81 0.456 29 1.568 85 0.030 25 0.272 24 1.1709
A5 0.900 00 0.402 68 0.555 43 2.382 95 0.028 69 0.258 22 1.1186
C0 1.000 00 0.000 00 0.000 00 0.000 00 0.019 06 0.246 57 1.0826
C1 0.997 92 0.055 86 0.084 03 0.045 61 0.019 00 0.245 84 1.0798
C2 0.983 33 0.157 64 0.237 16 0.366 82 0.018 58 0.240 64 1.0601
C3 0.950 00 0.271 45 0.408 37 1.113 03 0.017 60 0.228 61 1.0146
C4 0.933 33 0.312 46 0.470 06 1.492 36 0.017 11 0.222 51 0.9915
C5 0.900 00 0.380 06 0.571 77 2.262 85 0.016 10 0.210 09 0.9447
Note. The first column labels each model, while the second and third columns give, respectively, the ratio of polar to
equatorial axes and the angular velocity of the star. In the fourth column, the rotation rate is compared to the Kepler
velocity K that represents the mass shedding limit. The ratio between the rotational kinetic energy and gravitational
potential energy T/|W| is given in the fifth column. In the sixth and seventh columns, we show the moment of inertia
of the proton and neutron fluids, respectively, while in the eighth column we provide the stellar mass. All quantities are
given in dimensionless units, where G is the gravitational constant, ρ 0 represents the central mass density and Req is the
equatorial radius.


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
 The dynamics of pulsar glitches 1069
In this paper, we determine a sequence of rotating stars, whose which leads to the following expression for the initial fluid angular
non-rotating member is model III of Prix & Rieutord (2002). For velocities:
our polytropic models, we use γ n = 1.9 and γ p = 1.7, while the 
p  
2−γ
coefficients kx are given, in units GReq ρ0 x , by kn = 0.682 and = , (90)
  obs
kp = 3.419, respectively. Recall that G, Req and ρ 0 are the gravita-
tional constant, the equatorial radius and the central mass density, 1
respectively. Imposing β-equilibrium on the corotating background n = − (Ip p + I ), (91)
In
model, we can determine the proton fraction as (Prix & Rieutord

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

2002; Passamonti et al. 2009) where I = I p + I n .
For any corotating background, we can determine the non-

