You are on page 1of 19

Superuid helium interferometry:

an introduction

Eric Varoquaux
CNRSUniversite Paris-Sud, Laboratoire de Physique des Solides,
Batiment 510, F-91405 Orsay Cedex, France
and
Commissariat `a l

Energie Atomique, Service de Physique de l

Etat Condense,
Batiment 772, Centre de Saclay, F-91191 Gif-sur-Yvette Cedex, France
Abstract
These lecture Notes describe, at an introductory level, how superuids can be used
to measure absolute rotations. To make them self-contained to some degree, I rst
introduce briey the two-uid model for superuid helium and the concept of su-
peruid order parameter. These ideas, which were put forward for the superuid
heliums, are now widely used, in particular for the BEC gases which are the main
topic of this Volume. They are presented in the somewhat dierent perspective of
helium physics.
A second part will deal with the Josephson eects, the real engine behind super-
uid interferometry. Theses eects were predicted in the early sixties for supercon-
ductors and were promptly observed in the laboratory [1]). It was quickly realised
that they would also exist in superuids [2] but the search took longer and conclu-
sive experiments were performed in the eighties only in the B-phase of superuid
3
He [3]. How these experiments are done, and how they can be used to measure
the rotation of the Earth by superuid interferometry is surveyed in the last two
Sections.
1 Helium as a quantum liquid
1.1 Bose-Einstein condensation
Helium remains liquid at absolute zero under moderate pressure. This anomaly
is due to the fact that it is a rare gas element with weak long-distance inter-
atomic forces and with a light atomic mass. The hard core of the potential
(the size of the atomic cloud, 2.6

A for
4
He) causes the atoms to bump into one
Preprint submitted to Elsevier Preprint 21 December 2000
another, caging them into cells of size approximately equal to the interatomic
spacing a (3.6

A for
4
He at zero pressure).
"!
#
"!
#
q
"!
#
q
-
2.6

A
Q
Q
Q

Q
Q
Q
3.6

A
Fig. 1.
This caging of the atoms is a consequence of strong
correlations between neighbours. It also causes each
of them to have a large zero point energy. An esti-
mate of this residual kinetic energy is given by
K.E. =

2
2 m
_
2
a
_
2
= 17 K.
This large kinetic energy balances out most of the
attractive potential energy. It pushes the atoms
away from one another and is responsible for the low density of helium (as far
as condensed matter goes). Low density helium does not solidify; crystalline
order does not set in at very low temperature [4].
This poses the problem of how to full Nernsttheorem which states that en-
tropy vanishes at absolute zero. The solution was proposed in 1938 indepen-
dently by F. London and L. Tisza as explained in Londons book [4]: ordering
takes place in momentum space.
Liquids that live at T = 0 because of quantum mechanical eects are called
quantum uids. Quantum uids,
4
He,
3
He, but also the inside of neutron stars,
dier from cold atomic gases in a number of ways. They are thermodynam-
ically stable, provided the proper pressure and temperature environment is
maintained. They do not reside in traps and can be fairly extended. More
importantly, their number densities is much higher, of the order of 10
22
per
cm
3
for helium, 10
38
per cm
3
for neutron stars.
For a free ideal gas, the BEC temperature is given by
k
B
T
c
=
2
2
m
_
N
V
1
(3/2)
_
3/2
, (3/2) = 2.612 .
With V/N = (1/3.6)
3

A
3
and the atomic mass of
4
He, T
c
= 3.13 K. This
value is surprisingly close to the temperature for the onset of superuidity in
4
He, T

= 2.17 K. One would indeed think that the free gas formula would
be a poor approximation for a dense medium. In particular, the bare atomic
mass m should be heavily renormalised by the strong local correlations which
dress the atom. The above numerical agreement is not a sure proof by itself
that superuid
4
He undergoes a Bose-Einstein condensation at T

