Superuidity

E. V. Thuneberg
Department of Physical Sciences, P.O.Box 3000, FIN-90014 University of Oulu, Finland
(Dated: June 8, 2012)
PACS numbers: 67.40.-w, 67.57.-z, 74., 03.75.-b
I. INTRODUCTION
Fluids (gases and liquids) are distinguished from solids
by the property that they can ow. In almost all cases
there is viscosity associated with the ow. Due to vis-
cosity, the ow energy is gradually dissipated into heat.
Contrary to this common situation, there is a special
class of uids, which can ow without viscosity. These
are called superuids and the phenomenon is called super-
uidity. As a concrete example, consider a ring-shaped
container lled with superuid, see Fig. 1. Once the uid
is put into circular motion, it will continue to circulate
and no energy is dissipated. The ow can continue as
long as the conditions for superuidity are satised.
Superuids show many spectacular phenomena, which
are discussed in sections IV-VII. Before going into these
we discuss the systems where superuidity occurs (Sec.
II) and the microscopic basis of superuidity (Sec. III).
While Sec. III gives deeper insight, it is not absolutely
necessary for understanding the phenomena in Secs. IV-
VII.
II. OCCURRENCE
Superuidity occurs only in certain substances under
special conditions. As a rst case we discuss liquid he-
lium. Under standard pressure and temperature helium
is a gas. It liquies at temperatures around 4 kelvin.
Cooling further down, it enters the superuid phase at
temperatures around 2 kelvin, depending on pressure.
The phase diagram of natural helium at low temperatures
is shown in Fig. 2. Natural helium consists essentially of
isotope
4
He.
Helium has another stable isotope,
3
He. At tempera-
tures below a few kelvin, its behavior is radically dierent
from the isotope
4
He. It also becomes superuid, but at
temperatures that are a factor of one thousand smaller
than for
4
He. The phase diagram of
3
He at low temper-
atures is shown in Fig. 3.
3
He has three dierent super-
uid phases, A, A
1
, and B. The A
1
phase only appears
in magnetic eld, and therefore is not visible in Fig. 3.
Superuidity is closely related to superconductivity.
Superconductivity means that electric current can ow
without resistance. This phenomenon appears at low
temperatures in many metals like Al, Sn, and Nb. It
arises from resistanceless motion of the conduction elec-
trons in a metal. Therefore, superconductivity can be
understood as superuidity of the conduction electrons.
v
s
FIG. 1: Once generated, the circulation of a superuid (with
velocity v
s
) persists as long as the experiment can be contin-
ued.
4
3
2
1
0
0 1 2 3 4 5 6
superfluid
normal liquid
p
r
e
s
s
u
r
e

(
M
P
a
)
temperature (K)
(bcc)
gas
solid
FIG. 2: Phase diagram of
4
He at low temperatures.
4
He
remains liquid at zero temperature if the pressure is below
2.5 MPa (approximately 25 atmospheres). The liquid has a
phase transition to a superuid phase, also known as He-II,
at the temperature of 2.17 K (at vapor pressure).
Part of the discussion in this article applies also to su-
perconductivity, but there are dierences caused mainly
by two reasons. (a) Electrons have electric charge and
therefore their motion is essentially coupled with mag-
netic elds. (b) The crystal lattice of the ions constitutes
a preferred frame of reference, which does not exist for he-
lium liquids. Superconductivity is discussed extensively
in other articles of this Encyclopedia.
With laser cooling it is possible cool certain atomic
gases like
87
Rb,
7
Li,
23
Na, and
1
H to very low tempera-
tures. At low temperatures Bose-Einstein condensation
takes place in the gas. This state has many properties
that are similar to superuidity, although it is not a ther-
modynamically stable state, and therefore the ow can-
not last for ever. Most of the discussion in this article
applies also to condensed gases. An important dier-
2
0 1 2 3
0
1
2
3
4
normal fluid
temperature (mK)
p
r
e
s
s
u
r
e