−1
(γp kp )Np Nn −Np corotating corrections to the mass density δρ x , the chemical po-
xp = 1 + μ̃ , (85)
(γn kn )Nn tential δ μ̃x and the gravitational potential δφ via the Yoshida &
Eriguchi (2004) approach, see Passamonti & Andersson (in prepa-
where N n and N p are the neutron and proton polytropic indices of ration) for more details. In a spherical coordinate basis, the initial
the EoS, defined as Nx = (γx − 1)−1 . For this model, a sequence velocity fields are then given by
of rotating stars can then be determined via the self-consistent field
method (Passamonti & Andersson, in preparation; Passamonti et al. δvxi = −x ϕ i . (92)
2009). We refer to these stars as models C, in order to distinguish
This initial data is close to that considered in the bulk dynamics
them from the models B used by Passamonti et al. (2009). See
model of the previous section. Hence, we can meaningfully com-
Table 1 for some of the rotating configurations for this model.
pare the results of our time-evolutions to the averaged results. This
comparison will give us a better idea of the validity of the solid-
body approach. Of course, by solving the hydrodynamics problem,
5.1 Glitch initial data
we will be able to proceed beyond the averaged model and discuss
According to the two-fluid model, a glitch brings a star with an other interesting issues.
initial velocity lag between protons and neutrons to a corotating
configuration. The observed glitch jump is very small, / 
1, and can be associated with angular momentum transfer from 5.2 Spin-up time
the superfluid neutrons to the proton component. Regardless of the Let us first compare the mutual friction damping extracted from
physical mechanism that generates the original velocity difference, the time-evolution to the body averaged analytical formula (69).
we can describe the initial conditions for a glitch as an axisymmet- Considering the sequence of non-stratified models A, we test the
ric configuration where protons and neutrons rotate with a small dependence of τ on the mutual friction strength B, the background
relative velocity. If we assume that this relative rotation is aligned angular velocity , the moment of inertia ratio I p /I and the entrain-
with the ‘background’ rotation axis, we can use the perturbative ment parameter ε̄. For the corotating background A models given
approach developed by Yoshida & Eriguchi (2004) to determine the in Table 1, we determine the initial condition for the time-evolution
difference between the initial non-corotating configuration and the code that corresponds to a ‘typical’ glitch jump / = 10−6 .
final corotating background. The related neutron velocity lag is then given by equation (91). It
As we have already discussed, the observed variation of the star’s is worth noting that the method used to construct the initial data is
angular velocity during a glitch can be associated with the proton linear in the two fluid velocities. This means that the results can be
velocity. In fact, due to the magnetic field coupling between the rescaled to other values of the glitch size.
crust and core protons, we can assume that all charged particles If we exclude the mutual friction term, the numerical evolutions
corotate. Meanwhile, the velocity of the superfluid neutrons must preserve the initial lag between protons and neutrons. At the same
be determined from the conservation laws. If we assume that the time, the initial conditions excite low level oscillations. This is
angular momentum of the two-fluid system is conserved during a expected as we are mapping a non-corotating axisymmetric config-
glitch, then the initial relative amplitude between the proton and uration on to an axisymmetric corotating background. When mutual
neutron perturbations can be determined from friction is switched on, the countermoving motion is damped, lead-
 ing to a corotating configuration on a time-scale of the order of τ .
Jx = 0, (86) We determine the spin-up time-scale by monitoring the ϕ compo-
x i
nent of the velocity difference wpn , assuming that it depends on time
−t/τ
where the perturbed angular momentum of the x component is given as wpn = wpn (t = 0) e
ϕ ϕ
. By taking the natural logarithm, we can
by extract τ from a linear fit of the time evolved data.
In Figs 1 and 2, we show the agreement of the numerical results
Jx = Ix x + Ix , (87) with the analytical formula (69). In the left-hand panel of Fig. 1, we
and I x is the moment of inertia of the x fluid in the corotating consider several rotating A models with vanishing symmetry energy
configuration, which rotates with angular velocity . We determine and entrainment and with constant proton fraction xp = I p /I = 0.1.
the perturbation I x from The models have different mutual friction strength, controlled by
 r the parameter B that we take to be constant throughout the star.
Ix = δρx (r  sin θ)2 dr , (88) The expected linear dependence of τ on the inverse of the mutual
0 friction parameter is confirmed by the hydrodynamical simulations.
where δρ x is the Eulerian perturbation of the mass density. From In the right-hand panel of Fig. 1, we fix instead the mutual friction
equation (86), we then have strength, B = 0.1, a rather large value, and study three sequences
of rotating models with different proton fraction, respectively, xp =
In n + Ip p + (In + Ip ) = 0, (89) 0.01, 0.05 and 0.1. The mutual friction damping time exhibits the


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
1070 T. Sidery, A. Passamonti and N. Andersson
1000

Model A 10 xp = 0.01

xp = 0.1 xp = 0.05

σ = 0.0 xp = 0.1
ε = 0.0
100

1/2
1/2

τ (G ρ0 )
τ (G ρ0 )

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

1
Rp / Req = 0.998

10 Rp / Req = 0.984
Model A
Rp / Req = 0.950
B = 0.1
Rp / Req = 0.933
σ = 0.0
Rp / Req = 0.900 ε = 0.0

1 0.1
10 100 1000 1/2 10
1/B (G ρ0 ) / Ω

Figure 1. This figure shows the spin-up time τ as a function of the inverse of the mutual friction parameter B (left-hand panel) and the background rotation
rate of the star  (right-hand panel) for the sequence of rotating models A. The solid lines show the behaviour predicted by equation (69), while the symbols
(see legend) represent the values of the spin-up time extracted from the hydrodynamical simulations. The physical quantities are given in dimensionless units
by using the gravitational constant G and the central mass density ρ 0 . The axes use logarithmic scales. All the five models A1–A5 shown in this figure have
both vanishing symmetry energy term σ and entrainment parameter ε̄. In the left-hand panel, the proton fraction is fixed to xp = 0.1 and the mutual friction is
varied. Meanwhile, in the right-hand panel, we show three sequences of rotating stars with the same mutual friction strength B = 0.1, but with three different
values of proton fraction, namely xp = 0.1, 0.05, and 0.01. In all cases, the numerical values of spin-up time show a good agreement with the analytical
result.