.
The existence of a Bose-Einstein condensate has been conrmed directly by
the observation of a characteristic hump in the neutron diraction spectrum
[5]. The dependence of the BEC temperature on density has been measured
2
recently by making helium less dense and more gas-like by diluting it in an
extremely ne porousmedium matrix and measuring the onset of superuidity
with the help of a torsional oscillator [6]. Both types of experiments are dicult
and conceptual diculties persist. But it remains that superuidity possesses
the hallmarks of BEC and it is widely admitted that the lack of viscosity
exhibited by superuid ow is associated with the existence of a condensate
[7].
1.2 The two-uid model
At the BEC transition, superuidity sets in. The question arises of how to
describe such a state. Landau proposed in 1941 [8] that the uid behaves as
if made up of a superuid fraction which moves without friction with velocity
v
s
and carries no entropy, and a normal uid which carries entropy and
interacts with the surrounding walls of the cell. The normal fraction moves
with velocity v
n
and consists of the elementary excitations of the system which
are thermally excited at non-zero temperatures. Phonons, that is sound waves
in the liquid, are part of this elementary excitation gas and do contribute to
the entropy (and the specic heat) of the liquid as they do in a solid. However,
they do not account for all of it and Landau postulated that a second type
of elementary excitations must exist in liquid
4
He that he called rotons. The
energy spectrum of the combined excitations, phonons plus rotons, has the
shape shown in Fig. 2.
The energy spectrum of elementary exci-
Fig. 2. Energy spectrum
tations in
4
He has been measured by neu-
tron scattering over the years to a very
high accuracy [9]. A most remarkable fea-
ture is the sharpness of the neutron scat-
tering peak: as envisioned by Landau, the
elementary excitations are very long lived
and well-dened. In a normal liquid, such
as
3
He at 0.3 K, phonons and rotons also
exist but the energy spectrum is very broad:
excitations can be created with a wide range
of energies and momenta and the uid in-
teracts readily with any external pertur-
bations, in particular with walls and mov-
ing objects. This causes viscous damping. The sharpness of the excitation
energy spectrum prevents superuid
4
He to couple to the sample holder and
accounts for the lack of viscosity in superuid ow. There is no nal state in
the uid which a low-energy excitation can be scattered into, at least for ow
velocities appreciably smaller than the limit set by the dashed line in Fig.2.
3
Dissipation cannot occur. It should be appreciated that Landaus criterion for
superuidity makes no reference to the existence of a condensate; the linkage
is indirect as will appear below.
A two-uid hydrodynamics emerges from this model, a ow of viscous uid
with density
n
at velocity v
n
, and an inter-penetrating ow of non-viscous
uid with density
s
=
n
at velocity v
s
, being the total density of the
uid. At T = 0, there are no thermally excited elementary excitations,
n
= 0
and
s
is equal to the total density of the liquid. The ow of v
s
is that of
an inviscid uid and is irrotational: v
s
= 0. The two-uid hydrodynam-
ics equations can be written down with the help of the conservation laws for
mass, momentum, energy, ... and using Galilean invariance [8]; v
s
obeys the
Euler equation as does the ow velocity of an ideal inviscid uid. The stan-
dard derivation can be found, for example, in Landau and Lifshitzs course on
hydrodynamics [10].
In the two-uid model, the superuid state is described with the help of two
additional variables,
s
and v
s
, which we shall nd again in the superuid
order parameter below.
1.3 The superuid order parameter
For a nearly-ideal BEC gas with a number density of atoms n, the order pa-
rameter can be written as a complex number

nexp (i) as comes out in a
natural way from the microscopic description of a near-ideal Bose gas [11]. The
spirit of the denition of the order parameter for a dense, strongly interacting,
non-uniform system, as given by Penrose and Onsager [12], is to look at the
large-scale correlations in the superuid. In a usual uid, the correlations de-
crease rapidly as r-r

increases. A superuid can sustain a persistent current:


large scale correlations should be strong so that, when a particle is deected
at r by an obstacle and kicked out of the condensate, a sister particle is imme-
diately relocated in the condensate at r

with no loss of order in momentum


space. Such correlations are described by the density matrix,
(r, r

) =

n
|

(r)|
n

n
|(r

)| ,

(r) and (r

) being the boson creation and annihilation eld operators,


|
n
a complete set of eigenstates of the system and | the state in which
the average is expressed, which we shall take as the ground state |
0
. Among
the intermediate states |
n
, those of special relevance to the kicking-out and
relocation process discussed above are those which connect the ground state
with N bosons to the ground state with N 1 bosons. So we give special
4
attention to the following matrix element:
(r) =
0
(N 1)|(r)|
0
(N) . (1)
The (condensate) ground state is occupied by a macroscopic number of parti-
cles N
0
N, so that

(r)(r) = N
0
/V ; (r) is large in absolute value and
independent of r (to the extent that n
0
= N
0
/V is constant in space). The
density matrix can be split into two parts as follows [13]:
(r, r

) =

(r)(r

) +

other matrix elements .


The summation over all the remaining contributions is of order n
0
because
it spreads over many excited (non k=0) states. These excited states are not
macroscopically populated and only have short range coherence between them:
their sum decays as r-r

becomes large. The term

(r)(r

) in the density
matrix is equal to n
0
and remains constant as r-r

becomes large; it describes


the long-range correlations in the condensate. Its existence is the reason for
which Penrose and Onsager chose to describe superuid order by the following
quantity
(r) =
_
n
0
(r) e
i(r)
.
The dierence with the near-ideal BEC case mentioned above is that n
0
may
be appreciably smaller than n. The incoherent terms in the density matrix
contribute to the total density, (r, r) = n
0
+

