Strong-coupling and the Stripe phase of 3 He

Joshua J. Wiman and J. A. Sauls

Department of Physics & Astronomy,
Northwestern University

arXiv:1605.01047v1 [[Link]-con] 3 May 2016

(Dated: May 4, 2016)

Thin films of superfluid 3 He were predicted, based on weak-coupling BCS theory, to
have a stable phase which spontaneously breaks translational symmetry in the plane of
the film. This crystalline superfluid, or stripe phase, develops as a one dimensional
periodic array of domain walls separating degenerate B phase domains. We report
calculations of the phases and phase diagram for superfluid 3 He in thin films using
a strong-coupling Ginzburg-Landau theory that accurately reproduces the bulk 3 He
superfluid phase diagram. We find that the stability of the Stripe phase is diminished
relative to the A phase, but the Stripe phase is stable in a large range of temperatures,
pressures, confinement, and surface conditions.

Keywords: Superfluid 3 He, Phase transitions, Confined quantum liquids

I. INTRODUCTION

The theoretical prediction of a crystalline superfluid, or

stripe phase, that spontaneously breaks translational
symmetry in thin films of 3 He [1], along with advances in
nanoscale fabrication and experimental instrumentation
[2], has renewed interest in the properties of superfluid
3
He in thin films and confined geometries. In the weakcoupling limit of BCS theory the Stripe phase is predicted to be stable in a large region of temperature and
pressure for films of thickness D 700 nm. However,
recent experiments on 3 He confined in slabs of thickness
D 700 nm and D 1080 nm have failed to detect evidence of the Stripe phase [2].
A limitation of the Vorontsov and Sauls theory is that
it does not include strong-coupling corrections to the
BCS free energy. In bulk 3 He, weak-coupling theory predicts a stable B phase at all temperatures and pressures;
however, the A phase is found to be stable experimentally at Tc and pressures above pPCP 21.22 bar, with a
first-order transition at TAB < Tc to the B phase. Theoretically accounting for the stability of the A phase requires including next-to-leading order corrections to the
full free energy functional, i.e. corrections to the weakcoupling functional [3]. While these strong-coupling corrections are largest at high pressures, they remain significant even for p 0 bar[4]. Thus, for superfluid 3 He
confined within a film, it is to be expected that strongcoupling effects will increase the stability of the A phase
relative to both the B- and Stripe phases, which could diminish, or even eliminate, the experimentally accessible
region of the Stripe phase.

jjwiman@[Link]
sauls@[Link]

In this paper we report our study of the A-Stripe and

Stripe-B superfluid transitions using a Ginzburg-Landau
(GL) functional that incorporates strong-coupling corrections to the weak-coupling GL material coefficients and
accurately reproduces the bulk superfluid 3 He phase diagram [5]. Within this strong-coupling GL theory we calculate the superfluid order parameter and phase diagram
as a function of pressure, temperature, confinement, and
surface conditions.

II. GINZBURG-LANDAU THEORY

The general form of the p-wave, spin triplet order parameter for 3 He is given by the mean-field pairing self
energy, which can be expanded in the basis of symmetric Pauli matrices (S = 1) and vector basis of orbital
momenta (L = 1),
p) =
(

Ai (i y ) pi ,

(1)

where p is the direction of relative momentum of the

Cooper pairs defined on the Fermi surface, and Ai are
the elements of a 3 3 complex matrix,

Axx Axy Axz

A = Ayx Ayy Ayz ,
Azx Azy Azz

(2)

that transforms as a vector under spin rotations (with

respect to ) and (separately) as a vector under orbital
rotations (with respect to i). We choose aligned spin and
orbital coordinate axes.

2
A. Free energy functional

To determine the order parameter and the phase diagram of 3 He in a film geometry, we solve the EulerLagrange equations of the Ginzburg-Landau functional,
subject to relevant boundary conditions, and calculate
the order parameter and the stationary free energy. The
GL functional is defined by bulk and gradient energies

Z
[A] =

~
dR

with temperature dependent strong-coupling corrections,

and is supplemented by boundary conditions that we can
tune from maximal to minimal pair-breaking [5].

The Ginzburg-Landau free energy functional is expressed in terms of invariants constructed from the order
parameter matrix, A, and is given by [6]

2




2
(T )T r AA + 1 T r(AAT ) + 2 T r(AA )

(3)







+ 3 T r AAT (AAT ) + 4 T r (AA )2 + 5 T r AA (AA )

K2 (j Aj k Ak )

TAB

30

1
0.0

24

p (bar)

(4)

21wc = 2wc = 3wc = 4wc = 5wc ,

(5)

 
N
(0)
1
7
(3)
,
(6)
1wc =
(kB Tc )2 30 8
7(3)
K1wc = K2wc = K3wc =
N (0) 02 ,
(7)
60
and determined by the normal-state, single-spin density
of states at the Fermi energy, N (0), the bulk transition
temperature, Tc , and the Fermi velocity, vf . The Cooper
pair correlation length 0 ~vf /2kB Tc varies from 0 '
770
A at p = 0 bar to 0 ' 160
A at p = 34 bar.

