HAL Id: jpa-00246815
https://hal.archives-ouvertes.fr/jpa-00246815
Submitted on 1 Jan 1993
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
kink-bearing nonlinear one-dimensional scalar systems
D. Grecu, A. Visinescu
To cite this version:
D. Grecu, A. Visinescu. On the lattice corrections to the free energy of kink-bearing nonlinear one-dimensional scalar systems. Journal de Physique I, EDP Sciences, 1993, 3 (7), pp.1541-1550.
�10.1051/jp1:1993199�. �jpa-00246815�
J.
Phys.
I France 3(1993)
1541-1550 JULY 1993, PAGE 1541Classification
Physics
Abstracts05.90
On the lattice corrections to the free energy of kink-bearing
nonlinear one-dimensional scalar systems
D. Grecu and A. Visinescu
Department
of TheoreticalPhysics,
Institute of AtomicPhysics,
P-O- BoxMG-6, Magurele, Bucharest,
Romania(Received
5October1992,
revised 25February1993, accepted
26 March1993)
Abstract The effective potential obtained by
Trullinger
and Sasaki isdiscused.Using
asymp-totic methods from the theory of differential equations
depending
on alarge
parameter, the lattice corrections to the kink and kink-kink contributions to the free energy are calculated.The results are in
complete
agreement with a first order correction to the energy of the static kink.1 Introduction.
The influence of lattice discreteness on the
properties
of nonlinearsystems having
kink solutionswas
investigated by
several authors[1-10].
These studies havepointed
out alarge variety
ofeffects, including
modification of kinkvelocity
and its form andleading
sometimes to thepinning
of the kink on the lattice. But on the other side numericalinvestigations
[9] have revealed the existence of narrowsolitary
wavespropagating
in certain discrete lattices without energy loss. There is also a close connection of theseproblems
with thecomplete integrability
of
specific
models in nonlinear continuum and discretesystems [11-13].
Since the
pioneering
paper of Krumhansl and Schrieffer[14]
the roleplayed by
theelementary
excitations of kinktype
in thethermodynamics
of one-dimensional nonlinearsystems
wasinvestigated by
many authors([15-20]
and the referencestherein).
As concerns the influence of lattice discreteness on thesethermodynamic properties
this has received a smaller attention[20-23].
The class of i-Dsystems mostly
discussed is describedby
the Hamiltonian(in
the notations of CKBT[15])
H =
fliA()#)
+)I#;+1-#;)~
+Wlvl#,)) Ii)
where the
nonlinearity
entersonly througb
thepotential V(#),
assumed to have at least twodegenerate
minima withnonvanishing
curvature. It is the merit ofTrullinger
and Sasaki[21]
to have shown that the first lattice corrections are taken into account if the
potential V(#)
isreplaced by
an effectivepotential
where d = ~° is the mean width of the static kink. In the
present
paper we intend tocomplete
wo
their results and determine also the lattice corrections to the multi-kink contributions to the free energy.
The
partition
function and the free energy areusually
calculatedusing
the transferintegral operator (TIO)
method. In thethermodynamic
limit the lowesteigenvalues
of TIOplay
the mostimportant
role. In thedisplacive/continuum limit,
«1,
it ispossible
to transform the Fredholmintegral equation representing
the TIO into a differentialequation.
This iseasily
done when the
two-body
interaction in(1)
has an harmonic character (~[17, 19].
One obtainsexPl-7Vl#))
exP()D~) Aim#)
=
exP1-7in) All#). 13)
Here 7
=
fllAw],
m*=
(flAcowo)~,
Am is theeigenvector
of TIOcorresponding
to theeigen-
value en andin
= en + ln~'~. (3a)
7 7
The left hand side of
(3)
contains theproduct e~ e~
with A and B twononcommuting
opera-tors,
and in order to obtain the claimed differentialequation
we have toput
it ase~,
where theoperator
C can be written in terms of A and Busing
the Baker-Hausdorff formula[20-22].
