HAL Id: jpa-00247152
https://hal.archives-ouvertes.fr/jpa-00247152
Submitted on 1 Jan 1995
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.
Enhancement of Persistent Currents by Hubbard Interactions in Disordered 1D Rings: Avoided Level
Crossings Interpretation
Michael Kamal, Ziad Musslimani, Assa Auerbach
To cite this version:
Michael Kamal, Ziad Musslimani, Assa Auerbach. Enhancement of Persistent Currents by Hubbard
Interactions in Disordered 1D Rings: Avoided Level Crossings Interpretation. Journal de Physique I,
EDP Sciences, 1995, 5 (11), pp.1487-1499. �10.1051/jp1:1995212�. �jpa-00247152�
Classification
Physics
Abstracts72.10-d 71.27+a 72.15Rn
Enhancement of Persistent Currents by Hubbard Interactions in Disordered ID Rings: Avoided Level Crossings Interpretation
Michael
Kamal(~),
Ziad H.Musslimani(~)
and AssaAuerbach(~)
(~) Goldman, Sachs &
Co.,
85 BroadSt.,
New York, NY 10004, USA(~) Department
ofPhysics,
Technion-Israel Institute ofTechnology,
Haifa 32000, Israel(Received
26January1995, accepted
2July 1995)
Abstract. We study effects of local electron interactions on the persistent current of one dimensional disordered
rings.
For dilferent realizations of disorder we compute the ensemble average current as a function of Aharonov-Bohm flux to zeroth and first orders in the Hubbard interaction. We find thon thepersistent
current is enhancedby
on site interactions.Using
an avoided levelcrossings approach,
we deriveanalytic
formulas whichexplain
the numencal results ai weak disorder. The sameapproach
alsoexplains
the opposite effect(suppression)
found forspinless
fermion models with intersite interactions.1. Introduction
Recent
experiments
have found that small isolated metallicrings,
threadedby magnetic flux,
carry
persistent
currents.Although
the existence of this effect had beenanticipated
theoreti-cally
for many years[1-4],
themagnitude
of the observed current was found to be muchforger
thon
expected.
Experiments performed
onmoderately
disorderedrings
with small transverse width bave measured themagnetic
response of an ensemble of copperrings [si,
of isolatedgold loops [6],
and of a
dean,
almostone-dimensional, sample
[7].Trie ensemble
averaged persistent
current was found to bave anamplitude
of orderlo~~evf IL
and
periodicity
4lo/2, (where
L is the circumference of thering,
vf is the Fermivelocity,
e isthe electron
charge,
and 4lo " hle
is the fluxquantum).
Thesingle loops experiment reported finding
anunexpectedly large
current of order evfIL,
the free electronvalue,
anddisplaying
periodicity
in flux withperiod 4lo.
Theoretical
approaches
con be classified intonon-interacting
andinteracting
electrons ap-proaches.
Noninteracting
theories which take into account the effects of disorder and devia- tions from aperfect
ID geometry [8],predict
values of the averagepersistent
current that are smaller thonobserved,
but still within one or two orders ofmagnitude
of trieexperimentally
determined value of trie average current. More
surprisingly,
the theoretical estimates for thetypical
value of thepersistent
current, measured in thesingle loop experiment,
areperhaps
Q
Les Editions dePhysique
1995two or three orders of
magnitude
smaller than thereported
values. It seems thatagreement
betweentheory
andexperiment
worsens withincreasing
disorder. Thissuggests
thatperhaps
interaction corrections are
quantitatively important
in themoderately
disorderedregime.
A first guess would be that interactions enhance the
impurity scattering
and further s~tppress thepersistent
current, ashappens
for the conductance of the disorderedluttinger liquid [9].
However,
this intuitiveanalogy
may bequite misleading.