0.2

34

In the weak-coupling limit the GL material parameters

are given by
wc (T ) = 13 N (0)(T /Tc 1) ,

K3 (k Aj j Ak )

0.2

18

isc

K1 k Aj k Aj

pPCP

0.4
12

0
0.0

0.5

1.0

Choi et al
1.5

T (mK)

0.6

Tc

2.0

2.5

0.8

12

18

24

30

34

p (bar)

FIG. 1 (Left) Bulk phase diagram with lines showing the

measured phase transitions and shading showing the calculated regions of phase stability based on GL theory. The
isc coefficients are from Choi et al[4] and are plotted in the
right panel.

B. Strong-coupling corrections

The fourth order parameters that enter the GL free

energy functional are modified by next-to-leading order
corrections to the full Luttinger-Ward free energy functional [3]. These corrections scale as isc iwc (T /TF )
near Tc . Combining the isc with the weak-coupling
coefficients in the bulk GL functional yields the critical
pressure, pPCP , above which the A phase is stable relative
to the B phase. For p > pPCP the temperature scaling
of the strong-coupling corrections relative to the weakcoupling parameters breaks the degeneracy in temperature between the A and B phases at the critical pressure and accounts for the pressure dependence of the A-B
transition line, TAB (p), and thus an accurate bulk phase
diagram [5]. The resulting strong-coupling parameters
are given by
i (T, p) = iwc (p, Tc (p)) +

T
isc (p) .
Tc

(8)

Figure 1 shows the experimental bulk superfluid phase

diagram as well as the phase diagram calculated from
strong-coupling GL theory using isc coefficients obtained based on analysis of selected experiments by Choi
et al. [4]. These coefficients differ substantively from
those calculated from strong-coupling theory based on a
quasiparticle scattering amplitude that accounts for the
normal Fermi liquid properties of 3 He. Figure 2 shows
the bulk phase diagram calculated using the isc from
Sauls & Serene[7]. This set of coefficients has a higher
polycritical pressure than experiment; however, the pressure dependence of the isc represents the expectation
based on strong-coupling theory dominated scattering
from ferromagnetic spin-fluctuations. Below p = 12 bar
the isc are extrapolated to zero at a negative pressure
corresponding to Tc = 0 [5].

3
0.2

34

1.0

Unscaled GL

30

0.0

Scaling

0.8

Quasiclassical
Vorontsov & Sauls

0.2

0.4

0.6

T /Tc

18

isc

p (bar)

24

12

0.4
6

Sauls & Serene

0
0.0

Stripe

0.6

0.5

1.0

1.5

2.0

T (mK)

2.5

0.8

12

18

0.2
24

30

34

p (bar)

0.0

FIG. 2 (Left) Bulk phase diagram where the shaded regions

represents the phases calculated from GL theory with the
isc of Sauls & Serene [7]. These strong-coupling corrections
are plotted in the right panel.

10

12

14

16

18

20

D/0

FIG. 3 Comparison of the phase diagrams calculated within

weak-coupling quasiclassical theory (blue and orange lines),
weak-coupling GL theory (dashed lines), and weak-coupling
GL theory with D rescaled by GL (T )/ (T ) (solid black
lines).

C. Boundary Conditions

Confinement is represented in the GL theory through

boundary conditions. For infinite, planar surfaces there
are two limiting cases: maximal pairbreaking, due to
the retroreflection of quasiparticles[8], and minimal pairbreaking, corresponding to specular reflection [9]. For a
surface on the xy plane with 3 He filling z > 0, maximal
pairbreaking is defined within GL theory by

Ai z=0 = 0 i {x, y, z} ,
(9)
while minimal pairbreaking is defined by

Az z=0 = 0 ,


z Ax z=0 = z Ay z=0 = 0 .

