Digital Object Identier (DOI) 10.

1140/epjc/s2003-01192-6
Eur. Phys. J. C (2003)
THE EUROPEAN
PHYSICAL JOURNAL C
Magnetic collapse of a neutron gas:
Can magnetars indeed be formed?
A. Perez Martnez
1
, H. Perez Rojas
1,2,3
, H. J. Mosquera Cuesta
3,4
1
Grupo de Fsica Teorica, ICIMAF, Calle E No. 309, 10400 La Habana, Cuba
2
High Energy Physics Division, Department of Physics, University of Helsinki and Helsinki Institute of Physics, P.O. Box 64,
00014 Helsinki, Finland
3
Abdus Salam International Centre for Theoretical Physics, P.O. Box 586, Strada Costiera 11, Miramare 34100, Trieste, Italy
4
Centro Brasileiro de Pesquisas Fsicas, Laboratorio de Cosmologia e Fsica Experimental de Altas Energias, Rua Dr. Xavier
Sigaud 150, CEP 22290-180, Urca, Rio de Janeiro, RJ, Brazil
Received: 25 November 2002 / Revised version: 25 February 2003 /
Published online: c Springer-Verlag / Societ` a Italiana di Fisica 2003
Abstract. A relativistic degenerate neutron gas in equilibrium with a background of electrons and protons
in a magnetic eld exerts its pressure anisotropically, having a smaller value perpendicular to than along
the magnetic eld. For critical elds the magnetic pressure may produce the vanishing of the equatorial
pressure of the neutron gas. Taking this as a model for neutron stars, the outcome could be a transverse
collapse of the star. This xes a limit to the elds to be observable in stable neutron star pulsars as
a function of their density. The nal structure left over after the implosion might be a mixed phase of
nucleons and a meson condensate, a strange star, or a highly distorted black hole or black cigar, but
not a magnetar, if viewed as a superstrongly magnetized neutron star. However, we do not exclude the
possibility of superstrong magnetic elds arising in supernova explosions which lead directly to strange
stars. In other words, if any magnetars exist, they cannot be neutron stars.
1 Introduction
Gravitational collapse occurs in a body of mass M and
radius R when its rest energy is of the same order as its
gravitational energy, i.e., Mc
2
GM
2
/R. We would like
to argue that for a macroscopic magnetized body, e.g.,
composed of neutrons in an external eld

B ([

B[ B),
new physics arises and a sort of collapse occurs when its
internal energy density U is of the same order as its mag-
netic energy density

/

B, where

/ is the magnetiza-
tion. This problem is interesting in the context of both
cosmology and astrophysics, as for instance in the study
of objects such as neutron stars. A gas of neutral parti-
cles having an anomalous magnetic moment (as a model
for neutron stars (NSs); here we assume, as usually, a
background of electrons and protons in equilibrium,
which is demanded by Paulis principle to guarantee neu-
tron stability), when placed in extremely strong magnetic
elds, has a non-linear (ferromagnetic) response to the
external eld and is also unstable due to the vanishing
of the transverse pressure for surface elds strong enough
(B
surf
10
16
G). In this phenomenon quantum eects
play an essential role due to the coupling of the particles
spins to the microscopic eld

B seen by the particles (spin-
polarization). For elds of this order of magnitude there
are values of the density for which the magnetic energy of
the system becomes of the same order of magnitude as its
total energy. Under these physical conditions any struc-
ture of superdense matter composed of neutral particles
having a magnetic moment may undergo a transverse col-
lapse when its pressure perpendicular to

B vanishes. This
implosion is driven by the same mechanism described in
[1] for charged particles.
We present in this paper, which is an elaborated ver-
sion of [2], the main ideas concerning the role of ultra-
strong magnetic elds in a gas of neutral particles: The
standard model of NSs, as envisioned by Duncan and
Thompson in their model of magnetars [3]. The funda-
mental result that we obtain shows that a NS, i.e., a neu-
tron gas permeated by a superstrong magnetic eld, is
unstable and must collapse. This result seems to ban the
possibility of the formation of magnetars. We stress in
this respect the fact that similar results were obtained
by two dierent groups: Khalilov [4] and Ghosh, Mandal
and Chakrabarty [5]. Although both teams of researchers
arrived at the same conclusion as we did, quite dierent
approaches were pursued.
This paper is organized as follows: Sect. 2 reviews the
concept of anisotropic pressure in self-gravitating systems
like NSs. In Sect. 3 are given the tools to construct the
energy-momentum tensor of a neutron gas. Section4 dis-
cusses the main dierences between the classical and quan-
tum collapse of a gas conguration in an approximation-
independent way, based on the sign of the electromagnetic
2 A. Perez Martnez et al.: Magnetic collapse of a neutron gas
response of the medium to the external applied eld. In
Sect. 5 the dynamics of the neutron gas composing a NS
is studied in the presence of a magnetic eld, followed in
Sect. 6 by the derivation of its thermodynamical poten-
tial and magnetization. Section7 discusses the conditions
for the collapse to take place, and an application of this
physics to the stability analysis of the proposed magnetars
is given. The main result of this paper shows that such ob-
jects should not form if they are envisioned as a standard
neutron gas in Fermi beta equilibrium as is claimed in the
original idea introducing the concept of a magnetar. Some
closing remarks and further potential applications of this
theory are part of the nal section.
2 Anisotropic pressures
in self-gravitating systems
The issue of local anisotropy in pressures was extensively
reviewed by Herrera and Santos [6] in a general relativistic
approach. These authors present several physical mecha-
nisms for its origin in both extremely low and very high
density systems, which may include astrophysical compact
objects. In the case of highly dense systems, it was pointed
out that exotic phase transitions could occur during
gravitational collapse, which is exactly the problem we are
concerned with in this paper. However, the more funda-
mental problem regarding the appearance of anisotropic
pressures in strongly magnetized compact stars was left
open. More recently, Mak and Harko [7] presented a class
of exact solutions of Einsteins equations corresponding to
anisotropic stellar congurations which can describe real-
istic neutron stars.
We want to provide a more detailed description of the
arising of anisotropic pressures in a relativistic system like
a NS, an essential point in understanding the physics be-
hind the problem of stability of ultramagnetized compact
stars. We shall give rstly general arguments to support
our view, and then we concentrate on the specic calcu-
lations, in the one-loop approximation, of the thermody-
namic potential of the neutron star conguration and its
magnetization, the properties from which the most crucial
conclusions can be drawn.
To x ideas, we shall work in the grand canonical en-
semble, and we are considering a subsystem, some region
inside the star. Such a subsystem is under the inuence
of the magnetic eld created by the rest of the system,
which we name

H ([

H[ H). The response of the sub-

system is to polarize itself creating a magnetization

/
in the medium (the neutron gas) satisfying the relation

H =

B 4

/. Obviously, inside the subsystem the mi-
croscopic eld is

B =

H + 4

/, since

B (also called
magnetic induction) and the electric eld

E are the true
elds acting on the electric charges and magnetic dipoles
of elementary particles [8, 9]. (Note, however, that Lan-
dau in [10] uses the notation

H to denote the magnetic
eld in vacuum). The eld

B, as well as

E, satises the
Maxwell equations for particles in vacuum. In what fol-
lows, when we are to discuss the dynamics of the particles
in the neutron gas, we will sometimes refer to

B as the
external magnetic eld, as it is usually called in quantum
eld theory and astrophysics. For an external distant ob-
server,

B =

H (in Gaussian units), since the magnetiza-
tion is assumed to exist only inside the star. We emphasize
that actually

B and

H are external elds within dierent
contexts:

H is external to the subsystem object of study
in the grand canonical ensemble, whereas

B is external to
any particle chosen in the subsystem (it feels, in addition
to

H, the contribution from the magnetization eld 4

/
due to the other particles of the subsystem).
In the case of a gas of electrically charged particles in
an external constant magnetic eld

B, in classical electro-
dynamics, it is the Lorentz force

F = ev

B/c that is the
source of an asymmetry in the pressure components paral-
lel and perpendicular to

B. By writing ev =

jV , where
V = dx
1
dx
2
dx
3
, calling f
i
= F
i
/V the ith component
of the force density, and substituting

j = c

/, one
has
f
i
= (
i
/
s
)B
s
+ (
s
/
i
)B
s
. (1)
By multiplying by V = dx
1
dx
2
dx
3
and assuming
B
s
= B
s3
and /
i
/x
3
= 0 (actually we also have
/
i
= /
i3
), only the rst term in (1) remains non-
zero, and one recovers the expression for the Lorentz force,
which is obviously perpendicular to the eld

B. For the
corresponding pressure it yields P
L
=

/

B. This is
a classical eect and obviously P
L
must be added to the
usual kinetic isotropic pressure P
0
, so that the total trans-
verse pressure becomes P

= P
0
+P
L
. As in classical elec-
trodynamics, by the Lenz law,

/ is opposite to

B (spin
eects are neglected); then

/

B < 0, and P
L
> 0. The
opposite case occurs when

/ is parallel to

B, which oc-
curs in the quantum case, i.e., when spin eects are taken
into account. We should be alert, however, that in the def-
initions and derivations that follow use will be made of the
magnitudes [

B[ and [

/[ / of the two vectors

B and

/ instead of the vectors themselves.