This~
formula turns to be a mixed
expansion
in the smallparameters j
and ~=
(m*)
2 andconsequently
is very convenient for theproposed
purposes. The firstlatZce co$ection
is ob-tained if one retaines
only
the termsproportional
with ~~Contrary
theopinion
ofTrullinger
d2and Sasaki
[21]
the same result is found if one work with tbesymmetrised
ornon-symmetric
form of TIO. It ispossible
to showthrough
a set of transformations andneglecting system- atically higher
powers ini2 j,
that theeigenvalues
ofTIO,
with the first lattice correctionsincluded,
can be found from theequation [21] (~)
~~fi
+(in l~a(#))tvn
= 0.
(4)
2m*
d#
The result is very
general
and valid for any kind ofpotential
functionV(#).
In the next section the
leading
terms as anasymptotic expansion
of the lowesteigenvalues
of(4)
will be found in the lowtemperature limit,
m* » 1. The methodused, briefly
sketched in theAppendix,
allows us to make a clear distinction between the contributions ofphonons
(~)
As was shownby Guyer
and Miller [17] even for morecomplicated
interactions between nearestneighbours
it ispossible
to write a formalexpression
for thepartition
function and the free energyusinj
a cumulantexpansion.
The
discrepancy
between the results of [21] and [20] comes from aninadequate
treatment of a certainequation
obtained in an intermediate step of theproof
in ourprevious
paper [20].N°7 LATTICE CORRECTIONS TO KINK SYSTEMS 1543
and of the various kink sectors.
Explicit expressions
for the lattice corrections to the kink and kink-kink contributions to the free energy will begiven.
The last section is devoted to a discussion of the results in thespirit
of the CKBT[15] phenomenology
ofindependent phonons
and renormalized kinks. As it is known
[15, 16,
19] thisphenomenology
is exact in the lowtemperature
limit of the continuum models and the results ofTrullinger
and Sasaki[21]
and those of thepresent
paperstrongly support
theconjecture
that it remains valid even when the first lattice corrections are taken into account.2. Lattice corrections to the free energy.
Using (3a)
the free energy per unitlength
becomeswhere To is the lowest
eigenvalue
ofequation (4).
In the low
temperature
limitfl
»1,(m*
»1)
there are several ways to findapproximate
solutions of thisequation,
all included in so-called class of theimproved
WKB methods[15-20, 23, 24] (and
the referencestherein).
We shall use the methoddeveloped by
us in a series of papers[20, 23, 24]
which has theadvantage
to make a clear distinction between the variouscontributions to the free energy:
phonons, i-kink,
2-kink and soon sectors. The method is
based on well known results from the
theory
ofasymptotic
solutions of second order differentialequations depending
on alarge parameter [25]
an ispresented
in theAppendix.
As we are interested in theleading
term in m* the exactequation (4)
isapproximated by
thecomparison equation (A.9).
WithVea replacing
thepotential
V in the definition(A.8)
of the constant a the first term in theasymptotic expansion
of the lowesteigenvalue
of the isolatedpotential
well isgiven by
~° 2~$ ~ ~~2~
~~~/*~'
~~~Then the
corresponding
contribution to the free energydensity 16)
writes' l ~~~~~°°~~
~
~d ~ ~d~~
~~~
and is
easily
identified with the first terms in the seriesexpansion
in powers ofl~/d~
of the exact free energy of anindependent
gas ofphonons
with thedispersion
relationw~(k)
=w( 1+4~ ~ sin~
~~(8)
2
Higher
order terms in(m*)~~/~, neglected
in(6),
willcorrespond
to anharmonic interactions between thephonons [23].
The presence of the other
degenerate
minima determines asymmetrical splitting
of eacheigenvalue
ofthe isolatedpotential
well. In the case ofthe#~
model we obtainEo
-io
"Eo+
to where
~ 12
~°
@
~ 4d~~
~ ~~~
with c
= I. As is
explained
in theAppendix
the value of u is determined from theboundary
conditions(A,ll) imposed
on theeigenfunctions
of thecomparison equation.