D. Extrapolating GL theory to low temperatures

Ginzburg-Landau theory is only expected to be accurate in the vicinity of Tc . This is easily seen in the order
parameter amplitude, 2 1 T /Tc , which varies linearly in T down to T = 0; whereas the weak-coupling
BCS order parameter saturates at low temperatures. In
confined 3 He, this difference is reflected in the characteristic length scale for variations of the order parameter,
which in GL theory is


(10)

These boundary conditions may be extended by interpolating between the two extremes. In particular, Ambegaokar, de Gennes, and Rainer (AdGR) showed that
diffuse scattering from an atomically rough surface leads
to a GL boundary condition in which the transverse orbital components of the order parameter are finite at
the surface, but extrapolate linearly to zero a distance
bT = 0.540 past the boundary. Thus, we introduce more
general boundary conditions defined by

Az z=0 = 0 ,


1
z Ax z=0 =
Ax z=0
bT


1
z Ay z=0 =
Ay z=0 ,
(11)
bT
where bT = b0T 0 is the extrapolation length. The parameter b0T is allowed to vary from b0T = 0, maximal
pairbreaking, to b0T , minimal pairbreaking. The
film geometry consists of two infinite coplanar surfaces
separated by a distance D with 3 He filling the region
between them. The boundary conditions in Eq. 11 are
imposed at z = D/2.

GL (T ) =

7(3)/20
1 T /Tc

1/2 

~vf
2kB Tc


.

(12)

In weak-coupling BCS theory, the characteristic length

scale is
(T ) =

~vf
,
10BCS
B (T )

(13)

which is significantly larger than GL (T ) at low temperatures. In order to more accurately extrapolate the spatial
variations of the order parameter, as well as the confinement phase diagram, to lower temperatures we rescale
the film of thickness in the GL equations D D(T )
with
D(T ) = D(Tc )

GL (T )
,
(T )

(14)

where D(Tc ) = D is the thickness of the film and D(T )

is a rescaled thickness used within the GL theory calculation. Figure 3 shows the effect of this rescaling on
the weak-coupling GL theory phase diagram for the region of stability of the Stripe phase in comparison to
the Stripe phase region obtained in weak-coupling quasiclassical theory [1]. Rescaling lengths in the GL theory
in terms of (T ) gives a more accurate representation

4
Axx

1.1

Ayy

Axz

0.05

1.0

0.00

0.9

0.05

10

10

x/0

10

z/0

10

10

x/0

Azz

x/0

Azx

10

10

x/0

10

10

x/0

10

0.8

FIG. 4 Stripe phase order parameter for specular surfaces

as functions x and z for D = 120 , p = 3 bar, T = 0.5Tc ,
and calculated period L 23.60 . The amplitudes are
scaled
in units of the bulk B phase order parameter, B =
p
|(T )|/6(12 + 1/3345 ).

of the confinement phase diagram than simple extrapolation of the GL results to low temperature. The deviations
that remain reflect the non-locality of the quasiclassical
theory for inhomogeneous phases for T  Tc .
III. STRIPE PHASE

The Stripe phase spontaneously breaks translational

symmetry in the plane of the film. We assume it does so
along the x axis, leaving the order parameter translationally invariant along the y direction. Broken translational
symmetry leads to a new length scale, L, which is the
half-period of the Stripe phase order parameter; L is an
emergent length scale, which varies with temperature,
pressure, film thickness, and the surface boundary condition, and must be determined by numerical minimization
of the GL free energy in parallel with the self-consistent
determination of the order parameter.

A. Order parameter

The Stripe phase is predicted to be stable in superfluid

He films of thickness D 100 [1]. In weak-coupling
theory this phase appears as a second order transition
between the Planar and B phases, and for D . Dc2
130 , corresponds to a periodic array of degenerate Bphase domains separated by domain walls [1].
For broken translational symmetry along the x axis the
residual symmetry of the Stripe phase is defined by the
point group,
3

L
S
L
S
H = {e, cL2x cS2x } {e, xz
xz
}{e, xy
xy
}
L
S
{e, xz
xz
}{e, ei cL2z }T ,

(15)

where cL2x is an orbital space rotation about the x axis,

S
xz
is a spin space reflection about the xz plane, and T
is the operation of time reversal. Based on this residual
symmetry group we can simplify the form of the order

f bulk/fB

f /fB

0.10

f grad/fB

0.85

0.75
10

0.8

0.0

0.95

z/0

z/0

x/0

10

10

x/0

10

0.05

0.00
10

x/0

10

FIG. 5 Energy density of the Stripe phase with specular surfaces for D = 120 , p = 3 bar, T = 0.5Tc , and calculated
period L 23.60 . The energy density f is scaled by the unconfined bulk energy density fB = 12 (T ) B (T )2 < 0, and is
also shown separated into bulk and gradient contributions.