3 The energy-momentum tensor
of a neutron gas
Based on more fundamental grounds, one may write the
general structure of the energy-momentum tensor of a
neutron gas in an external eld

B in the same way as
one does to construct general tensors, like the polariza-
tion operator tensors [11], for instance. In an external
eld F

, in addition to the basic 4-velocity vectors of

the medium, u

, and the particle momentum k

, we have
two extra vectors F

, F
2

, to form a basis of inde-

pendent vectors (in what follows we shall use the nota-
tion F
2

= F

). From them we may build a set of

basic tensors which, together with the tensors

, F

,
F
2

, serves as a basis in terms of which we can expand

any tensor structure related to the particle dynamics
in particular, the energy-momentum tensor. However, to
get rid of tensor structures containing o-diagonal terms,
A. Perez Martnez et al.: Magnetic collapse of a neutron gas 3
which would correspond to unwanted shearing stresses in
the uid rest frame u = (0, 0, 0, u
4
), we exclude some of
them, i.e., k

, k

, F

, u

, or any of their com-

binations. By following the arguments used in [11], we
conclude that we are left in the present case with three
basic tensors:

, F
2

, u

, to describe the dynamics of

such a neutron gas. Thus, the structure of such an energy-
momentum tensor is then expected to be of the form
T

= a

+bF
2

+cu

, (2)
where , = 0, 1, 2, 3, and a = p is the isotropic pressure
term, b = //B and c = U + p. In the present case the
second of these tensors can be written in a simpler form
as F
2

= B
2

. The tensor (2) has then the spatial

eigenvalues p B/, p B/, p, and the time eigenvalue
c = U p, since u

= 1. These eigenvalues exhibit

the anisotropy in the pressures perpendicular and parallel
to B.
In dealing with a quantum gas in an external eld we
shall assume that the sources of the eld

H are either clas-
sical (currents) or are due to quantum eects. Although it
is out of the scope of this paper to discuss the mechanism
for producing such a eld, we suggest, however, some vi-
able sources able to induce a self-consistent eld, as for
instance a condensate of the vector meson , the neutron
spinspin ferromagnetic [12] coupling (see below), or even
by diquark [13] condensation.
As in the case of the electron in an atom the basic
dynamics in our present case is described by the Dirac
equation in an external eld, in place of the Lorentz force.
According to the Ehrenfest theorem, classical dynamics is
contained in the adequate average of quantum dynamics.
But quantum dynamics leads also to several new phenom-
ena not having a classical partner. After solving the Dirac
equation one gets the energy eigenvalue spectrum [14].
This energy depends on the microscopic magnetic eld B
through some interaction term in the initial Lagrangian.
These energy eigenvalues, after the quantum statistical
average, determine the thermodynamic properties, such
as the neutron gas pressure. If the coupling constant is
turned to zero, the particles would not feel any pressure
coming from the external eld. Thus, since the classical
Maxwell stress tensor of the eld H does not express by
itself the interaction with the particles, i.e., it expresses
the momentum and energy of the external eld, we have
no need to add it to the expression (4) below. The ten-
sor (4) already contains the basic tensor structures of the
problem, including the Maxwell stress tensor of the eld
B, which depends on the external eld H and the magne-
tization [

/[.
The total external eld B contributes with virtual par-
ticles, expressed through the EulerHeisenberg vacuum
terms arising in the regularization of the quantum vac-
uum terms [15]. These vacuum terms appear in the cal-
culation of the basic statistical quantity, the thermody-
namical potential:
1
ln:, where :, the grand
partition function, is built up on the particle spectrum.
Our thermodynamical potential is the sum of two terms,
=
st
+
V
, the nite statistical term
st
plus the vac-
uum eld contribution
V
, which is divergent. In the pro-
cess of regularization, it absorbs the classical eld energy
density term B
2
/8. We observe here that Landau, on p.
69 of [16], uses the specic term thermodynamic potential,
denoted by in that reference, to denote what in west-
ern literature is known as the Gibbs free energy, which is
denoted by G. Our thermodynamic potential is just what
Landau denes as a new thermodynamic potential ,
but we have taken it per unit volume, which is dependent
on T and in the absence of an external eld, and in our
case is dependent also on B. Observe that = F G.
In the zero eld case, it would be = P, where P is
the isotropic pressure. Due to the spatial anisotropy intro-
duced by the magnetic eld B, the pressures are not the
same in all directions, and only in the direction parallel to
B it acquires the value (B) = P
3
(see below).
The coupling of the spin dipole moment of neutrons in
an external magnetic eld B produces a loss of rotational
symmetry of the particle spectrum (in what follows we will
consider B along the x
3
axis). From the spectrum, which is
expressed in terms of B, by the standard methods of nite
temperature quantum eld theory we obtain the thermo-
dynamical potential (per unit volume) = (B), and
from it all the thermodynamic properties of the system,
in particular its magnetization, as is done by Huang [17]
(p. 237), which is the statistical average /= (/B).
Notice that this denition is consistent with our con-
vention by which thermodynamical quantities are dened
in terms of the microscopic magnetic eld B acting on
the particles as an independent variable (see for instance
[9]), instead of using the quantity H. Thus, the thermody-
namical variable conjugate to the magnetic eld B is the
quantity /(B), the system magnetization as introduced
above. We may write then
=
_
/dB P
0
, (3)
where P
0
= (0) is the term corresponding to the zero
magnetic eld pressure. One must emphasize that in the
quantum relativistic case depends on B non-linearly.
From the explicit expression for given in another sec-
tion below, one nds that the dependence of the energy
spectrum on the particle momentum is not rotationally
invariant. This fact determines a reduction of the symme-
try of the otherwise isotropic thermodynamic properties
of the system such as the pressure, which is expected to
be axially symmetric for the reasons pointed out above.
By using the Green function method it is found that
the energy-momentum tensor of matter in an external con-
stant magnetic eld obeys the general structure (2), and
we have [1]
T

=
_
T/T +

r
/
r
_

4
(4)
+ 4F

/F
2

,
where r runs over the species involved. Below we will take
r = n, p, e to describe the neutron, proton and electron
component of the star gas. Expression (4) in the zero eld
limit reproduces the usual isotropic energy-momentum
4 A. Perez Martnez et al.: Magnetic collapse of a neutron gas
tensor T

= P

(P +U)
4

4
of a perfect uid. From
(4) the spatial components are T
33
= P
3
= , T
11
=
T
22
= P

= B/. The time component T

00

U
r
= TS
r

r
N
r

r
, in which U
r

r
N
r

r
are quantities of the same order of magnitude. We shall
assume that inside a stable NS there is locally a bal-
ance between the mechanical stresses (and forces) from
the pressure exerted by the quantum gas and the gravi-
tational stress, i.e., the gravitational force per unit area,
exerted by the star mass, as dened below by (6) and (7).
By abusing a bit of the phrasing we shall refer in this pa-
per to this last equation as the gravitational pressure,
in an attempt to turn to more meaningful discussions re-
garding eects of the gas pressure and those stemming
from its own gravity. In the case of a neutron star the cor-
rect treatment of this problem requires general relativistic
corrections [18]. However, for the general purpose of this
paper we consider here the problem in a rough manner,
in a similar fashion as it was done in the white dwarf case
with zero magnetic eld [17], in Newtonian gravity. If dS
i
is a surface element and dx
i
an increase in the coordinate
i = 1, 2, 3, one can express the balance between the work
done on the star and the variation of its gravitational en-
ergy, E
g
GM
2
R
1
, under a small change of volume V
by
_ _
T
ij
dS
j
+
E
g
x
i
_
dx
i
= 0 , (5)
where the quantity in parentheses must be zero; it de-
scribes the balance of forces. We will use now cylindrical
coordinates with the z axis parallel to B, and assume that
the shape of the deformed star can be (very roughly) ap-
proximated by the simplied model of a cylinder of height
Z and radius r

