Conductance of Interacting Quasi-One-Dimensional Electron Gas with a Scatterer
A. V. Borin1, 2 and K. E. Nagaev1, 2

arXiv:1404.0473v2 [cond-mat.mes-hall] 12 Jun 2014

1

Kotelnikov Institute of Radioengineering and Electronics, Mokhovaya 11-7, Moscow, 125009 Russia
2
Moscow Institute of Physics and Technology, Institutsky per. 9, Dolgoprudny, 141700 Russia
(Dated: May 19, 2018)
We calculate the conductance of a quantum wire with two occupied subbands in a presence of
a scatterer taking into account the interaction between electrons. We extend the renormalizationgroup equation for the scattering matrix of the obstacle to the case of intersubband interactions,
ﬁnd its ﬁxed points, and investigate their stability. Depending on the interaction parameters, the
conductance may be equal to 0, e2 /h, or 2e2 /h per spin projection. In some parameter ranges, two
stable ﬁxed points may coexist, so the ultimate conductance depends on the properties of the bare
scatterer. For spinful electrons, the conductance of the wire may nonmonotonically depend on the
Fermi level and temperature.
PACS numbers: 72.25.-b, 73.23.-b, 73.63.Rt

I.

INTRODUCTION

Transport properties of one-dimensional (1D) quantum
conductors are of signiﬁcant interest because they represent strongly correlated systems even for a weak electron
- electron interaction1 . Despite this, the conductance
of a uniform 1D wire smoothly connected to the leads
is not changed by the interaction and equals 2e2 /h2,3 .
Things are diﬀerent if the wire contains an inhomogeneity that can backscatter electrons. In the case of repulsive interaction, even a weak backscattering results in a
power-law temperature dependence of conductance and
its complete suppression at zero temperature4,5 . Qualitatively, this behaviour may be explained as follows. The
scattering oﬀ the inhomogeneity results in Friedel oscillations of electron density with a period (2kF )−1 and the
corresponding oscillations of potential energy of electrons
in the wire. As the interaction with potential oscillations
couples the electron states with +kF and −kF , it would
lead to a formation of the gap at the Fermi level and
a completely insulating state if the oscillations were of
constant amplitude. However the amplitude falls oﬀ inversely proportionally to the distance from the inhomogeneity, and this results only in a power-law suppression
of the conductance6 in long quantum wires at low temperatures. This suppression was observed in Refs. 7 and
8. On the other hand, the attractive interaction must
result in a perfect transmission of the barrier4 .
The properties of quantum wires with two populated
transverse-quantization levels are even more complicated
because of interaction between electrons in the diﬀerent subbands. There are theoretical predictions9–11 that
these electron systems may exhibit zigzag ordering similar to a Wigner crystal and several gapped phases may
exist in them under diﬀerent conditions. These phases
result in diﬀerent low-temperature behaviour of conductance of a wire with an impurity9,12 .
The above theoretical considerations are based on the
Luttinger-liquid model13 , which allows one to exactly
take into account the electron-electron forward scattering in the homogeneous wire. The elementary excitations

in this model are of bosonic rather than fermionic type
(plasmons and spinons). However the electron scattering
oﬀ an inhomogeneity cannot be treated exactly in this
model. In addition, it assumes a perfectly linear electron
spectrum and cannot be used if the distance from the
Fermi level to the bottom of the subband is comparable
with the interaction energy. The goal of this paper is to
use an approach similar to Refs. 14–17 and consider the
scattering oﬀ an inhomogeneity in a two-subband wire
in the fermionic representation. We start from the exact scattering states of noninteracting electrons in a wire
with the impurity and derive the renormalization-group
(RG) equations for its scattering matrix in the real space
to one-loop order by assuming the interaction to be weak.
Then we ﬁnd the stationary points of these equations and
investigate their stability. In the most general case, the
interaction is described by three parameters, which correspond to the coupling in each of the two subbands and the
intersubband coupling. For any set of parameter values,
there is at least one stable stationary point. Depending
on these values, it corresponds to 0, 1, or 2 conductance
quanta, so each of the two conducting channels is either
fully open or fully closed at this point. In certain ranges
of parameters, two diﬀerent stable ﬁxed points may coexist. This indicates that the ultimate conductance depends on the scattering matrix of the bare impurity. If
one of the channels tends to become fully transmitting,
and the other - fully reﬂecting, a nonmonotonic temperature dependence of the conductance may take place. Our
results diﬀer from those of Refs. 9 and 12.
The paper is organized as follows. In Section II, we
present the derivation of RG equations for the scattering
matrix of the inhomogeneity for arbitrary number of conducting channel and arbitrary parameters of intra- and
interchannel interaction. In Section III, we consider the
particular case of a quantum wire with two conducting
channels and ﬁnd the ﬁxed points of the RG ﬂow. We investigate the behaviour of this ﬂow near the ﬁxed points
and draw the diagram of their stability in the interactionparameter plane. In Section IV, we consider the interaction parameters for realistic systems with spinful elec-

2
trons. In Section V, the temperature dependence of the
conductance is discussed, and Section VI presents the
conclusion.
II.