k=0
n
k
: the condensate is
depleted from all the particles which have non-zero momentum because of
inter-particle collisions and which populate the states k with distribution n
k
.
While this depletion is a small eect in low density atomic gases, it is large in
liquid helium. Penrose and Onsager have noticed that this depletion is easily
calculated for a gas of hard sphere bosons such as the one pictured in Fig. 1.1;
they have found that, due to hard sphere repulsion, only about 10 % of the
helium atoms can be in the condensate.
The strong depletion of the condensate raises the following question: how is
it that the condensate fraction is only 10 % while the superuid fraction in
the two-uid model is 100 % at T = 0? Simply because these are not the
same quantities. The superuid density stands in fact for the inertia of the
superuid fraction, as measured with a torsional oscillator for instance. This
quantity is dierent from the density of bosons in the macroscopically occupied
quantum state seen as a hump in the neutron diraction spectrum. When the
superuid is set into motion, the condensate enforces long-range order and
drags the excited states along through the short-range correlations; there is
entrainment of all the atoms in the uid by the condensate. Microscopic theory
is needed to describe this process in detail, as done for instance by Ceperley
[14] using path integral Monte Carlo methods.
5
But, even though the above question can be settled, we still have the problem
that the Gross-Pitaevskii (GP) equation, which applies so wonderfully well
to the BEC gases [11], has the ambiguity in helium of being derived for the
order parameter

n
0
e
i
and of being applied to

s
e
i
in order to agree
with the two-uid hydrodynamics, as mentioned below. This unsatisfactory
state of aairs cannot be resolved. The GP equation, in spite of its many very
useful features when applied to helium as reviewed for instance by Berlo and
Roberts [15], is simply not a good microscopic description of a dense superuid.
It takes no account of nite-range hard-core interactions and disregards the
full hierarchy of N-body collisions. It provides a phenomenological description
of superuid helium with a remote connection only to rst-principles theory.
Following this phenomenological bend, we shall consider below the quantity
(r) =
_

s
(r) e
i(r)
(2)
and treat it as an order parameter with the properties of a macroscopic
wavefunction. The particle current density is therefore given by
j = (/2i) [

(r)(r) (r)

(r)] = (/2i)
s
(r) (r) . (3)
From Eq.(3), the superuid velocity appears in a natural way as the gradient
of the macroscopic phase
v
s
=

m
. (4)
It can be checked that, when the order parameter (2) is plugged into the GP
equation, the equations of motion of the superuid component in the two-uid
model are recovered (see, for instance, [11] or [15]). In addition, the charac-
teristic length in the GP equation governs the healing length over which the
order parameter grows from zero, right at a solid wall for instance, to its value
in the bulk of the superuid. It is, however, a phenomenological parameter
whose value in
4
He, 2.5

A, can be derived from a number of experiments [16].
The healing length is a temperature dependent quantity and diverges at T

.
The situation with
3
He, which is a liquid of fermions, is slightly dierent.
Superuidity in such a case is of the BCS type (BardeenCooperSchrieer),
bosons appearing as a result of Cooper pair formation. The hard core repulsion
in helium is strong and prevent pairing to form in an s-state for which the
probability of overlap of the two atoms in the pair is too large. Pairing occurs in
a p-state, with orbital quantum number L = 1. The antisymmetry of the pair
wavefunction then requires S = 1. The order parameter reects the internal
structure of the pair and is a 33 matrix [17] to account for the three azimuthal
spin and orbit states. However, for the B-phase, we can disregard this internal
structure because it averages out when there is no strong perturbations of
the superuid (no strong heat current, ow, applied magnetic eld, ...). The
6
B-phase is said to be pseudo-isotropic and its hydrodynamics, in this simple
situation, reduces to that of superuid
4
He. The healing length is given by the
BCS theory as = v
F
/, v
F
being the Fermi energy and the superuid
gap above the Fermi sea. It is much larger than in
4
He: at T and P = 0,
= 650

A.
2 The Josephson eects
Josephson predicted the eects which bear his name in 1962 for supercon-
ductors within the framework of the BCS theory. As discussed below, there
are two distinct phenomena, ac and dc, which take place between two pieces
of superconducting material which are brought into weak interaction with
one another, for instance by allowing the quantum tunnelling of Cooper pairs
through a thin oxide barrier. The existence of these eects was promptly con-
rmed experimentally [1]. The search for analogous eects was then started
in superuids [2], but it took until the late-eighties to nally observe them
in liquid helium [3]. Interference eects which have a strong kinship with the
Josephson eects have been seen in the BEC gases [18].
2.1 N and are canonically conjugate
Let us follow Andersons simple approach [1] and construct a wave function
with a given number N of particles by applying boson creation operators to
the vacuum state |0:
|N =

1
(r
1
)

2
(r
2
)...