16 1000

Model A Model C
14
Rp / Req = 0.998 xp(0) = 0.1

12 Rp / Req = 0.984 ε = 0.0

xp = 0.1 100
10
1/2
1/2

B = 0.1
τ (G ρ0 )
τ (G ρ0 )

8 σ = 0.0

6
10 Rp / Req = 0.998

Rp / Req = 0.984
4
Rp / Req = 0.950
2

0 1
0.4 0.6 0.8 1 1.2 1.4 1.6 10 100 1000
1-ε 1/B

Figure 2. In this figure, we compare the numerical spin-up time τ with the analytical formula (69). The values extracted from the numerical code are shown as
open symbols (see legend), while the solid lines denote the analytical τ . In the left-hand panel, we consider the two models A1 and A2, with proton fraction xp =
0.1, vanishing symmetry energy σ = 0 and constant mutual friction parameter B = 0.1. The dependence of the numerically determined τ on the entrainment
parameter agrees very well with the analytical result. In the right-hand panel, we show (on a log-log scale) the damping time τ as a function of B−1 for the
three models C1–C3 with vanishing entrainment. The agreement between the numerical and analytical spin-up times is still good, although less accurate than
for the A models. See Fig. 3 and the main text for further discussion.

expected linear dependence on −1 . For models A1 and A2, we Next, we consider the stratified models C. The aim is to establish
test also the dependence of τ on the entrainment parameter ε̄, see to what extent equation (69) still provides accurate results for the
Fig. 2. In this case, the other stellar parameters are, respectively, spin-up time. On the one hand, one may not expect this to be the case
xp = 0.1, B = 0.1 and σ = 0. The linear dependence on 1 − ε̄ since the various parameters in the model are no longer uniform.
is clearly confirmed by the evolutions. According to equation (69), On the other hand, the simple prescription could still work pro-
the damping time should not depend (explicitly) on the symmetry vided that the parameters are interpreted in a body-averaged sense.
energy, σ . We have carried out simulations with different σ to We consider initial configurations with / = 10−6 and deter-
confirm this result. mine the non-corotating corrections using the method discussed in


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
 The dynamics of pulsar glitches 1071
10 est, this problem is unfortunately beyond the reach of our current
computational technology.
9 C1
A1
We will focus on the gravitational signal associated with the
8
C2 l = 2 axisymmetric oscillations that are excited in our glitch evolu-
A2 tions. At the linear perturbation level, the initial data excites a num-
7
C3 ber of the neutron star’s oscillation modes. Hence, a key question
6 A3 concerns which modes we expect to be present in the gravitational
Δ τ / τ x 100

signal. For a single fluid star, the general mode classification is based

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

5
on the main restoring force that acts on the displaced fluid elements
4 (Cowling 1941). In this work, we consider ‘acoustic modes’ that are
mainly restored by pressure variations. Since any perturbation of a
3
spherical star can be decomposed in vector harmonics, an oscillation
2 mode can be labelled by the harmonic indices (l, m) associated with
the spherical harmonics Ylm (θ, φ). This description can be extended
1
to rotating stellar models, as the modes can be tracked back to the
0 non-rotating limit. For any value of (l, m), the oscillation modes can
10 100 1000 be ordered by the number of radial nodes in their eigenfunctions.
1/B For acoustic modes, we then have the fundamental mode l f with no
Figure 3. In this figure, we estimate the agreement between the analytical
nodes and the series of pressure modes l pi with i nodes.
spin-up time τ from equation (69) and the values extracted from the hy- Our glitch model leads to axisymmetric initial data. Therefore,
drodynamical code. We show the relative deviation τ /τ as a function of we can only excite the family of axisymmetric modes, with m = 0.
the inverse of the mutual friction parameter B. The horizontal axis is on The quadrupole, l = 2, oscillations are expected to be dominant in
logarithmic scale. The results for models A and C are shown as filled and the gravitational signal. However, in rotating stars the gravitational-
open symbols, respectively (see legend for details). wave spectrum can also contain l = 0 ‘quasi-radial’ oscillation
modes. In the non-rotating limit, these become purely radial modes,
which do not generate gravitational radiation. The quasi-radial fun-
Section 5.1. For the three rotating models C1–C3, the dependence
damental mode will be denoted by F and its i overtones by Hi .
of the numerical damping time on the mutual friction parameter B
In a two-fluid model, the oscillation spectrum is richer, as the dis-
is shown in the right-hand panel of Fig. 2. Again, the results agree
placed fluid elements can now oscillate in phase and counterphase.
well with the analytical formula (69). However, a closer examina-
The comoving degrees of freedom generate oscillation modes that
tion of the data reveals that equation (69) is more accurate for the
are similar to those of a single fluid star, and are referred to as
non-stratified A models. For the first three rotating models of the
‘ordinary modes’. The countermoving degree of freedom produce a
two sequences A and C, we show in Fig. 3 the relative deviation
new class of modes known as ‘superfluid modes’. In our discussion,
between the numerical and analytical spin-up time for different mu-
ordinary and superfluid modes will be labelled with an upper index,
tual friction strengths. Comparing models with the same axis ratio,
for instance the l = 2 fundamental ordinary mode will be expressed
the error is generally smaller for models A (filled symbols).
as 2 f o , while the superfluid mode as 2 f s .
In general, we find that the agreement between the numerical
In a two-fluid star, the gravitational radiation is entirely generated
and analytical spin-up time is better for slowly rotating models that
i by the comoving degree of freedom (Andersson et al. 2009a). We
have strong mutual friction. In these cases, the damping of wpn is
determine the gravitational strain from the standard quadrupole
less contaminated by the excitation of axisymmetric oscillations,
formula (Thorne 1980):
and the numerical extraction of τ is more accurate.
20
G 1 d2 I
+ =
h20 TθθE2,20 , (93)
c4 r dt 2
5.3 Gravitational waves where the (l, m) = (2, 0) pure spin tensor harmonic is given by