(
M
P
a
)
solid
superfluid B phase
superfluid A phase
FIG. 3: The phase diagram of
3
He at low temperatures. Note
that the temperature is in units of millikelvin. Two superuid
phases of
3
He, A and B, are shown. (Figure based on data by
DS Greywall.)
ence is that instead of container walls for helium liquids,
one has to consider the conning potential of the gas,
which can be generated either magnetically (by eld gra-
dients) or optically (by laser beams). There is a separate
article about Bose-Einstein condensation (Encyclopedia
reference Bose-Einstein condensation).
Superuidity is expected to occur also in astrophysical
objects. The neutron liquid in a neutron star is believed
to be in a superuid state. This has been suggested as
an explanation for the observed sudden changes in the
rotation velocity of pulsars.
III. MICROSCOPIC ORIGIN
In short, superuidity can be explained as a quantum
mechanical eect that shows up on a macroscopic scale.
Quantum mechanics is crucial in understanding the mi-
croscopic world. It explains that electrons in atoms have
only discrete energies. There is no friction on the atomic
scale, and the electrons can circulate the nucleus without
losing energy.
We know that quantum mechanics rarely shows up
on macroscopic scale. Instead of quantum mechanics,
macroscopic objects obey the rules of classical physics.
The reason is that a macroscopic sample consists of large
number of particles and, instead of individual particles,
one can only observe their average behavior. Usually the
particles are in dierent quantum states, and an average
over them obeys classical laws of physics. Examples of
these laws are the Navier-Stokes equations for uids and
Ohms law for electrical conduction.
Superuidity is an exception to this general rule. In
superuids a macroscopic number of particles is in the
same quantum state. It follows that summing over par-
ticles does not lead to averaging, but produces a macro-
scopic wave function.
Further analysis depends essentially whether the par-
ticles are bosons or fermions.
4
He and the gas atoms
listed above are bosons, whereas
3
He atoms, electrons,
and neutrons are fermions.
Consider a particle with mass m and momentum p.
Its energy is E = p
2
/2m. Its state is represented by the
single-particle wave function
(r) =
1

V
exp(
i

p r), (1)
where V is the volume of the system and h = 2 the
Planck constant. The wave function of a many-body sys-
tem (r
1
, r
2
, . . .) is more general and depends on the
coordinates r
i
of all particles.
We now assume the particles are bosons. This means
that the total wave function must be symmetric when
exchanging any pair of particles. In the case of two par-
ticles this means (r
1
, r
2
) = (r
2
, r
1
). Further we as-
sume that there is no interaction between the bosons. It
can be shown that the occupation of the lowest energy
state becomes macroscopic, if the temperature T is less
than
T
BE
=
h
2
2mk
B

N
2.612V

2/3
. (2)
Here N is the number of particles and k
B
the Boltzmann
constant. This is known as Bose-Einstein condensation.
The wave function (1) of the lowest energy state (p = 0)
becomes macroscopic. At zero temperature, all particles
are in this state.
While the ideal gas model explains Bose-Einstein con-
densation, it is quite insucient in other respects. The
interactions between particles are essential for the system
to show superuidity. In interacting system the macro-
scopically occupied state (r) need not be the lowest
energy state, and thus the macroscopic wave function
can be nontrivial. In bose gases (
87
Rb etc.) the inter-
actions are weak, and a quantitative description can be
achieved by the relatively simple Gross-Pitaevskii equa-
tion. In
4
He the interactions are much stronger, and a
quantitative theory is not easily achieved.
Let us now turn to fermions. The fermions have spin,
which has to be described by an additional index s. Here
we need to consider only spin-half particles, where s takes
two values, s =
1
2
. The wave function of a fermion
system is (r
1
, s
1
, r
2
, s
2
, . . .), and it has to be antisym-
metric in the exchange of any pair of particles. For a
two-particle state this means
(r
1
, s
1
, r
2
, s
2
) = (r
2
, s
2
, r
1
, s
1
). (3)
This implies that the occupation of any single-particle
state only can be zero or one. This is known as the Pauli
exclusion principle. Thus macroscopic occupation of a
single-particle state (1) is not possible.
Superuidity in a fermion system can appear as a result
of an attractive interaction between particles. Such an
interaction can cause formation of pairs. Each pair has to
satisfy the antisymmetry condition (3). However, a pair
3
is a unit that behaves like a boson. In particular, it is
not excluded that several pairs are in the same pair state.
Superuidity in fermion systems can be understood as a
macroscopic occupation of a single pair state.
The spin part of the pair wave function has four dif-
ferent possibilities. These can be classied as a singlet
state
(4)
(which is a compact notation for
s
1
,1/2