N
(r
N
)|0 .
Multiply the eld operators by an identical phase factor:
|N = e
i

1
(r
1
)e
i

2
(r
2
)...e
i

N
(r
N
)|0 = e
iN
|N ,
which implies that i|N/ = N |N. This simple argument can be ex-
tended to coherent superpositions of states with dierent numbers of particles
[19]. Thus, it is established, at least for that class of states, that particle num-
ber and phase are canonically conjugate variables with the consequence that
they obey the following uncertainty relation:
N 1 .
For systems with xed number of particles, the phase is undetermined. For
systems with a condensate where particles can be kicked in and out of the
condensate, particle number can uctuate locally and the phase becomes a
7
meaningful quantity. In fact, if we assume that the particle number uctu-
ations are of order

N, then those of the phase are of order 1/

N. For N
large, both the phase and the particle number are well-dened quantities, their
relative uctuations being of order 1/

N.
2.2 The ac-Josephson equation
We now turn to the time evolution of and N which can be expressed quite
generally by:
i

N =[H, N] = i
H

,
i =[H, ] = i
H
N
,
H being the Hamiltonian of the system.
These equations hold for the operators N and but these quantities can also
be treated to a very good approximation as c-numbers because, as discussed
above, their relative quantum uncertainties are very small. Let us then take
an average over a volume of superuid which is small compared to the size
of the sample but still contains a large number of atoms. This coarse-grained
averaging procedure is well suited to a dense system such as helium (dense
compared to the cold atomic gases). The rate of change of the phase is thus
given, writing H as E, the energy, by

t
=
E
N
= , (5)
being the chemical potential. This relation is the ac Josephson relation.
For superconductors (for which it was initially derived), = 2eV , 2e being
the electrical charge of the Cooper pair, V the applied electric potential. For
superuids, = v
a
P, v
a
being the atomic volume for
4
He, twice the atomic
volume for
3
He. The contribution of the entropy to the chemical potential,
ST, is negligible at low temperature.
The ac Josephson relation applies more readily to phase and pressure dier-
ences. In particular, when applied to the gradient of the phase, it can be cast,
using Eq.(4), into the Euler equation:
v
t
+(v
a
P +
1
2
m
a
v
2
) = 0 ,
m
a
being the atomic mass of the eective boson.
8
The second Heisenberg equation of motion, that for

N, expresses particle
number conservation:

N
t
=
E

. (6)
What we have just done is simply to reproduce the equations for the motion
of the superuid component in the two-uid hydrodynamics from the fact
that N and are canonically conjugate. However, as stressed by Anderson
[2], the range of validity of Eq.(5) is quite wide and it will apply even when
hydrodynamics is expected to break down as for tunnelling supercurrents. In
the same kind of situations, to which we turn below, the internal energy E
depends in a non-trivial way on , as may be expected from Eq.(6).
2.3 The Josephson supercurrent
When applied between two regions of the super-
Fig. 3. Weak link
uid, Eqs.(5) and (6) describe the supercurrent
owing from one region to the other. This sit-
uation becomes especially interesting when the
two regions, the two superuid baths, are well
separated and only weakly coupled to one an-
other so that a well dened phase dierence be-
tween them can be sustained.
For superconductors, this was the case studied
by Josephson, the weak link being provided by a
thin layer of insulating oxide through which the
Cooper pairs could tunnel quantum-mechanically.
For superuids, the only practical weak link so
far is a microscopic aperture. As is well known
in superconductivity, weak links, or micro-bridges,
lead to the same kind of eects as tunnel junc-
tions [20].
Let us consider superow through such a micro-aperture, as pictured in Fig. 3.
For simplicity, we restrict the problem to one dimension along z . We describe
the barrier (the weak link) as a square potential wall of height U over length
l. We want to compute the areal current density through the barrier and we
shall take care of the nite lateral extent of the weak link at the end of the
calculation.
In the bulk of the uid, the wave function corresponding to a state with energy
E is taken as a plane wave with identical amplitudes A =
1/2
s
on both sides
9
of the barrier, but with phases which dier by . Here, we make explicit use
of the denition (2) of the order parameter.
Inside the barrier, A is severely depressed: this is a consequence of the as-
sumption that the barrier is a weak link between the two baths. Hence, the
interactions within the uid can be neglected (for instance, the V
0
(

) term
in the GP equation becomes small). The equation of motion then reduces to:
i

t
=

2
2m
a

2
+ U , U > E ,
and also has a plane wave solution exp{i(Et/kz)}. The momentum takes
two values corresponding to the two possible directions of propagation:
k

= (i/)
_
2m
a
(U E) .
Let b = /
_
2m
a
(U E) : the barrier height is characterised by a penetration
length. The wave function inside the barrier can be found by standard methods
[21] to be:
(z) =
A
sinh(l/b)
_
sinh
_
z
b
_
e
i
sinh
_
z l
b
__
.
The modulus of midway in the barrier is such that:

(l/2) (l/2) =
A
2
2 cosh
2
(l/2b)
[1 + cos ] . (7)
Knowing the wavefunction, we can compute the current density, Eq.(3), in a
straightforward manner. The total current through a micro-aperture of eec-
tive cross section s is found to be:
J =j s =
s
2i
A
2
sinh
2
(l/b)
__
sinh
_
z
b
_
e
i
sinh
_
z l
b
__

_
(e
i
/b)cosh
_
z
b
_

1
b
cosh
_
z l
b
__
complex conjugate
_
=J
c
sin () , with J
c
=
A
2
s
b sinh(l/b)
. (8)
Eq.(8) describes the dc Josephson eect. Although this equation has been
obtained here in a simplied manner, it is nearly identical to the result of
much more involved theories [22,23].
The supercurrent J is periodic by 2 in as it can be expected to be since
changing the phase by 2 on one side of the barrier must leave the overall
10
physical situation unchanged. It vanishes for = not because the velocity,
, goes to zero but because
s
inside the barrier, which is proportional to
A
2
sin()/, does. The modulus of the wave function at the midpoint in
the barrier, Eq.(7), vanishes: superuidity is actually destroyed at that point,
which is why the supercurrent goes to zero and the phase can slip by 2.
The critical current J
c
decreases exponentially with l/b. The assumption of
weak coupling implies that l/b is large. Weak coupling is achieved in practice
for a micro-aperture by making its diameter (or the shorter lateral size if it is
not round) of the order of a few healing lengths. The length along the stream
in the aperture l must also be of the same order so that can remain of
the order of .
If the coupling is not weak, a more elaborate calculation is necessary: the sine
function is replaced by a more general periodic function f
2
(), the current-
phase relation of a non-ideal weak link. Often, this relation is not even single-
valued and, when the phase is varied, the current jumps discontinuously from
one determination to another: the weak link is then said to be hysteretic.
This behaviour is due to the nucleation of vortices and is accompanied by
dissipation while the ideal Josephson case where f
2
() is a sine function is
dissipation-less [20].
Since the healing length is very small for
4
He, it cannot be expected to exhibit
a near-ideal Josephson eect in the micro-apertures that can be manufactured
at present, except very close to the point where the experiment has been
conducted recently [24]. The experiments which have rst shown the existence
of this eect have been carried out in
3
He [3].
3 The superuid Helmholtz resonator
We now turn to a brief description of the experimental conrmation of the
existence of a near-ideal dc Josephson eect in superuids.
To carry out the experiment, one must rst have a superuid sample, that is,
one must cool
4
He below T

= 2.17 K, or
3
He to less than 10
3
K at zero pres-
sure. This is done in a nuclear demagnetisation cryostat which can maintain
sub-milliKelvin temperatures for several weeks in a row. The experimental cell
itself is immersed in the main superuid bath. It consists of a small chamber
with two openings to the main bath, as depicted in Fig. 4. One opening is the
micro-aperture in which all the action takes place because the ow velocity is
highest there, the other is a longer and wider channel in which the superow
is subcritical and which provides a separate path to equilibrate temperature,
pressure and phase dierences in a controlled way. This parallel channel also
11
provides a closed path in the superuid as seen in Fig. 4 whose importance
will appear later.
One side wall of the chamber is a
Fig. 4. Twoaperture Helmholtz resonator
exible membrane: it is made of Kap-
ton, a polyimide material which does
not become sti at low temperature.
This membrane is used to drive the
ow of superuid through both open-
ings: it is coated with a thin alu-
minium lm and can be driven elec-
trostatically. Since the aluminium
coating is superconducting, its mo-
tion can be tracked by placing a su-
perconducting coil close to it and making a very sensitive electrodynamic mi-
crophone, capable of reading out position changes with a resolution of 710
5
nm/

Hz. It is important to realise that this device, with the restoring force
provided by the exible membrane and the inertia of the accelerated uid in
the two vents, behaves as a resonator, a exible wall Helmholtz resonator.
The operation of such resonators has
Fig. 5. Total current vs phase
been described in a number of publi-
cations, for instance [25,26]. Since the
current-phase relation J() is non-
linear (as well as periodic and possibly
multi-valued), the resonator response to
the drive is also non-linear. Its motion
can be understood with the help of the
diagram in Fig. 5 which shows the res-
onator characteristics where J() is
taken as a sine function. The volume
swept by the membrane in its motion
corresponds to the combined ow through
the two channels. Upon increasing the
drive level, hence = v
a

