Вы находитесь на странице: 1из 128

PHENOMENOLOGY OF SUPERSOLIDS

By
CHI-DEUK YOO
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
2009
1
c _ 2009 Chi-Deuk Yoo
2
To my family
3
ACKNOWLEDGMENTS
I am greatly indebted to my advisor, Professor Alan T. Dorsey, for his guidance,
encouragement, and patience he demonstrated throughout my work. Without his support
this work would not have been possible.
I would like to thank Professor P. J. Hirschfeld, Professor Y.-S. Lee, Professor K.
Matchev, Professor M. W. Meisel, Professor S. R. Phillpot, and Professor A. Roitberg
for valuable discussion and support. I also thank Professor M. H. W. Chan of the
Pennsylvania State University and Professor H. Kojima of the Rutgers University for
sharing their valuable experimental data.
Finally I thank my parents, Hae-Seun Yoo and Hee-Sook Yoo, for their unagging
interest, support, and encouragement. Most of all I want to thank my dearest wife, Sae-il,
and children, Dan and Seul, for standing beside me with endless love and trust.
4
TABLE OF CONTENTS
page
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
CHAPTER
1 INTRODUCTION TO SUPERSOLIDS . . . . . . . . . . . . . . . . . . . . . . . 12
1.1 History of Supersolids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2 Torsional Oscillator Experiments on
4
He Solid by Kim and Chan . . . . . . 13
1.3 Recent Theoretical and Experimental Works . . . . . . . . . . . . . . . . . 15
2 VISCOELASTIC SOLIDS: ALTERNATIVE EXPLANATION OF NCRI . . . . 28
2.1 Equation of Motion for the Torsional Oscillator . . . . . . . . . . . . . . . 28
2.2 Properties of Viscoelastic Solids under Oscillatory Motion . . . . . . . . . 30
2.2.1 Innite Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.2 Finite Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.3 Innite Annulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3 Torsional Oscillator with Viscoelastic Solids . . . . . . . . . . . . . . . . . 34
2.4 Possible Connection between Anomalies in Shear Modulus and NCRI . . . 37
3 NON-DISSIPATIVE HYDRODYNAMICS OF A MODEL SUPERSOLID . . . . 49
3.1 Variational Principle in Supersolids . . . . . . . . . . . . . . . . . . . . . . 49
3.1.1 Introduction to the Variational Principle in Continuum Mechanics . 49
3.1.2 Isotropic Supersolids . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.1.3 Anisotropic Supersolids . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.1.4 Quadratic Lagrangian Density of Supersolids . . . . . . . . . . . . . 62
3.2 Collective Modes and the Density-Density Correlation Function . . . . . . 66
4 DISSIPATIVE HYDRODYNAMICS OF A MODEL SUPERSOLID . . . . . . . 72
4.1 Andreev and Lifshitz Hydrodynamics of Supersolids . . . . . . . . . . . . . 73
4.2 Density-Density Correlation Function and its Detection . . . . . . . . . . . 77
4.2.1 Normal Fluids and Superuids . . . . . . . . . . . . . . . . . . . . . 78
4.2.2 Normal Solids and Supersolids . . . . . . . . . . . . . . . . . . . . . 79
5 DYNAMICS OF TOPOLOGICAL DEFECTS IN SUPERSOLIDS . . . . . . . . 85
5.1 Vortex Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.2 Dislocation Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5
6 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
APPENDIX
A CALCULATION OF BACK ACTION TERMS . . . . . . . . . . . . . . . . . . 105
A.1 Innite Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
A.2 Finite Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
A.3 Innite Annulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
B VARIATIONAL PRINCIPLE IN SUPERSOLIDS WITH THE ROTATIONAL
VELOCITY OF SUPER COMPONENTS . . . . . . . . . . . . . . . . . . . . . 111
C STATIC CORRELATION FUNCTIONS OF ISOTROPIC SUPERSOLIDS . . . 114
D KUBO FUNCTIONS AND CORRELATION FUNCTIONS . . . . . . . . . . . 115
E CALCULATION OF THE DENSITY-DENSITY CORRELATION FUNCTION 117
F DERIVATION OF AN EFFECTIVE ACTION FOR EDGE DISLOCATIONS . 119
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
BIOGRAPHICAL SKETCH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6
LIST OF TABLES
Table page
2-1 Fitting parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7
LIST OF FIGURES
Figure page
1-1 Resonant period change in temperature . . . . . . . . . . . . . . . . . . . . . . . 20
1-2 Resonant period change in temperature for various concentration of
3
He
impurities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1-3 Resonant period and the amplitude of oscillation of annular cell (panel A) and
of blocked annular cell (panel B) as a function of temperature . . . . . . . . . . 22
1-4 Annealing eect in the resonant period . . . . . . . . . . . . . . . . . . . . . . . 23
1-5 Annealing eect in the inverse of Q-factor . . . . . . . . . . . . . . . . . . . . . 24
1-6 Specic heat peaks of
4
He solid with dierent concentrations of
3
He impurities . 25
1-7 Shear modulus of
4
He solid as a function of temperature . . . . . . . . . . . . . 26
1-8 Shear modulus change for various concentrations of
3
He impurities . . . . . . . . 27
2-1 Schematic illustration of TO and geometry of a torsion cell . . . . . . . . . . . . 39
2-2 Eective moment of inertia of an innite cylinder of viscoelastic solid as a
function of the driving frequency . . . . . . . . . . . . . . . . . . . . . . . . . 40
2-3 Eective damping coecient of an innite cylinder of viscoelastic solid as a
function of the driving frequency . . . . . . . . . . . . . . . . . . . . . . . . . 40
2-4 Displacement vector in a half cycle for 1/
E
. . . . . . . . . . . . . . . . . . 41
2-5 Displacement vector in a half cycle for = 3/
E
. . . . . . . . . . . . . . . . . . 41
2-6 Displacement vector in a half cycle for = 4/
E
. . . . . . . . . . . . . . . . . . 42
2-7 Eective moment of inertia of a nite cylinder of viscoelastic solid as a function
of the driving frequency with /
E
= 1/100 . . . . . . . . . . . . . . . . . . . 42
2-8 Eective damping coecient of a nite cylinder of viscoelastic solid as a
function of the driving frequency with /
E
= 1/100 . . . . . . . . . . . . . . 43
2-9 Eective moment of inertia of an innite annulus of viscoelastic solid as a
function of the driving frequency with /
E
= 1/1000 . . . . . . . . . . . . . . 43
2-10 Eective damping coecient of an innite annulus of viscoelastic solid as a
function of the driving frequency with /
E
= 1/1000 . . . . . . . . . . . . . . 44
2-11 F(x) of nite cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2-12 F(x) of innite annulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
8
2-13 Resonant period of the blocked capillary sample of BeCu TO . . . . . . . . . . . 45
2-14 Inverse of Q-factor of the blocked capillary sample of BeCu TO . . . . . . . . . 46
2-15 Resonant period of the annealed blocked capillary sample of BeCu TO . . . . . 46
2-16 Inverse of Q-factor of the annealed blocked capillary sample of BeCu TO . . . . 47
2-17 Resonant period of the constant temperature sample of BeCu TO . . . . . . . . 47
2-18 Inverse of Q-factor of the constant temperature sample of BeCu TO . . . . . . . 48
4-1 Brillouin spectrum of liquid argon . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4-2 Brillouin spectra of
4
He superuid . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4-3 Density-density correlation functions of isothermal and isotropic normal solids
and supersolids) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4-4 Splitting of the Rayleigh peak due to the defect diusion mode of a normal
solid into the Brillouin doublet of the second sound modes . . . . . . . . . . . . 84
5-1 Cut for an edge dislocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
9
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulllment of the
Requirements for the Degree of Doctor of Philosophy
PHENOMENOLOGY OF SUPERSOLIDS
By
Chi-Deuk Yoo
May 2009
Chair: Alan T. Dorsey
Major: Physics
We investigate the phenomenological properties of supersolids - materials that
simultaneously display both crystalline order and superuidity. To explain the recent
observation in the torsional oscillator experiments on
4
He solid by Kim and Chan we
adopt a viscoelastic solid model which is characterized by a frequency-dependent complex
shear modulus. In this model, we found that a characteristic time scale which accounts for
dissipation in solids grows rapidly as the temperature is reduced, and results in a decrease
in the resonant period and a peak in the inverse of Q-factor. We also briey discuss the
possible relation between the torsional oscillator results and the anomalous increase of
shear modulus obtained by Day and Beamish.
In a related study, we employ a variational principle together with Galilean covariance
and thermodynamic relations to obtain the non-dissipative hydrodynamics and an
eective Lagrangian density for supersolids. We study the mode structure of supersolids
by calculating the second and fourth sound speeds due to defect propagation. We also
calculate the density-density correlation function of a model supersolid using the
hydrodynamics of Andreev and Lifshitz, and propose a light scattering experiment to
measure the density-density correlation function (which is related to the intensity of
scattered light). We nd that the central Rayleigh peak of the defect diusion mode of a
normal solid in the density-density correlation function splits into an additional Brillouin
doublet due to the longitudinal second sound modes in supersolid phase.
10
Finally, we study the dynamics of vortices and dislocations in supersolids by using
the derived Lagrangian for supersolids. We obtain the eective actions for vortices and
dislocations in two-dimensional isotropic supersolids emphasizing the dierences from the
dynamics in superuids and solids. As a result we obtain the frequency-dependent inertial
masses for slowly moving vortices and dislocations.
11
CHAPTER 1
INTRODUCTION TO SUPERSOLIDS
1.1 History of Supersolids
After Kapitza [1], and Allen and Misener [2] simultaneously discovered the
superuidity of He II at a temperature around 2.17 K in 1938, there were theoretical
speculations about the coexistence of crystalline order and superuidity in matter. In 1956
Penrose and Onsager studied the possibility of such a supersolid phase of matter using the
density matrix formalism, and concluded that supersolids could not exist [3]. However, in
1969 several novel theoretical proposals for a supersolid phase were made. Andreev and
Lifshitz proposed the possibility of a condensation of zero-point defects in a
4
He solid.
Every solid contains defects: vacancies, interstitials, and so on. Classically these defects
are considered to be objects localized at the lattice sites. However, at low temperatures,
due to quantum uctuations, defects in
4
He solid vibrate from the lattice sites and become
mobile. They called these quantum excitations zero-point defectons. They generalized
the two-uid model developed by Landau for superuids to solids with defects, and
obtained a new collective mode (the propagation of defects) at zero temperature [4].
One year later, Chester suggested that a system of interacting bosons can exhibit
both crystalline order and Bose-Einstein condensation at the same time [5]. Also, Leggett
suggested that the non-classical rotational inertia (NCRI) of liquid He II may be
observed in the solid phase. NCRI of He II can be explained by the two uid model.
When a vessel containing He II is rotated, the absence of viscosity of the superuid part
causes only the normal part to be dragged by the wall of container. Thus, the eective
moment of inertia of He II is less than that of normal helium liquid. Leggett predicted an
supersolid fraction
s
/ of 310
4
, and a simple and direct experiment of rotating a solid
was suggested to detect it [6].
Thereafter, Saslow [7] and Liu [8] improved Andreev and Lifshitzs calculation of
hydrodynamic modes. In Ref. [9], the supersolid fraction for a fcc lattice as a function
12
of the ratio between the localization parameter and the lattice constant was obtained.
Fernandez et al. obtained another upper bound for supersolid fraction for the hcp lattice
of 0.3 [10]. Other feasible systems that can exhibit supersolidity were also proposed
theoretically: vortex crystals in type-II superconductors [11], Wigner crystals formed by
excitons in electron-hole bilayers [12], and cold atoms in optical lattices [13].
However, experimental searches for possible signatures of supersolidity of solid
4
He
were not successful prior to 2004. These include sound speed experiments [14], mass ow
experiments [15], and torsional oscillator (TO) experiments [16, 17]. The early experiments
searching for the supersolid phase are summarized in Ref. [18].
1.2 Torsional Oscillator Experiments on
4
He Solid by Kim and Chan
In 2004, Kim and Chan reported two TO experiments that may have shown the
superuid phase of
4
He solid [19, 20]. Both experiments observed drops of resonant period
in the solid phase of
4
He, which might be an indication of the NCRI proposed by Leggett.
The TO used by Kim and Chan consist of a Be-Cu torsion bob and a Be-Cu torsional
rod which also was used to introduce
4
He into the torsion bob. In Ref. [19], the torsion
bob contained a porous medium (Vycor glass), while in Ref. [20] the experiment was
performed with bulk
4
He conned in an annular channel. They used pressures of 62 bar
and 51 bar for
4
He in porous media and bulk
4
He, respectively. They then electrically
drove the oscillator and measured the resonant period at a xed temperature. The
characteristic dependence of resonant period on the temperature is shown in Figs. 1-1
and 1-2A. The drop of resonant period occurred below the critical temperature around
T
c
= 175 mK for in porous media and T
c
= 250 mK for bulk helium.
The resonant period of a TO without dissipation is given by
P = 2
_
I/k, (11)
where I is the moment of inertia of the torsion bob with
4
He, and k is the torsional spring
constant. Using Eq. 11 Kim and Chan interpreted their results of the decrease in the
13
resonant period as a change in the moment of inertia of the
4
He solid, assuming that the
torsion constant k remained constant. This would mean that a part of the mass of
4
He
solid is decoupled from the oscillatory motion, indicating the supersolid phase. The NCRI
fraction (NCRIF) which is dened as the ratio of the superuid density
s
to the total
density of a bose solid, is related to the relative change in the total moment of inertia
[21],

s
(T)

=
I(T
0
) I(T)
I(T
0
) I
empty
=
P(T
0
) P(T)
P(T
0
) P
empty
, (12)
where T
0
is the onset temperature, and I
empty
the moment of inertia of the empty TO. The
largest observed NCRIFs are about 0.5 % and 1.7% for
4
He in Vycor glass and bulk
4
He,
respectively.
Kim and Chan also performed several control experiments to support their
interpretation. First, they investigated on the eect of the critical velocity, and found
that, in both experiments, the drop in period decreases with increasing rim velocity. Kim
and Chan estimated the critical velocity, at which the NCRI disappears, to be 300 m/s
for
4
He in the Vycor glass and 420 m/s for bulk
4
He (Figs. 1-1 and 1-3A). Second, they
repeated the same experiment with solid
3
He, which is a fermionic solid; consequently, no
Bose condensation is possible. With solid
3
He they did not observed any change in the
resonant period. However, an important and interesting feature is found that increasing
the concentration of
3
He impurities in solid
4
He increases the onset temperature and
broadens the change in the resonant period (Fig. 1-2). Third, taking advantage of the cell
geometry of the bulk
4
He sample they inserted a barrier into the annulus channel around
which superow is blocked, and found that the resonant period is signicantly reduced
(Fig. 1-3B). Kim and Chan concluded that this is due to the disruption of superow
around the annulus. Finally, they measured the amplitude, which is related to dissipation,
and observed a broad minimum over the range of temperatures where the resonant
period dropped. The minimum in the amplitude has the same trend as the rim velocity,
Fig. 1-3A.
14
Two years after the initial reports several groups replicated the TO result [2226].
Rittner and Reppy reported rst the annealing eect on TO results [22] and the sample
preparation eect [24]. Both signatures of TO experiments - a drop in the resonant period
and a peak in the inverse of Q-factor - disappeared by annealing the
4
He solid (Figs.
1-4 and 1-5). In Ref. [24] they reported the eects of quenching the sample by cooling
it rapidly, the eect of which is to make a solid of poor quality with a large number of
defects. They reported NCRIFs as high as 20%. Based on their results, Rittner and
Reppy suggested that extended defects such as dislocations and grain boundaries play an
important role in understanding the TO results.
On the other hand, Aoki et al. studied the frequency dependence of the NCRI using
a double torsional oscillator. They added another dummy cell concentrically above the
torsion bob with solid
4
He. This allowed them to investigate the NCRI of the same sample
with two dierent frequencies: the resonant frequency of the out-of-phase mode was a little
more than a twice that of the in-phase mode. They found no frequency dependence to
the onset temperature of NCRI. In addition to this, Aoki et al. found a hysteresis that
depends on the rim velocity: at T = 19 mK the NCRIF increased upon lowering the rim
velocity, but saturated at the maximum value as the rim velocity was again increased.
1.3 Recent Theoretical and Experimental Works
Kim and Chans results have revived both theoretical and experimental interest in
supersolids [27]. Reviews on both recent theoretical and experimental works can be found
in Refs. [28] and [21]. Saslow has suggested that one should use a three-uid model,
instead of the two-uid model of Landau, to correctly describe the supersolid. In his
model the mass density and the mass current contain an additional term consisting of the
lattice velocity and the lattice mass density [29]. Ceperley and Bernu calculated exchange
frequencies in perfect bulk hcp
4
He using the Path Integral Monte Carlo (PIMC) method
and concluded that superuidity would not be observed in a perfect crystal [30]. Prokofev
and Svistunov have found a similar conclusion that zero-point defects were necessary
15
for
4
He solid to be a supersolid by using a coarse-graining procedure [31]. However,
Boninsegni et al. obtained a large activation energy of 13 K for vacancies and of 23 K
for interstitials, and suggested that point defects are unlikely to be present in the low
temperature range of experimental conditions [32].
On the experimental side, several results that are unfavorable to the supersolid
interpretation are reported. Day et al. performed experiments of mass ow through small
capillaries with solid
4
He in Vycor glass [33] and in bulk [34]. One expects a persistent
mass ow in a supersolid; however, they detected no mass ow in either experiments. On
the contrary, a mass ow in solid
4
He was detected by Sasaki et al. [35] and by Ray and
Hallock [36]. Sasaki et al. observed a ow in
4
He solid on the melting curve with grain
boundaries (a poor quality crystal). For a single crystal (good quality crystal) no ow
was detected. Initially they suggested that the ow took place through grain boundaries,
but further experiments showed that mass could ow along the channel between a grain
boundary and a wall [37]. Ray and Hallock injected superuid through one line into a
cell lled with solid
4
He, and detected a change on the other line. They have found a
mass ow in solid
4
He at pressure o the melting curve [36]. Similarly, superuidity in
grain boundaries [38] and in screw dislocations [39] was studied using PIMC simulations.
Pollet et al. found that superuid is formed within grain boundaries in solid
4
He at
temperature around 0.5 K. Boninsegni et al. found superudity along the core of a screw
dislocation in a
4
He solid at zero temperature. On the other hand, there are two neutron
scattering experiments to measure the condensate fraction [40] and to detect changes in
the Debye-Waller factor [41]. Neither experiment showed any evidence for the existence of
a supersolid phase in solid
4
He.
It is well known that the transition of
4
He from the normal uid to superuid is
a second order phase transition accompanied with a -anomaly in the specic heat at
the transition temperature, e.g. see Ref. [42]. Dorsey et al. suggested that there should
be a -anomaly in the specic heat if the supersolid transition is of second order [43].
16
They also found that the -anomaly will be smeared out due to the inhomogeneity of
solid. Therefore, nding a feature (possibly a cusp) in the specic heat would support
the supersolid interpretation of the observed NCRI. Following the TO experiments,
Clark and Chan carried out a measurement of the specic heat in solid
4
He [44]. They
measured down to a temperature of about 80 mK, and did not observed any signature.
In contrast to this null result, Lin et al. reported a peak in the specic heat at about
T = 75 mK [45]. In this second experiment a silicon sample cell was used instead of an
aluminum cell; silicon has a smaller heat capacity and higher thermal conductivity at
low temperatures than aluminum. Thus, they could measure the specic heat down to
a temperature about 30 mK, and they observed deviations from the T
3
Debye specic
heat. Figure 1-6 shows the observed peaks in specic heat of solid
4
He with various
concentrations of
3
He impurities after subtracting the contributions of the empty cell,
phonons, and
3
He impurities. In addition, the constant specic heat term, which might
be due to the mobility of
3
He impurities, was also found in the 10 p.p.m. and 30 p.p.m
samples (the inset of Fig. 1-6). It is found that the height of peaks is about 20 J mol
1
K
1
, and does not depend on concentration of
3
He impurities. Lin et al. estimated the
NCRIF to be about 0.06 % which is comparable to one of their TO results [45]. Finally
they concluded that the observed peaks in specic heat measurements are another possible
signature of the supersolid phase transition, in addition to their TO results.
Another interesting and important experiment on solid
4
He was performed by Day
and Beamish [46]. They measured directly the shear modulus at low frequencies and
amplitudes using two piezoelectric transducers lled with solid
4
He. One transducer was
used to apply a shear stress while the other detects the induced shear deformation. They
found that the shear modulus of solid
4
He increased by about 10% upon lowering the
temperature (Fig. 1-7). Day and Beamish explained their observation using the adsorption
(desorption) of
3
He impurities into (from) a dislocation network. The adsorbed
3
He
impurities pin dislocations at low temperatures, increasing the shear modulus. If
3
He
17
impurities evaporate from dislocations by thermal uctuations, the dislocations become
mobile, reducing the shear modulus. The anomalous increase in the shear modulus of
4
He
solid manifested surprisingly similar behaviors to the TO results, such as the dependence
on the maximum amplitude of applied shear stress and on the concentration of
3
He
impurities. In Fig. 1-8 the reduced changes of the shear modulus for dierent values of
3
He
impurity concentration are shown as a function of temperature, and in comparison with
the reduced NCRI. Moreover, Day and Beamish also studied the resonance in the cavity.
They monitored the resonant frequency and the Q-factor, and found similar behavior to
the TO experiments: the resonant frequency increased as the temperature was lowered,
accompanied with a peak in the inverse Q-factor. The two measurements are very similar,
suggesting that they are closely related. This would mean that dislocations and grain
boundaries present in solids might be responsible for both the shear anomaly and the
NCRI [47].
On the other hand, several alternative explanations for the observed NCRI of
4
He
solid have been proposed as well. Dash and Wettlaufer gave an argument that there exists
a thin layer of liquid helium between the wall and the helium solid, and they showed
that the slippage between them could be responsible for the NCRI [48]. Nussinov at el.
proposed a glass model for the
4
He solid [49]. In their model, solid
4
He was assumed to be
in glassy phase at low temperatures, and they studied its eect on the TO experiments.
Remarkably they could nd the reasonable agreement with the experiment of Ref. [22].
Finally, Huse and Khandker proposed a phenomenological two-uid model for a supersolid
with a temperature dependent coupling constant [50] to explain the TO results.
To summarize, the existence of a supersolid phase in solid
4
He has remained
controversial, both theoretically and experimentally. The explanation for the observed
TO experiments is not complete, opening possibilities ranging from a supersolid transition
to a possibly already known mechanical eect such as dislocation unbinding. In this work
we will focus on
18
nding an alternative model to explain the TO result, such as a viscoelastic solid
model;
deriving the hydrodynamic equations for a supersolid, and obtaining the
hydrodynamic mode and new sound speeds;
proposing another method to detect the supersolid phase. We believe that light
scattering measurements can give some important information about detecting the
second sound modes;
studying the dynamics of vortices and dislocations in a supersolid.
19
Figure 1-1. Resonant period change in temperature. The empty cell data and lm data
are shifted up by 4,260 ns and 3290 ns, respectively. P

= 971, 000 ns.