parameter for the Stripe phase to

Axx 0 Axz
A(x, z) = 0 Ayy 0 ,
Azx 0 Azz

(16)

where the remaining five components are functions of x

and z, and are all real due to time reversal symmetry.
The spatial dependences of the self-consistent order
parameter components for the Stripe phase at pressure
p = 3 bar, T /Tc = 0.5, thickness D = 120 with specular
surfaces are shown in Fig. 4. Note that the calculated
half period is L 23.60 , and that the dominant components are the diagonal elements, Axx , Ayy and Azz .
The latter exhibits a domain wall separating degenerate
B-like order parameters with sgn(Azz ) = 1. The pairbreaking of Azz on the boundaries is alleviated by the
large off-diagonal component, Azx , at the junction with
the domain wall. The remaining symmetry allowed amplitude, Axz , clearly exhibits the symmetry with respect
to cL2x cS2x , but is smaller by an order of magnitude.
The stability of the Stripe phase results from a tradeoff between the lowering of the energy at junctions where
the surfaces intersect the domain wall (note the gradient energy in Fig. 5) and the cost in energy, away from
the film surface, due to the suppression of the order parameter along the domain wall. The total condensation
energy density, with separate bulk and gradient energy
densities, is shown in Fig. 5.
B. Variational Model

The magnitude of the half-period of the Stripe phase,

L, is most easily determined using a variational form of
the order parameter; L is a minimum at the Stripe-Planar
transition and diverges at the Stripe-B transition. At the
Stripe-Planar transition, and for specular boundaries, L
may be derived from the variational order parameter,

xx 0
0
A(x, z) = 0 yy 0 ,
(17)
Azx 0 Azz
where Azx = zx cos(x/L) sin(z/D) and Azz =
zz sin(x/L) cos(z/D). At the Stripe-Planar transi-

yy = xx ,

zx  xx ,

and

zz  xx .

(18)

After spatially averaging and dropping terms greater

than second order in zx and zz the resulting GL functional reduces to,

Q/Q0

tion we assume that

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.2

0.0
0.0

0.4
0
3
6
9

bar
bar
bar
bar
0.2

D = 11 0
0.4

0.6

0.8

D = 11.5 0

0.2

1.0

0.0
0.0

T /Tc

0.2

0.4

0.6

0.8

1.0

T /Tc

K23 zx zz
(19)
Fvar = 22xx + 4P 4xx

 2DL 2

K123 D + K1 L2

+ P 2xx + 2
+ 2zx
4
4D2 L2




K1 D2 + K123 L2
2
2
2
+ zz
,
+ 12 xx +
4
4D2 L2
where ijk... = i +j +k +..., Kijk... = Ki +Kj +Kk +...
and P = 12 + 1/2345 determines bulk free energy of
the Planar phase. Minimizing Fvar with respect to 2xx
gives,
2xx =

||
2 12
2
zx zz
.
2P
8
8P

(20)

The reduced free energy functional then simplifies to

2 K23 zx zz
2
(21)

4P
2DL
 

K123 D2 + K1 L2
+ 2zx 2
4D2 L2
 



P 12
K1 D2 + K123 L2
+ 2zz
.
+ 2
4P
4D2 L2

Fvar =

The last three terms in Eq. 21 determine when nonzero

values of zx and zz are favorable and the StripePlanar instability occurs. At the instability



2 P
zx
(T ) = 2 2
2DLK23
(22)
D L (P 12 )
zz
)



zx 2
2
2
2
2
+ D K123 + L K1
+(D K1 + L K123 ) .
zz
Minimizing Fvar with respect to the ratio zx /zz gives
DLK23
zx
= 2
.
zz
D K123 + L2 K1

(23)

Combining Eq. 23 with Eq. 22 yields the Planar-Stripe

instability temperature, TPS , as a function of D and L.
Optimizing TPS with respect to the Stripe phase period
yields,
s
K123
L=
D,
(24)
|K23 K1 |
which for weak-coupling
values of K1 , K2 , and K3 , re
duces to L = 3D.

FIG. 6 Temperature and pressure dependence of the

wavenumber Q for film thicknesses D = 11 0 , with no Stripe
to B transition (left panel), and D = 11.5 0 , with a Stripe to
B transition (right panel). The onset of the Stripe transition
is based on the Planar-Stripe instability, i.e. omitting the A
phase.