s
2
,1/2

s
1
,1/2

s
2
,1/2
) and three triplet states, which can be cho-
sen as
+ , i( + ), + . (5)
Let us rst study the case of spin singlet (4). The pair
wave function in this case is assumed to be of the form
(r
1
, s
1
, r
2
, s
2
) = (
r
1
+r
2
2
)(r
1
r
2
)( ) (6)
where we have separated the orbital wave function to a
center of mass part and a relative part . The singlet
spin state (4) is antisymmetric in the exchange of the
two spins. In order to satisfy pair antisymmetry (3), the
corresponding orbital part (r
1
r
2
) has to be symmet-
ric. In most superconductors the pair wave function is of
the form (6). In majority of them (Al, Sn, Nb, . . . )
is approximately independent of the direction of r
1
r
2
.
This is called s-wave pairing in analogy with s, p, d,
etc. atomic orbitals. In high-T
c
superconductors there is
strong evidence of d-wave symmetry of (Encyclopedia
reference High T
c
superconductors).
Another alternative is that the spin state of a pair is
triplet (5). This case is realized in
3
He (and possibly in
some superconductors). In
3
He the orbital wave function
is of p type. There are three degenerate p-wave states p
x
,
p
y
and p
z
. The pair wave function can be written as
(r
1
, s
1
, r
2
, s
2
) =
3

j=1
3

=1

j
(
r
1
+r
2
2
)p
j
(r
1
r
2
)i

2
.
(7)
Here i

2
denotes the same three spin states as in Eq.
(5), but expressed using Pauli spin matrices
i
.
The macroscopic wave function of bosons is called or-
der parameter, since it describes ordering of the particles
and it vanishes in the normal uid phase. For fermions
the same role is played by the center of mass part of the
pair function. This is the soft degree of freedom, which
can change as a function of time and location, whereas
the other parts in the pair wave function (6)-(7) are xed.
We see that the order parameter in
4
He and in most su-
perconductors is a complex-valued scalar , but in
3
He
it is a 3 3 matrix
j
.
Quantitative theory of fermion superuids is based
on the Bardeen-Cooper-Schrieer theory of super-
conductivity (Encyclopedia reference BCS theory).
Many properties can also be described by the sim-
pler Ginzburg-Landau theory (Encyclopedia reference
Ginzburg-Landau theory and vortex lattice).
IV. HYDRODYNAMICS
Many properties of superuids can be understood in
terms of the two-uid model. The basic assumption is
that the liquid consists of two parts. These are called the
superuid and normal components. The current density
j can be represented as a sum
j =
s
v
s
+
n
v
n
. (8)
Here
s
and v
s
are the density and velocity of the su-
peruid component and
n
and v
n
are the corresponding
quantities for the normal part. The liquid density is the
sum of the two densities, =
s
+
n
. The superuid
component can ow without viscosity and it carries no
heat or entropy. Moreover it is curl free,
v
s
= 0. (9)
(This is valid only in uncharged superuids.) The normal
component behaves more like a usual viscous uid.
The two-uid model can be justied from the micro-
scopic theory discussed in Sec. III. The superuid com-
ponent corresponds to particles in the macroscopic wave
function, and the normal component to particles in other
single-particle states. The densities of the two compo-
nents depend on temperature. With increasing temper-
ature
s
(T) drops continuously from
s
(0) = and van-
ishes at the superuid transition temperature T
c
.
The two-uid model can explain many properties of su-
peruids. In particular, the existence of frictional forces
depends on the type of experiment. On one hand, the
ow in a ring-shaped container persists because it is only
the superuid component that ows (Fig. 1). On the
other hand, nonvanishing viscosity is measured with a
rotational viscometer, where the superuid is placed be-
tween two coaxial cylinders that rotate at dierent angu-
lar velocities. Here the normal component is driven into
motion and causes dissipation.
Superuids show peculiar mixing of thermal and me-
chanical properties. Consider superuid in a channel
which is heated at one end, see Fig. 4. The superuid
component is attracted to the hot region because the
chemical potential is lower there. As a consequence a
pressure dierence appears. This drives the normal com-
ponent in the direction of decreasing temperature and
convects the heat away from the source. Assuming the
geometry does not allow net mass transfer, the mass
transported by the normal and superuid components
in opposite directions are equal in magnitude.
In addition to usual sound wave, superuids have an-
other propagating mode. This second sound is an oscil-
lation where normal and superuid components move in
opposite directions. This leads to oscillation of temper-
ature whereas the density remains nearly constant. Sec-
ond sound can be generated by heating the superuid pe-
riodically, and standing waves of temperature have been
demonstrated experimentally.
In addition to the mass current j, there can be persis-
tent spin currents. This is possible in superuids whose
4
v
s
v
n
T+T
p+p
p
T
J =