_
Pdt (be-
cause of the ac Josephson relation), the
resonance amplitude increases in a stepwise manner: the current climbs an ad-
ditional step for each increase of the (ac) peak excursion of by 2. For a
smoothly varying J(), the steps are rounded; in the hysteretic case, they
become steep and kinky. This response pattern to drive level is exactly simi-
lar to the staircase pattern of rf-SQUIDs. Two-aperture Helmholtz resonators
are the superuid analogues of electrodynamic SQUIDs which have become
widely used, in particular to measure very small magnetic elds.
12
The experiments have been conducted both in
4
He and in
3
He with micro-
apertures in the shape of rectangular slits 0.3 5 m
2
micro-machined in a
0.2 m thick nickel foil. In
4
He, the size of the aperture is large compared to
the healing length (except very close to the -point) and the eects which are
observed are due to the nucleation of vortices. These eects are only remotely
connected to the Josephson eects but they do involve the ac Josephson re-
lation. The current-phase relation in this case is highly non-ideal: it consists
of straight segments extending from J
c
to J
c
, parallel to one another and
shifted in phase by 2.
In superuid
3
He in the Bphase at
Fig. 6. Staircase patterns in
3
He-B
zero pressure, the current phase re-
lation was observed in 1987 [3] to be
very nearly a sine function close to
the superuid transition temperature
T
c
as deduced from the shape of the
staircase patterns. A set of such stair-
case patterns is shown in Fig. 6. The
resonance peak amplitude is plotted
in arbitrary (instrumental) units in
terms of the drive level, also in arbi-
trary units. The four curves for T/T
c
= 0.69, 0.77, 0.83 and 0.89 from top to
bottom all start from zero with no ap-
plied drive; they are shifted along the
ordinates for clarity. At T = 0.89 T
c
,
the modulation in the response is quite
smooth and regular; it is weak be-
cause the critical current is small close
to T
c
. As the temperature is reduced, the critical current, marked by the ap-
pearance of the rst step when the drive level is increased, grows and J() is
seen to gradually become less ideal, and even hysteretic below 0.8T
c
. These
observations, which were conrmed in 1997 by a dierent method at Berkeley
[27], show the existence of an ideal dc Josephson eect in superuids.
4 Rotation sensors
Ever since the Foucault pendulum experiment in 1851, physicists have been
fascinated by measurements of the rotation of the Earth, although just gazing
at the stars gives enough evidence of it (at least, since Galileo). Such an
interest arises because these measurements illustrate the basic foundations of
mechanics, and of optics and light waves with the Sagnac eect [28], and,
more recently, of the properties of systems at the crossing between matter and
13
waves, electrons [29], neutrons [30], BEC atoms [31] and superuid helium
[32]. Pushing to extremely high sensitivity, they can even provide a test of
Einsteins General Relativity [33].
Let us consider a loop (e.g. a hollow torus) lled with liquid helium that
we cool through the transition temperature. When superuid ordering takes
place, it does so in the inertial frame of reference, exactly in the same way as
the Foucault pendulum starts oscillating in that same frame when the piece
of string which holds it back is burnt. This means that liquid helium actually
sets into motion with respect to the Earth-bound laboratory (provided the
torus is properly oriented with respect to the axis of rotation of the Earth).
This eect, predicted by Landau, has been observed by Hess and Fairbank
and is one of the hallmarks of superuidity [7]. In most experiments however,
it is blurred by stray vorticity [32].
In the experiments described here, the spurious eect of vorticity is subtracted
by orienting the loop which picks up the rotation so that it cuts either no ux
from

(the rotation axis of the Earth lies in the plane of the loop), or
maximum ux. A schematic view of the experiment is given in Fig. 7. The
whole cold part of the apparatus is rotated about its (vertical) axis. The
experimental results are then checked to be independent of the arrangement
of the stray vorticity in the cell by changing this arrangement (shaking the
cryostat!) and repeating the measurements. The liquid motion is detected by
using a two-aperture Helmholtz resonator of the kind described in the previous
Section.
If a rotation-induced supercurrent is present in the micro-aperture, to which
corresponds a phase dierence

, the operating point of the resonator is


shifted (refer to Fig. 5) and the small-signal resonance frequency changes. This
method applies to the case of
3
He which is more favourable [34] because the
non-linear device is fully non-dissipative in the ideal Josephson regime. The
intrinsic noise is expected be of the order of . A residual rotation noise of
7 10
8
(rad/s)/

Hz, or 10
3

Hz, obtained in preliminary measure-


ments has been reported at the 22
nd
International Low Temperature Physics
Conference in Finland in 1999 [35]. This gure is undoubtly not the ultimate
noise level of the superuid
3
He rotation-sensor. The sensitivity can be in-
creased by enlarging the pick-up loop, which at present has an area of 6 cm
2
.
The long-term stability of superuid gyrometers is intrinsically very good
quantum-mechanical motion is drift-free and the practical limitations are
being currently explored.
But, as an inquisitive reader will notice, the device which has just been de-
scribed seems to be very little quantum-mechanical. All what appears to be
needed is a viscousless uid with a well-dened critical velocity. But this is
precisely where the quantum mechanics lies, in the existence of a condensate
14
Fig. 7. Resonator with rotation pick-up loop
which exhibits these very features.
To see in more details how this comes about, let us consider the velocity
circulation along a closed contour lying in the uid along the pick-up loop. In
the inertial frame (that which is xed with respect to the distant stars), this
quantity is dened as:
=
_
loop
v dl .
In the laboratory frame, which is rotating with respect to the inertial frame
with velocity , the expression of the uid velocity at point r becomes v=
v +r. The circulation becomes, along the same contour as above,