(we could equally well use as a simple

model a cylinder of height Z ended by two half-spheres of
radius r

; then (6) and (7) below would dier by unimpor-

tant numerical factors). We assume also that the changes
of the cylinders volume V and surface S are small,
so that the total surface of the star S = 2r

(r

+ Z)
is approximately constant. Then we have dS

= 2r

dz,
dS
3
= 2r

dr

and integrating we get the equilibrium

between the gravitational and gas pressures P

= P
g
and P
33
= P
g33
, where
P
g
=
1
2r

Z
E
g
r

(6)
and
P
g33
=
1
r
2

E
g
z
. (7)
Equations (6) and (7) are to be interpreted respec-
tively as the transverse and longitudinal gravitational
pressures in the sense dened above. By assuming that
E
g
r

and
E
g
z
are quantities of the same order, to preserve
the balance of transverse pressures when P

decreases
(due to the increase of the product B/), the quantity
1/2r

Z must decrease. This is achieved whenever Z in-

creases faster than r

decreases (all this can be seen by as-

suming that the total surface S = const and dr

/d < 0,
where is a parameter like time). In that case the outcome
is a stretching of the body along the direction of B. Thus,
the anisotropy of the pressures in our problem suggests
that matter in the body is also distributed anisotropically,
which leads to prolate isobaric surfaces
1
.
The condition P

= 0 implies also that U B/. As

P
3
> 0, an instability arises in the system leading to a
transverse collapse. Thus, as pointed out in the introduc-
tion, new physics arises in connection with the vanishing
of the transverse pressure, P

= 0. This peculiar behavior

will be discussed in the forthcoming sections.
By writing /= (B H)/4, one may formally write
=
1
8
B
2
+
1
4
_
HdB P
0
. We recall that as
FG, the last expression is consistent with what would be
obtained in the classical non-relativistic case [19], where
F = F
0
+
_
HdB/4 is the Helmholtz free energy and
G = F +
_
/dB + P
0
= G
0
+ B
2
/8 is the Gibbs free
energy. Because of our denition of /, the last term is
given in terms of B and not in terms of H. As we have
T
33
= =
1
8
B
2

1
4
_
HdB +P
0
, (8)
and also
T
11
= T
22
= B/=
1
8
B
2
+
1
4
_
BdH+P
0
, (9)
it is straightforward to see that the spatial components
of the energy-momentum tensor T
ij
(i, j = 1, 2, 3) can be
rewritten as
T
ij
= P
0

ij
T
M
ij
(B, H) +o(B)
ij
, (10)
where o(B)
ij
=
1
4
_
B
i
B
j

1
2
(B
2
)
ij

is the Maxwell
stress tensor for the microscopic eld B, and T
M
ij
(B) =
1
4
_
H
i
B
j

__
BdH
_

ij

is the Minkowski tensor for rel-

ativistic non-linear media (see below). This last term re-
duces to the usual expression in the non-relativistic limit if
H depends linearly on B [9]. The denition (10) expresses
the total pressure as a sum of an isotropic pure mechanical
pressure (independent of the electromagnetic eld) plus a
pressure coming from the Minkowski tensor due to the in-
teraction of the external eld H with the microscopic eld
B, plus the Maxwell stress tensor of the microscopic eld
B. If H = 0, B = 4/ and T
ij
= P
0

ij
+o(B)
ij
. In this
case the magnetic eld is kept self-consistently.
At this point we want to refer to the recent paper by
Khalilov [4] where a problem similar to the present one
is studied. However, the expression for the stress tensor
in a medium is taken as given only by the linear approx-
imation of the Minkowski tensor term. This approach is
not justied in the relativistic case. Further, the Maxwell
tensor of the microscopic eld B and the isotropic P
0

ij
terms are omitted. This leads the author to wrongly con-
clude that the collapse occurs like in the classical case (see
below), that is, along the

B eld, in contradiction to our
present results and those of [2, 1]. A consistent approach,
1
This is an eect opposite to the oblateness of the Sun,
Earth and other planets due to the eective decrease of the
transverse gravitational force by the centrifugal force
A. Perez Martnez et al.: Magnetic collapse of a neutron gas 5
as we have followed here, and discussed in the accompany-
ing paper [20], leads to opposite results compared to those
of [4] in what concerns the spatial direction of collapse, al-
though the fundamental issue regarding the collapse of the
neutron star remains taking place in both theories.
Note, in addition, that [6, 7], where the problem was
also studied, did not take into account the dynamical ef-
fects of the strong (and superstrong) magnetic elds sup-
posed to exist in the core of canonical NSs; see for instance
[1, 2224]. Thence, it is the contention of this paper to ad-
dress this open issue. The novel results obtained here point
to the occurrence of new physics and processes in the (rel-
ativistic) astrophysics of compact objects that were not
manifest in previous papers.
4 Classical versus quantum collapses
In [1] we found that a relativistic degenerate electron gas
placed in a strong external magnetic eld B is conned
to a nite set of Landau quantum states. As the eld is
increased the maximum Landau quantum number is de-
creased favoring the arising of either a paramagnetic or a
ferromagnetic response through a positive magnetization
/, up to the case in which only the ground state is occu-
pied. The gas then becomes topologically one-dimensional,
and in consequence the pressure transverse to the eld
vanishes for elds B = // [1]. Thus, the electron gas
becomes unstable due to the decrease of the transverse
pressure for elds strong enough, and the outcome is a
collapse.
For neutrons the magnetization is always positive (see
the arguments below) and non-linear, which leads to a
sort of ferromagnetic behavior. For elds strong enough
the pressure transverse to the eld, P

= B/, is
considerably decreased and may vanish. If we assume that
extremely magnetized NSs, as the Duncan and Thomp-
son magnetars [3], have elds H 10
15
G, and that inside
the star B increases by following a dipole law B(r

) =
B
surf
/r
3
, we expect near its surface magnetic elds rang-
ing from 10
16
10
17
G [3, 27] up to values of order 10
20
G in
its core [23], where the eld is maintained self-consistently,
i.e., H = 0. For elds of this order of magnitude super-
dense matter composed of neutral particles having a mag-
netic moment may undergo a transverse collapse since P

vanishes. As discussed below, the emerging physics seems

to ban the possibility of magnetar formation. The outcome
of such a collapse might be a compact star endowed with
a canonical magnetic eld, as discussed below.
In the classical case in which the response of the
medium is due to the Lenz law, the magnetization is oppo-
site to the external eld H and it may happen that /< 0.
This also occurs in the diamagnetic case. Then H > B and
P

> P
3
. Note that the opposite occurs in some perme-
able materials where /> 0 and H = B 4/ is small
in comparison to either / and B; this is due to ferro-
magnetic eects which have a quantum origin, as in the
neutron gas.
In the case of a classical magnetized gas, as P

> P
3
,
this leads to the Earth-like oblatening eect described
above. But opposite to this, for the critical quantum con-
guration of the NS gas the coupling of the spin mag-
netic dipole with the magnetic eld

B plays the main
role, and /> 0 (see (21) below and the subsequent dis-
cussion where this is shown explicitly), leading to ferro-
magnetic eects. The situation then is reversed and P

is
smaller than P
3
in the amount B/ and it vanishes for
1
8
B
2
=
1
4
_
BdH + P
0
leading, conversely, to a prolate
conguration.
In classical electrodynamics [10] it is suggested that
the total pressure is given by the sum of the Maxwell
stress tensor o

plus an isotropic pressure (P

0
) term. In
the case of a constant magnetic eld parallel to the x
3
axis, the total pressure tensor reads T
ij
= P
0

ij
+ o
ij
or
P
3
= P
0
B
2
/8 and P

= P
0
+B
2
/8.
As pointed out before, the body deforms under the
action of these anisotropic pressures. If the longitudinal
pressure decreases, the body attens along the magnetic
eld [21]. Thus, in this pure classical case, for the extreme
limit of attening, P
3
= 0 and P