So far we have discussed ‘global’ dynamics, e.g. how the two com- 1 15 2
ponents in the system relax to a corotating configuration due to TθθE2,20 = sin θ. (94)
8 π
the mutual friction. We now turn to the actual hydrodynamics, and
The mass quadrupole moment for the two-fluid star is defined by
consider to what extent this kind of glitch event radiates gravi-
(Andersson et al. 2009a)
tational waves. We have already established that these events are    
unlikely to be strong emitters of gravitational radiation. However, π π
I 20 = 8 dr δρ r 2 P 20 = 8 dr (δρn + δρp )r 2 P 20 ,
it seems inevitable that they should radiate at some level and it is 15 15
important to establish what that level may be. In particular, since (95)
there are a number of glitching pulsars in the Galaxy. The energy where P20 is the associated Legendre polynomial:
involved in glitches indicates how energetic ‘typical’ events in a
mature neutron star may be. In that way, these events provide inter- 3 cos2 θ − 1
P 20 = . (96)
esting benchmarks for gravitational-wave modelling. It is, however, 2
important to state already from the outset that we are not expecting In equation (93), the numerical calculation of the second-order
the mechanism that we are considering here to lead to detectable time derivative of the quadrupole moment may lead to inaccura-
signals. This is essentially because of the high level of symmetry in cies. However, the gravitational-wave extraction can be improved
the initial and final configurations, and the fact that we are assum- by using the dynamical equations and replacing the time deriva-
ing global vortex unpinning. The result may be quite different if we tive by spatial derivatives (Finn & Evans 1990). In our pertur-
were to model localized unpinning events. Although of great inter- bative analysis, we have found accurate results already with the


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
1072 T. Sidery, A. Passamonti and N. Andersson
momentum-formula, where only first time derivatives appear. More solve the stationary equations for the background model. We will
details and tests will be given by Passamonti & Andersson (in consider the case of a large glitch, where p / = 10−6 . This
preparation). means that, due to angular momentum conservation, the neutron
We rewrite equation (93) as follows: fluid slows down with n / = −1.11 × 10−7 for model A2
and n / = −7.74 × 10−8 for model C2. Note that we use a
G sin2 θ  20
+ = 4
h20 Ax , (97) first-order perturbative framework, where for a given corotating
c r x background model the results of the time-evolutions are linear with
where the quantity A20 respect to the parameter p /. Hence, the gravitational-wave