Reprinted by permission from Macmillan Publishers Ltd: Nature [E. Kim and
M. H. W. Chan, Nature 427, 225 (2004)], copyright (2004).
20
Figure 1-2. Resonant period change in temperature for various concentration of
3
He
impurities. P

= 971, 000 ns. Reprinted by permission from Macmillan


Publishers Ltd: Nature [E. Kim and M. H. W. Chan, Nature 427, 225 (2004)],
copyright (2004).
21
Figure 1-3. Resonant period and the amplitude of oscillation of annular cell (panel A)
and of blocked annular cell (panel B) as a function of temperature.

is the
resonant period at T = 300 mK. From E. Kim and M. H. W. Chan, Science
305, 1941 (2004). Reprinted with permission from AAAS. Copyright (2004) by
AAAS.
22
Figure 1-4. Annealing eect in the resonant period. P

= 5.428053 ms. Reprinted gure 3


with permission from A. S. C. Rittner and J. D. Reppy, Phys. Rev. Lett. 97,
165301 (2006). Copyright (2006) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.97.165301).
23
Figure 1-5. Annealing eect in the inverse of Q-factor. Reprinted gure 4 with permission
from A. S. C. Rittner and J. D. Reppy, Phys. Rev. Lett. 97, 165301 (2006).
Copyright (2006) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.97.165301).
24
Figure 1-6. Specic heat peaks of
4
He solid with dierent concentrations of
3
He
impurities. Open squares are of 10 p.p.m., blue triangles 0.3 p.p.m., and red
circles 1 p.p.b. The inset shows the data of the 10 p.p.m. sample before
subtracting a constant term (dotted line) of 59 J mol
1
K
1
. Reprinted by
permission from Macmillan Publisher Ltd: Nature [X. Lin, A. C. Clark, and
M. H. W. Chan, Nature 449, 1025 (2007)], copyright (2007).
25
Figure 1-7. Shear modulus of
4
He solid as a function of temperature. Data are shifted for
better clarity. Reprinted by permission from Macmillan Publisher Ltd: Nature
[J. Day and J. Beamish, Nature 450, 853 (2007)], copyright (2007).
26
Figure 1-8. Shear modulus change for various concentrations of
3
He impurities. Reprinted
by permission from Macmillan Publisher Ltd: Nature [J. Day and J. Beamish,
Nature 450, 853 (2007)], copyright (2007).
27
CHAPTER 2
VISCOELASTIC SOLIDS: ALTERNATIVE EXPLANATION OF NCRI
2.1 Equation of Motion for the Torsional Oscillator
In this section we investigate alternatives to the supersolid explanation for the
observed decrease in the resonant period of TO experiments. We believe that it is
worthwhile to speculate on these possibilities for several reasons. First, the NCRI is
accompanied by an increase in the damping, signaled by a peak in the inverse of quality
factor. A true supersolid transition in three dimensions will not involve any change in
damping. Second, the supersolid interpretation of the TO results is not supported by other
experiments as described in Chapter 1. These include the lack of pressure driven mass
ow, an anomalous increase of shear modulus with its dependence on the concentration
of
3
He impurities similar to NCRIF [46], and annealing eects of NCRI related to sample
preparation [22, 24]. These eects suggest that the elastic properties of solid
4
He must be
well understood because they might be responsible for the NCRI.
At the rst step of this chapter, we start describing the dynamics of an empty TO.
Let us assume that the torsion bob is completely rigid. The rigid body motion of the
torsion bob results in a constant moment of inertia I
osc
. The torsion rod is assumed to
be massless with a spring constant k that provides a restoring force proportional to the
angular displacement. Then the mode frequency of the undamped harmonic oscillator is

empty
=
_
k/I
osc
. Let M
ext
(t) be the external torque applied on the torsion bob producing
angular displacements (t) from the equilibrium position. Since the motion of the TO is
along the axis of the torsion bob, (t) is sucient to describe the dynamics of the TO. If
the motion is damped, the equation of motion for small of the torsion bob is
_
I
osc
d
2
dt
2
+
osc
d
dt
+ k
_
(t) = M
ext
(t), (21)
where
osc
is the dissipation coecient. With damping the resonant frequencies become
=

2
empty


2
osc
4I
2
osc
. (22)
28
Now lets ll the torsion cell with solid
4
He. If solid
4
He is perfectly rigid, it is
expected that the total moment of inertia of the TO will be increased by the moment of
inertia of solid
4
He (I
He
) with a no-slip boundary condition between the solid
4
He and the
torsion cell. In this case, the mode frequency of undamped harmonic oscillator becomes

0
=
_
k
I
tot
, (23)
where I
tot
= I
osc
+ I
He
. Therefore, the principal eect of loading solid
4
He in the torsion
cell is to simply increase the total moment of inertia by a constant I
He
. However, if we
assume that solid
4
He is not completely rigid, we need to be careful in describing the
dynamics of the TO with solid
4
He. In fact, angular displacements of the torsion bob
induce a shear stress in the solid
4
He, and drag it along into motion. The generated stress
causes elastic shear deformations to propagate with a nite velocity throughout the solid
4
He. The deviation from the rigid body motion results in an eective moment of inertia
which depends upon the driving frequency. Moreover, damping processes present in every
solid produce an eective frequency-dependent damping coecient.
As discussed in Chapter 1, the modication of Eq. 21 for a TO with solid
4
He was
made rst by Nussinov et al. [49] by adding the back action torque M(t) due to solid
4
He:
_
I
osc
d
2
dt
2
+
osc
d
dt
+ k
_
(t) = M
ext
(t) M(t). (24)
The motion of solid
4
He aects the dynamics of the TO by exerting back again an torque
M(t) on the torsion bob. Following Nussinov et al. [49] the torque exerted by the solid
4
He on the TO is taken to be related to the angular displacement through a linear
response function as
M(t) =
_
dt

g(t, t

)(t

). (25)
The response function g(t) is referred to as the back action term by Nussinov et al. [49].
Let us assume time translational symmetry so that g(t, t

) depends on t t

. Fourier
29
transforming Eq. 25, we obtain
g() = M()/(). (26)
Additionally, we assume that the response of (t) to the external torque M
ext
(t) is linear
and related by the susceptibility (t); in Fourier space, we have

1
() = I
osc

2
i
osc
+ k g(), (27)
where we have used Eq. 26. The zeros of the susceptibility,
1
( ) = 0, give the resonant
period
P =
2
'[ ]
, (28)
and the quality factor of the TO
Q
1
=
2[ ]
'[ ]
. (29)
Therefore, if the back action g(), and therefore the torque, depend on the temperature
T, the back action term contains all the information of the dynamics of the TO lled with
solid
4
He. In the following section we calculate the back action terms by modeling solid
4
He as a viscoelastic solid.
2.2 Properties of Viscoelastic Solids under Oscillatory Motion
In this section we study the general dynamical properties of a viscoelastic solid under
shear oscillation. A viscoelastic solid is a material that possesses both elastic and viscous
properties: its response to external disturbances becomes liquid-like or solid-like depending
upon the perturbing frequency. For simplicity we only consider isotropic viscoelastic solids
without pressure gradients.
In TO experiments, the shapes of torsion cells vary from a simple cylindrical cell
to complicated ones such as a blocked annulus. In this work we consider, for simplicity,
torsion cells with cylindrical symmetry. In Fig. 2-1 we show the geometry of a nite
cylinder torsion cell of radius R and height h. When a shear stress is applied to a
viscoelastic solid, due to the cylindrical symmetry, the only displacement is in the
30
azimuthal direction (u

), and its magnitude depends on the distances from and along


the axis of oscillation. In this geometry the divergence of the displacement eld vanishes.
For large driving frequencies, viscoelastic solids respond as viscous uids to the shear
stress, and the Navier-Stokes equation for the velocity eld v

(t) is suitable to describe its


dynamics

t
v

=
_

2
r
+
1
r

1
r
2
+
2
z
_
v

, (210)
where is the mass density and is the shear viscosity. By contrast, the elastodynamic
equation for the displacement u

(t) describes the solid-like dynamics for small frequencies

2
t
u

=
_

2
r
+
1
r

1
r
2
+
2
z
_
u

, (211)
where is the shear modulus. Equation 211 predicts the transverse sound speed of
c
T
=
_
/. Identifying v

(t) =
t
u

(t), and combining these two equations together we


have the Voigt model, also called the Kelvin model, of viscoelastic solids [51, 52]

2
t
u

= ( +
t
)
_

2
r
+
1
r

1
r
2
+
2
z
_
u

. (212)
From Eq. 212 we also identify a relaxation time dened as = /. When
1, a viscoelastic solid responds elastically whereas for 1 it responds viscously.
Equation 212 can be generalized to a viscoelastic solid model of a form like Eq. 211 with
a frequency dependent complex shear modulus () with the following properties: for
small
lim
0
() =
0
, (213)
lim
0
[()]

=
0
, (214)
and for large
lim

'[()] =

, (215)
lim

()

= i

. (216)
The shear modulus of the Voigt model of viscoelastic solids is () = i.
31
We study the dynamical properties of viscoelastic solids by solving Eq. 212 for
the displacement eld, given an oscillatory boundary condition. From the obtained
displacement eld we calculate the shear stresses (
r
and/or
z
) on the boundary, and
the torque M(t). Then the eective description of a system in oscillatory motion can be
investigated by analyzing the eective moment of inertia
I
e
() =
1

0
'
_
M(t) exp(it)
_
, (217)
and the eective damping coecient

e
() =
1

_
M(t) exp(it)
_
, (218)
where
0
is the initial angular displacement. The viscoelastic solid model predicts that, for
small ,
e
() vanishes and I
e
() becomes the moment of inertia of rigid body I
RB
. In
the following we present the results of calculation for three dierent geometries: innite
cylinder, nite cylinder and innite annulus (Appendix A for details).
2.2.1 Innite Cylinder
When an innite cylinder of viscoelastic solid of radius R is oscillating with a
frequency , its eective moment of inertia and the eective damping coecient are
I
e
() = I
RB
'
_
H
_

E

1 i
_
_
, (219)

e
() = I
RB

_
H
_

E

1 i
_
_
, (220)
where I
RB
= R
4
h/2,
H(x) =
4J
2
(x)
xJ
1
(x)
, (221)
and a characteristic time
E
= R/c
T
. In Figs. 2-2 and 2-3 we show the eective moment
of inertia and the eective damping coecients, respectively. There are elastic resonances
that appear as peaks in the damping coecients and an eective moment of inertia that
increases and then decreases to a negative value. To understand the negative eective
32
moment of inertia consider the displacement eld with = 0,
u

= R
0
J
1
(r/c
T
)
J
1
(
E
)
exp(it). (222)
The resonant frequencies are given by the zeros of J
1
(
E
) in Eq. 222. In Fig. 2-4 we
show how the normalized displacement vector evolves during a half cycle when 1/
E
.
In such a limit, the motion becomes purely like a rigid body motion (straight lines in Fig.
2-4). Increasing from 0, the deviation of the amplitude of displacement eld from the
rigid body motion becomes apparent, developing an eectively larger out-of-phase motion
with the applied shear stress (Fig. 2-5). As a result, elastic solids have an eectively
larger moment of inertia than the rigid body. When passes through the rst resonant
frequency, the direction of the displacement eld changes. Figure 2-5 is a schematic
showing the in-phase motion, thus the apparent moment of inertia becomes negative. We
estimate the rst resonance frequency
1
3.83c
T
/R = 0.2 MHz for c
T
300 m/s and
R 0.5 cm. In experimental conditions, driving frequencies ( 1 KHz) are well below the
rst resonance frequency to observe elastic resonances. The eect of viscosity (for large )
is to smear out the resonances.
2.2.2 Finite Cylinder
The eective moment of inertia for a nite cylinder of viscoelastic solid of radius R
and height h is
I
cyl
e
()
I
RB
=
8

m=1
'
_
_
1
(2m1)
2


2
R
2
h
2

2
m
_
H(i
m
)
_
+
8R
2
h
2

m=1
'
_
1

2
m
_
, (223)
where
m
=
_
(2m1)
2

2
R
2
/h
2

2
E
/(1 i) and I
RB
= R
4
h/2. On the other
hand, its eective damping coecient is

n cyl
e
()
I
RB
=
8

m=1

_
_
1
(2m1)
2


2
R
2
h
2

2
m
_
H(i
m
)
_
+
8R
2
h
2

m=1

_
1

2
m
_
. (224)
Figures 2-7 and 2-8 illustrate the eective moment of inertia and the eective damping
coecient of nite cylinder, respectively. The presence of the top and bottom of the nite
33
cylinder makes the mode structure of the elastic oscillation more complicated because
there is one more degree of freedom along the z-axis. Consequently, the position of the
resonant frequencies changes for dierent dimensions of the cylinder. Increasing the height
h or the radius R results in decreasing the resonant frequency.
2.2.3 Innite Annulus
In the case of an innite annular viscoelastic solid of inner radius R
i
and outer radius
R, the eective moment of inertia is
I
inf ann
e
() = '
_
2hR
2
q
E
_
AJ
2
(q
E
R) + BN
2
(q
E
R)
_
_
'
_
2hR
2
i
q
E
_
AJ
2
(q
E
R
i
) + BN
2
(q
E
R
i
)
_
_
, (225)
and the eective dampting coecient is

inf ann
e
() =
_
2hR
2

q
E
_
AJ
2
(q
E
R) + BN
2
(q
E
R)
_
_

_
2hR
2
i

q
E
_
AJ
2
(q
E
R
i
) + BN
2
(q
E
R
i
)
_
_
, (226)
where I
RB
= h(R
4
R
4
i
)/2 and
A =
R
i
N
1
(q
E
R) RN
1
(q
E
R
i
)
J
1
(q
E
R
i
)N
1
(q
E
R) J
1
(q
E
R)N
1
(q
E
R
i
)
, (227)
B =
R
i
J
1
(q
E
R) RJ
1
(q
E
R
i
)
N
1
(q
E
R
i
)J
1
(q
E
R) N
1
(q
E
R)J
1
(q
E
R
i
)
. (228)
In Figs. 2-9 and 2-10 we show the eective moment of inertia and the eective damping
coecients of the innite annulus of viscoelastic solid, respectively. We observe the same
trend of resonant frequencies as in the previous cases that resonant frequencies decrease as
the gap of the annulus is increased.
2.3 Torsional Oscillator with Viscoelastic Solids
Now we study the TO dynamics by modeling solid
4
He as a viscoelastic solid.
Our approach is to assume that solid
4
He at low temperatures, in the range that TO
34
experiments are done, becomes an isotropic viscoelastic solid which has only two complex
elastic moduli dependent on frequency. This is a simplication: solid
4
He has a hcp
crystalline structure, and has ve independent elastic moduli.
In Section 2.2, we investigated the dynamical response of an isotropic viscoelastic
solid to oscillatory disturbances by studying its eective moment of inertia and eective
damping coecient. Our main conclusion is that both quantities change with the driving
frequency and a relaxation time . Hence, the resonant frequency and the quality factor
of the TO depend on the relaxation time as well.
As explained in Section 2.1, to get the resonant period and the quality factor for each
case, we need to nd the poles of Eq. 27 employing the back action term of a viscoelastic
solid with a given geometry [Appendix A for the calculation of g()]. But under all
experimental conditions, the wavelength of the transverse sound mode is much larger than
the typical dimensions of torsion cells ([q
E
R[, [q
E
h[ 1). In this regime, the back action
terms, Eqs. A8, A29 and A38, can be cast into a single form
g() =
2
I
He
+
R
2

4
I
He
F(h/R)
24(1 i)
, (229)
where a function F(x) is dened such that for an innite cylinder F
inf cyl
= 1, for an
innite annulus
F
inf ann
=
1
R
2
(R R
i
)
2
(R + R
i
)
2
R
2
+ R
2
i
, (230)
and for a nite cylinder
F
n cyl
(x) =
192x
2

m=1
1
(2m1)
4

H
_
(2m1)
x
i
_
, (231)
where

H(x) H(x) 1. In Figs. 2-11 and 2-11 we show F
inf ann
(R/R
i
) and F
n cyl
(h/R),
respectively. Note that 0 F
inf ann
, F
n cyl
1.
Now we are in a position to calculate the resonant period and the inverse of the
quality factor of the TO. With the back action term, Eq. 229, the susceptibility of
35
Eq. 27 reduces to

1
() I
tot

2
i
osc
+ k
R
2

4
I
He
F(h/R)
24(1 i)
. (232)
For simplicity we assume that there is no damping from the oscillator (
osc
= 0). In fact,
this assumption is acceptable because for a high quality TO (Q 10
6
) the damping of the
oscillator (
osc
) can be neglected. Since I
He
/I
osc
10
3
under experimental conditions, the
contribution due to viscoelasticity can be treated as a perturbation from the rigid body
motion, whose resonant frequency is
0
given in Eq. 23. We expand the poles about
0
such that
1
( =
0
+
1
) = 0 with
1

0
. Then we obtain

1
=
R
2
I
He
F(h/R)
48I
tot

2
0
(1 i
0
)
. (233)
In the viscoelastic solid model, the resonant period and the inverse of quality factor of TO
are, using Eqs. 28 and 29,
P P P
0

R
2

0
I
He
F(h/R)
24I
tot
1
1 +
2

2
0
, (234)
and
Q
1

R
2

2
0
I
He
F(h/R)
24I
tot

1 +
2

2
0
, (235)
where P
0
= 2/
0
. Equations 234 and 235 are our central results in this chapter.
To interpret these results, rst notice that when passes through 1/
0
, there would
be a peak in Q
1
and a drop in the resonant period whose sizes are given by
Q
1
max
= P/P
0
=
R
2

2
0
I
He
F(h/R)
48I
tot
. (236)
Therefore, the viscoelastic model predicts that Q
1
max
/(P/P
0
) = 1. However, the TO
experiments showed that the maximum dissipation Q
1
max
is lower than (P/P
0
). The
typical experimental value varies around 0.1 [24, 53] reaching up to 0.65 [22] and down
to 0.01 [54]. This implies that we can only t either the peak in Q
1
max
or the decrease
in P taking the shear modulus as a tting parameter. Other quantities such as the
36
radius and moment of inertia are controllable by the experimental setup. Second, since
the sizes of F
n cyl
and F
inf ann
are less than the unity (Figs. 2-11 and 2-12), the reduction
of dimension of the system from the innite cylinder geometry results in decreased period
shifts and sizes of peak in Q
1
. Consequently, in tting the experimental data, the shear
modulus (tting parameter) of the nite cylinder turns out to be greater than that of
the innite cylinder by a factor of 1/F. Third, following Nussinov et al. [49], we take the
relaxation time as
=
0
exp
_
E
0
k
B
T
_
, (237)
and use
0
and E
0
as tting parameters. As the temperature is lowered below E
0
/k
B
, the
relaxation time becomes larger than 1/
0
, and solid
4
He behaves like a viscous uid. In
Figs. 2-13 - 2-18 we show change in the resonant period and in the inverse of quality factor
measured in Clark et al. [53] and their tting using Eqs. 234 and 235, respectively. We
have chosen to t Q
1
max
using the innite cylinder model. In Table 2-1 we list the shear
modulus, the activation energy E
0
, and
0
used in tting. As mentioned earlier, the change
in the resonant period of the viscoelastic model only accounts for about 10% of what is
actually observed. This might imply that the unexplained part is caused by a supersolid
transition. Moreover, it is useful to study the dynamical eect of dislocations on the
shear modulus of normal solids because the motion of dislocations can change the elastic
properties of regular crystals, and determine the characteristic time of the viscoelastic
solid model (e.g., Ref. [55]).
2.4 Possible Connection between Anomalies in Shear Modulus and NCRI
In this section, we now briey discuss a possible relation between TO experiments and
shear modulus experiments. In Section 2.3, we have studied how the viscoelastic model
explains the observed signatures in the resonant period and Q
1
of TO experiments. We
have shown that the relaxation time that decreases exponentially in temperature is
responsible for TO results with a constant shear modulus of solid
4
He. However, Day
and Beamish reported about 10% increase of shear modulus at the same temperature
37
range that TO experiments are performed [46]. We focus only on elasticity because in the
range of the employed frequency to measure the shear modulus the viscosity of solid
4
He
could not be probed simultaneously. As we discussed in Chapter 1, Day and Beamish also
investigated the resonance eect in the cavity where the apparatus to measure the shear
modulus was embedded. They found behaviors that resemble the NCRI. However, the
relation between the dissipation peak observed in the cavity resonance and the viscosity
present in the gap between two transducers is not yet clear.
Let us examine rst the eective moment of inertia of an innite cylinder. For small
, we have I
inf cyl
() I
RB
[1 + (
2
R
2
/24)] without viscosity. The second term is the
correction to the rigid body value due to a nite shear modulus. Since the correction term
in I
inf cyl
() is inversely proportional to , at a xed frequency I
inf cyl
() decreases as ()
increases. Therefore an increase of shear modulus will enhance the change in resonant
period. This could be a connection between the increasing shear modulus observed by Day
and Beamish and NCRI.
We calculate the actual change in the resonant period induced by an increase of shear
modulus. The resonant period of TO for an elastic solid of a nite shear modulus can be
easily obtained by setting to zero in Eq. 234. The result is
P() = P
0
_
1 +
R
2