= P
0
+B
2
/8, and the
body would collapse as a disk or a ring perpendicular to
the eld. Starting from general relativistic considerations
has been reported [22] the existence of a maximum mag-
netic eld for the case having stationary congurations of
NSs. (We interpret that result as indicating the occurrence
of a classical collapse for elds larger than the maximum.)
This eld induces a toroidal conguration, which is topo-
logically equivalent to a ring. In the quantum case, for
degenerate fermions, as / > 0, it is P

= B/
which is decreased by increasing B. As the NS is in equi-
librium under the balance of neutron and gravitational
pressures, the body stretches along the direction of the
magnetic eld. Thus, for any density there are values of
the eld B high enough so that these pressures cannot
counterbalance each other leading to a collapse perpen-
dicular to the eld for P

= 0. This collapse would leave

as a remnant nucleons plus a BoseEinstein-like conden-
sate, a hybrid or strange star
2
with canonical magnetic
eld [28], or a distorted (cigar-like) black hole.
Our previous considerations are approximation inde-
pendent. In order to discuss a specic model, we shall
start by computing the free particle spectrum for neutral
particles in a magnetic eld.
5 The neutron gas in a magnetic eld
For free neutrons in a magnetic eld B we have the Dirac
equation for neutral particles with an anomalous magnetic
moment [14]:
(

+m+ iq

) = 0, (11)
where

=
1
2
(

) is the spin tensor, and F

is
the electromagnetic eld tensor describing B. By solving
this equation we get the eigenvalues [14]
E
n
(p, B, ) =
_
p
2
3
+ (
_
p
2

+m
2
n
+qB)
2
, (12)
2
The reader is referred to the interesting publications of [50]
6 A. Perez Martnez et al.: Magnetic collapse of a neutron gas
where p
3
, p

are respectively the momentum components

along and perpendicular to the magnetic eld B, m
n
is
the neutron mass, q = 1.91M
n
(M
n
is the nuclear mag-
neton), = 1, 1 are the
3
eigenvalues corresponding
to the two orientations of the magnetic moment (paral-
lel and antiparallel) with regard to the eld B. Here we
make the assumption that the magnetic moment remains
constant for large magnetic elds. Actually, this is not so.
Radiative corrections change its value as a function of B.
This problem is, however, beyond the scope of the present
approach.
From (12) we see that although the Hamiltonian is
linear in B, the eigenvalues depend on B non-linearly.
This makes the relativistic thermodynamic and electro-
magnetic properties of the system of neutrons very dier-
ent from the non-relativistic case. For instance, all ther-
modynamic quantities, , / (and in consequence, H =
B4/), and N are also non-linear functions of B. The
expression (12) also shows manifestly the change of spher-
ical to axial symmetry with regard to momentum compo-
nents. This anisotropy in the dynamics is expected to be
reected in an anisotropy in the thermodynamic properties
of the system, as it is expressed by the dierence between
the transverse and longitudinal pressures discussed earlier
on an approximation-independent basis.
The partition function : = Tr() is obtained from the
density matrix describing the model =
e

d
3
x(H(x)

r
N
r
)
[31]. Here
j
(j = 1, 2, 3) are the
chemical potentials associated with the lepton, baryon
and electric charge conservation, so that
n
=
2
,
p
=

2
+
3
,
e
=
1
+
3
and

=
1
. The thermodynam-
ical potential can be written as =
1
ln:. Usually
the eigenvalues of 1 contain the contribution from neu-
trons, protons, electrons and some meson species, and the
densities are N
r
= /
r
, where r = n, p, e, We
call =

r

r
, /=

r
/
r
the total thermodynamical
potential and magnetization, respectively.
A standard procedure is to work in the mean eld ap-
proximation in which the meson elds , , are taken
as constant, as done in [23, 24], through which the mass
spectrum of the baryons is corrected and strong repul-
sive interactions between them are found. However, for
simplicity we will keep the spectra in the tree approxi-
mation to obtain the one-loop approximation for , and
neglect the statistical contribution from meson terms in
as compared with those of fermions (for them m
i

10
3
10
4
) except for elds B
<

B
c
= m
2

/e 10
20
G,
since the contribution of the vector meson condensate
to / becomes relevant, and in analogy with W

s [1],
leads to a self-consistent spontaneous magnetization B =
4/ = 2eN

_
m
2

eB, where N

is the condensate
density. However, for such elds the magnetic pressure,
_
/dB /B = B
2
/8, overwhelms the kinetic pres-
sure term P
0
(of order 10
36
dynes/cm
2
) leading to P

< 0,
and the star is denitely unstable: it collapses. This mech-
anism is valid for other quasi-particle vector boson con-
densates, as diquarks, which may be formed in the medium
even for smaller values of the eld B [28, 13].
6 Thermodynamical potential
of a neutron gas
The Green function method [29] leads to a general expres-
sion for the relativistic thermodynamical potential. At the
one-loop level, where no radiative corrections are consid-
ered, it is a generalization of the usual non-relativistic for-
mula because of the fact that antiparticles must also be in-
cluded. Particles and antiparticles contribute with chemi-
cal potentials of opposite sign, leading to sums or integrals
over the quantum numbers involved, of terms containing
the product of the logarithms of (e
(
n
E
n
)
+ 1). In this
case, when no external elds are present, a divergent term
accounting for the vacuum energy appears which is usu-
ally subtracted [29]. In the presence of an external eld,
however, a term accounting for the vacuum contribution,
must also be included: the EulerHeisenberg energy of a
vacuum in an external eld [11, 15, 31, 30]. One can obtain
an expression for the thermodynamic potential of the neu-
tron gas in the one-loop approximation:
n
=
sn
+
Vn
,
with

sn
=
1
4
2

=1,1
_

0
p

dp

dp
3
ln
_
f
+
(
n
, )f

(
n
, )

, (13)
where f

(
n
, ) = (1 + e
(E
n

n
)
) accounts, respec-
tively, for the contribution of particles and antiparticles.
The expression for the vacuum term thus reads

Vn
=
1
4
2

=1,1
_

0
p

dp

dp
3
E
n
, (14)
which is divergent. In the appendix we will show how to
regularize this expression, and how to obtain the analog to
the EulerHeisenberg energy of vacuum due to the neutron
contribution.
After integrating by parts in (13) its evaluation be-
comes easier. The Fermi distributions, which arise by dif-
ferentiating f

with respect to p
3
, are n

= 1/(1 +
e
(E
n

n
)
). In the degenerate case this expression reduces
in n

= ( E
n
) and n
+
= 0, since in that case only
particles contribute to . The resulting expression splits
itself in two terms where the integrals are bounded by the
Fermi surfaces E
n
( = 1) = 0. These surfaces have
axial symmetry, and thus, the Fermi momentum is not a
denite number, given only in terms of
n
and m
n
, but
on the opposite, it has innite values. Thence, we have

sn
=
1
4
2

=1,1
_

0
p

dp

p
3
dp
3
E
n
( E
n
()) . (15)
The function bound these integrals in the intervals
p
3F
p
3
p
3F
, where p
3F
=
_

2
(
_
p
2

+m
2
n
+y)
and 0 p

_
( y)
2
m
2
n
. After some transforma-
tions, it yields
A. Perez Martnez et al.: Magnetic collapse of a neutron gas 7

sn
=
0

=1,1
_
xf
3
12
+
(1 +y)(5y 3)xf
24
+
(1 +y)
3
(3 y)
24
L
yx
3
6
s
_
, (16)
where x =
n
/m
n
, (m
n
(x1) is the usual Fermi energy),
and y = qB/m. We dene the functions f f(x, y) =
_
x
2
(1 +y)
2
, s s(x, y) = (/2 sin
1
(1 +y)/x),
L L(x, y) = ln(x + f(x, y))/(1 + y). The functions
f = p
F
/m
n
, where p
F
=
_