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

x is given by (Passamonti & Andersson, in
preparation) strain can be rescaled to any desired glitch magnitude. Furthermore,
 π/2  R   for slower rotating models that have the same glitch size, p /,
d δvxθ ∂P 20 we expect the perturbations and the gravitational strain to exhibit
Ax ≡ 8π
20
sin θ dθ r drρx δvx P + r
3 r 20
.
dt 0 0 2 ∂θ a quadratic dependence on the background rotation rate . Our
(98) numerical simulations reproduce this behaviour when the stars have
The energy radiated in gravitational waves can be determined from small rotational deformations, as in the case of the A1–A2 and C1–
the following equation (Thorne 1980): C2 models. Already for models A3 and C3, this scaling with 2
is less clear and obviously it is not expected to hold for more
     2
1 G ∞  d3 I 
20 2
2 G ∞  dA 20  rapidly rotating stars. In conclusion, from the evolutions of the
Erad =   dt = dt
32π c5 −∞  dt 3  15 c5 −∞  dt  A2 and C2 models, we can easily estimate the gravitational strain
 emitted by other slowly rotating models with different glitch size
16 2 G ∞ 2 20 2 and background rotation.
= π 5 ν |Â | dν , (99)
15 c 0 In Fig. 4, we show the characteristic strain hc for the A2 and C2
models with p / = 10−6 . The results refer to a star with mass
where A20 = A20n +Ap and  is its Fourier transform. The energy
20 20
M = 1.4 M , radius Req = 10 km and with a low level of mutual
spectrum is then given by
friction. We consider an evolution that lasts for ∼27.25 ms and ex-
dE G 16 tract the signal at 1 kpc.2 This is roughly the distance to the Vela
= 5 π2 ν 2 |Â20 |2 , (100)
dν c 15 pulsar. The first key result in Fig. 4 is that the gravitational signal of
and the characteristic gravitational-wave strain is defined as the C2 model is about 10 orders of magnitude larger than that of the
(Flanagan & Hughes 1998) A2 model. This is an enormous difference, given that these ought
 √   to be the same kind of events. The difference is due to the presence
G 2 1 dE G 32 ν 20 of composition gradients in the C models. In a stratified model, the
hc (ν) ≡ = |Â |. (101)
c3 π d dν c4 15 d co- and countermoving degrees of freedom are coupled during the
In our analysis, we focus on the two rapidly rotating models A2 and evolution. This coupling is crucial, since only the comoving motion
C2, which rotate at a significant fraction of the mass shedding limit generates gravitational radiation. The initial data for the pre-glitch
/K = 0.23 (see Table 1 for more details). For a typical neutron lag between neutrons and protons, in accordance with equations
star with mass M = 1.4 M and radius Req = 10 km, the rotation (90) and (91), represent a counter flow. Hence, in the non-stratified
period of the A2 and C2 models is about P  3 ms. These models A models, these conditions generate a purely countermoving motion
are therefore rotating much faster than all known glitching pulsars. between the two components that does not produce any gravitational
We will demonstrate that the generation of gravitational waves is signal at all. The fact that the strain of model A2 is not completely
very small even for these relatively rapidly rotating models. The zero in the left-hand panel of Fig. 4 is likely due to numerical errors.
result can be considered as an upper limit for real glitching neutron In fact, we have established that the level of radiation decreases with
stars. Moreover, we will show that the gravitational signal of slower increased resolution. The result shown for model A2 corresponds
rotating models can be determined by a simple rescaling of the A2 to an initial non-corotating configuration where the variation of the
and C2 signals. total rotational kinetic energy is Erot /Erot  10−20 . If we increase
At the linear perturbation level, the entrainment parameter ε̄ the precision of the self-consistent field method that provides the
can be chosen independently from the background model. Recent initial data we can lower this value and consequently the gravita-
work suggests that this parameter can assume values in the range tional signal converges to zero. Moreover, the numerical noise in
0.2 ≤ ε̄ ≤ 0.8 (Chamel 2008). The effect of the entrainment on the simulation excites some oscillation modes that are related to the
the oscillation spectrum has been extensively studied by Prix & comoving motion. In the left-hand panel of Fig. 4, we identify the
Rieutord (2002) and Passamonti et al. (2009). The oscillation modes fundamental l = 2 mode 2 f, the first two pressure modes 2 p1 and 2 p2 ,
that are mainly affected by ε̄ are the so-called ‘superfluid’ modes, and the quasi-radial fundamental mode F with its first overtone H1 .
which are associated with the countermoving degree of freedom. Let us contrast the results for the non-stratified A2 model with the
Simple relations between the frequencies of the superfluid acous- results for model C2. In this case, the chemical coupling between
tic and inertial modes and the entrainment parameter ε̄ have been
obtained by Andersson et al. (2009a), Passamonti et al. (2009) and
2 It is worth noting that we have evolved the system for approximately
Haskell, Andersson & Passamonti (2009). In this work, we show
10 rotation periods. We did not extend the evolutions because, in reality,
results only for the ε̄ = 0.5 case, as apart from the spectral prop-
one would expect the coupling of the two components to start playing a
erties discussed above, we did not find any qualitative difference in
role at this stage. If this were not the case and the oscillation prevailed
simulations using other values of this parameter. for the entire time-scale of gravitational-wave damping, then the f-mode
With the method discussed in Section 5.1, we set up initial data may last a few seconds. This would be about a factor of 100 longer than
that mimic the configuration of an axisymmetric glitch. The initial our evolution, meaning that the effective gravitational-wave strain could
value for the proton and neutron angular velocity can be determined increase by perhaps an order of magnitude. It would still be too weak to be
from equations (90) and (91), once we fix the glitch size / and detectable.