While the conductancedepends
on the Fermi surface
properties,
1-e-velocity
andscattering
rates, thepersistent
current isa sum of contributions from ail
occupied
states.Also, by
Galileaninvariance,
aninteracting
electronliquid
without disorder carries the same current as the noninteracting
gas, eventhough
its Fermi
velocity
may be renormalized. Thus it isplausible
that in the disorderedsystem,
interactions could work both ways, 1-e- either suppress or enhance thepersistent
currents under conditions which need to beexplored.
Exact
diagonalization
studies ofspmless
fermions with intersite interactions on smallrings [10, iii,
have found that weak interactions red~tce thepersistent
current below itsnoninteracting
value. The correlation between the non
interacting persistent
current and the first order(in interactions)
corrections have been evaluatednumerically
for the Hubbard model withspin by Ramin,
Reulet and Bouchiat(RRB) [13]
for various disorder realizations. RRB found that thenon
interacting
and interaction correction have the samesign,
which agrees with exact results for the disordered Hubbard model foundby
Giamarchi andShastry [12]. They
alsoexplain
thiseffect
by estimating
theground
state energy correction. In this paper[14]
we shall gain moreinsight
into thepuzzle
of interaction correctionsby arriving
at ananalogous expression
for the interaction correction to thepersistent
current and itsexplicit dependence
on disorderstrength.
In Section 2 we
diagonalize
thetight binding
Hamiltoniannumerically
for different realizations ofdisorder,
andcompute
thepersistent
current to zeroth and to first order in the Hubbard interactions. We find that interactions enhance thepersistent
current. In Section 3 we present the Avoided LevelsCrossing (ALC) theory,
whichprovides
ananalytic approximation
to thenumerical results. This is an
expansion,
at weakdisorder,
ofnearly degenerate eigenstates
at fluxes close to
points
of time reversaisymmetry.
The correlation between the zeroth and first order currents isexplained by
a commonmechanism,
1e. the avoided level crossings.The
opposite (suppression)
effect forspinless fermions,
is alsoexplained by
the ALCtheory
inSection 4. We conclude
by
a summary and future directions.2. Numerical Perturbation
Theory
We consider a
tight-binding
Hamiltoman on aperiodic
chair withrepulsive
on-site Hubbard interaction:H =
Ho
+ U~j
niinii,
il
1
Ho
=~je~~ + %)cÎ+i~cm
+h-c-j
+~ eicl~cm. 12)
m m
where
c)~
creates an electron at site1 withspin
a. 0ur unit of energy is thehopping
energy. fiis the dimensionless on-site disorder energy
uniformly
distributed in the interval[-W/2,W/2].
The chair has L sites,
N~ electrons,
the fluxthrough
its center isgiven by
4l. With theseconventions,
the energy spectrum isperiodic
in the enclosed flux withpenod #o,
and thecurrent is given
by
1(#)
=~
E(#), (3)
à#
W=0
,--,__ ,,-,,,
"'~
io(~)
W=4~/~o
Fig.
l. The ensembleaveraged noninteracting
current (ID) as a function of theapplied
flux forseveral
strengths
of disorderedpotential:
W=0,o-à,
I.o, 2.0 and 4.0. AIT curves are for a half-filled Iattice of six sites.where
Ej4l)
=
Eo14l)
+Ei14l)
+OjU~). j4)
Eo
is the exactground
state energy ofHo,
whosesingle
electroneigenstates
are determinednumerically. El
is the first order correction inU,
which is givenby
El
= U~
niiniii
nia =
l4~olCÎaCialifo) là)
where
~o
is the Fockground
state ofHo
Thusby diagonalizing Ho
we canreadily
obtainf(#)
=Io(#)
+Ii (#)
+tJ(U~ (6)
The effects of disorder on the ensemble
averaged
noninteracting
currentIo (#)
can be seen mFigure 1,
where it is shown as a function of flux for a half filled lattice of sixsites,
for different disorderstrengths
W. Disorder smoothens and reduces triemagnitude
of ID, and for W » it is dominatedby
the first harmonicsin(2ir#/#o).
In
Figure 2,
the average value ofIi (#)
isplotted.