Although the Planar to Stripe transition is interrupted

by a first-order transition to the A phase, the StripePlanar instability determines the scale of the half period, L, and the temperature region where the Stripe
phase is expected to be stable.
The half-period defines

the wavenumber, Q0 = / 3D, of the single-mode instability at TPS . The wavenumber varies with the film
thickness, D, and temperature. Figure 6 shows the temperature dependence of Q for two values of the film thickness starting from the Planar to Stripe instability at TPS ,
i.e. omitting the A phase. The stability of the A-phase
relative to the Planar phase changes the Stripe instability to a first-order transition at a lower temperature TAS .
For D = 11 0 the stable region of Stripe phase persists
to T = 0, while for D = 11.5 0 there is a Stripe to B
phase transition at a temperature, TSB < TPS . In both
cases the wavenumber decreases (L increases) as T drops
below TPS , with Q 0 (L ) as T TSB . Strong
coupling corrections to the free energy lead to a modest
increase the period of the Stripe phase away from the
Stripe to B transition; however, the transition temperature, TSB , is sensitive to pressure (strong-coupling) as
shown in the right panel of Fig. 6.

IV. STRIPE PHASE STABILITY

The most prominent effect of strong-coupling corrections to the weak-coupling BCS theory in bulk superfluid
3
He is the stability of the A phase above pPCP = 21.22 bar.
In sufficiently thin films, the A phase is energetically stable relative to the B phase even in weak-coupling theory,
and is degenerate with the Planar phase [1; 10] Strongcoupling corrections favor the A phase over the Planar
phase, leading to a stable A-phase in thin films at all pressures. Since the Stripe phase can be understood as a periodic array of degenerate B phase domains separated by
time-reversal invariant domain walls, one expects strongcoupling to favor the A phase near the Planar-Stripe instability line. Indeed the A phase suppresses the Planar

6
6

1.00
5

B
4

T /Tc

p/bar

0.75

0.50
1

D = 700 nm

Sauls & Serene

Choi et al
0
0.0

0.25

0.4

0.6

T /Tc

0.00

0.2

(b

ar
)

3
8

12

16

20

D/0

FIG. 7 Pressure-temperature-confinement phase diagram for

the film with minimal pairbreaking boundaries and experimental strong-coupling corrections. The A phase is stable
everywhere not excluded by the Stripe and B phases.

to Stripe instability temperature. However, the Stripe

phase is found to be stable over a wide range of temperatures and pressures.
Figure 7 shows the phase diagram for minimal pairbreaking (specular) surfaces at pressures from 0 to 12 bar,
with the Stripe phase onsetting at temperatures above
0.5Tc . The accuracy of the strong-coupling GL theory is expected to diminish at very low temperatures;
therefore we show results for low and intermediate pressures for which the A- to Stripe transition onsets above
0.5Tc . Note that at T = 0 the strong-coupling GL corrections vanish, and the phase boundaries are determined
by weak-coupling theory at T = 0 and thus pressure independent. This is an artefact of the temperature scaling
of the strong-coupling GL parameters. It is known that
there are residual strong-coupling corrections at the few
percent level in the limit T = 0 [11].
A striking difference between the two sets of strongcoupling parameters shown in Figs. 1 and 2 is evident
at low pressures. The isc from Choi et al. [12] are nonmonotonic between p = 0 and p = 12 bar, which leads
to maximal stability of the Stripe phase at p 3 bar. In
contrast the theoretically calculated strong-coupling corrections are monotonic functions of pressure and predict
maximal stability of the Stripe phase at p = 0 bar and
decreasing stability with increasing pressure.

A. Pressure-Temperature Phase Diagram

Although a number of experiments have been reported

on superfluid 3 He in planar geometries, of particular interest are those involving slabs of thickness D 700 nm
and D 1080 nm, which are in the range of confine-

0.8

1.0

0
0.0

0.2

0.4

0.6

0.8

1.0

T /Tc

FIG. 8 Pressure-temperature phase diagram for a film of

thickness D = 700 nm with minimal pairbreaking (specular)
boundary conditions. The A phase is stable everywhere in
the white region below the bulk transition temperature. The
larger yellow circles are data for the A-B transition based on
NMR from Levitin et al obtained with 4 He preplating [2].

ment where the Stripe phase is expected to be stable.