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
 The dynamics of pulsar glitches 1073
-35
10
Model A2 Model C2
2 o
-26 f 2 o
10 p1 P = 3.0 ms
P = 3.07 ms
2
f ε = 0.5
-36
xp= 0.1
−6
10
2 s s ΔΩp/ Ω = 10
ε = 0.5 f F 2 o
p2
F σ = 0.0 o
-27 F

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

−6
10
ΔΩp/Ω = 10 2 s
p2

hc
hc

-37 s
10 o
H2 H1 o
H3
s
2 H2
p1 2
H1 p2 -28
10
-38
10
d = 1 kpc d = 1 kpc
-29
0 2 4 6 8 10 12 10 0 2 4 6 8 10 12 14 16 18
ν [ kHz ] ν [ kHz ]

Figure 4. This figure displays the gravitational-wave signal generated by our hydrodynamical glitch simulations for the two models A2 (left-hand panel) and
C2 (right-hand panel). On the horizontal and vertical axes, we plot the oscillation frequencies and the characteristic strain extracted at a distance of 1 kpc from
the source. We consider a neutron star with typical mass M = 1.4 M and radius Req = 10 km. Models A2 and C2 then correspond to stars with rotation
period P = 3.07 and 3.00 ms, respectively. The other stellar parameters are ε̄ = 0.5 for the entrainment and σ = 0 for the symmetry energy. In the case of
model A2, the proton fraction is constant, xp = 0.1, while model C2 is stratified with central proton fraction xp (0) = 0.1. The initial configuration corresponds
to a large glitch with p / = 10−6 , as described in the main text. We run the simulation for about 27.25 ms, i.e. about nine rotation periods, and neglect the
mutual friction force. From the displayed results, the strong effects of the stratification on both the oscillation spectrum and gravitational-wave amplitude are
evident.

the two fluids introduces a comoving motion already in the initial evolution requires additional assumptions, most likely, concerning
data. The evolutions then generate a larger gravitational strain and the repinning of vortices. Understanding this phase better, e.g. con-
several oscillations modes, like the fundamental l = 0 and 2 modes necting it to the two-fluid hydrodynamics and the averaged forces
and their respective overtones. In particular, in the right-hand panel that act on the vortices, is an important challenge for the future. It
of Fig. 4, we note that both ordinary and superfluid modes are seems clear that vortex creep will play a central role (Anderson &
present in the gravitational radiation. This is due to the coupling Itoh 1975; Alpar et al. 1984b, 1989, 1993; Link et al. 1993), but
of the degrees of freedom (the oscillation modes are no longer this mechanism has not yet been discussed in terms of the macro-
purely co- or countermoving). The mode frequencies of the non- scopic hydrodynamics. This issue needs to be addressed if we are
rotating model C0 have been compared with the results of Prix & to develop more detailed models of glitch dynamics. We definitely
Rieutord (2002). The two results agree to better than 1.4 per cent need to move beyond phenomenology.
(Passamonti & Andersson, in preparation). We have identified the As a first step towards hydrodynamic glitch modelling, we have
oscillation modes of the C2 model by carrying out simulations with extended the recent linear perturbation evolution code of Passamonti
different values for the entrainment ε̄ and tracking the superfluid et al. (2009) to include the mutual friction and the perturbed gravita-
modes as the parameter changes. To this end, we have also used the tional potential. Initiated with perturbations that represent two fluids
analytical formulae determined by Passamonti et al. (2009). rotating uniformly at different rates, the numerical code shows how
the system relaxes to corotation. We have analysed this relaxation in
detail and demonstrated that the behaviour is accurately described
6 CONCLUDING REMARKS by the phenomenological model, at least for non-stratified stellar
models. When the star is stratified (e.g. has varying composition),
We have discussed the dynamics of pulsar glitch events from two,
the relaxation deviates from the simple model. This is as expected,
complementary, points of view. First we constructed a simple model
since the global model was derived under the assumption of uni-
based on global ‘averaging’ of the standard two-fluid equations in-
form parameters. Of course, the numerical evolutions provide us
cluding the mutual friction due to superfluid vortices. This analysis
with a useful tool for studying the behaviour of more complex stel-
provides a more detailed derivation of the phenomenological rela-
lar models. In addition, our time-evolutions provide a first insight
tions that have been used in many discussions of glitches. In par-
into the excitation of neutron star oscillations by glitches. Our re-
ticular, our final relations clarify how the spin-up time depends on
sults show that a set of axisymmetric modes are excited by the glitch
key parameters like the entrainment. The derivation also highlights
initial data. These modes will radiate gravitational waves,3 and it is
the various assumptions and the restricted validity of the model.
Anyway, for typical values of the parameters (see Section 4.3), our
model has a glitch rise time shorter than the upper bound set by 3 In principle, the induced oscillations may also lead to variations in the
current observations. The model provides a useful description of electromagnetic signal. However, in order to quantify this effect, one would
the actual glitch event, but it does not account for the subsequent need a more detailed analysis of the coupling between the motion of the
long-term relaxation (on a time-scale of days to months) of the sys- crust and the magnetosphere. Such estimates are beyond the scope of the
tem. A key conclusion from our discussion is that the late stages of present analysis, but it is worth noting that the oscillation modes that we