There are several features of the first order interaction correction that deserve comment.First,
the mostimportant
observation is that itgeneraiiy
enhances the noninteracting
current for ail values of disorderstrength
and flux. That is to say: there is apositive
correlation between the noninteracting
current and the first ordercorrection,
jiojù)iijù))~,~ à
o.ii)
which is in
agreement
with the numerical results obtainedby
RRB for the same mortel[13].
This result is fourra to holà for ail reahzations of disorder which we have used.
Second,
in the limit of weakdisorder, Ii(#)
becomessingular
atdiscontinuity points
ofIo(#),
which are atW=1
Iii~)
w=~
~/~o
Fig.
2. The first order ensembleaveraged interacting
current(Ii)
as a function of theapphed
flux for disorderedpotential
values: W= I.o, 2.0 and 4.0. As before, ail curves are for a half-filled Iattice of six sites.fluxes
(m
+) )#o, (m#o
for even(odd)
number of filledorbitals,
where m areintegers. Finally,
we see the first order current also becomes dominated
by
its first harmonic atlarge
disorder.To
study
thescaling properties
of the current withsystem
size and disorderstrength,
it is convenient to characterize thestrength
of bothIo
andfi by
theiramplitude
at #=
#o/4.
And while it does Dot capture thesingular
nature offi (#)
near theregime
of very weakdisorder,
this characterization is still useful tostudy
thescaling properties
of the first order current.Previous studies have found
[4, loi
bothnumerically
andanalytically,
that theamplitude
of thenoninteracting
current,averaged
overdisorder,
behaves like:iloi*o/4)1
")) eXPi-L/1).
18) By fitting
to this functional form we extracted the localizationlength (
for differentsystem
sizes and disorder values.
Figure
3 shows the values of the localizationlength
obtained for sizesL=6,10,14,20
and25,
athalf-filling, averaged
over many(up
to fourthousand) impurity configurations
for each value of W. We findgood agreement
between Dur mferred value of the localizationlength
and the knownasymptotic
form m the weak disorder limit:(
=105/W~,
valid when W < 2ir for
large systems
athalf-filling.
There is less
agreement
in thestrongly
locahzed limit where the localizationlength
is known to behave like:(
=
(In(W/2) 1)~~.
We find a better fittaking (
=
Il In(W°.~/2.5)
for the L = 25 data. Thisdiscrepancy
could be due to the small sizesconsidered,
or a breakdown ofequation (8)
when(
is of orderunity.
We have calculated theamplitude fi(#
=#o/4)
asfunction of
strength
ofdisorder,
characterizedby
thescaling
parameterL/(,
for system sizesof
L=6,10,14,20,25.
At weak disorder(large (),
theamplitude
increases with thestrength
of disorderachieving
its maximum value at some intermediatestrength
of disorder(
m L).
In thisweakly
disorderedregime,
asingle scaling
function could be used to descnbe the results:)~
~AL/()/L. 19)
8.0
60
Înf
4.02.0
o-o
~ ~ ~ ~
W
Fig.
3. TheIogarithm
of the IocalizationIength (In(), implied by equation (8),
isplotted
as a function ofstrength
of disorder(W)
for half-filled rings with N=6, 10, 14, 20 and 25 sites. Data for the different system sizescollapse
weII onto asurgie
curve.This behavior is similar to that of the
amplitude
of thenoninteracting
currentgiven by (8).
Upon increasing
theimpunty scattering further,
the first order current decreases with disorder.When the
single partiale
wavefunctions aresufficiently locahzed, fi
cari be descnbedby
adifferent
scaling
form:Î~
'~
~~~~~~'
~~~~where g is a
decreasing
function of itsargument. Equation (10) suggests
that in the localized regimefi
dominatesfo
forlarge
systemsizei.
However for localizeddoubly occupied single
electron states
l~fiai~fiai lu ~niiniil ~fiai~fiai)
~(ii)
~
Thus even for weak
interactions,
the interaction corrections may belarger
than the non inter-acting
levelspacings
which go as1/L.
This invahdatespert~trbation theory
in U in the iocaiizedregime.