= +
_
loop
r dl .
By rearranging the triple product and introducing the vector area S spanned
by the (oriented) contour, we nd

= +2 S . The circulation changes by


2 S because of the rotation. The corresponding change of the phase around
the superuid loop is
= (m
a
/)2 S. (9)
Let us now take a dierent point of view and derive the phase shift due to
rotation in another way by looking at the propagation of the phase information
around the loop. Let us consider for simplicity a circular loop, of radius R.
This loop can be the superuid loop as we discuss at present, or the loop
formed by the arms of a ring-laser.
The time taken for a perturbation in the phase to travel around the loop is
= 2R/v, v being the velocity of propagation. Possible choices for this
velocity could be the Fermi velocity in
3
He if quasi-particles are considered, or
the sound velocity if the change in the phase corresponds to a local variation
15
of the uid velocity, or the velocity of light if electromagnetic disturbances or
light waves are involved. The exact choice will turn out to be irrelevant and
we do not specify v any further at this point.
If, during that time , the loop rotates at angular ve-
Fig. 8.
locity , the weak link, referring to Fig. 8, moves by
= (R) = 2
R
2
v
=
2S
v
.
In the case of a ring laser, the beam-splitting prism
moves by the same quantity, which induces a phase-
shift
= k = (2k/v) S , (10)
k being the laser light wave number. The comparison between (9) and (10)
tells us that the superuid helium condensate behaves as a giant matter wave
obeying the de Broglie relation, k = m
a
v.
One could argue that laser physics is dierent from superuid physics and
that this comparison is purely formal. We know on other grounds that this
is not the case, for example through the analogy with superconductivity for
which the relationship between the BCS theory and electromagnetism is well
spelled out. The recent work with BEC gases also illustrates very clearly the
wave-like properties of atomic condensates. With superuid helium, we are
led to mix this standpoint with uid mechanics and we nd a rather extreme
case of giant matter waves.
We can go one step further in the discussion of the quantum character of
these superuid helium interferometry experiments and mention a nal point:
where does the interference set in? The detection of phase shifts along the
loop makes use of the dc Josephson eect, which results from the beats of the
wave-functions from both sides of the weak link as described by Eq.(7). This
interference eect is a full-edged quantum-mechanical eect and represents
a key feature of the
3
He gyrometer [35]. For the
4
He case, [32], the quantum
features are buried deeper and lie in the nucleation of quantised vortices which
then go on behaving as purely classical objects. Thus, not only do these experi-
ments provide a laboratory demonstration of Einsteins Equivalence Principle,
which is the actual basis of the Sagnac eect, and of de Broglies duality of
waves and particles in condensed matter, but they also provide an illustration
of how closely the borderline between quantum and classical physics can be
approached.
To summarise:
16
Helium can remain liquid at absolute zero because of quantum eects: or-
dering takes place in momentum space as a Bose-Einstein condensation.
The condensate formation is described by a superuid order parameter
which can be chosen on phenomenological grounds to reproduce Landaus
two-uid model: =

s
exp(i).
Josephson eects take place between two weakly coupled baths of superuid.
They oer a concrete example of quantum eects in hydrodynamics and
provide a sensitive interferometric mean of measuring small changes of .
Superuid interferometry can be used to measure rotations with high reso-
lution in a drift-free manner. The present sensitivity is 7 10
8
(rad/s)/