Levitin et al. [2] (RHUL group) used transverse NMR
frequency shifts to determine transition temperatures in
these cells. They did not find NMR evidence of the
Stripe phase. These experiments were done both with
and without preplating the surfaces of the slab with 4 He,
the presence of which greatly increases the specularity of
the surface. Without the 4 He present, the RHUL group
reported large suppression of the onset of the superfluid
transition - a suppression larger than that predicted theoretically for maximally pairbreaking retro-reflective surface scattering. The explanation or origin of this anomalous suppression is currently lacking. Thus, we focus on
the measurements done with 4 He preplating, which exhibit minimal Tc suppression, and may be modeled theoretically with minimal pairbreaking boundary conditions
(specular scattering).
Calculations of the phase diagram for D = 700 nm are
shown in Fig. 8. The A phase onsets at the bulk Tc .
There is an A to Stripe transition followed by the Stripe
to B transition. For both sets of strong-coupling parameters, the Stripe phase is predicted to be stable at
low pressures and at experimentally accessible temperatures. Although the stability of the A phase is maximal
with specular boundary conditions, the calculated A-B or
A-S phase transition occurs at significantly higher temperature than that reported by the RHUL group. The
discrepancy is sufficiently large that it is well outside uncertainties in the magnitude of the strong-coupling parameters based bulk A- and B phase free energies. Based
on our calculations accessing the Stripe phase would be
optimal for pressures between p = 1 and p = 1.5 bar.
For the thicker slab geometry, D = 1080 nm, shown in
Figure 9, the Stripe phase is predicted to have a negligible region of stability in the pressure-temperature plane
based on the parameters from Choi et al. [12], and
only a small window of stability at the lowest pressures
based on the theoretically calculated strong-coupling parameters.

7
6

1.0

D = 1080 nm

0.70

0.75

0.80

0.85

T /Tc

0.90

0.95

1.00

0
0.65

bT = 0.54
bT = 0.0

0.6

Choi et al
0
0.65

0.8

Sauls & Serene

S
0.70

0.75

0.80

0.85

0.90

0.95

T /Tc

p/bar

A-Normal Transition

0.4

1.00

T /Tc

FIG. 9 Pressure-temperature phase diagram for a film of

thickness D = 1080 nm with minimal pairbreaking boundaries. The A phase is stable everywhere not excluded by the
Stripe and B phases.

B. Effects of Surface Conditions on the Phase Diagram

We use the variable boundary conditions in Eqs. 11 to

investigate the sensitivity of the Stripe phase to surface
disorder. Figure 10 shows the temperature-confinement
phase diagram at p = 3 bar for maximal (b0T = 0), diffuse
(b0T = 0.54), and minimal (b0T = ) pairbreaking boundary conditions. Maximal stability of the Stripe phase
occurs for minimal pairbreaking, i.e. specular surfaces,
as shown by the blue region of stable Stripe phase. Note
that for diffuse scattering the region of Stripe phase stability does not differ significantly from that for specular
boundary scattering. Conversely, for maximal pairbreaking the Stripe phase exists only in the vicinity of T = 0.
V. NMR SIGNATURES OF THE STRIPE PHASE

Nuclear magnetic resonance (NMR) spectroscopy of

the 3 He order parameter is based on resonance frequency shifts originating from the Cooper pair contribution
R 3 to the nuclear magnetic dipole energy, D =
d r fD [A], which evaluated to leading order in A is
V


fD = gD |T rA|2 + T rAA ,
(25)
where gD = 22 2A /2A is the nuclear dipole coupling,
is the 3 He nuclear gyromagnetic ratio, is the nuclear
magnetic susceptibility of normal 3 He, and A is the A
phase longitudinal NMR resonance frequency. The dipole
energy, of order gD 2A , lifts the degeneracy of relative
rotations of the spin- and orbital state of the Cooper
pairs.
NMR spectroscopy is based on the NMR frequency
shift, = L , resulting from the dipolar torque
acting on the total nuclear magnetization. The shift depends in general on the orientation of the NMR field, H,
the initial tipping angle, , generated by the r.f. pulse,
and particularly the spin- and orbital structure of the order parameter. We use the reduction of Leggetts theory
of NMR in 3 He proposed by Fomin [13], valid for intermediate magnetic fields, A  L  , where L = H

bT =

0.2

S
0.0
8

10

12

14

16

18

20

D/0

FIG. 10 Temperature-confinement phase diagram for films

at p = 3 bar with the Choi et al. strong-coupling corrections.
Results for three boundary conditions are shown: minimal
pairbreaking, b0T (solid); diffuse, b0T = 0.54 (dashed);
and maximal pairbreaking, b0T = 0 (dotted). For diffuse and
maximal pairbreaking, the suppression of the A to Normal
phase transitions are also shown.

is the Larmor frequency [13]. The key approximation is

the first inequality which provides a separation of fast
and slow timescales for the spin dynamics. The second inequality allows us to neglect the deformation of
the order parameter by the Zeeman field. Similarly, for
inhomogeneous states we use the separation of length
scales for spatial variations of the Stripe phase, of order
L D 1 m, both
psmall compared to the dipole coherence length, D gD /K1 20 m. The spin degrees
of freedom of the order parameter cannot vary on length
scales shorter than the dipole coherence length D . Thus,
for L  D the nuclear spin dynamics is determined by
the spatially averaged dipole energy. An exception to
this spatial averaging occurs near the Stripe-B transition
where the period of the Stripe phase diverges. In this
limit the dipolar energy varies on sufficiently long spatial
scales that the spin dynamics is determined by a spatially varying dipolar potential. Combined with Fomins
formulation, the separation in scales for spatial variations
of the orbital and spin components of the order parameter allows us to calculate the nonlinear NMR frequency
shifts for the inhomogeneous phases of the thin film as
described in Ref. [5].
A. Translationally invariant Planar-distorted B phase