3. Avoided Level
Crossings Theory
Here we will discuss how the numerical results of the
previous
section can be understood interms of avoided level
crossings (ALC)
at weak disorder.First,
we denve the ALCapproxima-
tion for the
noninteracting
currentfo.
3.1. ÀLC THEORY FOR
Îo.
InFigure
4 we can see atypical
spectrum for atight-binding
JOURNAL DBFHYSIQUBL -T. 5,N°ii NOWMBBR 1995 59
z-o
i o
e~(§j)
o-o-i o
-z,o
~
i.5 -1 o -o 5 O.O o 5 o 5
ié/4Ùo
Fig.
4.Energy
Ievel spectrum of thetight-binding Hamiltoman(12)
as a function ofapphed
flux for a N=6ring
m the absence of disorder(solid Iines)
and for weak disorder(dotted Iines).
Hamiltonian of a 6 site
ring,
as a function of theapphed
flux. In the absence of disorder(W
=0)
theeigenenergies
areen(#)
= -2 cos
() in
+
)))
,
(12)
o
with
period #o
" 2ir. At
points
of time reversaisymmetry,
i-e- where # is aninteger multiple
of4lo/2,
levelcrossings
occurs between states ofopposite angular
momenta. Thenoninteracting persistent
current is a sum over the currents carriedby
ailoccupied levels,
fo(4l)
=
(
=
~j (en(4l), (13)
where n,s are the orbital and
spin index, respectively.
Thenonmteracting
currentfo(4l)
will be a smooth function of 4l away from thepoints
of levelcrossings. By symmetry
of # --4l,
any
pair
of levels cross withopposite slopes,
and thus ifthey
are bothfully occupied
their contribution to the total current cancels. Theonly nonvamshing
contribution comes from atopmost
level which is notcompensated by
itspartner.
Then the currentchanges
signabruptly
as the
occupation
moves from one branch to another. Thus for an odd number offully occupied
orbitals the current
discontinuity
occur at #=
(m +1/2)#o,
otherwise the discontinuities willoccur at #
=
m#o.
In between the discontinuities the current varieshnearly
with theflux,
whichexplains
thepenodic
sawtoothshape
forfo
as seen inFigure
1.Introducing
a smallamount of disorder lifts the
degeneracy
at the crossingsby
opening small gaps. This reducesthe
persistent
currentfo
smce theoccupied
levels have smallerslopes
near the former crossing points. Thisexplains
the behavior shown mFigure
1: weakimpunty scattenng
softens thediscontinuities and leads to an overall reduction of the
magnitude
of current. We cariquantify
this observation
by examining pairs
of levels with momenta eh which cross at #= 0. We
specialize
to the case of an eveii number of filled levels withequal occupations
for bath spin directions. Theunperturbed
energies are givenby
f+k(4l)
=
-2cos1+k
+ ~~L#o
~(14)
with k
=
2irm/L,
for apositive integer
m. Consider the effect of a weak randompotential
fi which lifts thedegeneracy
in the 2 x 2subspace
ofk, -k,
Ho
-~Îl ~lji~
lis)
~~~~~
j~
_~ =
( exp(-12kzi)ei. l~~~
L
i=1
From now on we will omit the
subcripts k,
-k offÙ.
Trie
eigenvalues
and normalizedeigenfunctions
of(là)
ai-e:e+iù)
.ifi+
=
fi (A+eik~
+B+e-ik~)
e-iù) ifi-
=fi (A-eik~
+B-e-ik~) (ii)
~vhere:
~~
4(Ù(~
~~~
~ ~ ~
fk
~-k +
IIe-k ~k)~
+~ÎÎ~Î~
~ '~y ~
~~~~ A
,
(18)
f-k fk +
IIe-k ~k)~
+4(V(~
and,
ek + e-k +
~/(ek e-k)~
+4(Ù(2
f+ "
,
(19)
2 Whicii
SatisfY.
~~j~)
+~~iù)
=
e~i#)
+e-ki*).