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074
1074 T. Sidery, A. Passamonti and N. Andersson
important to establish if the associated signals may be observable Alpar M. A., Cheng K. S., Pines D., 1989, ApJ, 346, 823
with future detectors. In this respect, our results are quite pes- Alpar M. A., Chau H. F., Cheng K. S., Pines D., 1993, ApJ, 409, 345
simistic. In the cases that we have considered, the gravitational-wave Anderson P. W., Itoh N., 1975, Nat, 256, 25
signal is too weak to be detectable (even with a third generation of Andersson N., Comer G. L., 2001, Phys. Rev. Lett., 87, 241101
Andersson N., Comer G. L., 2006, Class Quantum Gravity, 23, 5505
detectors).4 However, it is not clear that this is the final say on
Andersson N., Comer G. L., Langlois D., 2002, Phys. Rev. D, 66, 104002
the matter. One should keep in mind that the gravitational-wave
Andersson N., Comer G. L., Prix R., 2003, Phys. Rev. Lett., 90, 091101
strain differs enormously for our two model configurations. The Andersson N., Sidery T., Comer G. L., 2006, MNRAS, 368, 162
non-stratified model does not (in principle) radiate at all, while the

Downloaded from https://academic.oup.com/mnras/article/405/2/1061/1181494 by Universidad de Santiago de Chile user on 02 December 2020