For the#~
modelwe
get
~-;xv
-1 =
2j%
exp 21J +u(1
+ln2) -(1+ 2u)In(1+ 2u) (10)
r(-u)
2This
expression
is valid inleading
order in(m*)~~/~,
but in any order in u, and is thestarting point
to find multi-kink contributions to the free energy. Here J is anintegral
definedby
J =
/~'/fid# (11)
o
with pi the left
turning point
of the isolatedpotential
well. As we are interested in theground
state the
quantity
I from(11)
is connected to uby
~
2~ ~ l)
2Ii
+
2uc)
=
fi (i
24d2l~(12)
where
io
is nowindependent
on the latticespacing
I. It is seen that Jdepends
onl~/d~
boththrough
theintegrand
and the upper limit ofintegration. Using (12)
and theexpression
ofl~a
theturning points
of the#~ model,
calculated in the orderO(l~/d~),
aregiven by
12 14
p],~
= i ~fi
+~fi (~j
+fi)
+
O(~). (13)
In the same order
J =
b £"' Ii &<~l /(Pi <2)(Pi <2) d<
+O(i~/d~). (14) Writing
~
~
~°
~~d2
~~ ~~~~
where now
Jo
andJi
does notdepend
any more onl~/d~,
weget
These
ntegralscan
be alculated interms of complete
elliptic ntegrals ofmodulus in the
2ViJo
t
io(1
+In$)
Co
2ViJi
t
~
+sic. (16)
N°7 LATTICE CORRECTIONS TO KINK SYSTEMS 1545
Introducing
these results into(10)
weget
@e ~~~~~~ ~~e ~
=
~~~
~ ~ ~
e~'~"
-u In(12flE)°~)
"@
(~~) ~~
r(-u)
~ ~ ~~~where
E)°~
=( Acowo
is the energy of the static kink.In the sine-Gordon case
(periodic potential)
eacheigenvalue
of the isolatedpotential
well issymmetrically splitted
in an allowed band of width2to
whereto
isgiven by (9)
with c=
2,
andu results from
boundary
conditionsimposed
on theeigenfunction
at thepoint #
= ir.
Finally
we
get
~;xv rj1
~~)
~ ~~l =
2@
~ exp 21J +2u(1
+ln2) In(1
+4u)
,
(10a)
~(~U) r( I)
2where the
integral
J is definedby
J
=
/ ~ fi~d#. (11a)
12
The
integral
J is calculated in the same linearapproximation
inj
and theleading
terms inio
is givenby
~
~~ ~~ ~~2
~
~
~~
~~2 ~
~~~~~
where
io
is defined in(12) being independent
on lattice corrections.Introducing (16a)
into(10a)
andusing
theexpression E)°~
=8Acowo
for the energy of the static kink one obtainse~~~~~~~~
~~e
41~?= l~
~;xu yji
~~)
~~~ ~,
=
2
e-~"
'~~fl~ke~"W II?)
SG~(-U) ~())
In order to find various kink contributions we have to
expand
theright
hand side ofequations (17)
in powers of u and to write alsou = uk + ukk +..
(18)
The
single
kink term uk comes from theexpansion
of~
r(-u) ~~~~~
(7
= 0.577 Euler'sconstant)
and one obtainseasily
bk =
@~-pE)°)j~_
,2 ~~~~
~~~~
~
48d2 (19)
~i4~
@~
p~io~~~ ,, i~lr
~ ~~'
ml jig)
sG
This result is in
complete agreement
with that foundby Trullinger
and Sasaki[21].
In
finding
the kink-kink contribution ukk we have to take into account that theexpansion (18)
is a formalexpansion
in powers ofe~fl~i~
andconsequently
ukk is of the same order asVI.
Then it iseasily
found thatukk "
u( (7
+In(12flE)°~) ~~~
+ir) (20) #~
24d
One sees that ukk
gets
animaginary part satisfying
the verysimple
relationImukk
"iru( (21)
This relation has been obtained
by
Zinn-Justin[26]
in hisanalysis
of the multi-instanton contributions inquantum
mechanics and istightly
related to the non-Borelsummability
of theRayleigh-Schr6dinger perturbation expansion
when tbepotential
hasdegenerate
minima.For the sine-Gordon model one obtains in
a similar way
~~
~~( (7
+ln4pE)o)
12 i~~~~~ ~
(~~~
~~Both
(19) #~
and(20)-
SG in the limit 12j - 0 are in
complete agreement
withprevious
d
calculations of the kink-kink sector contributions
[26-28].