2
0
I
He
F(h/R)
48I
tot
_
. (238)
One can verify this result nding the zeros of Eq. 232 without and . Lets consider a
small change in shear modulus from
0
. Then the fractional change in P() due to
becomes
P
P(
0
)

R
2

2
0
I
He
F(h/R)
48I
tot

0
. (239)
For Ref. [24], Eq. 239 predicts only 0.810
5
% decrease change in the resonant period
whereas the measured change is about 2.610
3
%. Therefore, we nd that an increasing
shear modulus accompanies a decrease in the resonant period of a TO; however, the elastic
solid model does not account for all the change observed in experiments.
38
Figure 2-1. Schematic illustration of TO and geometry of a torsion cell.
39
-8
-6
-4
-2
0
2
4
6
8
10
0 2 4 6 8 10 12 14 16 18 20
I
e
f
f

/

I
R
B

E
/
E
= 1
/
E
= 5
/
E
= 1/10
/
E
= 1/100
Figure 2-2. Eective moment of inertia of an innite cylinder of viscoelastic solid as a
function of the driving frequency .
0
10
20
30
40
50
60
0 2 4 6 8 10 12 14 16 18 20

e
f
f

E

/

I
R
B

E
/
E
= 1
/
E
= 5
/
E
= 1/10
/
E
= 1/100
Figure 2-3. Eective damping coecient of an innite cylinder of viscoelastic solid as a
function of the driving frequency .
40
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
u
(
r
)

/

R

0
r / R
t = 0
t = / 4
t = / 2
t = 3 / 4
t = /
Figure 2-4. Displacement vector in a half cycle for 1/
E
.
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0 0.2 0.4 0.6 0.8 1
u
(
r
)

/

R

0
r / R
t = 0
t = / 4
t = / 2
t = 3 / 4
t = /
Figure 2-5. Displacement vector in a half cycle for = 3/
E
.
41
-10
-8
-6
-4
-2
0
2
4
6
8
10
0 0.2 0.4 0.6 0.8 1
u
(
r
)

/

R

0
r / R
t = 0
t = / 4
t = / 2
t = 3 / 4
t = /
Figure 2-6. Displacement vector in a half cycle for = 4/
E
.
-4
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10
I
e
f
f

/

I
R
B

E
h / R = 0.5
h / R = 1
h / R = 2
Figure 2-7. Eective moment of inertia of a nite cylinder of viscoelastic solid as a
function of the driving frequency with /
E
= 1/100.
42
0
5
10
15
20
25
30
35
40
45
0 1 2 3 4 5 6 7 8 9 10

e
f
f

E

/

I
R
B

E
h / R = 0.5
h / R = 1
h / R = 2
Figure 2-8. Eective damping coecient of a nite cylinder of viscoelastic solid as a
function of the driving frequency with /
E
= 1/100.
-60
-40
-20
0
20
40
60
0 10 20 30 40 50 60 70 80 90 100
I
e
f
f

/

I
R
B

E
R
i
/ R = 0.9
R
i
/ R = 0.7
R
i
/ R = 0.5
Figure 2-9. Eective moment of inertia of an innite annulus of viscoelastic solid as a
function of the driving frequency with /
E
= 1/1000.
43
0
100
200
300
400
500
600
700
800
900
0 10 20 30 40 50 60 70 80 90 100

e
f
f

E

/

I
R
B

E
R
i
/ R = 0.9
R
i
/ R = 0.7
R
i
/ R = 0.5
Figure 2-10. Eective damping coecient of an innite annulus of viscoelastic solid as a
function of the driving frequency with /
E
= 1/1000.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 2 4 6 8 10 12 14 16 18 20
F
h/R
Figure 2-11. F(x) of nite cylinder.
44
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
F
R
i
/ R
Figure 2-12. F(x) of innite annulus.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
0
5
10
15


P

[
n
s
]
T [K]
Clark et al.
infinite cylinder model
finite cylinder model
Figure 2-13. Resonant period of the blocked capillary sample of BeCu TO. Experimental
data were adapted with permission from A. C. Clark, J. T. West, and M. H.
W. Chan, Phys. Rev. Lett. 99, 135302 (2007). Copyright (2007) by the
American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.99.135302).
45
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2


(
Q

1
)


1
0
6
T [K]
Clark et al.
infinite cylinder model
finite cylinder model
Figure 2-14. Inverse of Q-factor of the blocked capillary sample of BeCu TO.
Experimental data were adapted with permission from A. C. Clark, J. T.
West, and M. H. W. Chan, Phys. Rev. Lett. 99, 135302 (2007). Copyright
(2007) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.99.135302).
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
0
1
2
3
4
5
6
7
8
9
10


P

[
n
s
]
T [K]
Clark et al.
infinite cylinder model
finite cylinder model
Figure 2-15. Resonant period of the annealed blocked capillary sample of BeCu TO.
Experimental data were adapted with permission from A. C. Clark, J. T.
West, and M. H. W. Chan, Phys. Rev. Lett. 99, 135302 (2007). Copyright
(2007) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.99.135302).
46
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
0
1
2
3
4
5
6
7
8
9
10


(
Q

1
)


1
0
7
T [K]
Clark et al.
infinite cylinder model
finite cylinder model
Figure 2-16. Inverse of Q-factor of the annealed blocked capillary sample of BeCu TO.
Experimental data were adapted with permission from A. C. Clark, J. T.
West, and M. H. W. Chan, Phys. Rev. Lett. 99, 135302 (2007). Copyright
(2007) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.99.135302).
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
0
1
2
3
4
5
6


P

[
n
s
]
T [K]
Clark et al.
infinite cylinder model
finite cylinder model
Figure 2-17. Resonant period of the constant temperature sample of BeCu TO.
Experimental data were adapted with permission from A. C. Clark, J. T.
West, and M. H. W. Chan, Phys. Rev. Lett. 99, 135302 (2007). Copyright
(2007) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.99.135302).
47
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
0
0.5
1
1.5
2
2.5
3
3.5
4


(
Q

1
)


1
0
7
T [K]
Clark et al.
infinite cylinder model
finite cylinder model
Figure 2-18. Inverse of Q-factor of the constant temperature sample of BeCu TO.
Experimental data were adapted with permission from A. C. Clark, J. T.
West, and M. H. W. Chan, Phys. Rev. Lett. 99, 135302 (2007). Copyright
(2007) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevLett.99.135302).
Table 2-1. Fitting parameters.
sample [g/cm s
2
]
0
[s] E
0
/k
B
[mK]
blocked capillary 0.34 10
8
2.86 158
annealed blocked capillary 0.75 10
8
0.21 395
constant temperature 1.53 10
8
3.72 166
48
CHAPTER 3
NON-DISSIPATIVE HYDRODYNAMICS OF A MODEL SUPERSOLID
3.1 Variational Principle in Supersolids
In this chapter we investigate the hydrodynamics of a supersolid whose properties
are governed by conservation laws and broken symmetries. As we discussed in Chapter 1,
the hydrodynamic equations of motion for supersolids are rst derived by Andreev and
Lifshitz [4]. In their model point defects, such as interstitials and vacancies, in bosonic
solids undergoes a Bose-Einstein condensation and a crystal becomes a supersolid. The
hydrodynamics of supersolids was studied further by Saslow [7] and Liu [8].
We use the variational principle to derive a Lagrangian for the supersolid and the
non-dissipative hydrodynamic equations of motion. The variational principle was often
used in the literature to obtain the hydrodynamics of various continuum systems: normal
uids [5658], superuids [56, 5963], normal solids [57, 64], liquid crystals [65], and so on.
Let us start this section by giving a plain example of variational principle applied to ideal
uids to show the simplicity of the method.
3.1.1 Introduction to the Variational Principle in Continuum Mechanics
We write a Lagrangian density L for isentropic ideal uids, which are irrotational,
inviscid, and incompressible, in the Eulerian description
1
as follows
L
IF
=
1
2
v
2
U
IF
(), (31)
where is the mass density, v the velocity eld, and U
IF
the internal energy density. The
internal energy density satises the thermodynamic relation
dU
IF
= d, (32)
1
The Eulerian description of the motion of continuum medium is a eld description
with a coordinate system xed in space. In this Eulerian description, the properties of
ow are functions of both space and time. An alternative is the Lagrangian description in
which the motion of individual particles is traced as a function of time.
49
where the chemical potential per unit mass. All the dynamical variables are taken to be
functions of the position x and time t.
The variational principle states that the equations of motion can be derived by
minimizing the action S of a Lagrangian density L
S =
_
dt
_
dx L, (33)
with respect to all the dynamical variables i.e., and v in this example. However,
the variation of the action with the Lagrangian density in Eq. 31 with respect to the
dynamical variables and v does not provide us with the right equations of motion - they
are not independent and restricted by side conditions such as conservation laws. The easy
way to see this is to take the variation of the action of Eq. 31 with respect to v: the
resulting equation of motion is a trivial and irrelevant one (v = 0). In order to overcome
this problem, the side constraints must be included into the Lagrangian density. As one
knows, for uids the total mass is conserved, and this conservation law is expressed in the
continuity equation

t
+
i
(v
i
) = 0. (34)
Since we are considering isentropic ideal uids, the mass conservation is the only
side condition to be taken into account. In fact the momentum is conserved as well.
Nonetheless, the momentum conservation law is not a side condition, but the byproduct
of the variational principle. The continuity equation is incorporated into the Lagrangian
density Eq. 31 by using a Lagrange multiplier :
L
IF
=
1
2
v
2
U
IF
() +
_

t
+
i
(v
i
)
_
. (35)
Then the equations of motion are obtained by taking variations of the action with respect
to and v. With Eq. 35 we obtain
v
i

i
= 0, (36)
50
1
2
v
2

U
IF


D
Dt
= 0, (37)
where D/Dt
t
+ v
i

i
. The other trivial equation of motion is the continuity equation
which one obtains by taking the variation with respect to the Lagrange multiplier .
Equation 36 implies that
v
i
=
i
, (38)
and this conrms that there is no vorticity ( v
s
= 0) for ideal uids, as expected.
We can express Eq. 37 in more familiar form by taking its gradient. Since in thermal
equilibrium the change in the chemical potential per unit mass is related to the change
in the pressure P by the Gibbs-Duhem relation as follows
d = dP, (39)
we obtain

Dv
i
Dt
=
i
P. (310)
Equation 310 is the Euler equation for ideal uids without external forces. Consequently,
we derived the hydrodynamics describing ideal uids using the variational principle: the
continuity equation and the Euler equation.
As we showed in the analysis above, the variational principle provides us with
the Lagrangian density for a continuum system; Eq. 35 is the Lagrangian density for
isentropic ideal uids. Moreover, as long as the boundary contributions are negligible in
the action, the analysis above is equivalent to that with a Lagrangian density
L
IF
=
t

1
2
(
i
)
2
U
IF
(). (311)
We have integrated by parts the action of Eq. 35, and replaced the velocity with
using Eq. 38. This Lagrangian density also can be derived from the time-dependent
Gross-Pitaevskii Lagrangian density for the superuid which is
L
SF
=
i
2
[

t

t

]
1
2m
(i
j
)

(i
j
)


g
2
(

)
2
, (312)
51
where = h/2 with h the Planck constant, and m the mass of the superuid component.
Taking =

ne
i
with the number density n, and the velocity v
i
= (/m)
i
, the
Lagrangian density Eq. 312 can be written as follows:
L
SF
=
i
2

t
n n
t


2
2m
n(
i
)
2


2
8mn
(
i
n)
2
n
g
2
n
2
. (313)
Since the rst term is a total derivative of an analytical eld, it does not contribute to the
dynamics and can be neglected. Hence we can identify that = mn and = (/m);
therefore, we have shown that the time-dependent Gross-Pitaevskii Lagrangian density is
the same as the Lagrangian density for isentropic ideal uids Eq. 311 with a particular
form of the internal potential energy density.
3.1.2 Isotropic Supersolids
Let us now turn on the variational principle applied for supersolids which have
both crystalline order and superuidity. In this part we start with the simple case of an
isotropic supersolid whose Lagrangian density in the Eulerian description is given by
L
SS
=
1
2

s
v
s
2
+
1
2
(
s
)v
n
2
U
SS
(,
s
, s, R
ij
), (314)
where
s
is the density of super-components, the total density, v
s
the velocity of
super-components, v
n
the velocity of normal components, s the entropy density, and
R
ij
R
j
/x
i
, (315)
the deformation tensor with R and x being Lagrangian and Eulerian coordinates,
respectively. The rst two terms in the Lagrangian density are the kinetic energy
densities of the super-component and normal components, and the last term is the
internal potential energy density. Because of the isotropy, the Lagrangian density of
an isotropic supersolid is very similar to the Lagrangian density of superuid, e.g.
the Lagrangian density of a superuid used by Zisel (Eq. 2.1 in Ref. [59]). The only
exception is that U
SS
depends on the deformation tensor. In contrast to uids, the elastic
52
energy which is proportional to the Eulerian strain tensor (
ij
R
ik
R
jk
)/2 arises for
solids. In linear elasticity, the Eulerian strain tensor turns into a more familiar form
(
ij
R
ik
R
jk
)/2 = (
i
u
j
+
j
u
i
)/2
i
u
k

j
u
k
/2 with the displacement vector u.
Given the Lagrangian density, Eq. 314, the total energy density for supersolids is
dened as the sum of the kinetic energy densities and the potential energy density
E
SS
=
1
2

s
v
s
2
+
1
2
(
s
)v
n
2
+ U
SS
(,
s
, s, R
ij
). (316)
This total energy density can be related to the energy density measured in the frame
where the super-component is at rest as follows:
E
SS
=
1
2
v
s
2
+ (
s
)(v
ni
v
si
)v
si
+ . (317)
As is a Galilean invariant, it must depend on Galilean invariant quantities [66]. Similar
to the two-uid model for superuid [66], has a thermodynamic relation [4]
d = Tds + d
ik
dR
ik
+ (v
ni
v
si
)d[(
s
)(v
ni
v
si
)], (318)
with
ik
the stress tensor. We now can obtain the thermodynamic relation for E
SS
by
dierentiating Eq. 317 and replacing Eq. 318 for d. The result is
dE
SS
= Tds
ik
dR
ik
(v
ni
v
si
)v
ni
d
s
+
_
+
1
2
(2v
n
2
2v
ni
v
si
+ v
s
2
)
_
d
+
s
v
si
dv
si
+ (
s
)v
ni
dv
ni
. (319)
This thermodynamic relation was used by Saslow [7] and by Liu [8] in deriving the
hydrodynamics of a supersolid provided that the chemical potential per unit mass is given
by

Saslow, Liu
= v
ni
v
si
+ v
s
2
/2. (320)
Finally from Eq. 316 and Eq. 319, we get the thermodynamic relation for U
SS
:
dU
SS
= Tds +
_
+
1
2
(v
ni
v
si
)
2
_
d
1
2
(v
ni
v
si
)
2
d
s

ik
dR
ik
. (321)
53
Note that the dierence between a superuid and an isotropic supersolid is the dependence
of the potential energy on the deformation tensor R
ij
.
As illustrated in Section 3.1.1, the dynamical variables of a supersolid in the
Lagrangian density, Eq. 314, are not independent each other. The side conditions
relating the dynamical variables must be included in the Lagrangian density in order to
derive the correct equations of motion. There are two important conservation laws for a
supersolid: the mass conservation law and the entropy conservation law. Once again the
momentum conservation law is not a side condition to be imposed, but a consequence of
the variational principle. The conservation of mass is expressed by the continuity equation

t
+
i
_

s
v
si
+ (
s
)v
ni
_
= 0, (322)
with the total current
j
i
=
s
v
si
+ (
s
)v
ni
. (323)
For the entropy conservation we have

t
s +
i
(sv
ni
) = 0. (324)
One should note that in the equation for the conservation of entropy, Eq. 324, only the
velocity of the normal component are involved because the entropy is carried solely by the
normal component. In addition to the two conservation laws above there is one more side
condition, called Lins constraint, to be included [56]:
D
n
R
i
Dt
= 0, (325)
where D
n
/Dt
t
+ v
ni

i
. Lins constraint for solids is an expression of the fact that the
Lagrangian coordinates, (i.e., the initial positions of particles) do not change along the
paths of the normal component. The same condition is also used for an isentropic normal
uid to generate vorticity whereas the entropy conservation equation produces vorticity for
the normal uid [56]. Lets incorporate these constraints into the Lagrangian density using
54
Lagrangian multipliers , and . Then we have
L
SS
=
1
2

s
v
s
2
+
1
2
(
s
)v
n
2
U(,
s
, s, R
ij
) +
_

t
+
i
_

s
v
si
+ (
s
)v
ni
_
_
+
_

t
s +
i
(sv
ni
)
_
+
i
_

t
(sR
i
) +
j
(sR
i
v
nj
)
_
, (326)
where we have used the Lins constraint combined with the conservation equation of
entropy making it in a form of continuity equation. It is also possible to use the continuity
equation, instead of the conservation equation of entropy, combining the Lins constraint.
We are now in a position to get the equations of motion. We take the variations of
the action with respect to all the dynamical variables. We obtain

1
2
v
n
2

U
SS


D
n

Dt
= 0; (327)

s
1
2
v
s
2

1
2
v
n
2

U
SS

s
(v
si
v
ni
)
i
= 0; (328)
s
D
n

Dt
+ R
i
D
n

i
Dt
+
U
SS
s
= 0; (329)
v
si

s
v
si

i
= 0; (330)
v
ni
(
s
)v
ni
(
s
)
i
s
i
sR
j

j
= 0; (331)
R
i
s
D
n

i
Dt

j
_
U
SS
R
ji
_
= 0. (332)
Obviously the variations with respect to the Lagrange multipliers recover the imposed
constraints, Eqs. 322 through 325. In the following we show that the derived equations
of motion can be rearranged to recover the hydrodynamics of a supersolid developed by
Andreev and Lifshitz [4], by Saslow [7], and by Liu [8].
55
We obtain rst the Clebsch potential representation [57] for the velocity of the
super-component, from Eq. 330,
v
si
=
i
, (333)
and the velocity of the normal component, from Eq. 331,
v
ni
v
si
=
s

s
_

i
+ R
j

j
_
. (334)
With these representations we nd rst that
v
s
= 0. (335)
Consequently the velocity of the super-component is irrotational as one expects for a
superuid without vortices: v
s
is only longitudinal. On the other hand, we also nd that
vorticity can be generated for v
n
:
v
n
=
_
s

s
_
+
_
sR
i

s
_

i
. (336)
In fact it is also possible to include systematically the transverse part of v
s
by introducing
a second Lins constraint (Appendix B).
The use of Eq. 333 in Eq. 328 corroborates one of the thermodynamic relations
given in Eq. 321,
U
SS

s
=
1
2
(v
ni
v
si
)
2
. (337)
Taking the gradient of Eq. 327 and using the Clebsh representation of v
s
, Eq. 333, we
get the Josephson equation for v
s
D
s
v
si
Dt
=
i
_
U
SS

i
_
U
SS

s
_
=
i
, (338)
where D
s
/Dt
t
+ v
si

i
. In deriving Eq. 338, we have used Eq. 337 and v
sj

i
v
sj
=
v
sj

j
v
si
because v
s
is irrotational.
56
The Euler equation for v
n
can also be derived. We take D
n
/Dt of Eq. 331, and use
the derived equations of motion to eliminate the Clebsch potentials of the relative velocity.
The resulting equation is the Euler equation for v
n
:
(
s
)
D
n
v
ni
Dt
= (
s
)
i
_
U
SS

_
(
s
)
i
_
U
SS

s
_

j
_
U
SS
R
jk
_
R
ik
s
i
_
U
SS
s
_

1
2
(
s
)
i
(v
nj
v
sj
)
2

s
+
j
(
s
v
sj
)
_
v
si

t
(
s
) +
j
_
(
s
)v
nj
_
_
v
ni
= (
s
)
i

j

jk
R
ik
s
i
T
1
2
(
s
)
i
(v
nj
v
sj
)
2

t
(
s
) +
j
_
(
s
)v
nj
_
_
(v
ni
v
si
), (339)
where we have used an useful identity of the convective derivative,
D(a
i
b)
Dt
=
i
b
Da
Dt
+ a
i
_
Db
Dt
_
a
j
b
i
v
j
. (340)
Finally the Josephson equation and the Euler equation can be put together into the
momentum conservation equation,

t
_

s
v
si
+ (
s
)v
ni
_
=
k
_

s
v
si
v
sk
+ (
s
)v
ni
v
nk

_
Ts
n
(v
nj
v
sj
)
2
_

ik

kj
R
ij
_
,
(341)
because
jk

i
R
jk
+
j

jk

i
R
k
=
j
(
jk
R
ik
). In deriving Eq. 341 we have used the mass
conservation equation and the thermodynamic equation for .
As shown in the example of ideal uids, the Lagrangian density for the isotropic
supersolid can also be reduced in a compact form by using the Clebsh representation of v
s
.
Neglecting the boundary contributions, the Lagrangian density Eq. 326 is equivalent to
57
the following Lagrangian density:
L
SS
=
1
2

s
v
s
2
+
1
2
(
s
)v
n
2
U
SS
(,
s
, s, R
ij
)
t

_

s
v
si
+ (
s
)v
ni
_

s
D
n

Dt
sR
i
D
n

i
Dt
. (342)
The next step is to use Eqs. 329 through 331 to replace v
s
, and . Then we obtain
the Lagrangian density for an isotropic supersolid
L
SS
=
t