~~°~We see that the contribution of the
occupied
orbitals toIo
isunchanged by
weak disorder since:~"~
"
~~~il ~~
"~~~~ù~~~ l~~~
Let us now use
(13)
to calculateIo(#)
for free electrons with weak disorderby summing
over
occupied
levels. We tilt ail levels up tokf
=
2irmf IL, plus
two electrons atkf
so thatan even number of orbital levels is filled.
Although
~ve restrict our calculations to an even number of filledorbitals,
the ALCtheory
cari beeasily
extended to ail otherfillings
as follows:we note that the non
interacting
wavefunctions have two-fold spmdegeneracy.
Anunpaired
spin
in an open orbital will netproduce
any interaction correction to thepersistent
current since itsdensity
interacts with anapproximately
uniform coredensity.
For odd number ofdoubly occupied orbitals,
theimportant
level1 crossmgs occur at # i+)#o,
rather than near# 1 0. The total
noninteracting
current isgiven by
the contribution from the filled level pairs,together
with the contribution of lowest m= 0
(nondegenerate)
level and thetopmost
orbital level at mf:10
(#)
"~j If
+If"°
+I~~
m
" ~~
~~
~~
S
~~~Z ~~
~Î~ ~W l~~~
Neglecting
corrections of order1/L~,
we obtain the expression:~ ~
8irsin(ki) ~
# #
/
# ~L(V(
~°~
~~
~~L#o #o~ #o #o
~4irsin(kf)
'for # E
[-#o/2, #o/2].
This expression for ID is valid aslong
as the energy scale of the disorderis much smaller than the level
spacing
at the Fermi level:iii
« 47rsinjkf) /L. j24)
Using
the relation betweenVi
and the mean freepath
l~i definedby
the one dimensional Bornapproximation,
i~i e
~~ ~~~~ ~~
(25)
(V(~L According
to(24),
the ALCapproximation
is vahd for8iri~i/L / 1, (26)
which is the
ballistic,
or delocalizedregime(~
In terms of l~i one can writeI01*)
~~)jl~f) _~
*~
#
/
4~ 2 ~o
#~ $ -)
~
~
*o 8irl~i (27)
In order to compare
(27)
to the numerical result forequation (2),
one needs to determme theparameter
l~i Since fluctuations in ID for different disorder realizations arelarge,
we determine l~iby fitting (27)
to the numerical ensembleaveraged
ID Once l~i is determined for aparticular
disorder
realization,
we use it to evaluatefi
as shown below.3.2. ÀLC THEORY FOR
Ii.
We shall nowproceed
to use the sameapproximation
toexplain
the behavior offi
(~(~)There
is no intermediate "diffusiveregime"
between the locahzed and ballistic regimes m onedimension [15].
(~ This
approach
is similar to the discussion of Aharonov-Bohm oscillations of theparticipation
ratiom disordered rings [16].
According
to(5),
the first order energy is a sum of thedensity squared
at aII sites,Ei(4l)
= U
~j
n~in~i(28)
where for the
filling
of an even number of orbitals perspin,
thesingle
spmdensity
is,n~s(«)
=
lino(1)i~
+iJfim,,-(i)i~
+~É~ ~ iifim,«(1)i~- (29)
We wiII now show that the first order energy,
Ei(4l)
= U
~j
n~in~i(30)
is enhanced near Ievel
crossings.
For a
pair
offully occupied
Ievels onehas, by unitary
lifim,+(1)l~
+lifim,-(1)l~
=lifik(1)l~
+titi-k(1)l~
=). (31)
Consequently,
for weakdisorder,
aIIoccupied Ievels,
except for the Iast one, contribute constant values to the totaldensity:
n4 =
~~~
+
(ifi-(1)(~ (32)
L
or m terms of the coefficients
A,
B(17)
n~ =
~ [ ~2k~,
L~
~
~
~/
+
~~~5~ ~~~~' L The first order energy is thus
given by (32),
Ei(4l)/U
=
~~~~
+
~~~~~~~~~~~
(34) Using (17)
and(25)
we con writeEl (4l) /U
m~~~
+ l +~~~~'
~,
(35)
L
~
2L L 4lo
~ ~
Dilferentiating (35)
with respect to fluxyields
Ii(«)/U
-
ll~il Ill
+l~l~l Ill j~ (36)
In
Figure 5, (36)
iscompared
to the numencal ensembleaveraged
result forIi
for vanous values of disorder. For each disorderrealization,
l~j is determinedby fitting
the numerical and ALC results for the averageIo.