Andersson N., Sidery T., Comer G. L., 2007, MNRAS, 381, 747
stratified model leads to a qualitatively interesting (albeit weak) Andersson N., Glampedakis K., Haskell B., 2009a, Phys. Rev. D, 79, 103009
signal. The enormous difference between these results shows that Andersson N., Glampedakis K., Samuelsson L., 2009b, MNRAS, 396, 894
we need to continue to refine our modelling. We obviously have to Avogadro P., Barranco F., Broglia R. A., Vigezzi E., 2008, Nucl. Phys. A,
account for the variation of composition throughout the star, and 811, 378
consider the fact that superfluid components will only be present in Carter B., Chamel N., 2004, Int. J. Mod. Phys. D, 13, 291
specific density regions. We also need to understand the nature of Carter B., Chamel N., Haensel P., 2005, Nucl. Phys. A, 748, 675
Chamel N., 2006, Nucl. Phys. A, 773, 263
the vortex pinning better. In our models, we have assumed that the
Chamel N., 2008, MNRAS, 388, 737
vortices unpin in a catastrophic global event. It is far from clear that
Chamel N., Carter B., 2006, MNRAS, 368, 796
this is the case in a real system. It could, for example, be that the Cowling T. G., 1941, MNRAS, 101, 367
unpinning is localized. This would make the event less symmetric Dodson R. G., McCulloch P. M., Lewis D. R., 2002, ApJ, 564, L85
which may enhance the gravitational-wave signal. We clearly need Donati P., Pizzochero P. M., 2006, Phys. Lett. B, 640, 74
to understand the actual mechanism that triggers the glitches better. Finn L. S., Evans C. R., 1990, ApJ, 351, 588
The superfluid instability discussed by Glampedakis & Andersson Flanagan É. É., Hughes S. A., 1998, Phys. Rev. D, 57, 4535
(2009) is interesting in this respect, but we need to study this mech- Glampedakis K., Andersson N., 2009, Phys. Rev. Lett., 102, 141101
anism in more detail to establish to what extent it can operate in a Gusakov M. E., Haensel P., 2005, Nucl. Phys. A, 761, 333
real neutron star. Hachisu I., 1986, ApJS, 62, 461
Haskell B., Andersson N., Passamonti A., 2009, MNRAS, 397, 1464
To make progress, we need to overcome a number of challenges.
Larson M. B., Link B., 2002, MNRAS, 333, 613
Yet, recent developments have provided us with interesting insights
Lattimer J. M., Prakash M., 2004, Sci, 304, 536
and (most importantly) computational technology that should allow Link B., 2003, Phys. Rev. Lett., 91, 101101
us to study much more realistic neutron star models in the not too Link B., Epstein R. I., 1996, ApJ, 457, 844
distant future. Link B., Epstein R. I., Baym G., 1993, ApJ, 403, 285
Link B., Epstein R. I., Lattimer J. M., 1999, Phys. Rev. Lett., 83, 3362
Lyne A. G., Smith F. G., Pritchard R. S., 1992, Nat, 359, 706
AC K N OW L E D G M E N T S Lyne A. G., Shemar S. L., Smith F. G., 2000, MNRAS, 315, 534
NA acknowledges support from STFC via grant number Melatos A., Warszawski L., 2009, ApJ, 700, 1524
PP/E001025/1. TS is supported by a post-doc fellowship from EU Melatos A., Peralta C., Wyithe J. S. B., 2008, ApJ, 672, 1103
Mendell G., 1991, ApJ, 380, 530
FP6 Transfer of Knowledge project ‘Astrophysics of Neutron Stars’
Passamonti A., Haskell B., Andersson N., 2009, MNRAS, 396, 951
(ASTRONS, MTKD-CT-2006-042722) at Sabanci University.
Peralta C., Melatos A., 2009, ApJ, 701, L75
Peralta C., Melatos A., Giacobello M., Ooi A., 2005, ApJ, 635, 1224
REFERENCES Peralta C., Melatos A., Giacobello M., Ooi A., 2006, ApJ, 644, L53
Prix R., 2004, Phys. Rev. D, 69, 043001
Alpar M. A., Anderson P. W., Pines D., Shaham J., 1981, ApJ, 249, L29 Prix R., Rieutord M., 2002, A&A, 393, 949
Alpar M. A., Langer S. A., Sauls J. A., 1984a, ApJ, 282, 533 Prix R., Comer G. L., Andersson N., 2002, A&A, 381, 178
Alpar M. A., Pines D., Anderson P. W., Shaham J., 1984b, ApJ, 276, 325 Prix R., Comer G. L., Andersson N., 2004, MNRAS, 348, 625
Ruderman M., 1976, ApJ, 203, 213
Shapiro S. L., Teukolsky S. A., 1983, Black Holes, White Dwarfs, and
consider are all in the kHz range (much faster than the spin-period) meaning Neutron Stars. Wiley Interscience, New York
that they would not be seen as ‘modulations’ of the pulsar signal. Sidery T. L., 2008, PhD thesis, Univ. Southampton
4 At first sight this conclusion seems at variance with the results of van
Thorne K. S., 1980, Rev. Mod. Phys., 52, 299
Eysden & Melatos (2008), who consider a different glitch scenario. In their van Eysden C. A., Melatos A., 2008, Class. Quantum Gravity, 25, 225020
(cylindrical) model problem, the gravitational-waves are associated with Warszawski L., Melatos A., 2008, MNRAS, 390, 175
the large-scale Ekman flow that results from a rotational lag between the Watts A. L., Strohmayer T. E., 2007, Ap&SS, 308, 625
crust and the core in the star. The two mechanisms are obviously different. Yoshida S., Eriguchi Y., 2004, MNRAS, 347, 575
In particular, in our case the event is impulsive and one would expect the
signal to be burst-like. We certainly cannot envisage the 14 d integration
suggested by van Eysden & Melatos (2008) to improve the signal-to-noise
ratio. Basically, the model parameters used by van Eysden & Melatos (2008)
seem rather optimistic. This paper has been typeset from a TEX/LATEX file prepared by the author.


C 2010 The Authors. Journal compilation 
C 2010 RAS, MNRAS 405, 1061–1074