Hz .
Phase coherence in a superuid arises from the existence of a single-particle
quantum state with a macroscopic occupation number. In the language of
hydrodynamics, for distances larger than the healing length, it boils down to
the existence of a velocity potential and to the quantisation of circulation.
Acknowledgements
The author wishes to acknowledge useful discussions on these problems with
O. Avenel, P. Hakonen and Yu. Mukharsky as well as their comments, and
those of R. Packard, on the manuscript.
References
[1] P.W. Anderson in Lectures on the Many-body Problem ed. E.R. Caianello, Vol.
2, (Academic Press, New-York, 1964), p.113.
[2] P.W. Anderson, Rev. Mod. Phys. 38 (1966) 298 and in Quantum Fluids, ed.
D.F. Brewer (North-Holland, Amsterdam, 1966), p. 146.
[3] O. Avenel and E. Varoquaux, Jpn. J. App. Phys. 26 (1987) 1798; Phys. Rev.
Lett. 60 (1988) 416.
[4] F. London, Superuids Vol. II, Macroscopic Theory of Superuid Helium (Dover,
New-York, 1964).
[5] P.E. Sokol, in Bose-Einstein Condensation, eds. A. Grin, D.W. Snoke, S.
Stringari (CUP, Cambridge, 1995) Ch. 4.
[6] J.D. Reppy, B.C. Crooker, B. Hebral, A.D. Corwin, J. He and G.M. Zassenhaus,
Phys. Rev. Lett. 84 (2000) 2060.
[7] A.J. Leggett, Rev. Mod. Phys. 71 (2000) S318.
[8] L.D. Landau, in Theory of Superuidity, I.M. Khalatnikov, (W.A. Benjamin,
New-York, 1965).
17
[9] K.H. Andersen, J. Bossy, J.C. Cook, O.G. Randl and J.L. Ragazzoni, Phys. Rev.
Lett. 77 (1996) 4043 and references therein.
[10] L.D. Landau and E.M. Lifschitz, Fluid Mechanics (Pergamon, London, 1958),
139.
[11] F. Dalfovo, S. Giorgini, L.P. Pitaevskii and S. Stringari, Rev. Mod. Phys. 71
(1999) 413 and this Volume.
[12] O. Penrose and L. Onsager, Phys. Rev. 104 (1956) 576.
[13] P. Nozi`eres and D. Pines, The Theory of Quantum Liquids, Vol. II (Addison-
Wesley, New-York, 1990) Ch.10.
[14] D.M. Ceperley, Rev. Mod. Phys. 67 (1995) 279.
[15] P.H. Roberts and N.G. Berlo, in Quantized Vortex Dynamics and Superuid
Turbulence, ed. C. Barenghi, (Springer, Berlin, to be published).
[16] G.G. Ihas and F. Pobell, Phys. Rev. A 9 (1974) 1278.
[17] A.J. Leggett, Rev. Mod. Phys. 47 (1975) 331.
[18] M.R. Andrews, C.G. Towsend, H.J. Miesner, D.S. Duirfee, D.M. Kurn and W.
Ketterle, Science 275 (1997) 637.
[19] P.W. Anderson, in Prog. Low Temp. Phys., Vol. V ed. C.J. Gorter (North-
Holland Amsterdam, 1967) p.1; see also A.J. Leggett, in Bose-Einstein
Condensation, eds. A. Grin, D.W. Snoke and S. Stringari (C.U.P., Cambridge,
1995) p. 452.
[20] K.K. Likharev, Rev. Mod. Phys. 51 (1979) 101.
[21] L.D. Landau and E.M. Lifschitz, Quantum Mechanics (Pergamon, London,
1958), 23.
[22] D. Rainer and P. Lee, Phys. Rev. B35 (1987) 3181.
[23] J. Kurkij arvi, Phys. Rev. B 38 (1988) 11184.
[24] K. Sukhatme, Yu. Mukharsky, T. Chui and D. Pearson, J. Low Temp. Phys.
(to be published).
[25] E. Varoquaux, O. Avenel and M. Meisel, Can. J. Phys. 65 (1987) 1377.
[26] D.R. Tilley and J. Tilley, Superuidity and Superconductivity, 3
rd
edition (Adam
Hilger, Bristol, 1990) p. 290.
[27] S. Backhaus, S.V. Pereverzev, A. Loshak, J.C. Davis and R.E. Packard, Science
278 (1997) 1435.
[28] See, for instance, G.E. Stedman, Rep. Prog. Phys. 60 (1997) 615); R. Anderson,
H.R. Bilger and G.E. Stedman, Am. J. Phys. 62 (1994) 975.
[29] F. Hasselbach and M. Nicklaus, Phys. Rev. A 48 (1993) 143.
18
[30] S.A. Werner, Class. Quantum Grav. 11 (1994) A207.
[31] B.P. Anderson and M.A. Kasevich, Science 282 (1998) 1686.
[32] O. Avenel, P. Hakonen and E. Varoquaux, Phys. Rev. Lett. 78 (1997) 3602; J.
Low Temp. Phys. 110 (1998) 709.
[33] See the Gravity Probe B web site at einstein.stanford.edu/index.html.
[34] In
4
He, J() is linear and the rotation-induced current is measured by
the unbalance of the critical velocity thresholds between both ow directions
through the aperture. The noise is due to the stochastic nature of vortex
nucleation. The rotation-sensing experiment was rst done in
4
He by O. Avenel
and E. Varoquaux, Proc. of the 21
rst
Int. Low Temp. Phys. Conf., Czech. J.
Phys. 46-S1 (1996) 3319. It was repeated by K. Schwab, N. Bruckner and R.E.
Packard, Nature, 386 (1997) 585. However, the spurious eect of stray vorticity
(see Ref.[32]) is not discussed in the latter work.
[35] Yu. Mukharsky, O. Avenel and E. Varoquaux, Physica B 280 (2000) 287.
19

You might also like