3
Concluding
remarks.According
to CKBTphenomenology
thethermodynamic properties
of the systems describedby
the Hamiltonians of the form(1)
are influencedby
the existence of the static kinks[15-19].
A
complete agreement
between thisphenomenology
and the exact results of TIO in the lowtemperature
limit is found if one takes into account thescattering
ofphonons
on the statickink, leading
to a renormalization of the kink energy.The results of
Trullinger
and Sasaki and of thepresent
paper areshowing
that lattice corrections areeasily
included into thethermodynamics
of thissystems. Although
acomplete proof
of a similar CKBTphenomenology
does not exist atpresent,
theexisting
resultsstrongly support
the idea that theirphenomenology
is still valid at lowtemperatures.
This comes both from thephonon part, which,
as mentionedabove, reproduces exactly
the first terms in the12
series
expansion
inj
of the free energy of apbonon
lattice gas, and from the kink contribution.As is seen from
(19)
in the kink contribution appear a lattice corrected kink energyEk
=E)°~ (
E)~~ (21)
where
E)°~
is the knownunperturbed
static kink energy and the correctionE)~~
for the SG and#~
model aregiven by
N°7 LATTICE CORRECTIONS TO KINK SYSTEMS 1547
E)~~
=E)°~, E)°~
=~A0woco (22) #~
These corrected values have been obtained also
by
other authors[5, 6],
and as will be shownbelow,
result from a verysimple perturbation theory.
Indeedexpanding
the field variable#(z
+I)
in aTaylor
series the discreteEuler-Lagrange equation
transforms into a fourth order differentialequation
~~~
~~~2 ~~
~~~ ~
~~~~where y
= z
Id
andVj
=(~'. Looking
for aperturbed
kink solution#
4k(Y)
"4[°~lY)
+)#[~~lY)1 124)
where
#)°~
is the kink solution of the continuumlimit,
the linearizedequation determining
~(l)
i~k
~~~i~~
~~~j~(0)~ ~(l) _d~~~~
dy~
~ ~dy~ j~$)
The
homogeneous
part of(25)
hasappeared
some years ago in thestudy
of the kinkdy-
namics in the presence ofperturbations
and has as solution the "translation mode"(Goldstone mode) [15, 16, 29].
Thenequation (25)
can be solvedby
the method of variation of constantsand the obtained results are in
complete agreement
with thosealready existing
in literature[4-7, 10]. Using
the sameTaylor expansion
for the field variable one can write down a cor-rected
integral
of energy.Introducing
in it the newexpansion
of#k (24)
theintegration
iseasily performed
and the result isjust
thatpresented
in(22).
Still unsolved remains theprob-
lem of calculation of renormalization of kinks due to thescattering
of latticephonons
on theperturbed kinks,
renormalization which determines the factormultiplying
theexponentials
inl19).
Acknowledgement.
The authors would like to thank Dr. Marianne Croitoru for a fruitful collaboration
during
many years.
Appendix.
The lowest
eigenvalues
of(4)
can befound,
in the lowtemperature
limit m* »I, using asymptotic
methods known from thetheory
of second order differentialequation depending
ona
large parameter [25].
Theequation
underinvestigation
is of thetype
where the
potential V(#)
has at least twodegenerate
minima andnonvanishing
curvature at the minima.The
problem
will be solved in two distinct steps. In thefirst,
based onLanger's
transforma-tion,
one looks for an uniform validexpansion
of the solution near the minimum. Due to thepeculiarities
mentionedabove,
this transformation is chosen in such a way togive
the harmonic oscillator behaviour as aleading
term. In the secondstep,
the existence of the other minima ofV(#)
are taken into accountusing symmetry properties
of the wave functions. Eacheigenvalue
of an isolated
potential
well isslightely splitted.