2
n
(m
n
+qB)
2
and we
call them the average Fermi momenta for = 1. We see
that m
n
qB behaves formally as a two-valued magnetic
mass.
In the zero eld limit one gets P
0
=
n
(y = 0),
where

sn
(y = 0) =
0
_
xf
3
0
12

xf
0
8
+
1
8
L
0
_
, (17)
where f
0
=

x
2
1 is the relative Fermi momentum
p
F
/m, and L
0
= ln(x + f
0
). The neutron vacuum term
(see the appendix) has an EulerHeisenberg-like form:

Vn
=
1
4
2
_

0
dyy
3
e
(m
2
n
+q
2
B
2
)y
[cosh(qBmy) 1 (qBmy)
2
/2!]
+
qB
2
2
_

0
dyy
2
_

0
dwe
[(w+m
n
)
2
+q
2
B
2
]y
[sinh(2qB(w +m
n
)y) (2qB(w +m
n
)y)
+ (2qB(w +m
n
)y)
3
/3!]. (18)
It can be shown (see the appendix) that the more sig-
nicant term in (18) is the rst one, which for elds of
order 10
17
G leads to
Vn
10
30
erg cm
3
and is negligi-
bly small as compared with
sn
up to B 10
18
G. Thus,
we neglect it in a rst approximation in what follows. We
must point out, however, that since neutrons have a quark
structure, a more fundamental quantity would be the vac-
uum quark contribution, whose order of magnitude is ex-
pected to be near
Vn
. Apart from this note, it should be
emphasized that the role of the vacuum cannot be ignored
for elds greater than 10
18
G.
From N
n
= /
n
one gets
N
n
= N
0

=1,1
_
f
3
3
+
y(1 +y)f)
2

yx
2
2
s
_
. (19)
In the limit B = 0 (19) reproduces the usual den-
sity of a relativistic Fermi gas at zero temperature, N
n
=
N
0
f
3
0
/3.
Having an equation relating the chemical potentials,
and demanding conservation of both baryonic number
N
n
+N
p
= N
B
and electric charge N
p
+N
e
= 0, in prin-
ciple one may solve exactly the problem in terms of the
external eld as a parameter. Nonetheless, we shall focus
our discussion on the properties of the equation of state.
Note in passing that our expressions for the spectra and
densities of neutrons and protons are similar to those of
[32] of a neutron gas in a magnetic eld, but we get dif-
ferent equations of state.
Finally, for the magnetization, given as /
n
=
n
/
B, we have
/
n
= /
0

=1,1

_
(1 2y)xf
6
(20)

(1 +y)
2
(1 y/2)
3
L +
x
3
6
s
_
,
where N
0
= m
3
n
/4
2
2.04 10
39
,
0
= N
0
m
n

3.0 10
36
, and /
0
= N
0
q 2.92 10
16
and one can
write /
n
= /
+
n
( = 1) /
+
n
( = +1), and obviously,
/
n
0.
We conrmed by explicit calculation that /is a non-
linear function of B and, in this sense, the magnetic re-
sponse is ferromagnetic. A fully ferromagnetic response
demands the inclusion also of the spinspin coupling con-
tribution. We briey discuss this point below.
To see why the magnetization is always positive for
the neutron gas note that the magnetic susceptibility =
/
n
/B can easily be obtained as
=
q/
0
2m
n

=1
_
xf + (1 +y)
2
L

, (21)
which for x > 1, y 1, and f real and positive, is >
0. This means that /
n
is an increasing function of B
(or y) under such conditions. As /
n
(y = 0) = 0 and
/
n
(y = 1) = 2/
0
(1 + ) > 0, this means that /
n
> 0
in the region between these two points, which is the one
of interest for us (the region I discussed below and shown
in Fig. 2).
The fact that we are summing over the magnetic mo-
ments oriented parallel ( = 1) and antiparallel ( =
+1) to B is similar to the well-known Pauli paramag-
netism in non-relativistic quantum statistics. We may con-
sider each term = 1 as representing a phase of the
system. One can draw in the x, y plane, for both x > 0,
y > 0, two regions limited by the lines x y = 1 and
x + y = 1 (see Fig. 1). The region I is bounded above
by the line x = y + 1, on which the contribution from
f( = +1), and all other terms containing = +1 vanish:
all magnetic moments are aligned parallel to B. Below,
such a region is limited by y = 0. For points in the region
x > y + 1 and y > 0 both terms containing = 1, +1
are non-zero and real. Thus, the quantities , N
n
, and /
are real and although most neutrons have their magnetic
moments along B, there are some having their magnetic
moments antiparallel to B. In this region, near the line
x = y + 1, is located the curve P

= 0, which we will
discuss below.
In the region II limited by x 1 y, y + 1 > x
y 1 only the term = 1 is real and the term = +1
becomes purely imaginary, and all neutrons would have
their magnetic moments parallel to B. The term = 1
contributes to , N
n
, and /. For x < y 1, there are
no physical solutions. For y = 0, the magnetic eld, and
8 A. Perez Martnez et al.: Magnetic collapse of a neutron gas
Fig. 1. This plot shows the regions in
the x, y plane where the neutron mag-
netic moments are oriented parallel or
antiparallel to the magnetic eld

B.
Special attention should be given to re-
gions I and II, where the solutions dis-
cussed in the text are valid
Fig. 2. The instability condition, P

= 0, in terms of the vari-

ables x, y. A nascent strongly magnetized neutron star having
a conguration such that its dynamical stage would be repre-
sented by a point above the central curve in this plot would be
unstable to transversal collapse, since P

0 there
in consequence, the magnetization / vanish. For y < 0
(magnetic eld along the negative y axis) we have the
situation reverse to the one described above (See Fig. 1).
With regard to the background of electrons and pro-
tons, we recall that the case of the electron gas was dis-
cussed in [1]. We expect that for magnetic elds of order
B
ce
= m
2
e
/e 10
13
G and densities around 10
30
cm
3
all
electrons are in the Landau ground state, and the system
show the instability which arises from the vanishing of
the transverse pressure. For densities and magnetic elds
above these critical values, the stability of the electron gas
is doubtful. Recently Ghosh, Mandal and Chakrabarty [5]
have pointed out strong arguments against the equilibrium
of an electron gas under decay in strongly magnetized
neutron stars. The situation may be modied if radiative
corrections are taken into account: if B
ce
= m
2
e
/e 10
13
G
is the usual QED critical eld, for larger elds of order
B

ce
4/B
ce
10
16
G the contribution of the electron
anomalous magnetic moment becomes signicant and the
problem cannot be satisfactory treated at the tree level.
However, starting from the results of [1, 5], one concludes
that the electron gas is hardly in equilibrium for elds be-
yond B
ce
. One possibility is the bosonization of the elec-
tron system, as has been recently suggested [33]. This may
be accomplished through the increase of the spinspin in-
teraction, leading to the formation of parallel spin electron
pairs, equivalent to charged vector bosons.
If we include both the normal and the anomalous mag-
netic moments for protons, one can give a formula for their
spectrum in the external eld B [14]:
E
p
=
_
p
2
3
+ (
_
2eBn +m
2
p
+q
p
B)
2
, (22)
We have q
p
= 2.79M
n
. For neutrons, the critical eld at
which the coupling energy of its magnetic moment equals
the rest energy is B
cn
= 1.57 10
20
G. For protons it
is B
cp
= 2.29 10
20
G. By dening x
p
=
p
/m
p
, y
p
=
q
p
/m
p
, b = 2e/m
2
p
, then y
p
= 2.79e/2m
2
p
. We mention
also g g(x
p
, B, n) =
_
x
2
p
h(B, n)
2
and h h(B, n) =
(

bBn + 1+y
p
B). Thus for the proton thermodynamical
potential we get
A. Perez Martnez et al.: Magnetic collapse of a neutron gas 9

p
=
eBm
2
p
4
2

_
x
p
g h
2
ln(x
p
+g)/h

, (23)
and for its density
N
p
=
eBm
p
2
2

g(x
p
, B, n) , (24)
while the magnetization is given by
/
p
=
em
2
p
4
2

_
x
p
g
_
h
2
+ (y
p
(25)
+ (bn/2

bBn + 1))
_
ln(x
p
+g)/h
_
,
where the coecients of these formulae are N
0p
= em
p
B/
2
2
4.06 10
19
B,
0p
= N
0p
m
p
B 6.1 10
16
B, and
/
0p
= N
0p
m
p
=
0
/B. The maximum occupied Lan-
dau quantum number n may be given as n
max
= (x
p

y
p
B)
2
1/bB. For B B
cp
, so that y
p
B 1, and
x
p
1, one can take approximately n
max
(x
2
p
1)/bB,
and for elds large enough n
max
= 0. We expect that from
the equation
n
=
p
+
e
, x
p
x
n
, the previous expres-
sion for the proton density N
p
decreases with increasing
B, favoring the inverse beta decay. For elds B m
p
/q
p
and x
p
1, n
max
1, and thus large Landau numbers
are again occupied. However, for x
n
, x
p
>