For non-equal-spin pairing (non-ESP) states, e.g. the

polar distorted B phase or the Stripe phase, the nuclear
magnetic susceptibility, , is suppressed relative to that
of normal 3 He, N . For all non-ESP phases, including
the Stripe phase, the susceptibility can be expressed as
B =

N
.
1 + 2 gz /N (h2zx i + h2zz i)

(26)

8
For a non-ESP superfluid phase of a 3 He film with the
~
magnetic field H||z,
for both the B and Stripe phases,
there are two possible dipole orientations corresponding
to different local minima in the dipole energy [14]. The
first orientation is a minimum of the dipole energy and

has positive frequency shift, which following Levitin et

al. we denote as the B+ state in the case of the translationally invariant B phase. The frequency shift for the
B+ state is obtained as


2

2 
hA2xx i hAxx Azz i2
hAxx Azz i2
2

+
2

Azz cos ,
2
2
+
hAxx i
hAxx i
=
gD

A2  hA A i 2
(A + A )2  cos ,
B
xx

xx

where h...i = (1/V )

and

R
V

d3 R . . . denotes spatial averaging,

!
hAxx Azz i 2 A2xx
hAxx Azz i + hA2xx i

1
cos =
2

zz

(28)

is the critical angle.

of the Planar-distored

Axial
 symmetry

B phase implies
A2yy = A2xx ; thus, only A2xx , A2zz , and hAxx Azz i
are non-zero. This NMR resonance is analogous to the
Brinkman-Smith mode in bulk 3 He-B, but with a positive frequency shift at small tipping angle and a shifted
critical angle.
The translationally invariant, but meta-stable, B
state, corresponds to a minimum of the dipole energy,
and has a frequency shift given by



2 

2

=
gD Axx + 2 Azz cos . (29)
B
This mode has a negative frequency shift at small tipping
angles and, unlike the B+ state, has no critical angle, and

1 2
=
gD
2 B
+

(Ayy + Axx )

1 2
=
gD
2 B

(Ayy Axx )

zz

(27)

cos < cos ,

therefore no deviation from cosine tipping angle dependence. The tipping angle dependences of both Planardistorted B phase states are shown in Fig. 11 plotted as
a function of cos . The positive (negative) shift at small
tipping angle is the signature of the of the B+ (B ) state
in the NMR spectra of the RHUL group [2]. These identifications are confirmed by nonlinear NMR measurements
[15] showing both the pure cosine tipping angle dependence of for the B state, and the kink in the
shift at the critical angle for the B+ state. Note that
for D = 120 at p = 3 bar there is a small slope to the
positive shift for cos > cos .
B. Nonlinear NMR shifts for the S Stripe phases

The breaking of both translational and rotational symmetry in the plane of the film by the Stripe phase leads to
a qualitatively different transverse NMR frequency shift
for the Stripe phase with relative spin-orbit rotation corresponding to a minimum of the dipole energy, i.e. the
S+ state,

(Ayy Axx )

The S+ phase is distinguished with respect to both the

bulk B phase and the Planar-distorted B+ phase by the
absence of a critical tipping angle. This results from spatial averaging over the period of the Stripe phase which
contains equal volumes of Azz > 0 and Azz < 0 giving

xx

cos cos ,

+8

A2zz

A2xz



+ A2zx
cos


.

hAxx Azz i = hAyy Azz i = 0.

By contrast the frequency shift of the metastable S
phase does not differ substantially from that of the B
phase,



(1 + cos ) 2 A2xx + 2 A2yy + 8 A2zz 4 A2xz + A2zx
cos

Note that the constant term

in the shift for the S state
proportional to the average (Ayy Axx )2 is absent for

(30)


. (31)

the B state; however, this constant shift is negligibly

(1010 Hz2 )

0.5

0.5

Stripe
B

0.3

0.3

0.1

0.1

0.1

REFERENCES

0.1

S , B
0.3
1.0

DMR-1508730).