Thesetunneling
terms aredirectly
related to the kink contribution to the free energy.Let pi and p2 be the two
turning points
of an isolatedpotential
well. One passes from the variables tV and#
to the new ones R and z.~ =
x~iR (A.2)
((z)
=f~ W@ d< (A.3)
x =
[j~[. jA.4)
Here
1'
=
~~,
and z is stillan undefined function of
#.
ThenIA-i)
transforms into dzd2~
j +
l~R
= 6R
IA-b)
where 6
=
-x~1~~~~ l'~
andl~
= 2m*. The function x is
subjected
to therequirement
to bed#
regular
and not to vanish in the interval of interest. In order to obtain the harmonic oscillator behaviour as theleading term,
one takesjl')~
=
4a~(1 z~). jA.6)
This is
a function with two
simple
zeroes and z= -1 is associated with the
turning point
#
= pi andz = +1 to
#
= p2. Now it ispossible
to determine the relation between the new variablez and the old one
#. Integrating IA-G)
weget
a(ir cos~~
z +z/~)
=
/~
Vd# IA.?)
p,
where the constant a is
given by
P~
a =
Vd#. (A.8)
lr
/,
It can be
easily proved
that for all thedomain,
even in theasymptotic region,
6-w
~,
and Itaking
into account thatl~
= 2m* »
1,
the first orderapproximation
can be found from thefollowing comparison equation [25]
~
+
4a~l~(1 z~)R
=
0, (A.9)
whose solutions are the functions of the
parabolic cylinder.
For the#~ model,
theonly
conve-nient solution
(decreasing exponentially
when z -cc)
is Whittaker's functionDv(y),
where y =2fi
z and u
= al
). (A.10)
N°7 LATTICE CORRECTIONS TO KINK SYSTEMS 1549
The
eigenvalues
of the isolatedpotential
well are obtained for u= n an
integer
number. Inleading
order in I one findsj(°)
g~ +n)
+o(j2) jA 11)
"
@
2Now,
due to existence of the otherdegenerate
minima of thepotential V(#)
eacheigenvalue
of the isolated well issplitted
into two very near levels.They
can be foundusing
theboundary
conditions for the wave function and its derivative
The solution
WI))
ofIA.1)
is related to the solutionR(z)
of(A.9) by Wi#)
=I(l~l Rlz)
,
and the
explicit
connection formula between the old variable#
and the new variable z is(asymptotic region
of~)
z~
m/~~ Vi d#
+ +ln2[z[
+O(z~~). (A.13)
a o 2
In
applying
theboundary
conditions(A.
ii we have to use theasymptotic expansion
ofDv(y)
1301
~~
/~
i /D"lY)
~'Y~~~~
~j_~~
~"'~p
~ ~lA.14)
The value of u is found from a
matching
between the dominant e /4 and the subdominant term /e~ 4 Then from
(A.12-A.14)
one obtains~-;xv
Pi-1 =
2j%
exp 21W/ d#
+u(1
+ In2) -(1
+2u) In(1
+2u)
r(-u)
2(A.15)
The sine-Gordon case can be treated in a very similar way, the solution for the
comparison equation (A.9) being,
in that case,expressed
more convenient in terms of Kummer's functionshaving
a definiteparity.
Eacheigenvalue
of an isolatedpotential
well issplitted
into a narrow allowed band. If thepotential V(#)
has a2ir-periodicity
the index ucorresponding
to the lower and upperboundary
of the lowest band follows from similar conditions as(A.12)
calculated in thepoint #
= ir.
Finally
a very similar relation with(A. lb)
isfound, namely
~-ixu rj1
~~)
~~
~~
rj_ I) /jl)
~2 2
x exp
21 /" W/ d#
+2u(1
+ In2) (l
+4u) In(1
+4u)) (A.15a)
j 2
Here
by +j
we have denoted the
turning points
of the isolatedpotential well,
centred around the minimum#
= 0. Also the definition of u is
slightly
modified and insted of(A.10)
we have 2u = alj.
To conclude we have to indicate the connection relation between thesymmetric
splitting
of eacheigenvalue
of the isolatedpotential well, Eo
~
io
"
Eo
+ to and thequantity
u. We
get
to =
flu IAiG)
where c
= I or 2 for the
#~
or the SG modelrespectively.