1, being quanti-
ties of the same order of magnitude, from the comparison
of
0p
, N
0p
, /
0p
with
0
, N
0
, /
0
we conclude that for
elds below 10
19
G, the dominant longitudinal pressure,
density and magnetization comes from the neutron gas.
7 Condition for zero transverse pressure
and collapse
In the electron gas case [1], the vanishing of the trans-
verse momentum can be guessed from the spectrum when
the whole system is conned to the Landau ground state.
The spectrum corresponds to a purely one-dimensional
system moving along the external eld, and the transverse
Fermi momentum is zero. In a similar way, in the neutron
gas case we observe that the threshold of zero transverse
pressure P

= 0 can be gured out from the spectrum

described by (12), since the contribution from the = 1
term is dominant (observe that the term with = +1 con-
tributes with a negative term to /). We shall consider on
the Fermi surface for = 1 the quantity
p
2
Fe
=
2
n
p
2
F3
m
2
n
=
_
_
p
2
F
+m
2
n
qB
_
2
m
2
n
, (26)
which we call the eective squared Fermi transverse mo-
mentum. If B 10
20
G, then q
2
B
2
2eBm
n
, and the
vanishing of p
Fe
is guaranteed if
p
F

_
2qBm
n
. (27)
The resulting Fermi surface would be equivalent to that
for one-dimensional motion, parallel to B, i.e., for parti-
cles having energy E
n

_
p
2
3
+m
2
n
, and in consequence
the transverse momentum (and pressure) vanishes. Notice
from (27) that for p
F
/m
n
10
1.5
one has qB/m
n
=
y 10
3
, which means elds of order 10
17
G. A more
accurate result is obtained, however, from the condition
T

= 0.
In Fig. 2 we have drawn the equation P
n
=
B/ = 0 in terms of the variables x, y. We observe that
there is a continuous range of values of x, starting from
x = 1.005 to x = 1.125 and from y = 0.001 to y = 0.1,
for which the collapse takes place. The latter values of y
means elds in the interval B 10
17
to 10
19
G. To these
ranges of x, y corresponds a continuum range of densi-
ties, from 10
2
N
0
(10
12
g/cm
3
) onwards. The transverse
compression of the whole mass of the star due to ux con-
servation leads to an increase of B and the mechanism of
collapse is enhanced.
7.1 The spinspin coupling contribution
Although we previously used the term ferromagnetic to
denote the non-linear response of the medium to the exter-
nal eld, what we have considered actually in our previous
calculations is the occurrence of relativistic Pauli para-
magnetism. We have not fully considered the spinspin
coupling, which would lead more denitely to Heisenberg
ferromagnetism, and it is physically reasonable to expect
its appearance in nuclear matter for densities high enough.
This problem has been studied in [34] via the interaction
through axial vector and tensor exchange channels. These
authors show that if the interactions are strong enough
and dier in sign, the system loses the spherical symme-
try due to a mechanism independent of the one discussed
in the present paper. In the case of a quark liquid [35]
the problem has been studied under one gluon exchange
interaction, and conditions for ferromagnetism may arise.
Although the problem requires further research, one
expects that if Heisenberg ferromagnetism is to appear, it
would increase the magnetization to /
s
= /, where
is the internal eld parameter. If >> 1 our previous es-
timates for the vanishing of the transverse pressure might
be largely exceeded, with the arising of a new spontaneous
magnetization /
s
/ = B/4. This would mean
that the magnetic eld B could be kept self-consistently
and our previous calculations would be a lower bound of
the fully ferromagnetic case. The vanishing of P

is ex-
pected to occur surely at values of B smaller than those
depicted in Figs. 2 and 3. As a rough estimation, if we as-
sume the exchange interaction J among neutrons to be of
the order of a few hundreds of MeV, and the number of
nearest neighbors to be z 10, by dividing their prod-
uct by the dipole interaction energy, say, for the core of
the star where N = N
0
, we get 10
4
. This means
B 4/
s
10
20
G. For such extremely large elds
the magnetic coupling of quarks with B would become of
the order of their binding energy through the color eld
producing a deconnement phase transition leading to a
10 A. Perez Martnez et al.: Magnetic collapse of a neutron gas
quark (q)-star, a pressure-induced transition to uds-quark
matter via ud-quark condensates, as discussed in [43, 44].
But elds of that order lead surely to the collapse of the
star and even to the instability of the vacuum.
8 Magnetar formation and stability
Next we briey review the basic ideas supporting the the-
ory of magnetars and then show why these hypothetical
objects cannot survive after reaching the claimed super-
strong magnetic elds. We then present prospectives for
a hybrid or strange star to appear as a remnant of the
quantum magnetic collapse of a NS. According to Dun-
can and Thompson [3], NSs with very high dipole sur-
face magnetic eld strength, B [10
14
10
15
] G, may form
when the (classical) conditions for a helical dynamo ac-
tion are eciently met during the seconds following the
core collapse and bounce in a supernova (SN) explosion
[3]. A newly born NS may undergo vigorous convection
during the rst 30 s following its formation [39]. If the NS
spins (dierentially) suciently fast (P 1 ms) the condi-
tions are created for the dynamo action to be built,
which may survive depletion due to turbulent diusion.
Collapse theory, on the other hand, shows that some pre-
supernova stellar cores could acquire enough spin so as to
rotate near their Keplerian equatorial velocity, the break-
up spin frequency:
K

_
_
2
3

3
G
N
M/R
3
_
1/2
, which im-
plies a period P
K
0.6 s, after core bounce. Under these
conditions, elds as large as [42, 3]
B 10
17
_
P
1 ms
_
G (28)
may be generated as long as the dierential rotation is
dragged out by the growing magnetic stresses. For this
process to eciently operate the ratio between the spin
rate (P) and the convection overturn time scale (
con
),
the Rossby number (R
0
), should be R
0
1. Duncan and
Thompson warned that R
0
1 should induce less eec-
tive mean-dynamos [3]. In this case, an ordinary dipole
B
D
[10
12
10
13
] G may be built by incoherent superpo-
sition of small dipoles of characteristic size [
1
3
1] km,
and a saturation strength B
sat
= (4)
1/2
/
con
10
16
G
may be reached at the surface during this early evolution
of the nascent neutron star. At such elds, the huge rota-
tional energy of a NS spinning at
NS
1 kHz, is leaked
out via magnetic braking and an enormous energy is in-
jected into the SN remnant. This process may explain the
power emitted by a plerion.
As shown above, at the end of the SN core collapse we
are left with a rapidly rotating NS endowed with an ex-
tremely strong magnetic eld (ESMF) strength and a large
matter density [10
14
10
12
] g cm
3
. As illustrated in
Fig. 1, those are the conditions for the quantum instabil-
ity to start to dominate the dynamics of the young pulsar.
At this stage, the magnetic pressure inwards may pass the
star energy density at its equator and the collapse becomes
unavoidable. As the collapse proceeds, higher and higher
densities are reached till the point the supranuclear den-
sity may reverse the direction of the implosion. A hybrid or
strange star (SS) may have formed. We explore next this
plausible outcome, among the other possibilities quoted
above. From that moment, the sound wave generated at
the core bounce builds itself into a shock wave traveling
through the star at V
SW
c/