We see that for weak disorder there is asatisfying agreement
between the ALC
approximation
and the numencal results. InFigure 5c,
the disorder is toolarge,
and the ALCapproximation
faitsbadly.
Equation (36) explams
both thepositive
correlation betweenIo
andIi
of(7).
Since m one dimension the mean freepath (l~j)
and the IocahzationIength (~)
aillerby
aproportionality
factor of order unity
[15], (36)
agrees with theempirical scahng
form(9) iii Î/U
+w
f(L/l~j) IL
which was found
numerically
at weak disorder.W=025
ii(~)/U
~)
w~ow = o-s
Ii(~)/u
b)
~/~oW = 1
~~ ~mo
Fig.
5.Comparison
betweennumerically
determined interactions correction Ii(dashed fines)
and theanalytic
ALC result(36) (sohd fines). a) c)
show results for three values of disorderstrength
W.loi(W)
are determined byfitting
equation(27)
to the numerical disorderaveraged Io(4l).
Error bars depict fluctuations of numerical Ii for different disorder reahzations.4. Difference of Hubbard Model and
Spinless
FermionsIn recent papers
[10, iii,
small disorderedrings
ofspinless
electrons have beenexactly diago-
nalized. The
persistent
currents have beencomputed
as a function of disorder and interactionstrength.
In contrast to our result for the Hubbardmortel,
interactions have been seen to reduce thepersistent
current at weak disorder. Here weapply
the ALCapproach
toexplain
this apparent dilference between the mortels.
The
spinless
Hamiltonian H~ is givenby
H~ =
Hi +~jU~n~nj,
~
Hi
=-~j[e~~+(~c)~~c~+h.c.]+ ~je~c)c~. (37)
~ ~
The non
interacting
current,Ii,
isgiven by
half the value ofequation (13).
In the ALCapproximation,
it isgiven by
half the value ofequation (27).
Following
theanalogous
derivation ofIi,
we use(33)
to obtain:E((4l)
=
4mf~U(0)
+2(A- (~(B- (~U(2kf), (38)
where
Ù(k)
=j ~exp(-ikxj)U~, (39)
J#,
which
implies
E((4l)
=4m)U(0)
+Û(2kf)
+~~~~'
~L 4lo
I((4l)
=
~~j)~ U(2kf)
1 +~~~~' ())
~ ~(40)
o L
o
For the
spinless
nearestneighbor
case studied in referenceil Ii El (4)
" u
~j
il~n~+1>(41)
~
the
corresponding
Fourier coefficient isU(2kf)
= 2U
cos(2kf). (42)
Above a
quarter filling, kf
> x/4,
U isnegative. Consequently Ii
has theopposite sign
to that ofIi
inequation (36),
and toIi.
That is to say, trie persistent current of thespinless
fermion mortel issuppressed by
the interactions which is magreement
with the results obtainedby
reference
Ill, 13].
The difference between the mortels con be attributed to the different effects of interstte interactions
(in
thespmless mortel),
versus local interactions(in
the Hubbardmortel).
5. Remarks and Conclusions
We have
investigated
the first order effect of Hubbard interactions on thepersistent
current inone dimensional disordered
rings.
The
findings
con be summarized as follows.Ii
was round ta correlate insign
withIo,
a fact which goescontrary
to the naïve intuition that interactions suppress currents(e.g.
asthey
do for theconductivity
of theLuttinger
mortel with animpurity [9]).