References
[II
Bak P., Rep,Prcgr. Phys.
45(1982)
587 and references therein.[2]
Aubry S.,
Solitons and CondensedMatter,
A.Bishop,
T. Schneider Eds.(Springer, Berlin)
Solid State Sci. 8(1978)
264.[3] Currie
J-F-, Trullinger S-E-, Bishop A-R-,
KrumhanslJ-A-, Phys.
Rev. 815(1977)
5567.[4] Andrew Combs
J., Yip S., Phys.
Rev, B 28(1983)
6873.[5] Ishimori
Y.,
MunakataT.,
J,Phys.
Sac.Jpn
51(1982)
3367.[6] Willis C., El-Batanouny M., Stancioif
P., Phys.
Rev. B 33(1986)
1904;Stancioif
P.,
WillisC., El-Batanouny M.,
BurdickS., Phys.
Rev. B 33(1986)
1912.[7]
Theodorakopoulos
N., Wunderlich W., Klein R., Solid State Commun. 33(1980)
213.[8] Peyrard
M.,
KruskalM-D-, Physica
13D(1984)
88.[9] Peyrard M., Pnevmaticos St., Flytzanis N., Physica 19D
(1986)
268.Flytzanis N.,
PnevmaticosSt., Peyrard M.,
J.Phys.A:
Math. Nucl. Gen. 22(1989)
783.[10] De Lillo S., Nuovo Cimento 100B
(1987)
105.[11] Ablowitz
M-J-, Segur H.,
Solitons and InverseScattering Transform,
SIAM J.Appl.
Math.(1981).
[12] Ablowitz
M-J-,
LadikJ.,
J. Math.Phys.
16(1975)
598; 17(1976)
loll.[13] Toda M.,
Theory
of Nonlinear Lattices(Springer, Berlin)
Solid State Sci. 20(1988).
[14] Krumhansl
J-A-,
SchrieiferJ-R-, Phys.
Rev. 811(1975)
3535.[15] Currie J-F-, Krumhansl
J-A-, Bishop A-R-, Trullinger S-E-, Phys.
Rev. B 22(1980)
477.[16]
Bishop A-R-,
KrumhanslJ-A-, Trullinger S-E-, Physica
ID(1980)
1.[17]
Guyer
R-A-, MillerM-D-, Phys.
Rev. A 17(1979)
1205.[18] Schneider
T.,
StollE., Phys.
Rev. B 22(1980)
5317.[19] De Leonardis
R-M-, Truliinger S-E-, Phys.
Rev. B 22(1980)
4558.[20] Croitoru
M.,
GrecuD.,
VisinescuA., Cionga V.,
Rev. Roum.Phys.
22(1984)
853.[21]
Trullinger
S-E-, Sasaki K., Physica 28D(1987)
181.[22] Weiss
G.H.~
Maradudin A.A.~ J. Math.Phys.
3(1962)
771.[23] Grecu D.~ Visinescu A.,
Phystca
44D(1990)
605j Rev. Roum.Phys.
35(1990)
45.[24] Grecu D.~ Visinescu
A.,
Rev.Roum; Phys.
35(1990)
3.[25] Hasan
Nayfeh A.,
Perturbation Methods(John Wiley,
New York~1973).
[26] Zinn-Justin J.~ Nucl.
Phys.
8192(1981)
125j 218(1983)
333j J. Math.Phys.
22(1981)
511j25
(1984)
549.[27]
Miyashita
T.~ MakiK., Phys.
Rev. B 23(1983)
6733j 31(1985)
1836.[28] Sasaki K.~
Prog.
Theor.Phys.
64(1982)
411j 70(1983)
593j 71(1984)
1169;Phys.
Rev. B 33(1986)
7743.[29]
Fogel
M-B-,Trullinger
S.E.~Bishop A-R-,
KrumhanslJ-A-, Phys.
Rev. 815(1977)
1578.[30] Abramowitz M.,
Stegun
I.A.~ Handbook of Mathematical Functions, Nat. Bur. ofStandards,
U-S- Handb.