3 kms
1
. Although the
ESMF strength could be quite large as long as the collapse
advances, the huge kinetic energy, E 10
5152
erg, the
mean energy obtained in calculations of energy release in
neutron star phase transitions to strange (twin) stars [36
38] and some prompt shock supernovae (E
ps
10
51
erg)
[39], carried away by the shock wave drives a kind of su-
pernova explosion inasmuch as in the quark nova model
and similar scenarios [38]. Such a huge ram pressure may
counterbalance the magnetic pressure, and even surpass
it, i.e.,

eject
V
2
SW

B
2
8
0
_
R
r
A
_
6
, (29)
at a location from the star center equivalent to the Alfven
radius of the magnetar
r
A
=
_
2
2
G
2
0
_
1/7
_
B
4
R
12
M

M
2
_
1/7
80 km. (30)
This radius is quite large, about 7 times the NS radius
(see Table 1). Here

M denes the accretion rate of the free-
falling overlaying material making up the NS crust, which
is left out when the transition occurs (see denitions and
further details in [38], and references therein). Therefore,
it is quite legitimate to expect that most of the magnetic
energy stored inside the magnetosphere to be drained out
of the Alfven radius. Notice in addition that the strange
star radius scales as R
SS
= R
NS
_

NS

SS
_
1/3
, where the den-
sities ratio reads

NS

SS
0.10.2. Other relations between
both stars can be obtained by using conservation laws or
appropriate scalings.
Then the ESMF lines are pushed out and nally bro-
ken, in a process inverse to the standard accretion one,
from r
A
onwards into the SN remnant surroundings, as a
violent explosion that dissipates a large part of the mag-
netic ux ( B
2
r
2
A
) and the energy trapped inside
the magnetar magnetosphere [28]. Energy from the mag-
netic eld can be dissipated via vacuum polarization and
electronpositron pair creation, as well as acceleration of
charged matter owing away ( synchroton and curvature
losses [40, 41] with the explosion and material trapped in
the star magnetosphere and near the Alfven radius. This
is analogous to the mechanism operating during a solar
are or a coronal mass-ejection, where the very high B
in a given Sun-spot is drastically diminished after aring
for a short period of time (see also [42]). In the Sun-spots
outbursts and coronal mass-ejection launches into space,
part of the Solar wind of charged and neutral particles
passes by the Earth [46, 47]. In the case of an implod-
ing NS, the phase of open magnetic eld lines over which
the strange star is acting as a propeller lasts for about
T
prop
E
spin
/L
prop
[10
2
10
3
] s, with L
prop
2

Mc
2
A. Perez Martnez et al.: Magnetic collapse of a neutron gas 11
Fig. 3. The curve P

= 0 in terms of
the neutron star mean density N and
the surface magnetic eld B
Table 1. Parameters used in modeling the phase transition of a neutron to a strange star,
as discussed in the text
Mass M Radius R Period P Magn. eld B Accretion rate
[M

] [km] [ms] [G] [ M

s
1
]
Neutron star 1.5 12.5 2.0 2 10
15
10
15
Strange star 1.5 9.5-10.0 0.5 2 10
13
10
5
the propeller luminosity, and E
spin
I
2
SS
the star rota-
tional energy. Thence, the large amount of matter ejected
from the strange star at such large velocities and the pairs
created, in the vacuum breakdown, drains out the dipole
eld of the remnant below the quantum electrodynamic
limit of B
ce
4.4 10
13
G [28].
To give an insight into this piece of the physics of the
problem, notice that once the propeller phase is over and
no more luminosity is coming from that mechanism, the
magnetic eld lines can recombine again if the energy re-
leased in this new stable phase is essentially the strange
star rotational dipole luminosity (as the one from a mil-
lisecond pulsar), which then becomes the stars dominant
mechanism of energy emission, that is,
4
B
2
SS
R
6
SS

4
SS
9c
3
= I
SS

SS
. (31)
For the parameters shown in Table 1 and the observed
luminosity from fast rotating pulsars, this relation implies
a new equilibrium magnetic eld B 10
1213
G, which is
below the quantum electrodynamics threshold.
Since all the dierential rotation has been dragged up
to build up the former ESMF, then nothing else remains
to make the magnetic eld grow to its pre-collapse value.
Thence no such ultrahigh B should reappear. We may
be left with a sub-millisecond strange star [43] or a hy-
brid star [44] with canonical eld strength, but not any
magnetar. We note in passing that the above theoreti-
cal result is attained on the standardized assumption that
the structure of the magnetic eld of the pulsar is dipolar.
This premise is underlying the claim by Kouveliotou et
al. [45] that a magnetar had been discovered in the soft
gamma-ray repeater source SGR 1806-20. Notwithstand-
ing, for other NS (multipolar or uniform) eld congura-
tions we do expect the overall behavior here discussed to
persist, since once the spinspin coupling is taken into ac-
count the unavoidable consequence is the appearance of
a ferromagnetic (axial) conguration or structure which
is dipolar in nature, and therefore the theory propounded
here still holds, because the magnetic eld could then be
amplied by a factor 10
4
), putting the nascent pulsar
above the threshold for stability, and the collapse ensues.
Rephrasing this, one can think of this theory as a eld-
conguration independent constraint on initial NS mag-
netic eld strengths.
In looking for other contexts far from those involving
compact remnant stars, we noticed that recent Tabletop
Astrophysics experiments performed by Wieman et al.
[48] have succeeded in rening sophisticated techniques to
switch atoms in a BoseEinstein condensate (BEC) [48]
12 A. Perez Martnez et al.: Magnetic collapse of a neutron gas
from states of implosion to states of re-expansion (or ex-
plosion). In certain cases they observed that some col-
lapses appear to be rather similar to microscopic super-
novae. The initial implosion is followed by an explosion in
which atoms are ejected as a hollow ball or in narrow jets,
like in the collapse and rebound of a exploding star that
forms characteristic expanding balls or streams of out-
owing gas. The Wieman team named the phenomenon
bosenovae because of the similarities with typical su-
pernovae. In fact, in both phenomena some material is
left after the implosion as a compact object. Because of
the similarity of the physics of BoseEinstein condensates
with the one we presented above, we are condent that
to this our theory can also be adapted to explain these
impressive results by Wieman et al. [48] and Ketterle and
Anglin [26]. The main feature of these oscillating Bose
Einstein condensates, said to be a scaled-down version of
either a neutron star or a white dwarf [48], is that for
some critical elds what appears is an attractive force be-
tween the atoms, and the condensate implodes and re-
bounds driven by some sort of internal negative pressure.
We claim in this paper (a detailed description of BECs
phenomenology is to be given in a work in preparation
[49]) that such a negative pressure could be explained in
the context of the theory introduced in this paper, since
a precise relationship between gas density and magnetic
eld strength in the BEC is settled by switching the atoms
between attractive and non-attractive states. Thence the
negative pressure acts as the equivalent of an attractive
force among the atoms directed towards the magnetic eld
axis, leading to the BEC implosion.
9 Conclusions
We conclude by claiming that if a degenerate neutron
gas is under the action of a superstrong magnetic eld
B
ce
<

B
<

B
cn
, for values of the density typical of NS
matter its transverse pressure vanishes, the outcome be-
ing a transverse collapse. This phenomenon establishes a
strong bound on the magnetic eld strength expected to
be found in any stable neutron star pulsar, regardless of
its initial eld conguration, and suggests a possible end-
point in the early evolution of highly magnetized neutron
stars. They could likely be a mixed phase of nucleons and
a

,
0
, K

, K
0
,

K
0
, ,

, meson condensate, a hybrid

or strange star, or a distorted black hole but not any mag-
netar at all. We point out, nonetheless, that if by any
mechanism a strange star could be formed directly in a su-
pernova explosion (which is uncertain, but not ruled out)
and if a vigorous dynamo action operates in the strange
quark matter bulk, pulsars with elds higher than B
ce
could still be formed. In other words, if any magnetars
exist, they cannot be neutron stars.
A Appendix
We use the integral representation
a
1/2
=
1/2
_

0
dx
x
2
(e
ax
2
1)
=
1/2
_

0
dyy
3/2
(e
ay
1). (32)
We regularize the divergent term dependent on a in
(32), by introducing a small quantity as the lower limit
in the integral, and neglecting the term independent of a
a
1/2
() =
1/2
_

dyy
3/2
e
ay
. (33)
By taking a(, ) = p
2
3
+ (
_
p
2

+m
2
n
+ qB)
2
and sub-
stituting in (15) and performing the Gaussian integral on
p
3
, one obtains