1
2
(
i
)
2
+
1
2
(
s
) (v
ni

i
)
2
f(,
s
, T, R
ij
), (343)
where f = U
SS
sT. This Lagrangian density used in the derivation of the hydrodynamics
of a supersolid has many other applications, including the calculation of correlation
functions and collective modes, and the study of the dynamics of dislocations and vortices.
3.1.3 Anisotropic Supersolids
Having completed the variational principle in an isotropic supersolid, let us extend
the analysis to an anisotropic supersolid. Because of the anisotropy, the density of the
super-component becomes a tensor
s
ij
which is another distinction of solids from uids.
Then the Lagrangian density for an anisotropic supersolid in Eulerian description is given
by
L
SS
=
1
2

s
ij
v
si
v
sj
+
1
2
(
ij

s
ij
)v
ni
v
nj
U
SS
(,
s
ij
, s, R
ij
). (344)
The internal potential energy satises the thermodynamic relation
dU
SS
= Tds +
_
+
1
2
(v
ni
v
si
)
2
_
d
ik
dR
ik

1
2
(v
ni
v
si
)(v
nj
v
sj
)d
s
ij
. (345)
The derivation of this thermodynamic relation is identical to the calculation of Eq. 321
for isotropic supersolids.
The constraints to be imposed for the anisotropic supersolid are
conservation of mass

t
+
i
_

s
ij
v
sj
+ (
ij

s
ij
)v
nj
_
= 0; (346)
58
conservation of entropy

t
s +
i
(sv
ni
) = 0; (347)
Lins constraint
D
n
R
i
Dt
= 0. (348)
We incorporate these constraints into the Lagrangian density using Lagrange multipliers ,
and . Then we have
L
SS
=
1
2

s
ij
v
si
v
sj
+
1
2
(
ij

s
ij
)v
ni
v
nj
U
SS
(,
s
ij
, s, R
ij
) +
_

t
s +
i
(sv
ni
)
_
+
_

t
+
i
_

s
ij
v
sj
+ (
ij

s
ij
)v
nj
_
_
+
i
_

t
(sR
i
) +
j
(sR
i
v
nj
)
_
. (349)
In the above Lagrangian density Lins constraint was introduced after combining it with
the equation of the entropy conservation. We calculate the equations of motion as follow

1
2
v
n
2

U
SS


t
v
ni

i
= 0; (350)

s
ij
1
2
v
si
v
sj

1
2
v
ni
v
nj

U
SS

s
ij

1
2
(v
sj
v
nj
)
i

1
2
(v
si
v
ni
)
j
= 0; (351)
s
D
n

Dt
+ R
i
D
n

i
Dt
+
U
SS
s
= 0; (352)
v
si

s
ij
(v
sj

j
) = 0; (353)
v
ni

i
+ R
j

j
=
1
s
(
ij

s
ij
)(v
nj
v
sj
); (354)
R
i
D
n

i
Dt

1
s

j
_
U
SS
R
ji
_
= 0. (355)
Additionally, the variations with respect to the Lagrange multipliers just reproduce the
imposed side conditions, Eqs. 346 through 348. First, the variation with respect to
v
si
leads again to Eq. 333. Next, the derivation of the Josephson equation, the Euler
59
equation for v
n
, and the momentum conservation equation for an anisotropic supersolid is
analogous to that for an isotropic supersolid. The results are: the Josephson equation

t
v
si
=
i

1
2

i
v
s
2
, (356)
the Euler equation
D
n
Dt
_
(
ij

s
ij
)(v
nj
v
sj
)
_
= s
i
_
U
SS
s
_

i
R
j

k
_
U
SS
R
kj
_
(
ij

s
ij
)(v
nj
v
sj
)
k
v
nk
(
jk

s
jk
)(v
nk
v
sk
)
i
v
nj
, (357)
and the equation of the conservation of momentum

t
_
(
ij

s
ij
)v
nj
+
s
ij
v
sj
_

j
_
R
ik

jk
_
+
j
_
v
sj
v
si
+ v
si
p
j
+ v
nj
p
i
_

i
_
Ts (v
nj
v
sj
)p
j
_
= 0,
(358)
where
p
i
(
ij

s
ij
)(v
nj
v
sj
). (359)
Note that this momentum conservation equation agrees with the non-dissipative
momentum current derived by Andreev and Lifshitz (Eq. 12 in Ref. [4]) except for the
nonlinear strain term that they neglected. Moreover, the derived momentum conservation
equation also can be mapped into Eq. 4.16 of Saslow [7] by taking v
s
as a Galilean
velocity, and Eq. 3.40 of Liu [8] with the vanishing super thermal current with the
chemical potential given by Eq. 320.
Similar to the isotropic supersolid case, the equivalent Lagrangian density of
anisotropic supersolids to Eq. 349 is
L
SS
=
t

1
2

s
ij

j
+
1
2
(
ij

s
ij
)v
ni
v
nj
(
ij

s
ij
)v
nj

i
f(,
s
ij
, T, R
ij
), (360)
60
where f U
SS
Ts. There are several works that have derived a Lagrangian density for
supersolid at zero temperature, that employ dierent methods. Son derived the Andreev
and Lifshitz nondissipative hydrodynamics and a Lagrangian density by using Galilean
invariance and symmetry arguments [67]. First, we can also make the connection to the
Lagrangian density derived by Son [67] by replacing v
n
with the displacement vector R
and the strain tensor R
ij
using the inverted Lins constraint,
v
ni
= R
1
ji

t
R
j
, (361)
where R
1
ji
x
i
/R
j
and R
ij
R
1
jk
=
ik
. Josserand et al. used the homogenization method
starting from the time-dependent Gross-Pitaevskii equation [68]. On the other hand, Ye
proposed a Lagrangian density for a supersolid introducing an arbitrary phenomenological
coupling constant between elasticity and superuidity in the Gross-Pitaevskii equation
[69].
At this point it is worthwhile to speculate about the possible connection between the
derived Lagrangian density and the time-dependent Gross-Pitaevskii Lagrangian density
coupled to an elastic eld. In extending the superuid Gross-Pitaevskii equation Eq. 312,
to a supersolid one must require Galilean invariance. The covariant form of the gradient of
the order parameter wave function is
i
i
i
i
m
t
u
j
+ i
k
u
j

k
, (362)
where m is the mass of the bosonic particle. However, the use of the covariant gradient
results in the coupling constant
s
between v
s
and v
n
while from the interaction term
in the Lagrangian density, Eq. 360, the coupling constant is (
ij

s
ij
) which is set
by conservation laws and Galilean invariance. Unfortunately the connection between the
variational principle and the Gross-Pitaevskii Lagrangian density is still unclear.
61
3.1.4 Quadratic Lagrangian Density of Supersolids
The quadratic Lagrangian is interesting because it gives us the linearized
hydrodynamics from which collective modes can be obtained. We study the dynamics
of a supersolid by considering small uctuations from a thermal equilibrium state (or
at T = 0). Unless we specify otherwise, in the remaining part of this chapter thermal
uctuations are excluded. We rst consider small uctuations in densities, denoted with ,
from constant equilibrium values, subscripted by zero, as
=
0
+ , (363)
and

s
ij
=
s
0ij
+
s
ij
. (364)
For the velocity of the super-component, we take a variation in the Clebsch potential so
that we have its non vanishing gradient and the time derivative:
i
and
t
. In addition,
we assume that the Lagrangian coordinates dier from the Euler coordinates by a small
displacement vector eld u:
R = x u. (365)
The deformation tensor then becomes
R
ij
=
ij
w
ij
, (366)
where w
ij

i
u
j
. From the inverted Lins constraint, Eq. 361, we obtain the linear
relation between the velocity of the normal component and the displacement vector,
v
ni
=
t
u
i
. (367)
Therefore, we found that in linear elasticity the velocity of the normal component is given
by the time derivative of the displacement vector. It becomes clear that Lins constraint
is the hydrodynamic equation for the elastic variables that arises from the translational
broken symmetries of solids rather than the conservation of the initial positions.
62
We now expand the Lagrangian density up to second order in small uctuations ,

s
ij
, and u:
L
quad
SS
=
0

t

0ij
w
ij

0

t

1
2

0
(
i
)
2


w
ij

w
ij

1
2

w
ij
()
2
+
1
2

n
0ij
(
t
u
i

i
) (
t
u
j

j
)
1
2

ij
w
lk

w
ij
w
lk
, (368)
where
n
0ij

0
ij

s
0ij
, and we have used
f

= +
1
2
(v
ni
v
si
)
2
, (369)
and
f
w
ij
=
ij
. (370)
In the expansion we have dropped constants which do not contribute to the equations of
motion. Additionally the terms proportional to f/
s
ij
are neglected because they are of
higher order:
f

s
ij
=
U
SS

s
ij
=
1
2
(v
ni
v
si
)
2
0. (371)
As a result, the dependence on
s
ij
is absent in the quadratic expansion of the
Lagrangian density, Eq. 368. Often the rst two terms in Eq. 368 are neglected because
they are total derivatives, and do not contribute to the equations of motion as long as
boundary contributions are not important. This is true as long as topological defects,
such as vortices or dislocations, are not present in the supersolid. In Chapter 5 we will
show that the rst term is responsible for the Magnus force acting on vortices [70] while
the second term gives the Peach-Koehler force on a dislocation [71, 72]. Since we are not
considering any topological defects in this chapter, we neglect these terms for now.
The linearized equations of motion are obtained taking the variations of the action of
Eq. 368 with respect to , , and u
i
:
()
=

w
ij

w
ij

w
ij
(
t
+
0
) (372)
63

t
+
n
0ij

i
(
t
u
j
) +
s
0ij

j
= 0 (373)
u
i

n
0ij

2
t
u
j


ji
w
lk

j
w
lk


w
ji

j

n
0ij

j
= 0 (374)
where we have used the identity
x
y

z
=
z
y

x
x
z

y
. (375)
Note that Eq. 373 is the linearized continuity equation. The other two equations of
motion are equivalent to Eq. 19 of Andreev and Lifshitz [4] when the Josephson equation,
Eq. 356, (
t
=
0
) is used.
Density uctuations in solids are caused by uctuations in either the lattice
displacement or the net defect density dened as [73, 74]

=
i

v
, (376)
where
i
is the density of interstitials and
v
the density of vacancies. The defect density is
conserved as long as surface eects are ignored; vacancies and interstitial are created and
destroyed by pairs in the bulk, but can be created or destroyed individually by migrating
to the surface. Now we can take

as the independent variable instead of . Since


=

w
ij

w
ij
+

w
ij

, (377)
Eq. 372 reduces to
=

w
ij

w
ij
+

w
ij

, (378)
where we have used Eq. 375 and
x
y

0
=
x
y

z
+
x
z

y
z
y

0
. (379)
In the higher order expansion of the Lagrangian density, the terms proportional to the
superuid density uctuation must be considered in Eq. 378. Following Zippelius et al.
64
[73], we can identify

w
ij

=
0

ij
, (380)
and

w
ij
= 1. (381)
The same relation was used by Ostlund et al. when they studied the hydrodynamics of
an anisotropic normal solid [75]. Taking the time derivative of Eq. 378 and using the
continuity equation Eq. 373, we get the linearized conservation equation of the defect
density

=
s
0ij
(
t
w
ij

j
), (382)
with the defect current proportional to the density of the super-component
j

i
=
s
0ij
(
t
u
j

j
). (383)
We found a very important feature: the defect current only arises if superuidity is present
with non-vanishing relative velocity.
Replacing Eqs. 372 and 382, into Eq. 374 and the time derivative of Eq. 373, we
get the following equations of motion

2
t

s
0ij

w
ij

s
0ij

2
t
u
j

s
0ij

w
lk

j
w
lk
= 0, (384)

n
0ij

2
t
u
j

_

w
ji

n
0ij

w
ij
_

ji
w
lk

n
0ij

w
lk

w
ij

lk
_

j
w
lk
= 0.
(385)
In the case where u = 0, we get the dispersion relation of the fourth sound modes, from
Eq. 384,

2
=
s
0ij

w
ij
q
i
q
j
, (386)
which is just the result obtained by Andreev and Lifshitz [4], since (/

)
w
ij
= (/)
w
ij
.
On the other hand, when defect uctuations are absent (

= 0), Eqs. 384 and 385


65
combine to become

2
t
u
i
=

ji
w
lk

j
w
lk

w
lk

i
w
lk

w
ij

j
w
kk
. (387)
Therefore, there are no additional sound modes without defects.
Another interesting case is vanishing the superuid density, i.e., normal solids. In
such a case, we get the same sound speed as Eq. 387 in the case in which the defect
density uctuations vanish because the second term in Eq. 385 does not contribute to the
sound speed, but rather to the diusion. Moreover, the conservation equation for defects,
Eq. 382, implies that
t

= 0 when
s
ij
= 0, which agrees with the dissipationless
description of supersolid: a defect current arises only when dissipation is taken into
account [73]. In Chapter 4 we study the dissipative hydrodynamics and calculate the
defect diusion coecient.
To conclude this section we rewrite the Lagrangian density, Eq. 368, in terms of the
defect uctuation

using Eq. 378:


L
quad
SS
=
0

t

0ij
w
ij
+
0
w
ii

t

n
0ij

t
u
i

t

1
2

s
0ij


w
ij

w
ij
+
0

w
ij
w
ii

1
2

w
ij

1
2

ji
w
lk

w
ij
w
lk
+
0

w
ij

w
ij
w
kk

1
2

2
0

w
ij
w
2
ii
+
1
2

n
0ij

t
u
i

t
u
j
, (388)
where we have introduced = +
0
t. In the following section, we study this Lagrangian
density considering a two-dimensional isotropic supersolid to calculate the hydrodynamic
modes and the density-density correlation function.
3.2 Collective Modes and the Density-Density Correlation Function
In Section 3.1.4 we have taken into account defect density uctuations, and recovered
the fourth sound speed of defects obtained by Andreev and Lifshitz [4]. In this section
we calculate the second sound speed and the density-density correlation function which
contains additional information on the second sound modes.
66
Hydrodynamic modes of a system can be inferred by simply counting the number of
conservation laws and broken symmetries of a system [76]. We start by enumerating the
conservation laws and broken symmetries of a three-dimensional supersolid. Since thermal
uctuations are not considered, the energy conservation law can be omitted; therefore,
there are only conservation laws of mass and three components of momentum. In addition
to these, there are three broken translational symmetries, and one broken gauge symmetry
due to the Bose-Einstein condensation. Thus, we have eight conservation laws and the
broken symmetries, and therefore eight hydrodynamic modes for the three-dimensional
supersolid without thermal uctuations. The corresponding hydrodynamic modes are two
pairs of transverse propagating modes, one pair of longitudinal propagating modes, and
one pair of longitudinal second sound modes. The appearance of these longitudinal second
sound modes is another denite signature of a supersolid, in addition to the -anomaly in
the specic heat [43], the NCRI [6], and so on.
In order to investigate the appearance of the second sound modes in a supersolid we
consider a two-dimensional isotropic supersolid. The reduction by one spatial dimension
results in only one pair of transverse propagating modes; the longitudinal part remains
intact. Note that the indices on the densities are lost because of the isotropy. We have the
following thermodynamic relations:

ji
w
lk

ji

lk
+ (
il

jk
+
ik

jl
), (389)

w
ij

=
ij
, (390)

w
ij
=

w
ij
=
1

2
0

, (391)
where is the isothermal compressibility at constant strain, is a phenomenological
coupling constant between the strain and the density, and

and are the bare Lame
coecients at constant density. Then, the Euclidean Lagrangian (with the imaginary time,
= it) of Eq. 388 by using the above representations of the thermodynamic relations
67
becomes
L
quad
E
=
0
w
ii

i
0
u
ii

+
n
0

u
i

i
+ i

+
1
2

s
0
(
i
)
2
+

u
ii

u
ii

+
1
2
1

2
0

2
+
1
2
_

+
1

4
0

_
u
2
ii
+ u
2
ij
+
1
2

n
0
(

u
i
)
2
, (392)
where u
ij
= (w
ij
+ w
ji
)/2, and we have neglected the linear terms which do not aect
the dynamics because they are the total derivatives of analytical variables. When Fourier
transformed Eq. 392 becomes
L
quad
E
=
1
2
_

(Q
n
) (Q
n
) u
L
(Q
n
)
_
A
_
_
_
_
_
_

(Q
n
)
(Q
n
)
u
L
(Q
n
)
_
_
_
_
_
_
+
1
2
_

n
0

2
n
+ q
2
_
u
T
(Q
n
)u
T
(Q
n
), (393)
where Q
n
(q,
n
) with Matsubara frequencies
n
, u
L
= (q u)/q with q = [q[,
u
T
= u (u
L
/q)q, and
A =
_
_
_
_
_
_
1

2
0


n
iq
_

1

n
q
2

s
0
i
n
q
s
0
iq
_

1

_
i
n
q
s
0

n
0

2
n
+ q
2
_
+
1

2
0

_
_
_
_
_
_
_
, (394)
where

+ 2 . As we discussed earlier, there is one pair of transverse sound modes
whose sound speed is
c
T
=

n
0
. (395)
In addition, we get one pair of longitudinal rst sound modes and another pair of
longitudinal second sound modes. Their sound speeds are
c
2
L
=

2
n
0
+
1
2
0

+
1
2

n
0
+
1

2
_
2

4
s
0

n
0
_

2
0

2
_
, (396)
68
and
c
2
2
=

2
n
0
+
1
2
0


1
2

n
0
+
1

2
_
2

4
s
0

n
0
_

2
0

2
_
. (397)
When the superuid density vanishes, i.e., for a normal solid, the second sound speed
c
2
vanishes. Therefore, the normal solid has only one pair of transverse sound modes and
one pair of longitudinal sound modes whose sound speeds become
c
T
=

0
, (398)
and
c
2
NS
=
1

0
_

+ 2 +
1

_
2, (399)
which agree with the results obtained by Zippelius et al. [73] after identifying

=

Zippelius
+ 2
Zippelius
+ 1/
Zippelius
, =
Zippelius
, = (
Zippelius
+ 1/
Zippelius
)/
0
, and
=
Zippelius
. Moreover, we recover the sound speed of the normal uid by assuming that
the Lame coecients and the coupling constant vanish.
On the other hand, the correlation functions can be easily derived from Eq. 393. We
obtain

(Q
n
)

(Q
n
)) = q
2

s
0

2
n

0
+ ( + 1 2
0
)q
2

A
, (3100)

(Q
n
)u
L
(Q
n
)) = iq

s
0

2
n
+ (1
0
)q
2

A
, (3101)
and
u
L
(Q
n
)u
L
(Q
n
)) =
1

2
0

2
0

2
n
+
s
0
q
2

A
, (3102)
where

A
=
n
0

4
n
+
_
+
n
0
_
1

2
__
q
2

2
n

s
0
_

2
0
_
q
4
. (3103)
69
Since the density uctuation is related to the defect density uctuation and the strain
tensor by Eq. 378, the density-density correlation function becomes
(Q
n
)(Q
n
)) = A
_
1
i
n
c
L
q

1
i
n
+ c
L
q
_
+ B
_
1
i
n
c
2
q

1
i
n
+ c
2
q
_
, (3104)
where
A = q

0

n
0
c
2
L

s
0

2c
L

n
0
(c
2
L
c
2
2
)
, (3105)
B = q

0

n
0
c
2
2

s
0

2c
2

n
0
(c
2
2
c
2
L
)
. (3106)
We can now perform the analytic continuation (i
n
= + i) from (Q
n
)(Q
n
)) to
(Q)(Q)), resulting in
(Q)(Q)) = A
_
1
c
L
q + i

1
+ c
L
q + i
_
+ B
_
1
c
2
q + i

1
+ c
2
q + i
_
,
(3107)
where Q (q, ).
Then, the response function can be obtained by taking the imaginary part of the
density-density correlation function Eq. 3107:

(q, ) = A
_
( c
L
q) ( + c
L
q)
_
B
_
( c
2
q) ( + c
2
q)
_
, (3108)
where we have used the identity
1

i
= P
1

+ i (

) . (3109)
It is easy to show that the response function satises the sum rules: the thermodynamic
sum rule
_

(q, )

(q), (3110)
where

(q) is the static density-density correlation function (Appendix C), and the
f-sum rule
_