(
s
v
s
+
n
v
n
)dz = 0
z
FIG. 4: A dierence in temperature generates ow of nor-
mal and superuid components in opposite directions, and a
pressure dierence appears. The viscosity of the normal com-
ponent causes v
n
to vanish at walls. The superuid velocity
v
s
(z) has to be constant in order to be curl free (9).
order parameter is more complicated than scalar (
3
He).
Spin current is described by a tensor j
spin
j
. The index
= x, y, z indicates the direction of spin angular momen-
tum that is owing, and j = x, y, z indicates the direction
of the ow. Even in equilibrium the order parameter of
3
He has nontrivial spatial variation called texture. This
is associated with persistent spin currents and, in case of
3
He-A, also with persistent mass currents.
V. QUANTIZATION OF CIRCULATION
Consider a superuid with order parameter . (As-
sume an uncharged superuid, can be either scalar or
matrix.) The superuid velocity v
s
can be expressed as
a function of the order parameter as
v
s
=

M
. (10)
Here (r) is the phase of the order parameter, (r) =
Ae
i(r)
, and the amplitude A is assumed constant. M
is the boson mass, i.e. the mass of a particle in a boson
superuid and the mass of a pair in a fermion superuid.
Eq. (10) can be justied starting from the expression of
current in quantum mechanics.
An alternative form of Eq. (10) is obtained by taking
line integral along a closed path,

v
s
dl = N
h
M
. (11)
Here we have used the property that is dened modulo
2, and N is an integer. Eq. (11) is known as quantiza-
tion of circulation. The curl-free condition (9) is a direct
consequence of Eq. (10) or (11).
Consider again superuid in a ring-shaped container
(Fig. 1). We can apply Eq. (11) to a path in the ring.
We see that, in addition to being persistent, the super-
uid velocity can only have discrete values. A similar
phenomenon in superconductors is ux quantization (En-
cyclopedia reference Flux quantization).
+2
v
s
r
FIG. 5: The vortex line and the magnitude of the velocity
eld (12) around it.
VI. ROTATING SUPERFLUID
Let us consider superuid in a container that is rotated
with angular velocity . The normal component will
follow this motion because of its viscosity. In equilibrium
it rotates uniformly with the container, v
n
= r. This
is not possible for the superuid component because it
has to be curl free (9). [Eq. (9) should be compared to
v
n
= 2.]
The rotating state of a superuid is most commonly
realized by vortex lines. On a path around the vortex
line, the phase changes by 2 (or an integral multiple
of it). This is illustrated in Fig. 5. Equivalently, the cir-
culation of superuid velocity (11) around the vortex line
is h/M. Assuming cylindrical symmetry, the phase is
the same as the azimuthal angle in the cylindrical coordi-
nate system (r, , z). The velocity eld can be calculated
from Eq. (10):
v
s
=