0.5

S+ , B +
0.0

cos

0.5

1.0

0.3
1.0

0.5

0.0

cos

0.5

1.0

FIG. 11 Transverse NMR frequency shifts as a function of

tipping angle at D = 120 , p = 3 bar, and T = 0.5 Tc , with
minimal pairbreaking for the B and S states.

small. Figure 11 shows the comparison between the

translationally invariant B NMR shifts and those for the
corresponding stable and metastable S Stripe phases.
The primary NMR signature of the Stripe phase is the
positive shift with an offset, a finite slope and the absence
of critical angle. This signature clearly differentiates the
S+ phase from the B states and the A phase.
VI. SUMMARY AND OUTLOOK

By formulating a GL theory that incorporates pressure

and temperature dependent strong-coupling corrections,
combined with temperature dependent rescaling of the
confinement length, D, we have greatly expanded the region of applicability of GL theory for calculations of the
properties of confined superfluid 3 He. Strong-coupling
corrections expand the region of stability of the A phase
and decrease the region of stability of the Stripe phase;
however, the Stripe phase remains stable in a large region of pressure, temperature, and confinement. The stability of the Stripe phase is insensitive to diffuse surface
scattering; the phase diagram for specular and fully diffusive scattering predict the Stripe phase to occur in nearly
equivalent regions of the phase diagram. Nonlinear NMR
measurements are probably the best means of detecting
the Stripe phase. The NMR signatures - positive shift
with no critical angle - differentiates the S+ phase from
the B and A phases.
ACKNOWLEDGMENTS

The research of JJW and JAS was supported by the

National Science Foundation (Grants DMR-1106315 and

[1] Crystalline Order in Superfluid 3 He Films, Phys. Rev.

Lett., 98, 045301, (2007), A. B. Vorontsov and J. A.
Sauls.
[2] Phase Diagram of the Topological Superfluid 3 He Confined in a Nano-scale Slab Geometry, Science, 340,
pp. 841844, (2013), L. V. Levitin, R. G. Bennett, A.
Casey, B. Cowan, J. Saunders, D. Drung, T. Schurig,
and J. M. Parpia.
[3] Free Energy of Superfluid 3 He, Phys. Rev. B, 13, 4745,
(1976), D. Rainer and J. W. Serene.
[4] Strong coupling corrections to the Ginzburg-Landau theory of superfluid 3 He, Phys. Rev. B, 75, 174503, (2007),
H. Choi, J. P. Davis, J. Pollanen, T. Haard, and W.
Halperin.
[5] Superfluid phases of 3 He in nanoscale channels, Phys.
Rev. B, 92, 144515, (2015), J. J. Wiman and J. A.
Sauls.
[6] Ginzburg-Landau theory of vortices in superfluid 3 He-B,
Phys. Rev. B, 36, pp. 35833597, (1987), E. Thuneberg.
[7] Potential Scattering Models for the Quasiparticle Interactions in Liquid 3 He, Phys. Rev. B, 24, 183, (1981),
J. A. Sauls and J. W. Serene.
[8] Surface states, Edge Currents, and the Angular Momentum of Chiral p-wave Superfluids, Phys. Rev. B, 84,
214509, (2011), J. A. Sauls.
[9] Landau-Ginzburg Equations for an Anisotropic Superfluid, Phys. Rev. A, 9, 2676, (1975), V. Ambegaokar, P.
de Gennes, and D. Rainer.
[10] Thermodynamic Properties of Thin Films of Superfluid
3
He-A, Phys. Rev. B, 68, 064508, (2003), A. Vorontsov
and J. A. Sauls.
[11] The Quasiclassical Approach to 3 He, Phys. Rep., 101,
221, (1983), J. W. Serene and D. Rainer.
[12] A1 and A2 Transitions in Superfluid 3 He in 98% Porosity
Aerogel, Phys. Rev. Lett., 93, 145302, (2004), H. C.
Choi, A. J. Gray, C. L. Vicente, J. S. Xia, G. Gervais,
W. P. Halperin, N. Mulders, and Y. Lee.
[13] Solution of spin dynamics equations for 3 He superfluid
phases in a strong magnetic field, J. Low Temp. Phys.,
31, pp. 509526, (1978), I. Fomin.
[14] On the Possibility of the Homogeneously Precessing Domain in Bulk 3 He-A, Euro. Phys. Lett., 21, 837, (1993),
Y. M. Bunkov and G. E. Volovik.
[15] Surface-Induced Order Parameter Distortion in Superfluid 3 He-B Measured by Nonlinear NMR, Phys. Rev.
Lett., 111, 235304, (2013), L. V. Levitin, R. G. Bennett,
E. V. Surovtsev, J. M. Parpia, B. Cowan, A. J. Casey,
and J. Saunders.