(q, ) =
0
q
2
. (3111)
70
We close this chapter with a summary of our main results. Based on the conservation
laws and the broken symmetries present in a supersolid, we derived the Lagrangian density
and the non-dissipative hydrodynamics, which agrees with the work done by Andreev
and Lifshitz [4], using the variational principle. Starting from the derived non-dissipative
hydrodynamics we calculated both the fourth sound speed and the second sound speed.
In addition, we obtained the density-density correlation function of a model supersolid
using the Lagrangian density, and found that each pair of longitudinal propagating modes
produces singularities in the density-density correlation function and a pair of -function
peaks in the response function.
71
CHAPTER 4
DISSIPATIVE HYDRODYNAMICS OF A MODEL SUPERSOLID
In Chapter 3 we obtained the non-dissipative hydrodynamics for a supersolid using
the variational principle, and showed that the equations of motion are those which
Andreev and Lifshitz had derived without dissipation [4]. Using the linearized equations of
motion we found that for a two-dimensional supersolid there are in total six propagating
modes: one pair of transverse sound modes, one pair of longitudinal rst sound modes,
and one pair of longitudinal second sound modes. The longitudinal propagating modes
appear in the density-density correlation function as double delta-function peaks located
at = cq with a sound speed c.
When dissipation is taken into account, the mode structure of the system changes:
due to dissipation some diusive modes appear, e.g., a thermal diusion mode arises
due to viscosity. On the other hand, dissipation damps the sound mode. The dispersion
relation of a propagating mode in the limit of small q is given by
= cq i
D
2
q
2
, (41)
where D is the attenuation coecient.
For a three dimensional normal solid there are ve conservation laws: mass, energy
and three components of momentum. In addition, the translational symmetries are
broken in three dierent directions, and there are eight hydrodynamic modes for a three
dimensional normal solid. The corresponding hydrodynamic modes are four transverse
sound modes, two longitudinal rst sound modes, one thermal diusion mode and one
defect diusion mode [76].
In contrast, as we discussed in Chapter 3 there are nine hydrodynamic modes for
a three-dimensional supersolid because the additional broken gauge symmetry due to
the Bose condensation. Because of this broken gauge symmetry one of the diusion
modes of the normal solid becomes a pair of propagating longitudinal second sound
72
modes. The defect-mediated supersolid proposed by Andreev and Lifshitz suggests
that the defect diusion mode of the normal solid becomes the second sound modes in
the supersolid phase. In this chapter we investigate the mode structure of a supersolid
including dissipation, obtain the sound speeds and their attenuation constants, and
describe a light scattering experiment to observe the conversion of the defect diusion
mode of normal solid into the second sound modes of supersolid.
4.1 Andreev and Lifshitz Hydrodynamics of Supersolids
The hydrodynamics for the supersolid with dissipation was rst derived by Andreev
and Lifshitz [4]. Their derivation starts by identifying nine hydrodynamic variables,
and writing down the conservation equations for the conserved densities and additional
equations for the broken symmetry variables. The mass, the momentum and the energy
are conserved:

t
+
i
j
i
= 0, (42)

t
j
i
+
k

ik
= 0, (43)

t
E +
i
Q
i
= 0, (44)
where
ik
is the stress tensor and Q the energy current. In addition to these conservation
laws, there are equations of motion for the broken symmetry variables. First, from the
broken translational symmetry in the three spatial directions we have

t
u
i
J
i
= 0, (45)
where J is an arbitrary function to be determined. For a perfect solid, i.e., solid without
any defects, the density change is linked to the lattice uctuation ( u). However,
Andreev and Lifshitz pointed out that the density uctuation is independent of the lattice
displacement due to defects; therefore, the density becomes a separate hydrodynamic
variable. Since the gauge symmetry of the Bose-Einstein condensate wave function is
73
broken, we have

t
v
si
+
i
= 0. (46)
The currents involved in the conservation laws and the arbitrary functions J and are to
be determined. In addition to these eight hydrodynamic equations, there is an additional
equation which is the entropy production equation

t
s +
i
_
sv
ni
+
q
i
T
_
=
R
T
, (47)
where q is the heat current and R is the positive denite dissipation function. The
redundancy condition, Eq. 47, combined with the thermodynamics and Galilean
covariance allows us to obtain the constitutive relations that relate currents, J, and
to hydrodynamic variables. The resulting reversible currents are

R
ij
=
_
Ts (v
nk
v
sk
)p
k
_

ij
+ v
sj
v
si
+ v
si
p
j
+ v
nj
p
i

ji

jk
w
ik
, (48)
J
R
k
= v
ni
w
ki
v
nk
, (49)

R
= +
1
2
v
s
2
. (410)
On the other hand, using the positiveness of R we obtain the dissipative currents:
q
i
=
ik

k
T T
ik

lk
, (411)

D
=
i
[p
i
(v
ni
v
si
)]
ki

i
v
nk
, (412)
J
D
k
=
kl

ml
+
kl

l
T, (413)

D
ki
=
kilm

l
v
nm

ki

l
[p
l
(v
nl
v
sl
)] , (414)
where
ik
is the thermal conductivity tensor,
iklm
, , and
ik
are the viscosity tensors,

ik
is the thermodiusion coecient tensor for the defects, and
ik
is the defect diusion
coecient tensor. Note that the dissipative currents are linked to the quantities of
opposite time-reversal properties. The reversible current has the same time-reversal
74
properties as the hydrodynamic variables whereas the dissipative current has opposite
time-reversal properties.
We are now in a position to investigate the mode structure of a supersolid with
dissipation. Since we are interested in the collective motion of defects rather than thermal
diusion, we will neglect thermal uctuations (
ik
=
ik
= 0). Also we consider, for
simplicity, an isotropic two-dimensional supersolid. As we discussed in Chapter 3, the
reduction of dimension for a solid with isotropy results in removing a pair of transverse
propagating modes. Due to isotropy, we have for the viscosity coecients

iklm
=
ik

lm
+
_

il

km
+
im

kl

2
3

ik

lm
_
, (415)

ik
=
ik
, (416)

ik
=
ik
. (417)
Then the hydrodynamic equations for a two-dimensional isotropic supersolid, including the
nonlinear term which was neglected by Andreev and Lifshitz, are

t
+
i
j
i
= 0, (418)

t
j
i
+
k

R
ik

k
v
nk

2
v
ni
+
i

k
_

s
(v
nk
v
sk
)
_
= 0, (419)

t
u
i
J
R
i

k

ki
= 0, (420)

t
v
si
+
i

R
+
i

k
_

s
(v
nk
v
sk
)
_

k
v
nk
= 0. (421)
We consider uctuations from the equilibrium values, Eqs. 363 through 366 and
=

w
ij
+

w
ij

w
ij
=
1

2
0

+ w
ii
, (422)
75

ij
=

ij

w
ij
+

ij
w
lk

w
lk
=
ij
+

ij
w
kk
+ (w
ij
+ w
ji
), (423)
where we have used the representations Eqs. 389 through 391 and the Maxwell relation
(/w
lm
)

= (
lm
/)
w
ij
, and linearized Eqs. 418 through 421. We next divide the
linearized hydrodynamic equations into the transverse and longitudinal parts. First, the
Fourier transformed equations of motion for the transverse part are

n
0

t
v
nT

2
u
T

2
v
nT
= 0, (424)

t
u
T
v
nT

2
u
T
= 0. (425)
Consequently, we get the transverse sound speed obtained in the previous section with the
attenuation constant D
T
= +
n
0
. Second, the longitudinal equations of motion are

t
+
s
0
v
s
+
n
0
v
nL
= 0, (426)

n
0

t
v
nL
+
_

n
0

2
0

n
0

2
u
L

2
s
0

s
2
0

2
v
nL

s
0

2
v
s
= 0, (427)

t
u
L
v
nL

2
u
L
= 0, (428)

t
v
s
+
1

2
0

+
2
u
L

2
v
nL

s
0

2
v
s
= 0, (429)
where
s
0
, and

+ . Sound speeds and the attenuation constants can
be obtained by calculating the dispersion relations after Fourier transforming Eqs. 426
through 429. We Laplace-Fourier transform them instead for later purposes. After
Laplace-Fourier transforming we nd
C(z, q)
_
_
_
_
_
_
_
_
_
(q, z)
v
nL
(q, z)
u
L
(q, z)
v
s
(q, z)
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
(q)
v
nL
(q)
u
L
(q)
v
s
(q),
_
_
_
_
_
_
_
_
_
, (430)
76
where C(z, q) is a 44 matrix given by
_
_
_
_
_
_
_
_
_
iz iq
n
0
0 iq
s
iq
_
1

2
0

n
0
_
iz + q
2 1

n
0
_

2
s
0

s
2
0

_
q
2 1

n
0
(
n
0
) q
2
s
0

n
0

iq 1 iz + q
2
0
iq
1

2
0

q
2
q
2
iz + q
2

s
0

_
_
_
_
_
_
_
_
_
.
(431)
From Eq. 431 we obtain two sound speeds c
L
, Eq. 396 and c
2
, Eq. 397, with two
attenuation constants, respectively,
D
L
=
1

n
0
(c
2
L
c
2
2
)
_
c
2
L
n
1
+ n
2
_
, (432)
D
2
=
1

n
0
(c
2
L
c
2
2
)
_
c
2
2
n
1
+ n
2
_
, (433)
where
n
1


2
s
0
+
n
0
+
s
0
(
n
0

s
0
), (434)
n
2

1

2
0

_
2
0

s
0
(
0
1) +
0

n
0
(
2
0

2
)
+
s
0

+
0

s
0
_

n
0

s
0
+
0
( 2
n
0
)
_

_
.
(435)
Now we can see that when
s
= 0, i.e. a normal solid, the second sound modes disappear
but there is the defect diusion mode with the diusion constant
D
2
=

2
0

0
c
2
L
, (436)
which agrees with the result obtained by Zippelius et al. [73].
4.2 Density-Density Correlation Function and its Detection
Light scattering measures uctuations in the local dielectric constant, which is a
function of the local mass density. The intensity of the scattered light is related to the
density-density correlation function [77]
I(q, ) S(q, ) = (q, ), (q, )) . (437)
77
Thus, light scattering provides a direct measure of the density-density correlation function
which can be calculated from the hydrodynamic equations using the Laplace transform
technique [74, 78]. We start this section by reviewing the light scattering experiments
done on normal uids and on the He II superuid. We then calculate the density-density
correlation function for a model supersolid, and describe a light scattering experiment for
detecting the supersolid transition.
4.2.1 Normal Fluids and Superuids
The density-density correlation function for a normal uid was rst calculated by
Landau and Placzek [79]:
_

(q, )

_
NF
=

P

T
_
1
c
v
c
P
_
D
th
q
2

2
+ (D
th
q
2
)
2
+

P

T
c
v
c
P
c
2
NF
q
4
D
NF
(
2
c
2
NF
q
2
)
2
+ (D
NF
q
2
)
2
+

P

T
_
c
v
c
P
1
_
(
2
c
2
NF
q
2
) q
2
D
th
(
2
c
2
NF
q
2
)
2
+ (D
NF
q
2
)
2
,
(438)
where c
v
is the volume specic heat at constant volume, c
P
the volume specic heat at
constant, the pressure. The thermal diusion constant is
D
th
=

c
P
, (439)
with the thermal conductivity. The sound speed of normal uid is c
2
NF
= (P/)
s
with
an attenuation constant
D
NF
=
1

_
4
3
+
_
+

c
P
_
c
P
c
v
1
_
, (440)
where is the shear viscosity and the bulk viscosity. The rst term in Eq. 438 is
a Lorentzian of width 2D
th
q
2
located at = 0 due to the thermal diusion: for each
diusion mode a Lorentzian peak (the Rayleigh peak) at the center appears. The
remaining two terms are contributions from the sound modes. The second term in
Eq. 438 is a pair of Lorentzians of width D
NF
q
2
located at c
NF
q (the Brillouin doublet).
The last term does not contribute much in the correlation function; however, its existence
78
is important for the sum rules [74, 78]. Figure 4-1 shows the spectrum of liquid argon at
T = 84.97 K measured by Fleury and Boon [80].
The density-density correlation function for a superuid was obtained by Hohenberg
and Martin using the two-uid hydrodynamics [81]. The rst sound contribution to the
density-density correlation function is [81, 82]
_

(q, )

_
1
SF
=
2D
2
1
c
2
NF
q
4
(
2
c
2
NF
q
2
)
2
+ (D
1
q
2
)
2
S
1
(q), (441)
where S
1
(q) = k
B
T
2
K
T
/ with the isothermal compressibility K
T
= 1/(/T)
P
, and
D
1
=
1

_
4
3
+
_
. (442)
The contribution of the second sound is [81, 82]
_

(q, )

_
2
SF
=
2q
2
[D

c
2
2
q
2
+ D
th

2
]
(
2
c
2
2
q
2
)
2
+ (D

+ D
th
)
2
q
4

2
S
2
(q), (443)
where S
2
(q) = (c
P
/c
v
1)S
1
(q) and
D

=

s

n
_

4
+ D
1
_
, (444)
where
1
,
2
,
3
and
4
(=
1
) are the coecients of second viscosity, and is the coecient
of rst viscosity. For temperatures above T

, we have
s
= 0, and Eq. 443 reduces to
the rst term in Eq. 438; therefore, the Rayleigh peak of the thermal diusion mode in
normal uid splits into a Brillouin doublet of the second sound modes. This splitting was
observed in the light scattering experiments by Winterling et al. [83] and by Tarvin et al.
[84] (Fig. 4-2).
4.2.2 Normal Solids and Supersolids
If the supersolid transition is of second order similar to the superuid transition, it
is natural to ask if the same observation of the light scattering repeats in the supersolid
transition. The most eective way to answer this question is to obtain the density-density
correlation function using the hydrodynamics for supersolids, Eq. 430. In order to obtain
79
the density-density correlation function, we calculate rst the density-density Kubo
function (Appendix D) from Eq. 431. The expression for the density-density Kubo
function can be found in Appendix E, Eq. E1. Then the correlation function

(q, )
can be obtained by taking the real part of Eq. E1:
_

(q, )

_
SS
=
iq
4
c
2
L
D
L
I
1
(q)
(
2
c
2
L
q
2
)
2
+ (q
2
D
L
)
2
+
(
2
c
2
L
q
2
)I
3
(q)
(
2
c
2
L
q
2
)
2
+ (q
2
D
L
)
2

iq
4
c
2
2
D
2
I
2
(q)
(
2
c
2
2
q
2
)
2
+ (q
2
D
2
)
2
+
(
2
c
2
2
q
2
)I
4
(q)
(
2
c
2
2
q
2
)
2
+ (q
2
D
2
)
2
, (445)
where I
1
(q), I
2
(q), I
3
(q), and I
4
(q) are given in Appendix E. The rst and third terms in
Eq. 445 give two Brillouin doublets centered at = c
L
q, and = c
2
q with width
D
L
q
2
and D
2
q
2
, respectively. The second and fourth terms in Eq. 445 are negligible near
the Brillouin doublets. The obtained density-density correlation function satises both the
thermodynamic sum rule, Eq. 3110, and f-sum rule, Eq. 3111.
On the other hand, we can show that when dissipation is neglected, the second and
fourth terms in Eq. 445 vanish, and delta functions are obtained from the remaining two
terms by taking the limit D
L
, D
2
0. Therefore, the susceptibility, Eq. 445, reduces to
the non-dissipative density-density correlation function, Eq. 3108.
Moreover, for the normal solid (
s
= 0), the second term in Eq. 445 vanishes so that
there is only one Brillouin doublet due to the longitudinal rst sound modes. At the same
time, the fourth term in Eq. 445 becomes a Lorentzian centered at = 0 which is the
Rayleigh peak due to the defect diusion. Therefore, we conclude that the same splitting
as in the transition from a normal uid to a superuid occurs in the transition from a
normal solid to a supersolid.
We now study qualitatively the density-density correlation function Eq. 445. Let us
set rst = = = 0 for simplicity. Then for small
s
, the density-density correlation
80
function reduces to
c
4
NS

0
D

(q, )


_
1 2
(x 1)
2
x
2

s
0

0
_
q
4

c
2
L
D
L
/c
2
NS
D

(
2

c
2
L
q
2

/c
2
NS
)
2
+ (

q
2

D
L
/D

)
2

_
(x 1)
2
x
2
2
_
(x 1)
3
(x 3)
x
3

x 1
x
2
y
_

s
0

0
_
q
2

(
2

c
2
L
q
2

/c
2
NS
)
(
2

c
2
L
q
2

/c
2
NS
)
2
+ (

q
2

D
L
/D

)
2
+
_
x 1 + 2
(x 1)
2
x
2

s
0

0
_
q
4

c
2
2
D
2
/c
2
NS
D

(
2

c
2
2
q
2

/c
2
NS
)
2
+ (

q
2

D
2
/D

)
2
+
_
(x 1)
2
x
2
2
_
(x 1)
3
(x 3)
x
3

x 1
x
2
y
_

s
0

0
_
q
2

(
2

c
2
2
q
2

/c
2
NS
)
(
2

c
2
2
q
2

/c
2
NS
)
2
+ (

q
2

D
2
/D

)
2
,
(446)
where x =
0
c
2
NS
, y = /
0
D

(D

/c
2
NS
) and q

(D

/c
NS
)q with the
defect diusion constant D

= / and the longitudinal sound speed of normal solid c


NS
,
Eq. 399. We show in Fig. 4-3 the normalized density-density correlation functions of a
normal solid and a supersolid. To do so, we have used the rst sound speed c
NS
= 550
m/s, the density = 0.19048 g/cm
3
, the isothermal compressibility = 0.29615 10
8
cm s
2
/g for solid
4
He [85, 86] at the molar volume 21 cm
3
/mole, the viscosity of
4
He uid
2 10
5
g/cm s, the typical wave number involved in light scattering q
1
= 100 nm, and
= 8 10
11
cm
3
s/g. In Fig. 4-4 we show the splitting of the Rayleigh peak due to the
defect diusion of normal solid into an additional Brillouin doublet of the second sound for
dierent values of supersolid fraction.
Therefore, we conclude that the light scattering on a solid
4
He could provide us with
very rich information about the supersolid phase analogous to the superuid phase. The
detection of the additional Brillouin doublet in the spectrum of scattered light will give
another signature for the existence of the supersolid. However, a detection of the diusion
mode of defects must be preceded by that of the propagating modes. The defect-mediated
supersolid predicted by Andreev and Lifshitz assumes the existence of a suciently large
number of defects. But the small number of thermal defects present in a solid
4
He [87] and
the large activation energies of vacancies and interstitials [32] are problematic in realizing
the proposed Brillouin doublet of the second sound modes.
81
Figure 4-1. Brillouin spectrum of liquid argon. Reprinted gure 3 with permission from P.
A. Fleury and J. P. Boon, Phys. Rev. 186, 244 (1969). Copyright (1969) by
the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRev.186.244).
Figure 4-2. Brillouin spectra of
4
He superuid. Reprinted gure 6 with permission from J.
A. Tarvin, F. Vidal, and T. J. Greytak, Phys. Rev. B 15, 4193 (1977).
Copyright (1977) by the American Physical Society
(http://link.aps.org/doi/10.1103/PhysRevB.15.4193).
82
0.06 0.04 0.02 0 0.02 0.04 0.06
0
200
400
600
800
1000
1200
1400
(D

/ c
NS
2
)
N
o
r
m
a
l
i
z
e
d

D
e
n
s
i
t
y

D
e
n
s
i
t
y

C
o
r
r
e
l
a
t
i
o
n

F
u
n
c
t
i
o
n


normal solid

s
/ = 0.1
Figure 4-3. Density-density correlation functions of isothermal and isotropic normal solids
(dashed line) and supersolids (solid line).
83
0.01 0.008 0.006 0.004 0.002 0 0.002 0.004 0.006 0.008 0.01
0
100
200
300
400
500
600
700
(D

/ c
NS
2
)
N
o
r
m
a
l
i
z
e
d

D
e
n
s
i
t
y

D
e
n
s
i
t
y

C
o
r
r
e
l
a
t
i
o
n

F
u
n
c
t
i
o
n


normal solid

s
/ = 0.01

s
/ = 0.02
Figure 4-4. Splitting of the Rayleigh peak (dashed line) due to the defect diusion mode
of a normal solid into the Brillouin doublet of the second sound modes.
84
CHAPTER 5
DYNAMICS OF TOPOLOGICAL DEFECTS IN SUPERSOLIDS
In Chapter 1 we discussed the growing theoretical and experimental interest in
dislocations and grain boundaries in solids and vortices in superuids, due to the
anomalous behaviors of solid
4
He at low temperatures. In this Chapter we study the
dynamics of vortices and dislocations present in supersolids. Our primary objective in this
chapter is to derive an eective action for vortices and dislocations in supersolids, and
study their properties such as the inertial mass of a topological defect.
We start with the Lagrangian density derived in Chapter 3, Eq. 368, for supersolids
without thermal uctuations:
L
quad
SS
=
t

ij
w
ij

t

1
2
(
i
)
2


w
ij

w
ij

1
2

w
ij
()
2
+
1
2

n
ij
(
t
u
i

i
) (
t
u
j

j
)
1
2

ij
w
lk

w
ij
w
lk
, (51)
where we have shifted = t and omitted, for simplicity, the subscript 0 used to
indicate the constant equilibrium values. The dynamical variables in Eq. 51 are , ,
and u
i
. As we discussed in Chapter 3 the rst two terms generally can be ignored because
they are total derivatives. However, when vortices and dislocations are considered, it is
not possible to simply drop these terms because of singularities due to defects. In fact as
we discuss in Sections 5.1 and 5.2, the rst term is responsible for the Magnus force for
vortices [70] and the second term the Peach-Koehler force for dislocations [71, 72]. Before
we proceed, we integrate out the density uctuations to obtain
L
quad
SS
=
t

ij
w
ij
+
1
2

w
ij
(
t
)
2

1
2

s
ij

j
+
1
2

n
ij

t
u
i

t
u
j

1
2

ij
w
lk

w
ij
w
lk

1
2
_

n
ij
+

w
ij

__

t
u
i

j
+ w
ij

_
, (52)
where we have dropped constants which do not aect on the dynamics, and used
Eqs. 375 and 379. In addition, for the interaction terms between superuidity
and elasticity we distinguished explicitly
t
u
i

i
from w
ij

t
, even though they can
85
be merged into one by integrating by parts (provided that u
i
and are analytic);
nevertheless, when topological defects are taken into account, the integration by parts
is no longer possible. Consequently, when one deals with partial derivatives associated
with vortices and dislocations, extra caution must be taken. It is worthwhile to note that
the density uctuation renormalizes the coupling constant between the elasticity and the
superuidity from
n
ij
in the case of constant density ( = 0) to
n
ij
(/w
ij
)

.
Therefore, when
n
ij
= (/w
ij
)