Mr

, (12)
where

is a unit vector in the azimuthal direction.
The structure of the rotating state is determined by
minimum of free energy. The rotation of the container is
taken into account by minimizing F = F
0
L , where
F
0
is the free energy functional in the stationary case and
L the angular momentum. In the two-uid model this
reduces to
F =

d
3
r
1
2

s
(v
s
v
n
)
2
+ constant. (13)
Thus the optimal solution corresponds to v
s
as equal as
possible to v
n
= r, but subject to condition (10).
This is achieved by a regular array of vortex lines. The
number of vortex lines n per unit area is determined by
the condition that the circulations of normal and super-
uid velocities are the same over an area containing many
vortex lines. This yields
n =
2M
h
. (14)
There are approximately 1000 vortex lines in a circular
container of radius 1 cm that is rotating 1 round per
minute.
5
Vortex lines in an uncharged superuid are analogous
to ux lines, which occur in type II superconductors (En-
cyclopedia reference Ginzburg-Landau theory and vor-
tex lattice). Flux lines of superconductors appear in
magnetic eld, which is analogous to rotation of an un-
charged superuid.
The velocity eld (12) of a vortex diverges at the vor-
tex line. Thus there must be a vortex core, where the
two-uid description is insucient. A nite energy in
the vortex core is achieved if the amplitude of the or-
der parameter vanishes at the vortex line. This is the
case for a scalar order parameter. For a matrix order
parameter it is not necessary that all components of the
matrix vanish at the line. Such vortex lines are realized
in superuid
3
He-B.
The quantization of the superuid velocity (11) is not
always true for uncharged superuids. This happens
when there is an additional contribution to the superuid
velocity (10) coming from the matrix form of the order
parameter. Such a case is realized in superuid
3
He-
A, and careful reanalysis of the rotating state is needed.
It turns out that, in addition to one-dimensional vortex
lines, the vorticity may be arranged as two-dimensional
vortex sheets and three-dimensional textures. All these
have been conrmed experimentally. In any case, a ho-
mogeneous rotation of the superuid is excluded.
VII. PHASE SLIP
Let us study superow in a channel under thermal
equilibrium (v
n
= 0). The maximum supercurrent is de-
termined by a process called phase slip. Consider that a
short piece of vortex line is nucleated at a surface on one
side of the channel. This vortex expands, goes through
the whole cross section of the channel, and nally dis-
appears on the other side. As a result of this process,
the phase dierence between the ends of the channel
has changed by 2. Part of the superuid kinetic energy
is dissipated in the motion of the vortex. This means
that the ow ceases to be dissipationless above a criti-
cal velocity for phase slips. Phase slips take rst place
in constrictions of the ow channel, where the superuid
velocity has its maximum value.
A special type of phase slip takes place in very short
constrictions, where Eq. (10) ceases to be valid. An ide-
ally short constriction shows the Josephson eect, where
the supercurrent J
s
depends on the phase dierence
as
J = J
c
sin(), (15)
and J
c
is a constant. Moreover, the time derivative of
is proportional to the dierence of the chemical potential
on the two sides of the constriction,
d
dt
=
2

. (16)
Combining the two equations, one sees that a constant
generates an oscillating current at the frequency
2/h.
The Josephson eect takes place in all superuids, and
has extensively been studied in superconductors (Ency-
clopedia reference Josephson Junctions and their appli-
cations). In helium superuids it is more dicult to
fabricate constrictions that are small enough, but this
has been achieved recently.
[1] F. London (1954) Superuids, vol. II, Wiley, New York.
[2] Andronikashvili EL and Mamaladze YuG (1966) Quan-
tization of Macroscopic Motions and Hydrodynamics of
Rotating Helium II, Reviews of Modern Physics 38, pp
567.
[3] Wilks J (1967) The properties of liquid and solid helium,
Clarendon, Oxford.
[4] Leggett AJ (1975) A theoretical description of the new
phases of liquid
3
He, Reviews of Modern Physics 47, p
331.
[5] Wheatley JC (1975) Experimental properties of super-
uid
3
He, Reviews of Modern Physics 47, p 415.
[6] Tilley DR and Tilley J (1990) Superuidity and Super-
conductivity, third edition, IOP publishing, Bristol.
[7] Vollhardt D and W ole P (1990) The superuid phases
of helium 3, Taylor & Francis, London.
[8] Donnelly RJ (1991) Quantized Vortices in Helium II,
Cambridge.
[9] Dobbs ER (2001) Helium Three, Oxford, Oxford.
[10] Lounasmaa OV and Thuneberg EV (1999) Vortices in ro-
tating superuid
3
He, Proceedings of National Academy
of Sciences USA 96, p 7760.
[11] Davis JC and Packard RE (2002) Superuid
3
He Joseph-
son weak links, Reviews of Modern Physics 74, p 741.