Vn
() =
1
4
2

=1
_

0
p

dp

dyy
5/2
e

p
2

+m
2
n
+qB

2
y
. (34)
By substituting z =
_
p
2

+m
2
n
+qB one is left with the
expression

Vn
() =
1
4
2
_

dyy
3
e
(m
2
n
+q
2
B
2
)y
cosh(qBmy)
+

=1
qB
4
2
_

dyy
2
_

m
n
+qB
e
z
2
ydz. (35)
By introducing the new variable w = z m
n
qB, the
second integral in (35) becomes
qB
4
2
_

dyy
2
_

0
e
[(w+m
n
)
2
+q
2
B
2
]y
sinh(2qB(w +m
n
)y). (36)
After subtracting from cosh(qBmy) and sinh[2qB(w+
m
n
)y] the rst two terms in their series expansion, one
can take 0 and obtain the nite expression (18). This
process is equivalent to the subtraction of divergent terms,
one of which is proportional to B
2
and absorbs the clas-
sical eld energy term B
2
/8.
It is not dicult to check that for elds B 10
20
G,
the rst term in (18) is the dominant one. Its rst contri-
bution after the series expansion of cosh(qBmy) is
q
3
B
3
m
3
n
/2
2
(m
2
n
+q
2
B
2
). For elds of order 10
17
G such
a term is of order 10
30
ergs/cm
3
, much smaller than
sn
.
But for elds near 10
20
G its contribution is comparable
to that of
sn
.
Acknowledgements. The authors thank J. Arponen, M.
Chaichian, J. Ellis, A. Green, K. Kajantie, C. Montonen, A.E.
Shabad, and A. Zepeda for useful comments and suggestions.
H.P.R thanks M. Virasoro, IAEA and UNESCO for hospitality
in ICTP. The nancial support of the Academy of Finland un-
der the project No. 163394 is gratefully acknowledged. HJMC
thanks Fundac ao de Amparo a Pesquisa do Estado de Rio de
Janeiro (Brasil) for a Grant-in-Aid. The authors thank the
anonimous referee for his many valuable criticisms and sugges-
tions, which led to this highly improved version of our former
manuscript.
A. Perez Martnez et al.: Magnetic collapse of a neutron gas 13
References
1. M. Chaichian, S. Masood, C. Montonen, A. Perez
Martnez, H. Perez Rojas, Phys. Rev. Lett. 84, 5261 (2000)
2. A. Perez Martnez, H. Perez Rojas, H. Mosquera Cuesta,
hep-ph/0011399
3. R.C. Duncan, C. Thompson, Ap. J. 392, L9 (1992)
4. V.R. Khalilov, Phys. Rev. D 65, 056001 (2002)
5. S. Ghosh, S. Mandal, S. Chakrabarty, astro-ph/0207492
(2002)
6. L. Herrera, N.O. Santos, Phys. Rep. 286, 53 (1997)
7. M.K. Mak, T. Harko, gr-qc/0110103
8. A. Sommerfeld, Electrodynamics (Academic Press, New
York 1952)
9. J.D. Jackson, Classical electrodynamics (J. Wiley and
Sons, New York 1966),
10. L.D. Landau, E.M. Lifshitz, The classical theory of elds
(Pergamon, New York 1970)
11. A.E. Shabad, H. Perez Rojas, Ann. of Phys. 121, 432
(1979)
12. P. Haensel, S. Bonazzola, Astron. Astrophys. 314, 1017
(1996)
13. J.E. Horvath, J.A. de Freitas Pacheco, Int. J. Mod. Phys.
D 8, 19 (1998)
14. V.G. Bagrov, D.M. Gitman, Exact solutions of relativistic
wave equations (Kluwer Academic Publ. 1990)
15. W. Heisenberg, H. Euler, Z. Phys. 98, 714 (1936)
16. L.D. Landau, E.M. Lifshitz, Statistical physics (Pergamon,
Oxford York 1958)
17. K. Huang, Statistical mechanics (J. Wiley and Sons, New
York 1963)
18. S. Weinberg, Gravitation and cosmology (J. Wiley and
Sons, New York 1972)
19. L.D. Landau, E.M. Lifshitz, Electrodynamics of continu-
ous media (Pergamon, New York 1963)
20. H.J. Mosquera Cuesta, H. Perez Rojas, A. Perez Martnez,
Anisotropic pressures in very dense matter, in preparation
21. K. Konno, T. Obata, Y. Kojima, Astron. Astrophys. 356,
234 (2000)
22. C.Y. Cardall, M. Prakash, J.M. Lattimer, Astrophys. J.
554, 322 (2001)
23. S. Chakrabarty, D. Bandyopadhyay, S. Pal, Phys. Rev.
Lett. 78, 2898 (1997)
24. A. Broderick, M. Prakash, J.M. Lattimer, Ap. J. 537, 351
(2000)
25. T. Chiba, Prog. Theor. Phys. 95, 321 (1996)
26. W. Ketterle, J. Anglin, Nature 216, 211 (2002)
27. V.N. Kondratyev, Phys. Rev. Lett. 88, 221101 (2002)
28. J.E. Horvath, Int. J. Mod. Phys. D 8, 669 (1999)
29. E.S. Fradkin, Quantum Field Theory and Hydrodynamics,
Proceedings of the P.N. Lebedev Institute No. 29, Consul-
tants Bureau (1967)
30. V. Weisskopf, Kong. Dans. Videns. Selskab, Math-fys.
Meddeltser, 14 (1936)
31. H. Perez Rojas, Acta Phys. Pol. B 17, 861 (1986)
32. I-S. Suh, G.J. Mathews, Ap. J. 530, 949 (2000)
33. S. Mandal, S. Chakrabarty, astro-ph/0209462 (2002)
34. T. Maruyama, T. Tatsumi, Nucl. Phys. A 693, 710 (2001)
35. T. Tatsumi, Phys. Lett. B 489, 310 (2000)
36. I.N. Mishustin, et al., Phys. Lett. B 552, 1 (2003)
37. G. Marranghello, C.Z. Vasconzellos, J.A. de Freitas
Pacheco, Phys. Rev. D 66, 064027 (2002)
38. H.J. Mosquera Cuesta, Ph.D. thesis (in Portugese), INPE,
Brazil, unpublished (1998). See also the paper in prepa-
ration: Gravitational waves, gamma-ray bursts and ul-
tra high energy cosmic ray acceleration during neutron
star phase transitions: An unied picture, H.J. Mosquera
Cuesta, J. A. de Freitas Pacheco, M. Dillig. The Quark
Nova model can be found in R. Ouyed, J. Dey, M. Dey,
Astron. Astrophys. 390, L39 (2002)
39. H.-Th. Janka, E. M uller, Astron. Astrophys. 306, 167
(1996)
40. B. Zhang, A.K. Harding, Astrophys. J. 535, L51 (2000)
41. B. Zhang, A.K. Harding, A.G. Muslimov, Astrophys. J.
531, L135 (2000)
42. W. Kluzniak, M. Ruderman, Ap. J. Lett. 505, L113 (1998);
M. Ruderman, L. Tao, W. Kluzniak, Ap. J. 542, 243
(2001)
43. J. Madsen, Phys. Rev. Lett. 81, 3311 (1998)
44. N.K. Glendenning, Compact stars: Nuclear physics, par-
ticle physics and general relativity (Springer, New York
1997)
45. C. Kouveliotou et al., Nature 393, 235 (1998); C. Kouve-
liotou et al., Astrophys. J. 510, L115 (1999)
46. E.N. Parker, Chin. J. Astron. Astrophys. 1, 99 (2001)
47. J. Chen, Spac. Sci. Rev. 95, 165 (2001)
48. S.L. Cornish et al., Phys. Rev. Lett. 85, 1795 (2000)
49. H.J. Mosquera Cuesta, H. Perez Rojas, A. Perez Martnez,
Exploding BoseEinstein condensate and critical magnetic
elds, in preparation
50. R. Xu, B. Zhang, G. Qiao, Astropart. Phys. 15, 101 (2001);
Astrophys. J. 552, L109 (1999)