We con understand this result
by observing
thatIi depends
oncharge
fluctuations which varystrongly
with flux neon thedegeneracy points,
which also determine thesign
andmagnitude
of
Io. Using
the avoided Ievelcrossings theory,
we obtainanalytical expressions
which fit the numerical results for bothIo
andIi
We con use the sametheory
toexplain why spinless
fermions with intersite interactions exhibitsuppression
of currents rather thon enhancement.We have found
numerically
thatIi (L If)
scalesdifferently
with the IocalizationIength
thonIo(L If),
it seems that effects of electron-electron interactions grow as disorder is increased.However,
first orderperturbation theory
in theweakly
disordered case isexpected
to holà aslong
as the interaction matrix elements do not exceed thesingle particle
Ievelspacings.
Thus weare restricted to the
regime
U < 1. We con draw on the exactdiagonalization
results[10, iii
where for weak
disorder,
the numerical currents are found to varyhnearly
with interactionstrength
in a sizeable regime. Thus we believe that first orderperturbation theory
should be valid forphysically interesting
interactionparameters.
It ,vould be very
satisfying
lia similaranalysis
coula beapplied
to theexperimentally
relevantcase of three dimensional
rings. Preliminary
resultsyield positive
correlations betweenIo
andIi,
but where asimple
mmdedapplication
of the ALCapproach
cannot descnbe the diffusiveregime of l~j < L. It would also be
important
to understand the effects of truelong
range Coulomb interactions with the screening andexchange
elfects which are absent in the Hubbard mortel.Acknowledgments
We thank E.
Akkermans,
R. Berkovits and Y. Gefen for useful discussions. Thesupport
of the Us~lsrael Binational ScienceFoundation,
the IsraeliAcademy
of Sciences and the Fund for Promotion of Research at Technion isgratefully acknowledged.
Part of this work has beensupported by
grant from USDepartment
ofEnergy,
DE-FG02-91ER45441. M.K. thanks theInstitute for Theoretical
Physics
at Technion for itshospitahty.
References
[ii Byers
N. andYang
C.N.,Phys.
Reu. lent. 7(1986)
46.[2] Bloch F.,
Phys.
Reu. A 137(1965)
787; 166(1968)
401.[3] Buttiker M.,
Imry
Y. and Landauer R.,Phys.
Lett. A 96(1983)
365.[4]
Cheung
H-F-, GefenY.,
Riedel E-K- and Shih W-H-,Phys.
Reu. B 37(1988)
6050.[Si Levy L.P-, Dolan G., Dunsmuir J. and Bouchiat H., Phys. Reu. Lent. 64
(1990)
2074.[6] Chandrasekhar V., Webb R-A-, Brady M.J., Ketchen M.B.,
Galager
W-J- and I<leinsasser A., Phys. Reu. Lent. 67(1991)
3578.[7]
Mailly
D.,Chapelier
C. and Benoît A.,Phys.
Reu. Lent. 70(1993)
2020.[8] Altshuler B., Gefen Y. and
Imry
Y.,Phys.
Reu. Lent. 66(1991)
88; vonOppen
F. and Riedel E.,Phys.
Reu. Lent. 66(1991)
84; Schmid A.,Phys.
Reu. Lent. 66(1991)
80; Akkermans E.,Eut.ophys.
Lent. 15(1991)
709.[9] Kane C.L. and Fisher
M.P.A., Phys.
Reu. lent. 68(1992)
1220.[loi
Abraham M. and Berkovits R.,Phys.
Reu. Lent. 70(1993)
1509.[iii
Bouzerar G., Poilblanc D. and Montambaux G.,Phys.
Reu. B 49(1994)
8258.[12] Giamarchi T. and
Shastry
S., preprint(1994).
[13] Ramm
M.,
Reulet B. and Bouchiat H., Phys. Reu. B 51(1995)
5582.[14] Much of this work is based on M. Kamal's Ph.D. Thesis, Harvard
University, July
1994(unpub- hshed).
[15]
Gogohn A.A., Phys. Rep.
86(1982)
1.[16] Carini