, the elastic part becomes completely decoupled from


the superuid part.
We consider a two-dimensional isotropic supersolid for simplicity. With the
assumption of isotropy and the representations of Eqs. 389 through 391, the Euclidean
action ( = it) of the Lagrangian Eq. 52 can be written as
S
E
=
1
2
_
d
_
d
2
x
_
2i

+ 2
ij
u
ij
+
2
(

)
2
+
s
(
i
)
2
+
n
(

u
i
)
2
+(

2
)u
2
ii
+ 2 u
2
ij
+ i(
n

2
)(

u
i

i
+ u
ii

)
_
, (53)
which is our starting point to derive the eective action for vortices and dislocations, and
study their dynamics in supersolids.
5.1 Vortex Dynamics
Let us rst consider vortices which can be characterized by the single-valued line
integral of along a closed path enclosing only one vortex of strength (charge) of some
integer e in units of h/m
_
d =
h
m
e, (54)
where m is the mass of particle of supersolid. For two-dimensional supersolids vortices are
point-like particles, and the line integral is taken in the counterclockwise direction around
the axis z perpendicular to the two-dimensional supersolid system. We now can separate
into two parts such that
=
S
+
V
, (55)
86
where
S
is the analytical part and
V
is the singular part due to vortices. Then we have
Eq. 54 along a path enclosing vortices
_
d =
_
d
V
=
h
m

, (56)
which implies

V
=
h
m

z
(2)
(x x

), (57)
where e

= 1 is the charge of the -th vortex located at x

() = [x

(), y

()]. In
addition, the velocity of the super-component can be written as the sum of a longitudinal
part
i

S
and a transverse part
i

V
. Since
2

V
= 0, for
V
we take the ansatz

V
=

=

m

arctan
_
y y

()
x x

()
_
, (58)
and, in the following, we derive an eective action for vortices in supersolids in terms of
vortex coordinates x

(). Lets rst insert Eq. 55 into Eq. 53. Then the action becomes
S
E
= S
1
+ S
2
, (59)
where
S
1
= i
_
d
_
d
2
x

V
, (510)
S
2
=
1
2
_
d
_
d
2
x
_

2
(

S
)
2
+
s
(
i

S
)
2
+
2
(

V
)
2
+
s
(
i

V
)
2
+2
2
(

V
)(

S
) +
n
(

u
i
)
2
+ (

2
)u
2
ii
+ 2 u
2
ij
+2i(
n

2
)

u
i

S
+ i(
n

2
)(

u
i

V
+ u
ii

V
)
_
. (511)
The linear term in u
ij
is dropped in S
E
because there are no dislocations. As shown in
Ref. [70], from S
1
one can derive a transverse force, known as the Magnus force, that
acts on a vortex. Since

i
()/x

i
() with s

i
() = dx

i
()/d being the
vortex velocity, S
1
can be rewritten in terms of the vortex velocity and an eective vector
87
potential,
S
1
= i

_
d
_
d
2
xA

i
[x

()]s

i
(), (512)
where we have dened
A

i
(x

) =
_
d
2
x

(x x

)
x

i
. (513)
Therefore, the -th vortex moves with a velocity s

() in an eective magnetic eld given


by
x
a A

(x

) = he

z/m. The Magnus force [70, 88] acting on the -th vortex is
then given by
F

Magnus
= i
he

m
s

() z, (514)
which shows a motion of vortex perpendicular to the applied force.
Now we eliminate the displacement vector u and the analytic phase
S
in S
2
by using
the equations of motion to express S
2
in terms of
V
. As usual, the equations of motion
can be calculated by taking variations of S
E
with respect to
S
and u. The decomposition
of u into the longitudinal part u
L
and the transverse part u
T
leads us to the equations of
motion:

u
T
+
2
u
T
= i
1
2
(
n

2
)

(
V
), (515)

S
+
s

S
+ i(
n

2
)
i

u
i
=
2

V
), (516)

u
L
+ (
2

2
)
2
u
L
+ i(
n

2
)

S
= i
1
2
(
n

2
)(

V
), (517)
where =

+ 2 . These equations of motion are inhomogeneous dierential equations
for u and
S
with source terms from
V
:

V
for longitudinal wave modes and
i

V
for
transverse wave modes. When
n
=
2
, there is no source term for the displacement
vector as expected because there is no coupling between superuidity and elasticity. In
this case, Eqs. 515 through 517 become ordinary wave equations. The equation of
motion for the transverse part has the same form as Eq. 10 of Ref. [89] in which the
vortex dynamics in a superconductor was studied. In their work, they estimated the size
of the shear deformation caused by a vortex which is moving with a velocity much less
88
than the transverse sound speed. They concluded that the shear deformation generated
by a vortex was negligible. Following Ref. [89] we can also estimate the maximal shear
deformation due to a vortex for solid
4
He:
u
max
T


m
He
(
n

2
)s

n
c
2
T
1.57 10
8
[m
2
/s]
(
n

2
)s

n
c
2
T
. (518)
Taking the phenomenological coupling constant to be zero, and the vortex speed to be
s c
T
300 m/s, the maximal shear deformation u
max
T
5 10
11
m which is negligible
similar to the conclusion obtained in Ref. [89].
We now solve for u
L
,
S
, and u
T
, in Fourier space, resulting in
u
L
(q, ) = q(
n

2
)

2
(
s
/
2
)q
2
2
A
(

V
)(q, ), (519)

S
(q, ) = i

2
+ (
2

2
/2 +
n
2
/2
2

n
)q
2

A
(

V
)(q, ), (520)
u
T
(q, ) =
(
n

2
)

2
+ q
2
(
i

V
)(q, ), (521)
where
A
is given by Eq. 3103, and (
i

V
)(q, ) and (

V
)(q, ) are the temporal
Fourier transforms of
(
i

V
)(q, ) = i
h
m

ik
q
k
q
2

e
iqx

()
, (522)
(

V
)(q, ) = i
h
m

ik
q
k
q
2

i
()e
iqx

()
, (523)
which are easily calculated from Eq. 58. By replacing Eqs. 519 and 521 into Eq. 511,
S
2
reduces to the following form:
S
2
=

s

2
2m
2

,
e

_
d F
0
(x

() x

())
+

2
2m
2

,
e

_
d
_
d

i
()s

i
(

j
G
2
(x

() x

),

; c
L
, c
2
)


2
2m
2

,
e

_
d
_
d

i
()s

j
(

j
G
2
(x

() x

),

; c
L
, c
2
)
+

2
(
n

2
)
2
8m
2

,
e

_
d
_
d

G
1
(x

() x

),

; c
T
), (524)
89
where

i
/x

i
() and

i
/x

i
(

), and we have dened some useful integrals


F
0
(x) =
_
d
2
q
1
q
2
e
iqx
, (525)
G
1
(x, ; c) =
_
d
2
q
1
q
2
e
iqx
_
d
2

2
+ c
2
q
2
e
i
, (526)
G
2
(x, ; c, c) =
_
d
2
q
1
q
2
e
iqx
_
d
2
a
2
+ bq
2
D
e
i
, (527)
with c and c being some sound speeds, and
D =
n
(
2
+ c
2
q
2
)(
2
+ c
2
q
2
). (528)
In Eq. 527, we have D =
A
, a =
n

s
+ (
n

2
)
2
/4 and b =
s
(
2

2
) + (
n

2
)
2

s
/4
2
for S
2
written in Eq. 524. Therefore, we derive the action by evaluating
the integrals dened in Eqs. 525 through 526. First, the term with F
0
in Eq. 524 is
local in time and has a logarithmic divergence [F
0
(x) ln(x)]. However, we show below
that this logarithmic divergence is canceled out with contributions from the other two
terms in Eq. 524. Second, we now evaluate G
1
by separating out the local term, resulting
in
G
1
(x, ; c) =
_
d
2
q
1
q
2
e
iqx
_
d
2
_
1
c
2
q
2

2
+ c
2
q
2
_
e
i
= ()F
0
(x) c
2
F
1
(x, ; c), (529)
where
F
1
(x, ; c) =
_
d
2
qe
iqx
_
d
2
1

2
+ c
2
q
2
e
i
=

c
_
c
2

2
+[x[
2
. (530)
Third, to calculate G
2
we divide it into two terms by expanding in partial fractions: one
with the sound speed c and the other with c
G
2
(x, ; c, c) =
A

n
F
2
(x, ; c) +
B

n
F
2
(x, ; c), (531)
90
where
A
ac
2
b
c
2
c
2
, (532)
B
a c
2
b
c
2
c
2
, (533)
F
2
(x, ; c) =
_
d
2
q
1
q
2
e
iqx
_
d
2
1

2
+ c
2
q
2
e
i
. (534)
The second term in Eq. 524 contains spatial derivatives of G
2
. The rst spatial
derivatives equal minus the Laplacian of G
2
because

k
=

k
. Next, in Ref. [90] it
is shown that the second spatial derivatives with vortex velocities can be converted into
temporal derivatives: from s

i
()s

j
(

j
to

. Hence, we can evaluate G


2
as well as
the third term in Eq. 524 by calculating
2

F
2
and
2
F
2
. In fact, they can be written in
terms of F
0
and F
1
:

2
F
2
(x, ; c) = F
1
(x, ; c), (535)
and

F
2
(x, ; c) = F
0
(x)() + c
2
F
1
(x, ; c). (536)
Using Eqs. 530, 535 and 536, we obtain
S
2
=

2
m
2
_

s
2
+
(
n

2
)
2
8
n

a
2
n
_

,
e

_
d F
0
(x

() x

())
+

2
A
2m
2

,
e

_
d
_
d

_
s

i
()s

i
(

) + c
2
L
_
F
1
(x

() x

),

; c
L
)
+

2
B
2m
2

,
e

_
d
_
d

_
s

i
()s

i
(

) + c
2
2
_
F
1
(x

() x

),

; c
2
)


2
c
2
T
(
n

2
)
2
8m
2

,
e

_
d
_
d

F
1
(x

() x

),

; c
T
), (537)
where
A =
[
n

s
+ (
n

2
)
2
/4]c
2
L

s
(
2

2
) (
n

2
)
2

s
/4
2

c
2
L
c
2
2
, (538)
91
and
B =
[
n

s
+ (
n

2
)
2
/4]c
2
2

s
(
2

2
) (
n

2
)
2

s
/4
2

c
2
2
c
2
L
. (539)
Note that the local term cancels out because a =
s

n
+ (
n

2
)
2
/4, and we nally get
the action for vortices in supersolids
S
E
= S
1


2
(
n

2
)
2
8m
2

n
c
T

_
d
_
d

c
2
T
_
c
2
T
(

)
2
+[x

() x

)[
2
+

2
A
2m
2

n
c
L

_
d
_
d

i
()s

i
(

) + c
2
L
_
c
2
L
(

)
2
+[x

() x

)[
2
+

2
B
2m
2

n
c
2

_
d
_
d

i
()s

i
(

) + c
2
2
_
c
2
2
(

)
2
+[x

() x

)[
2
. (540)
From the derived action we can infer that each wave mode coupling to vortices produces
a non-local Coulomb potential term. In particular, a wave mode perpendicular to the
vortex velocity generates an additional term which is proportional to the product of two
vortex velocities at dierent times. Without coupling, the elastic variable is not related
to the phase variable so that one can integrate out the elastic part in Eq. 511, and the
remaining part of action is of a superuid with vortices. In this case (
n
=
2
), we get
A = 0 and B =
n

s
, and the derived Lagrangian for vortices reduces to
S
E
= S
1
+

2

s
2m
2
c
2

_
d
_
d

i
()s

i
(

) + c
2
2
_
c
2
2
(

)
2
+[x

() x

)[
2
, (541)
which is the result obtained by Eckern and Schmid [90].
We calculate the vortex mass by taking the limit as c
2
(

)
2
[x

i
() x

i
(

)[
2
from the derived action, Eq. 540. Then we obtain a frequency-dependent mass of vortex
in a supersolid
M
vortex
=

2
2m
2
_

2
+
1
4
(
n

2
)
2
_
1

+
1

2

2
_
_
[ + ln()] , (542)
where = 0.5772 . . . is Eulers constant, and is a cut-o introduced to regulate the
divergence in temporal integrals. For the case in which = 0, the inertial mass of a vortex
92
reduces to
M
vortex

2
2m
2
_

2
+

n
2
4
_
1

+
1

_
_
[ + ln()] . (543)
Once again, without the coupling of superuid to the displacement vector we recover the
mass of vortex in superuid, obtained by Eckern and Schimd.
5.2 Dislocation Dynamics
Having completed the dynamics of vortices, let us now consider dislocations and study
their dynamics in supersolids by deriving an eective action. Analogous to vortices in
superuid, dislocations produce singularities breaking the discrete translational symmetries
in lattices: the line integral along a closed path enclosing a dislocation yields
_
du
i
= b
i
, (544)
where b is the Burgers vector. In a complete circulation around a dislocation a Burgers
vector causes a mismatch to the circulation in an ideal lattice by an amount of b. There
are two types of dislocation lines: edge dislocations and screw dislocations. If l is the
tangent unit vector to the dislocation line, for edge dislocations l b while for screw
dislocations l | b. Since we work with two-dimensional supersolids, there are only edge
dislocations because screw dislocations produce a displacement perpendicular to the plane
of the system.
The derivation of the eective action for dislocations is analogous to the calculation
done for vortices in Section 5.1. First we take into account dislocations by introducing
explicitly the non-analytical part due to dislocations in the displacement vector:
u = u
S
+u
D
, (545)
where u
S
is analytic and u
D
is singular. Consequently, the contour integral for a system of
dislocations results in
_
du
i
=
_
du
D
i
=

i
, (546)
93
where the summation is over all the enclosed dislocations. Thus we take the ansatz for u
D
u
D
i
=

i
2
arctan
_
y y

()
x x

()
_
, (547)
where b

is the Burgers vector of the -th edge dislocation located at x

() =
[x

(), y

()]. Then upon replacing the decomposition of the displacement vector into
Eq. 53, the action can be written as the sum of two parts:
S
1
=
_
d
_
d
2
x
ij
u
D
ij
, (548)
and
S
2
=
1
2
_
d
_
d
2
x
_

n
(

u
S
i
)
2
+ 2
n
(

u
S
i
)(

u
D
i
) +
n
(

u
D
i
)
2
+ u
S
ij

S
ij
+ u
S
ij

D
ij
+ u
D
ij

S
ij
+u
D
ij

D
ij
+ (

)
2
+
s
(
i
)
2
+ 2i

u
S
ii

+ i

u
D
i

i
+ i

u
D
ii

_
, (549)
where
u
S,D
ij
=

i
u
S,D
j
+
j
u
S,D
i
2
, (550)
and

S,D
ij
=

ij
u
S,D
kk
+ 2 u
S,D
ij
. (551)
In S
2
we also introduced some simple notations:


n

2
,
2
and

2
. Second, we remove the analytic variables u
S
and in the action by means of
the equations of motion. The equations of motion are obtained by taking the variations of
the action Eq. 53 with respect to and u
S
i
:

2

+
s

2
+ i

u
S
ii
=
i
2

i
(

u
D
i
)
i
2

u
D
ii
, (552)

u
S
k
+
j

S
kj
+ i

=
n

u
D
k
)
j

D
kj
. (553)
When the coupling between the superuidity and the elasticity is neglected (

= 0), u
S
and become completely decoupled from each other, and dislocations produce only elastic
94
deformations. We invert the equations of motion to obtain u
S
and , resulting in
u
S
i
(q, ) =
_
i
n
T
1
ij

2
q
j
T
1
i0
_

u
D
j
(q, ) + iq
k
T
1
ij

D
jk
(q, ) +

2
T
1
i0
u
D
jj
, (554)
and
(q, ) =
_
i
n
T
1
0j

2
q
j
T
1
00
_

u
D
j
(q, ) + iq
k
T
1
0j

D
jk
(q, ) +

2
T
1
00
u
D
jj
(555)
where
T
1
ij
=
_

ij

q
i
q
j
q
2
_
1

2
+ q
2
+
q
i
q
j
q
2

2
+
s
q
2

A
, (556)
T
1
0i
= T
1
0i
= i

q
i


A
, (557)
T
1
00
=

2
+
_

+ 2
_
q
2

A
. (558)
Now we use Eqs. 554 and 555 to write the action in terms of u
D
(along with
D
ij
). Then
the last step in deriving the action for dislocations is to use the ansatz for u
D
, Eq. 547,
and express the action in terms of the coordinates of dislocations (Appendix F for details).
The result is
95
S
2
=
(2)
2
2
_

+ 2 +

2
4
_

,
b

i
b

i
_
dF
0
(x

() x

())
+
(2)
2
2

,
b

i
b

i
_
d
_
d

_
s

j
()s

j
(

) + c
2
T
_
F
1
(x

() x

),

; c
T
)

(2)
2
2

,
b

i
b

j
_
d
_
d

_
s

k
()s

k
(

) + 4c
2
T
_

j
F
2
(x

() x

),

; c
T
)
(2)
2

,
b

i
b

j
_
d
_
d

i
()s

k
(

k
F
2
(x

() x

),

; c
T
)
+
(2)
2
2

,
b

i
b

j
_
d
_
d

k
()s

k
(

j
G
(1)
2
(x

() x

),

; c
L
, c
2
)
(2)
2

,
b

i
b

j
_
d
_
d

j
()s

k
(

k
G
(2)
2
(x

() x

),

; c
L
, c
2
)

(2)
2
2

,
b

i
b

i
_
d
_
d

j
G
(3)
2
(x

() x

),

; c
L
, c
2
)
+ 2(2)
2

2

,
b

i
b

j
_
d
_
d

j
G
(4)
2
(x

() x

),

; c
L
, c
2
), (559)
where G
(1)
2
, G
(2)
2
, G
(3)
2
and G
(4)
2
can be written in terms of F
2
by using Eqs. 531 through
533 with
a
(1)
=
n
_

+ 2
_
+

n

4
, (560)
b
(1)
=

+ 2

_

s
+

2
4
_
, (561)
a
(2)
=
n

+

n

4
, (562)
b
(2)
=
n

2
4
_

2
_
, (563)
a
(3)
=

2
+

2

+
_

2
+
s

n
_

2
4
2
, (564)
b
(3)
=

s

_

2
+

2
4
_

+ 2
_
_
, (565)
a
(4)
= 1, (566)
96
b
(4)
=

s

. (567)
In Eq. 559 the spatial derivatives of the function F
2
is involved. Since

i
= [x

i
()
x

i
(

)]/X
X
with X = [x

() x

)[, we have

j
F
2
(x, ; c) =

ij
1
c[[ +
_
c
2
[[
2
+[x[
2
+

c
x
i
x
j
[x[
4
(c[[
_
c
2
[[
2
+[x[
2
)
2
_
c
2
[[
2
+[x[
2
. (568)
As in the case of vortices, we obtained non-local Coulomb potentials for the propagating
modes existing in supersolids (the second and the penultimate terms in Eq. 559). It is
worthwhile to notice that the derived action for dislocations is anisotropic even though
the solid is isotropic. This is because of the vectorial characteristics of the singularity
(Burgers vector) due to dislocations which breaks the discrete symmetry of the lattice
and leaves the lattice anisotropic. In the case of vortices the eective action Eq. 540
remains isotropic because the singularity caused by a vortex is a scalar. In addition,
with dislocations a term local in time appears in the action. More interestingly, it does
not depend on the superuid density. Thus this local term seems to be intrinsic to the
dynamics of dislocations in normal solids; however, superuidity makes a contribution
through the coupling with dislocations (the term with

). If we neglect the interaction
between the superuidity and the elasticity (

= 0), the coecient of the local term


becomes
n
c
2
L
which could suggest that its origin is due to the longitudinal elastic sound
modes.
On the other hand, we can obtain the Peach-Koehler force from S
1
. Since the stress
tensor is constant, it can be taken out of the integral, and the integral of u
D
ij
becomes a
summation over each dislocation of a surface integral over the two sides of the cut along
which the singular displacement vector of the -th dislocation undergoes a discontinuity
of b

. The two surfaces of the cut are denoted by

on which u
D
= 0 and
+
on which
u
D
= b

, where d

i
and d
+
i
are given by [ z dx

()]
i
and [ z dx

()]
i
, respectively
97
(Fig. 5-1). Then we have
S
1
=

ij
_
d
_
u

i
[

+ u

i
[

__
[ z dx

()]
i
=

ij
_
db

j
_
[ z dx

()]
i
, (569)
where u

is the singular part of the -th dislocations displacement vector. The negative
of coecient of dx

() in the integrand of Eq. 569 is the force acting on the -th


dislocation. By the cyclic property we obtain
F

PK
= ( b
a
) z, (570)
which is the Peach-Koehler force perpendicular to the vector b

in the lattice plane


[71, 72].
Similar to vortices we now can dene an inertial mass for a slowly moving dislocation.
In the limit c
2
(

)
2
[x

i
() x

i
(

)[
2
, we obtain
M
disl
ij
= 2
3
[ + ln()]
_
2b
i
b
j

s
(

+ 2 )
2
_

n
(

+ )
2
+

2
4

2
_

ij
b
k
b
k
_
2
n
+

n
(

2

2
)
(

+ 2 )
2
+

2
2
s

2
+ 2
2
(

+ 2 )
2
__
. (571)
The eective mass of a dislocation is a tensor because of the reason discussed earlier. If
is neglected, we get
M
disl
ij
= 2
3

n
[ + ln()]
_
2b
i
b
j

s
(

+ 2 )
2
_

s
(

+ )
2
+

n
4

2
_

ij
b
k
b
k
_
2 +

2

2
(

+ 2 )
2
+

n
(

2
+ 2
2
)
2
s
(

+ 2 )
2
__
. (572)
An eective action for dislocations in normal solids can be obtained from Eq. 559 by
neglecting the coupling between superuidity and elasticity (

= 0). For a normal solid,


98
we get the mass of a dislocation
M
disl
ij
= 2
3

n
[ + ln()]
_
2b
i
b
j
_

+

+ 2
_
2

ij
b
k
b
k
_
2 +

2

2
(

+ 2 )
2
__
. (573)
In summary, in this chapter we have considered two basic topological defects (vortices
and edge dislocations) which could exist in a supersolid phase. We have derived the
eective actions in terms of the coordinates of topological defects. The inertial mass
associated with the kinetic energy of such defects coupling with the elastic deformation
and superuidity were obtained.
99
z
^
z
^
^
dx

^
dx

^
dx

-
-

+
z
^
Figure 5-1. Cut for an edge dislocation. On surface

u
D
= 0 whereas on
+
u
D
= b.
The unit normals of

are given by z d x

.
100
CHAPTER 6
CONCLUSION
We proposed a viscoelastic solid model as an alternative explanation for the recent
torsional oscillator observations on solid
4
He. We have rst studied the dynamical
response of isotropic viscoelastic solids with cylindrical symmetry to an oscillatory
shear stress using the no-slip boundary condition. We found the elastic resonance eect
in the eective moment of inertia and the eective damping coecients. At low driving
frequencies, the eective moment of inertia decreases as the shear modulus increases.
This leads us to a possible connection between the anomalous increase in shear modulus
obtained by Day and Beamish [34] and the NCRI observed by Kim and Chan [19, 20].
However, the quantitative estimate of the period shift due to the 10 % increase in the
shear modulus was found to be only about one hundredth of the the observed period shift
in TO experiments. We also found that the frequency-dependent complex shear modulus
of a viscoelastic solid can explain the results of the TO experiments. In our model, a
characteristic time, which is related to the viscosity of solids increases rapidly as the
temperature is lowered, and causes both the drop in the resonant period and the peak in
the inverse of Q-factor. Our viscoelastic model predicts Q
1
max
/(P/P
0
) = 1 as other
theoretical works [49, 50]; however, the experimental values of this ratio is less than the
predicted value varying from 0.01 to 0.65. The consequence of this discrepancy between
the theories and the experiments is that any theoretical model does not t both the period
shift and the change in the inverse of Q-factor. In this work, we found that we could
identify the change in the inverse of Q-factor, but only the 10 % of the resonant period
shift. This fact could suggest that the unexplained part of the period shift might be due to
supersolidity.
In the second part of work we investigated the hydrodynamic properties of a model
supersolid. We introduced the variational principle to derive a Lagrangian density as well
as the non-dissipative hydrodynamics for supersolids. One of our main results is that
101
the coupling constant between the elastic variable and the superuid, which is set by the
Galilean covariance, the conservation laws and the broken symmetries, is the density of the
normal component. Using the hydrodynamic equations of motion we calculated the second
and fourth sound speeds of defects for a model supersolid.
Next, we have calculated the density-density correlation function for an isotropic
supersolid in local thermal equilibrium using linear response theory. First, we found
that when dissipation is neglected, each pair of propagating modes whose sound speed
is c produces a -function pair located at = cq in the density-density correlation
function. Second, we have extended the study to include dissipation. Due to dissipation
the neglected defect diusion mode of a normal solid appears as a central Rayleigh peak
in the density-density correlation function. Analogous to the superuid transition of
liquid
4
He, we found that the Rayleigh peak due to the defect diusion mode of a normal
solid becomes a Brillouin doublet of the pair of longitudinal second sound modes in the
supersolid transition. We proposed a Brillouin light scattering experiment to observe this
splitting as an alternative way to detect the supersolid phase.
Finally, we have studied the dynamics of vortices and dislocations in supersolids
using the derived Lagrangian for supersolids. An eective action for vortices and another
for dislocations were obtained in terms of the coordinates of vortices and dislocations.
In the case of vortices, we found that each sound mode existing in a supersolid couples
with vortices, and generates a non-local Coulomb potential. Contrary to a superuid,
transverse propagating modes exist in a supersolid and these transverse elastic modes
contribute to the action for vortices as well. For dislocations each mode in supersolids
produces both Coulomb and non-Coulomb potentials which are non-local in time. Another
contrast to vortices is that a local term with a logarithmic divergence arises. As a result,
we calculated the frequency-dependent inertial mass of vortices and dislocations. The
vectorial nature of singularity of dislocations destroys the isotropy of solids and the
derived mass becomes a second rank tensor. In both cases the eective mass has a
102
logarithmic dependence on the frequency. On the other hand, we showed that the terms
linear in the displacement vector u and the superuid variable of the action, which are
usually neglected in studying the dynamics without topological defects because of their
total derivative form, are responsible for the Magnus force acting on vortices and the
Peach-Koehler force on dislocations.
This thesis lays the groundwork for several future studies. First, the viscoelastic
model for solid
4
He provided us with a relaxation time which might be related to
dislocations in solids. We hope that it would be possible to derive analytically the
relaxation time of viscoelasticity from the dynamic equations of dislocations. As suggested
by Day and Beamish, the dislocation motion is controlled by
3
He impurities in solid
4
He, and eectively changes the shear modulus and, possibly, the relaxation time of
solid
4
He. Consequently, we would understand better the connection between the shear
modulus experiment and the TO experiment. In this regard, the derived action of a model
(super)solid with dislocation will be useful because it describes the dislocation dynamics
in (super)solids. Second, we can extend the viscoelastic model for solid
4
He in a TO in
several directions. We can investigate the response of a supersolid in the TO experiment
by using the derived hydrodynamic equation of supersolids. We expect that when the
supersolid hydrodynamics is combined with the viscoelastic model, it would be possible
to t all the TO results more precisely. Alternatively, we hope that the model could be
improved by considering the inhomogeneity of the system and/or a dierent boundary
condition. We think that the local variation of the relaxation time could smear out one
of the TO responses of the homogeneous system, and the slip boundary condition would
provide us with a larger period shift. Finally, we showed that the variational method is
systematically eective in deriving the non-dissipative hydrodynamics and the Lagrangian
density of a continuum medium which is characterized with conservation laws and broken
symmetries. The derived Lagrangian density was useful in studying the dynamics of
topological defects. We plan to apply the variational principle in other systems; possible
103
systems include Wigner crystals formed by excitons in electron-hole bilayers, striped
superuids, and so on.
104
APPENDIX A
CALCULATION OF BACK ACTION TERMS
In this appendix we show the detailed calculation to get the back action terms for
dierent geometries of torsion cells. Taking u

(t) exp(it), Eq. 212 reduces to


_

2
r
+
1
r

1
r
2
+ q
2
E
+
2
z
_
u

= 0, (A1)
where q
E

_

2
/() with () = (1 i). In the following we solve Eq. A1 for
three dierent boundary conditions: innite cylinder, nite cylinder, and innite annulus.
For the coordinate system used in this appendix, refer to Fig. 2-1.
A.1 Innite Cylinder
First, we consider an innite cylinder of radius R. The boundary condition for u

is
R
0
e
it
for r = R. Since there is no z dependence for innite cylinder, Eq. A1 becomes

2
r
u +
1
r

r
u +
_
q
2
E

1
r
2
_
u = 0. (A2)
The dierential equation Eq. A2 can be reduced to the Bessel equation of order one.
Applying the boundary condition and the niteness at the center we get the displacement
eld
u

= R
0
J
1
(q
E
r)
J
1
(q
E
R)
e
it
, (A3)
with q
E
R =
E
. The obtained displacement eld u

yields a non-vanishing stress

r
(r, t) = ()
_
u

r

u

r
_
= q
E
()R
0
J
2
(q
E
r)
J
1
(q
E
R)
e
it
. (A4)
The corresponding total torque for a height h is
M
inf cyl
(t) =
2

0
I
RB
H(q
E
R)e
it
, (A5)
where the moment of inertia of rigid body I
RB
= R
4
h/2. Using Eqs. 217 and 218,
from Eq. A5, we nally get the eective moment of inertia
I
inf cyl
e
() = I
RB
'[H(q
E
R)], (A6)
105
and the eective damping coecient

inf cyl
e
() = I
RB
[H(q
E
R)]. (A7)
Also, we get the back action term using Eq. 26
g
inf cyl
() =
2
I
RB
H(q
E
R). (A8)
A.2 Finite Cylinder
For the second case, we consider a nite cylinder with radius R and height h. The
boundary condition for u

is
u

=
_

_
R
0
e
it
for r = R
r
0
e
it
for z = h/2.
(A9)
In this case we need to solve Eq. A1:
_

2
r
+
1
r

1
r
2
+ q
2
E
+
2
z
_
u

= 0. (A10)
We decompose u

into two parts: u

=
0
r + V (r, z) with V = 0 at the boundary. Then
Eq. A10 reduces to

2
r
V +
1
r

r
V +
_
q
2
E

1
r
2
_
V +
2
z
V = q
2
E

0
r. (A11)
The solution of Eq. A11 is
V (r, z) = q
2
E

0
_
dx

G(x, x

), (A12)
where the Greens function G(x, x

) satises the inhomogeneous dierential equation


_

2
x

1
r
2
+ q
2
E
_
G(x, x

) =
4
r
(r r

)(

)(z z

), (A13)
106
with G = 0 at the boundary. Using
(z z

) =
1

_

0
dk cos[k(z z

)], (A14)
(

) =
1
2

m=
e
im(

)
, (A15)
we can expand G(x, x

) as
G(x, x

) =
1
2
2

m=
_

0
dk e
im(

)
cos[k(z z

)] g
m
(k, r, r

). (A16)
Replacing this into Eq. A13, we get the modied Bessel equation of order m

for
g
m
(k, r, r

):

2
g
m
r
2
+
1
r
g
m
r

_
k

2
+
m

2
r
2
_
g
m
=
4
r
(r r

), (A17)
where k

2
k
2
q
2
E
and m

2
m
2
+ 1. The general solution, using the niteness condition
at the center, is
g
m
(k, r, r

) = I
m
(k

r
<
) [AI
m
(k

r
>
) + BK
m
(k

r
>
)] , (A18)
where r
<
= minr, r

and r
>
= maxr, r

. Then the use of the boundary condition at


r = r

,
dg
m
dr

dg
m
dr

=
4
r

, (A19)
yields the Greens function
G(x, x

) =
2

m=
_

0
dk e
im(

)
cos[k(z z

)]

I
m
(k

r
<
)
I
m
(k

R)
_
K
m
(k

R)I
m
(k

r
>
) I
m
(k

R)K
m
(k

r
>
)
_
. (A20)
Finally, using Eq. A12 we get the displacement eld
u

(r, z, t) =
0
re
it
+
4q
2
E

m=1
(1)
m+1
2m1
cos
_
(2m1)
h
z
_ _
r

2
m

RI
1
(
m
r)

2
m
I
1
(
m
R)
_
e
it
,
(A21)
107
where

m
=
_
(2m1)
2

2
h
2
q
2
E
. (A22)
First we see that we recover the results of the innite cylinder case taking the limit that
h . In this limit,
m
becomes iq
E
, and the rst term in the second squared bracket
cancels the rst term in Eq. A21, and the second term in the second squared bracket
becomes the displacement eld for innite cylinder Eq. A3.
There are two non-vanishing components of the stress tensor which are

r
(r, z, t) = ()
_
u

r

u

r
_
=
4()q
2
E

0
R

e
it

m=1
(1)
m+1
2m1
_
cos
(2m1)
h
z
_
I
2
(
m
r)

m
I
1
(
m
R)
,
(A23)

z
(r, z, t) = ()
u

z
=
4()q
2
E

0
h
e
it

m=1
(1)
m+1
sin
(2m1)
h
z
_
r

2
m

RI
1
(
m
r)

2
m
I
1
(
m
R)
_
.
(A24)
The total torque is
M(t) =
_
2
0
d
_
d
d
dz R
2

r
[
r=R
+ 2
_
2
0
d
_
R
0
r
2

z
[
z=h/2
=
8
2

2
I
RB
e
it

m=1
_
1
(2m1)
2

_
R
h
m
_
2
_
H(i
m
)

8
2
R
2

0
h
2
I
RB
e
it

m=1
1

2
m
, (A25)
where

m
R
m
=
_
(2m1)
2

2
R
2
h
2
q
2
E
R
2
. (A26)
Then, the eective moment of inertia for a viscoelastic cylinder is
I
cyl
e
()
I
RB
=
8

m=1
'
_
_
1
(2m1)
2


2
R
2
h
2

2
m
_
H(i
m
)
_
+
8R
2
h
2

m=1
'
_
1

2
m
_
, (A27)
108
and the eective damping coecient is

n cyl
e
()
I
RB
=
8

m=1

_
_
1
(2m1)
2


2
R
2
h
2

2
m
_
H(i
m
)
_
+
8R
2
h
2

m=1

_
1

2
m
_
. (A28)
The back action term is
g
cyl
() =
8
2

2
I
RB

m=1
_
1
(2m1)
2

_
R
h
m
_
2
_
H(i
m
) +
8
2
R
2
h
2
I
RB

m=1
1

2
m
. (A29)
A.3 Innite Annulus
We lastly consider an innite annulus of inner radius R
i
and outer radius R. In this
case the boundary condition for u

becomes
u

=
_

_
R
i

0
e
it
for r = R
i
R
0
e
it
for r = R.
(A30)
The dierential equation for u

is the same as that of an innite cylinder Eq. A2 whose


general solutions are J
1
(q
E
r) and N
1
(q
E
r). Since the niteness at the origin is no longer
necessary, the displacement eld is given by
u

=
0
AJ
1
(q
E
r) +
0
BN
1
(q
E
r), (A31)
where
A =
R
i
N
1
(q
E
R) RN
1
(q
E
R
i
)
J
1
(q
E
R
i
)N
1
(q
E
R) J
1
(q
E
R)N
1
(q
E
R
i
)
, (A32)
B =
R
i
J
1
(q
E
R) RJ
1
(q
E
R
i
)
N
1
(q
E
R
i
)J
1
(q
E
R) N
1
(q
E
R)J
1
(q
E
R
i
)
. (A33)
The non-vanishing stress and the total torque for this case are

r
= ()
_

r
u

r
_
= q
E

0
()e
it
_
AJ
2
(q
E
r) + BN
2
(q
E
r)
_
, (A34)
109
M(t) = h
_
2
0
d [
r=R
i
R
2
i
+ h
_
2
0
d [
r=R
R
2
=
2
0
hR
2
i

2
q
E
e
it
_
AJ
2
(q
E
R
i
) + BN
2
(q
E
R
i
)
_

2
0
hR
2

2
q
E
e
it
_
AJ
2
(q
E
R) + BN
2
(q
E
R)
_
. (A35)
The eective moment of inertia is
I
inf ann
e
() = '
_
2hR
2
q
E
_
AJ
2
(q
E
R) + BN
2
(q
E
R)
_
_
'
_
2hR
2
i
q
E
_
AJ
2
(q
E
R
i
) + BN
2
(q
E
R
i
)
_
_
. (A36)
The eective dampting coecient is

inf ann
e
() =
_
2hR
2

q
E
_
AJ
2
(q
E
R) + BN
2
(q
E
R)
_
_

_
2hR
2
i

q
E
_
AJ
2
(q
E
R
i
) + BN
2
(q
E
R
i
)
_
_
. (A37)
The back action term is
g
inf ann
() =
2hR
2
i

2
q
E
_
AJ
2
(q
E
R
i
) + BN
2
(q
E
R
i
)
_

2hR
2

2
q
E
_
AJ
2
(q
E
R) + BN
2
(q
E
R)
_
.
(A38)
110
APPENDIX B
VARIATIONAL PRINCIPLE IN SUPERSOLIDS WITH THE ROTATIONAL
VELOCITY OF SUPER COMPONENTS
In Chapter 3 we derived the hydrodynamics of a supersolid using the variational
principle, and showed that the velocity of the super-component is irrotational. In this
Appendix we show that it is possible to derive systematically the hydrodynamics of
a supersolid with the transverse part of v
s
which generated by vortices present in the
supersolid. The transverse part of v
s
can be obtained by imposing another Lins constraint
[61]:

t
+
1

s
v
si
+ (
s
)v
ni
_

i
= 0. (B1)
Then the Lagrangian of the isotropic supersolid with all the constraints, Eqs. 322, 324,
325 and B1 becomes
L
SS
=
1
2

s
v
s
2
+
1
2
(
s
)v
n
2
U
SS
(,
s
, s, R
ij
) +
_

t
+
i
_

s
v
si
+ (
s
)v
ni
_
_
+
_

t
s +
i
(sv
ni
)
_
+
_

t
() +
i
_

s
v
si
+ (
s
)v
ni
_
_
+
i
_

t
(sR
i
) +
j
(sR
i
v
nj
)
_
, (B2)
where , , , and are Lagrangian multipliers, and we have used the second Lins
condition combined with the mass conservation equation.
The equations of motion are calculated by taking the variations with respect to the
dynamical variables:

1
2
v
n
2

U
SS


D
n

Dt

D
n

Dt
= 0 (B3)

s
1
2
v
s
2

1
2
v
n
2

U
SS

s
(v
si
v
ni
)(
i
+
i
) = 0 (B4)
s
D
n

Dt
+ R
i
D
n

i
Dt
+
U
SS
s
= 0 (B5)
111
v
si

s
v
si

i

s

i
= 0 (B6)
v
ni
(
s
)v
ni
(
s
)
i
s
i
sR
j

j
(
s
)
i
= 0 (B7)
R
i
s
D
n

i
Dt

j
_
U
SS
R
ji
_
= 0 (B8)

t
+
s
v
si

i
+ (
s
)v
ni

i
= 0 (B9)
From Eq. B6, we obtain the following Clebsch potential representations for the
velocity of the super-component
v
si
=
i
+
i
. (B10)
Therefore, we get
v
s
= ,= 0. (B11)
On the other hand, the use of the Clebsch representation of v
s
leads us Eq. 334.
Taking the gradient of Eq. B3, we get the Euler equation for v
s
D
s
v
si
Dt
=
i
_
U
SS

i
_
U
SS

s
_

_
(v
s
v
n
) (v
s
)
_
i
=
i


s

_
(v
s
v
n
) (v
s
)
_
i
. (B12)
112
The Euler equation for v
n
can be derived following the same calculation done in previous
sections:
(
s
)
D
n
v
ni
Dt
= (
s
)
i
_
U
SS

_
(
s
)
i
_
U
SS

s
_

j
_
U
SS
R
jk
_
R
ik
s
i
_
U
SS
s
_

1
2
(
s
)
i
(v
nj
v
sj
)
2

s
+
j
(
s
v
nj
)
_
v
si

t
(
s
) +
j
_
(
s
)v
nj
__
v
ni
+

(
s
)
_
(v
s
v
n
) (v
s
)
_
i
.
= (
s
)
i

j

jk
R
ik
s
i
T
1
2
(
s
)
i
(v
nj
v
sj
)
2

t
(
s
) +
j
_
(
s
)v
nj
_
_
(v
ni
v
si
),
+

(
s
)
_
(v
s
v
n
) (v
s
)
_
i
. (B13)
However, the momentum conservation equation for this case in which v
s
= 0, is the
same as that of the case of the longitudinal v
s
, Eq. 341.
113
APPENDIX C
STATIC CORRELATION FUNCTIONS OF ISOTROPIC SUPERSOLIDS
Given Eq. 319, the thermodynamic relation of the static free energy density F
SS
=
E
SS
Ts is calculated
dF
SS
= sdT
ik
dR
ik
+ d. (C1)
Without thermal uctuations we expand the static free energy density F
SS
up to the
second order in uctuations in the density and the displacement vector u:
F
SS
=
1
2

w
ij
()
2
+

w
ij

w
ij
+
1
2

ij
w
lk

w
ij
w
lk
. (C2)
Using Eqs. 389 through 391 the free energy can be written as, in Fourier space,
F
SS
=
1
2
q
2
u
2
T
+
1
2
_
(q) u
L
(q)
_
B
_
_
_
(q)
u
L
(q)
_
_
_
, (C3)
where
B =
_
_
_
1

2
0

iq
iq q
2

_
_
_
. (C4)
Then the static correlation functions can be easily read out from Eq. C3. We obtain

(q) = (q)(q)) =

2
0


2
0

, (C5)

u
L

(q) = u
L
(q)(q)) =
i
2
0

q(
2
0

2
)
. (C6)
114
APPENDIX D
KUBO FUNCTIONS AND CORRELATION FUNCTIONS
In general, the Laplace-Fourier transformations of hydrodynamic equations of a
physical system will have a form:

(q, ) = A

(q, )

(q, t = 0), (D1)


where

are some hydrodynamic variables and A

is an arbitrary matrix which can be


obtained from the hydrodynamic equations. The index indicates the hydrodynamic
variable

. On the other hand, the linear response theory tells us how

(q, ) is related
to

(q, t = 0) through the response function

[74],

(q, ) =
1
i
[

(q, )

(q)]
1

(q)

(q, t = 0). (D2)


Note that

(q) are the static susceptibilities which can be calculated from statistical
mechanics. Then, Eq. D1 and D2 imply that
1
i
[

(q, )

(q)] = A

(q, )

(q), (D3)
where we have used
1

.
Now, we dene a Kubo function K

[78] as
i
t
K

(x, x

, t) =
2

(x, x

, t), (D4)
where

(x, x

, t) is dened as
(x, x

, t) = 2i(t)

(x, x

, ), (D5)
115
where (t) is the step function. The Laplace transform of K

is
K

(x, x

, z) =
_

0
dte
izt
K

(x, x

, t)
=
1

d
i

(x, x

, )
( z)
=
1
iz
[

(x, x

, z)

(x, x

, t = 0)] .
(D6)
Therefore, using Eq. D3, the Kubo function can be obtained
K

(q

, ) =
1

(q, )

(q), (D7)
and the susceptibility

is related to this Kubo function through Eq. D4.


116
APPENDIX E
CALCULATION OF THE DENSITY-DENSITY CORRELATION FUNCTION
Using Eq. D7 the density-density Kubo function is obtained from Eq. 431
K

(q, z) =

(q)
k
B
T
C
1

(q, z) +

u
L

(q)
k
B
T
C
1
u
L

(q, z)
=

(q)
k
B
T
iz
3
+ b

z
2
+ d

q
2
z + e

q
2
Z
+

u
L

(q)
k
B
T
d
u
L
q
2
z + e
u
L
q
2
Z
, (E1)
where the static susceptibilities are given in Appendix C, and
Z (z
2
c
2
L
q
2
+ izq
2
D
L
)(z
2
c
2
2
q
2
+ izq
2
D
2
), (E2)
b

= q
2


s
0
(
n
0

s
0
)

n
0
q
2

2
s
0

n
0
q
2
, (E3)
d

= i

n
0
+ i, (E4)
e

=

s
0

n
0
q
2

s
0
q
2
+

s
0

n
0
q
2
, (E5)
d
u
L
= (
0
)q, (E6)
e
u
L
= i
s
0
_

n
0
_
q
3
+ i

s
0

n
0
_
2
0



0
(
0
2
s
0
)
_
q
3
. (E7)
We can rewrite the density-density Kubo function by doing partial fraction expansion.
Then each term in Eq. E1 can be separated into the rst sound part and the second
sound part as follows
a
ij
z
3
+ b
ij
z
2
+ d
ij
q
2
z + q
2
e
ij
(z
2
c
2
L
q
2
+ izD
L
q
2
)(z
2
c
2
2
q
2
+ izD
2
q
2
)
=
A
ij
z + B
ij
z
2
c
2
L
q
2
+ izD
L
q
2
+
D
ij
z + E
ij
z
2
c
2
2
q
2
+ izD
2
q
2
,
(E8)
where A
ij
, B
ij
, D
ij
and E
ij
can be written as a function of a
ij
, b
ij
, d
ij
and e
ij
, Eqs. E3
through E7:
A
ij
=
a
ij
[c
4
L
c
2
2
c
2
L
+ q
2
D
L
(D
L
c
2
2
D
2
c
2
L
)]
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
+
ib
ij
(c
2
2
D
L
D
2
c
2
L
) + d
ij
(c
2
L
c
2
2
) + ie
ij
(D
L
D
2
)
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
, (E9)
117
B
ij
=
ia
ij
c
2
L
q
2
(D
L
c
2
2
D
2
c
2
L
) + b
ij
c
2
L
(c
2
L
c
2
2
)
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
+
id
ij
c
2
L
q
2
(D
L
D
2
) + e
ij
[c
2
L
c
2
2
+ q
2
D
L
(D
2
D
L
)]
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
, (E10)
D
ij
=
a
ij
[c
4
2
c
2
L
c
2
2
+ q
2
D
2
(D
2
c
2
L

1
c
2
2
)]
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
+
+ib
ij
(c
2
L
D
2

1
c
2
2
) + d
ij
(c
2
2
c
2
L
) + ie
ij
(D
2

1
)
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
, (E11)
E
ij
=
ia
ij
c
2
2
q
2
(D
2
c
2
L

1
c
2
2
) + b
ij
c
2
2
(c
2
2
c
2
L
)
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
+
id
ij
c
2
2
q
2
(D
2

1
) + e
ij
[c
2
2
c
2
L
+ q
2
D
2
(
1
D
2
)]
(c
2
L
c
2
2
)
2
+ q
2
(D
L
D
2
)(D
L
c
2
2
D
2
c
2
L
)
. (E12)
The functions in the density-density correlation function, Eq. 27, are given by
I
1
(q) =

(q)A

+
u
L

(q)A
u
L
, (E13)
I
2
(q) =

(q)D

+
u
L

(q)D
u
L
, (E14)
I
3
(q) =

(q)B

+
u
L

(q)B
u
L
iq
2
D
L
I
1
(q), (E15)
I
4
(q) =

(q)E

+
u
L

(q)E
u
L
iq
2
D
2
I
2
(q). (E16)
118
APPENDIX F
DERIVATION OF AN EFFECTIVE ACTION FOR EDGE DISLOCATIONS
In this appendix we show the derivation of S
2
, Eq. 559, in detail starting with
Eq. 549 and the ansatz used for u
D
, Eq. 547. The use of the equations of motion
Eqs. 554 and 555 with Eqs. 556 through 558 leads us to the action in terms of the
singular displacement vector
S
2
= S
(1)
2
+ S
(2)
2
+ S
(3)
2
+ S
(4)
2
, (F1)
where
S
(1)
2
=

2
_
d
2
_
d
2
q
(2)
2
q
2

2
+ c
2
T
q
2
_

ij

q
i
q
j
q
2
_

u
D
j
(q, )

u
D
i
(q, )
+
1
2
_
d
2
_
d
2
q
(2)
2
q
i
q
j

A
_

n
(

+ 2 )(
2
+
s
q
2
)
+

2
4
_

2
+ (

+ 2 )q
2
_
_

u
D
j
(q, )

u
D
i
(q, ), (F2)
S
(2)
2
=
1
2
_
d
2
_
d
2
q
(2)
2
_
u
D
ij
(q, )
D
ij
(q, ) q
i
q
k
T
1
il

D
lk
(q, )
D
ij
(q, )
+

2
4

2
T
1
00
u
D
jj
(q, )u
D
ii
(q, ) + i

q
k
T
1
0j

D
jk
(q, )u
D
ii
(q, )
_
,(F3)
S
(3)
2
=
1
2
_
d
2
_
d
2
q
(2)
2

2
2
q
j

2
(

+ 2 )q
2

u
D
j
(q, )u
D
ii
(q, ), (F4)
S
(4)
2
=
_
d
2
_
d
2
q
(2)
2
q
j

il

q
i
q
l
q
2
_
1

2
+ c
2
T
q
2

u
D
l
(q, )
D
ij
(q, )
+
_
d
2
_
d
2
q
(2)
2
q
i
q
j
q
l
q
2

n

2
+ (
n

s
+

2
/2)q
2

u
D
l
(q, )
D
ij
(q, ).(F5)
119
Taking the ansatz, Eq. 547, one can easily calculate the spatial Fourier transforms:

j
u
D
i
(q, ) = i
jk
q
k
q
2

i
e
iqx
a
()
, (F6)

u
D
i
(q, ) = i
jk
q
k
q
2

i
s

j
()e
iqx
a
()
, (F7)

D
ij
(q, ) = i
ij

lk

q
k
q
2

l
e
iqx
a
()
+ i
q
k
q
2

_
b

ik
+ b

i

jk
_
e
iqx
a
()
. (F8)
Using these equations we get:
S
(1)
2
=
(2)
2

2

,
b

i
b

i
_
dF
0
(x

() x

())
+
(2)
2
2
_

a
1

n
_

,
b

i
b

j
_
d

j
F
3
(x

() x

())
+
(2)
2

2

,
b

i
b

i
_
d
_
d

_
s

j
()s

j
(

) + c
2
T
_
F
1
(x

() x

),

; c
T
)

(2)
2

2

,
b

i
b

j
_
d
_
d

_
s

k
()s

k
(

) + c
2
T
_

j
F
2
(x

() x

),

; c
T
)
+
(2)
2
2

,
b

i
b

j
_
d
_
d

k
()s

k
(

j
G
(1)
2
(x

() x

),

; c
L
, c
2
)

(2)
2
2

,
b

i
b

j
_
d
_
d

G
(1)
2
(x

() x

),

; c
L
, c
2
), (F9)
where G
(1)
2
is dened by Eq. 531 with
a
(1)
=
n
_

+ 2
_
+

n

4
, (F10)
b
(1)
=

+ 2

_

s
+

2
4
_
, (F11)
120
S
(2)
2
=
(2)
2
2
_

+ 2 +

2
4
_

,
b

i
b

i
_
dF
0
(x

() x

())

(2)
2
2
_

+ 2 +

2
4
_

,
b

i
b

j
_
d

j
F
3
(x

() x

())

(2)
2

2
c
2
T

,
b

i
b

j
_
d
_
d

j
F
2
(x

() x

),

; c
T
)

(2)
2
2

,
b

i
b

j
_
d
_
d

ij

j
_
G
(4)
2
(x

() x

),

; c
L
, c
2
)

(2)
2
2

,
b

i
b

j
_
d
_
d

ij

j
_
G
(5)
2
(x

() x

),

; c
L
, c
2
)
+
(2)
2
2

2
4

,
b

i
b

i
_
d
_
d

G
(6)
2
(x

() x

),

; c
L
, c
2
)

(2)
2
2

2
4

,
b

i
b

j
_
d
_
d

G
(6)
2
(x

() x

),

; c
L
, c
2
), (F12)
where G
(4)
2
, G
(5)
2
, and G
(6)
2
are dened by Eq. 531 with
a
(4)
= 1, (F13)
b
(4)
=

s

, (F14)
a
(5)
=
1

, (F15)
b
(5)
= 0, (F16)
a
(6)
=

n

, (F17)
b
(6)
=

+ 2

, (F18)
S
(3)
2
=
2

2
a
2

,
b

i
b

j
_
d

j
F
3
(x

() x

())

,
b

i
b

j
_
d
_
d

j
()s

k
(

k
G
(7)
2
(x

() x

),

; c
L
, c
2
)
+
2

,
b

i
b

j
_
d
_
d

G
(7)
2
(x

() x

),

; c
L
, c
2
), (F19)
121
where G
(7)
2
is dened by Eq. 531 with
a
(7)
=

n

, (F20)
b
(7)
=

+ 2

, (F21)
S
(4)
2
= (2)
2
_

+
_

,
b

i
b

j
_
d

j
F
3
(x

() x

())
(2)
2

,
b

i
b

j
_
d
_
d

j
()s

k
(

k
G
(8)
2
(x

() x

),

; c
L
, c
2
)
+(2)
2

,
b

i
b

j
_
d
_
d

j
G
(8)
2
(x

() x

),

; c
L
, c
2
)
(2)
2

,
b

i
b

j
_
d
_
d

i
()s

k
(

k
F
2
(x

() x

),

; c
T
)
(2)
2
c
2
T

,
b

i
b

j
_
d
_
d

j
F
2
(x

() x

),

; c
T
) (F22)
where G
(8)
2
is dened by Eq. 531 with
a
(8)
=
n
, (F23)
b
(8)
=
1

_

s
+

2
2
_
, (F24)
In Eqs. F9 through F22

G
2
is dened as

G
2
(x, ; c, c) =

n
F
2
(x, ; c) +

n
F
2
(x, ; c), (F25)
where

A
ac
2

b
c
2
c
2
, (F26)

B
a c
2

b
c
2
c
2
, (F27)
with a = b a(c
2
+ c
2
) and

b = ac
2
c
2
. In deriving Eqs. F9 through F22 one can nd
useful the fact that s

i
()

i
can be changed to

. Now we can combine S


(1)
2
, S
(2)
2
, S
(3)
2
,
and S
(4)
2
together by using Eqs. 529, 535 and 536. We nd that the local term with F
0
122
still remains unlike to the case of vortices, but the local term with F
3
cancels out. Then
we nally obtain the result that is given by Eq. 559.
123
REFERENCES
[1] P. Kapitza, Nature 141, 74 (1938).
[2] J. F. Allen and A. D. Misener, Nature 141, 75 (1938).
[3] O. Penrose and L. Onsager, Phys. Rev. 104, 576 (1956).
[4] A. F. Andreev and I. M. Lifshitz, Sov. Phys. JETP 29, 1107 (1969).
[5] G. V. Chester, Phys. Rev. A 2, 256 (1970).
[6] A. J. Leggett, Phys. Rev. Lett. 25, 1543 (1970).
[7] W. M. Saslow, Phys. Rev. B 15, 173 (1977).
[8] M. Liu, Phys. Rev. B 18, 1165 (1978).
[9] W. M. Saslow, Phys. Rev. Lett. 36, 1151 (1976).
[10] J. F. Fernandez and M. Puma, J. Low Temp. Phys. 17, 131 (1974).
[11] E. Frey, D. R. Nelson, and D. S. Fisher, Phys. Rev. B 49, 9723 (1994).
[12] Y. N. Joglekar, A. V. Balatsky, and S. D. Sarma, Phys. Rev. B 74, 233302 (2006).
[13] V. W. Scarola and S. Das Sarma, Phys. Rev. Lett. 95, 033003 (2005).
[14] G. A. Lengua and J. M. Goodkind, J. Low Temp. Phys. 79, 251 (1990).
[15] D. S. Greywall, Phys. Rev. B 16, 1291 (1977).
[16] D. J. Bishop, M. A. Paalanen, and J. D. Reppy, Phys. Rev. B 24, 2844 (1981).
[17] C. Lie-zhao, D. F. Brewer, C. Girit, E. N. Smith, and J. D. Reppy, Phys. Rev. B 33,
106 (1986).
[18] M. W. Meisel, Physica B 178, 121 (1992).
[19] E. Kim and M. H. W. Chan, Nature 427, 225 (2004).
[20] E. Kim and M. H. W. Chan, Science 305, 1941 (2004).
[21] S. Balibar and F. Caupin, J. Phys.: Condens. Matter 20, 173201 (2008).
[22] A. S. C. Rittner and J. D. Reppy, Phys. Rev. Lett. 97, 165301 (2006).
[23] M. Kondo, S. Takada, Y. Shibayama, and K. Shirahama, J. Low Temp. Phys. 148,
695 (2007).
[24] A. S. C. Rittner and J. D. Reppy, Phys. Rev. Lett. 98, 175302 (2007).
[25] A. Penzev, Y. Yasuta, and M. Kubota, J. Low Temp. Phys. 148, 677 (2007).
124
[26] Y. Aoki, J. C. Graves, and H. Kojima, Phys. Rev. Lett. 99, 015301 (2007).
[27] T. Leggett, Science 305, 1921 (2004).
[28] N. V. Prokofev, Advances in Physics 56, 381 (2007).
[29] W. M. Saslow, Phys. Rev. B 71, 092502 (2005).
[30] D. M. Ceperley and B. Bernu, Phys. Rev. Lett. 93, 155303 (2004).
[31] N. Prokofev and B. Svistunov, Phys. Rev. Lett. 94, 155302 (2005).
[32] M. Boninsegni, A. B. Kuklov, L. Pollet, N. V. Prokofev, B. V. Svistunov, and
M. Troyer, Phys. Rev. Lett. 97, 080401 (2006).
[33] J. Day, T. Herman, and J. Beamish, Phys. Rev. Lett. 95, 035301 (2005).
[34] J. Day and J. Beamish, Phys. Rev. Lett. 96, 105304 (2006).
[35] S. Sasaki, R. Ishiguro, F. Caupin, H. J. Maris, and S. Balibar, Science 313, 1098
(2006).
[36] M. W. Ray and R. B. Hallock, Phys. Rev. Lett. 100, 235301 (2008).
[37] S. Sasaki, F. Caupin, and S. Balibar, Physical Review Letters 99, 205302 (2007).
[38] L. Pollet, M. Boninsegni, A. B. Kuklov, N. V. Prokofev, B. V. Svistunov, and
M. Troyer, Phys. Rev. Lett. 98, 135301 (2007).
[39] M. Boninsegni, A. B. Kuklov, L. Pollet, N. V. Prokofev, B. V. Svistunov, and
M. Troyer, Phys. Rev. Lett. 99, 035301 (2007).
[40] S. O. Diallo, J. V. Pearce, R. T. Azuah, O. Kirichek, J. W. Taylor, and H. R. Glyde,
Phys. Rev. Lett. 98, 205301 (2007).
[41] E. Blackburn, J. M. Goodkind, S. K. Sinha, J. Hudis, C. Broholm, J. van Duijn,
C. D. Frost, O. Kirichek, and R. B. E. Down, Phys. Rev. B 76, 024523 (2007).
[42] J. A. Lipa, J. A. Nissen, D. A. Stricker, D. R. Swanson, and T. C. P. Chui, Phys.
Rev. B 68, 174518 (2003).
[43] A. T. Dorsey, P. M. Goldbart, and J. Toner, Phys. Rev. Lett. 96, 055301 (2006).
[44] A. C. Clark and M. H. W. Chan, J. Low Temp. Phys. 138, 853 (2005).
[45] X. Lin, A. C. Clark, and M. H. W. Chan, Nature 449, 1025 (2007).
[46] J. Day and J. Beamish, Nature 450, 853 (2007).
[47] A. T. Dorsey and D. A. Huse, Nature 450, 800 (2007).
[48] J. G. Dash and J. S. Wettlaufer, Phys. Rev. Lett. 94, 235301 (2005).
125
[49] Z. Nussinov, A. V. Balatsky, M. J. Graf, and S. A. Trugman, Phys. Rev. B 76,
014530 (2007).
[50] D. A. Huse and Z. U. Khandker, Phys. Rev. B 75, 212504 (2007).
[51] R. Darby, Viscoelastic Fluids: an Introduction to their Properties and Behavior (M.
Dekker, New York, 1976).
[52] R. S. Lakes, Viscoelastic Solids (CRC Press, Boca Raton, 1999).
[53] A. C. Clark, J. T. West, and M. H. W. Chan, Phys. Rev. Lett. 99, 135302 (2007).
[54] E. Kim and M. H. W. Chan, Phys. Rev. Lett. 97, 115302 (2006).
[55] M. C. Marchetti and K. Saunders, Phys. Rev. B 66, 224113 (2002).
[56] C. C. Lin, in Liquid Helium, edited by G. Careri (Academic Press, New York, NY,
1963), vol. 21 of Proceedings of the International School of Physics Enrico Fermi;
Course XXI, pp. 93146.
[57] R. L. Seliger and G. B. Whitham, Proc. R. Soc. London, Ser. A 305, 1 (1968).
[58] R. Salmon, Annu. Rev. Fluid Mech. 20, 225 (1988).
[59] P. R. Zilsel, Phys. Rev. 79, 309 (1950).
[60] H. W. Jackson, Phys. Rev. B 18, 6082 (1978).
[61] J. A. Geurst, Phys. Rev. B 22, 3207 (1980).
[62] A. J. Purcell, Phys. Rev. B 23, 5769 (1981).
[63] C. Coste, Eur. Phys. J. B 1, 245 (1998).
[64] D. D. Holm and B. A. Kupershmidt, Physica D 6, 347 (1983).
[65] D. D. Holm, arXiv.org:nlin/0103041 (2001).
[66] S. J. Putterman, Superuid Hydrodynamics (Northe-Holland Pub. Co., Amsterdam,
1974).
[67] D. T. Son, Phys. Rev. Lett. 94, 175301 (2005).
[68] C. Josserand, Y. Pomeau, and S. Rica, Phys. Rev. Lett. 98, 195301 (2007).
[69] J. Ye, Europhys. Lett. 82, 16001 (2008).
[70] X.-M. Zhu, Y. Tan, and P. Ao, Phys. Rev. Lett. 77, 562 (1996).
[71] M. Peach and J. S. Koehler, Phys. Rev. 80, 436 (1950).
126
[72] L. D. Landau and E. M. Lifshitz, Theory of Elasticity (Pergamon Press, New York,
1986), 2nd ed.
[73] A. Zippelius, B. I. Halperin, and D. R. Nelson, Phys. Rev. B 22, 2514 (1980).
[74] P. M. Chaikin and T. C. Lubensky, Principles of Condensed Matter Physics
(Cambridge University Press, Cambridge, UK, 1995).
[75] J. T. S. Ostlund and A. Zippelius, Ann. Phys. (NY) 144, 345 (1982).
[76] P. C. Martin, O. Parodi, and P. S. Pershan, Phys. Rev. A 6, 2401 (1972).
[77] B. J. Berne and R. Pecora, Dynamic Light Scattering: with Applications to Chem-
istry, Biology, and Physics (Wiley, New York, 1976).
[78] D. Forster, Hydrodynamic Fluctuations, Broken Symmetry, and Correlation Functions
(W. A. Benjamin, Advanced Book Program, Reading, 1975).
[79] L. Landau and G. Placzek, Phys. Z. Sowjetunion 5, 172 (1934).
[80] P. A. Fleury and J. P. Boon, Phys. Rev. 186, 244 (1969).
[81] P. C. Hohenberg and P. C. Martin, Ann. Phys. (NY) 34, 291 (1965).
[82] M. J. Stephen, The Physics of Liquid and Solid Helium (Wiley, New York, 1976),
chap. Brillouin and Raman Scattering in Helium, pp. 307348.
[83] G. Winterling, F. S. Holmes, and T. J. Greytak, Phys. Rev. Lett. 30, 427 (1973).
[84] J. A. Tarvin, F. Vidal, and T. J. Greytak, Phys. Rev. B 15, 4193 (1977).
[85] R. Wanner and J. P. Franck, Phys. Rev. Lett. 24, 365 (1970).
[86] S. B. Trickey, W. P. Kirk, and E. D. Adams, Rev. Mod. Phys. 44, 668 (1972).
[87] B. A. Fraass, P. R. Granfors, and R. O. Simmons, Phys. Rev. B 39, 124 (1989).
[88] P. Ao and D. J. Thouless, Phys. Rev. Lett. 70, 2158 (1993).
[89] E. M. Chudnovsky and A. B. Kuklov, Phys. Rev. Lett. 91, 067004 (2003).
[90] U. Eckern and A. Schmid, Phys. Rev. B 39, 6441 (1989).
127
BIOGRAPHICAL SKETCH
Chi-Deuk Yoo was born in Seoul, Korea and immigrated to Argentina during his
high school years. He nished, overcoming cultural shock, high school in Buenos Aires,
Argentina, and enrolled in the Physics Department of the University of Buenos Aires,
where he received a Licenciado en Ciencias Fsicas in 2002. In the same year he was
admitted to the Physics Department of the University of Florida. During the early years
at the University of Florida he was interested in high-energy theory, but in 2004 he joined
Prof. Alan T. Dorseys group to study theoretical condensed matter physics.
128

Вам также может понравиться