RG EQUATIONS

Consider a long wire of uniform cross-section with a
local inhomogeneity. The wire has N transverse quantum channels corresponding to transverse energies Ei and
transverse wave functions φi (y), so that the total energy
of an electron is the sum of transverse and longitudinal
2
energies E = Ei + 2 ki /2m. The inhomogeneity is located at x = 0 and described by the scattering matrix S,
which relates the incoming and outgoing waves in diﬀerent quantum channels. This matrix is unitary because of
current conservation, and we assume both time-reversal
symmetry and a symmetric obstacle, so
S=

r t
,
t r

j

dr′ G(r, r′ ) VH(ex) (r′ ) Ψα (r′ ), (8)
i

δΨα
i,H(ex) (r) =

and G(r, r′ ) is the Green’s function of the single-particle
Schr¨dinger equation. The density matrix n(r, r1 ) is cono
veniently expressed in terms of the scattering states ΨL
i
and ΨR . At zero temperature, it is given by
i

(1)

where the r and t are symmetric N ×N matrices of reﬂection and transmission amplitudes. The scattering state
with the electron incident in channel i from the left is
described by the wave function
ΨL (r, E) =
i

ˆ
ˆ
n(r, r1 ) = Ψ+ (r1 ) Ψ(r) is the electron density matrix,
and V (r − r1 ) is the potential of the electron-electron
interaction. Typically, it is the Coulomb interaction
screened by the gate. The coeﬃcient of spin degeneracy
gs appears only in the direct term because it involves
interactions between electrons with both spin directions
while the exchange interaction is possible only for electrons with the same spin projection.
If the electron–electron interaction is considered as a
perturbation, Eqs. (6) and (7) result in a correction to
the wave function (2) in the form δΨα = δΨα − δΨα ,
i
i,ex
i,H
where

Θ(E − Ej ) L
ψij (x, E) φj (y),
2π vj

EF

dE Ψα (r, E) [Ψα (r1 , E)]∗ .
i
i

n(r, r1 ) =
i,α

−∞

(9)

If Ψα are unperturbed by the interaction and given by
i
Eqs. (2) - (3), the electron density far from the inhomogeneity may be presented in the form

(2)
n(r, r) = n0 +

δnij (x) φi (y) φj (y),

(10)

ij

where
L
ψij (x, E)

=

δij eiki x + rji e−ikj x ,
tji eikj x ,

x<0
.
x>0

(3)

and vj = −1 ∂E/∂kj . The scattering state with the
electron incident from the right ΨR (r, E) is given by a
i
R
L
similar expression with ψij (x, E) = ψij (−x, E). These
states obey the normalization condition
dr Ψα∗ (r, E) Ψβ (r, E ′ ) = δαβ δmn δ(E − E ′ )
m
n

(4)

with α, β = L, R and form a full orthogonal basis for
single-particle wave functions.
A weak electron-electron interaction results in a scattering of electrons by the Friedel oscillations caused by
the inhomogeneity. To take it into account, we solve a
Schr¨dinger-type equation
o
2

−

2m

∇2 + Vef f (r) Ψ(r) = E ψ(r)

Vex (r) ψ(r) =

(11)

The subscript F shows that the corresponding quantity
is evaluated at the Fermi level. Apart from the uniform
background n0 , n(r, r) also contains rapidly oscillating
components with wave vectors kiF + kjF , which slowly
decay with the distance from the inhomogeneity as 1/|x|.
The presence of these components results in both interchannel and intrachannel backward scattering. Note that
components with wave vectors kiF − kjF vanish because
of the unitarity condition r+ t + t+ r = 0.
The Green’s function is conveniently expanded in
transverse wave functions as
gij (x, x′ ) φi (y) φj (y ′ ),

G(r, r′ ) =

(12)

i,j

(5)

and obtain δΨ(r) in the lowest order in the interaction.
The interaction potential induced by the Friedel oscillations is a sum of a direct and exchange terms Vef f (r) =
VH (r) − Vex (r), where
VH (r) = gs

δnij (x) = |rij |
√
viF vjF sin[(kiF + kjF )|x| + arg rij ]
.
×
viF + vjF
π|x|

dr1 V (r − r1 ) n(r1 , r1 ),

(6)

dr1 V (r − r1 ) n(r, r1 ) ψ(r1 ),

(7)

where
gij (x, x′ ) =

i

1
√
vi vj
′

×

tij ei|ki x−kj x | ,
xx′ < 0
′
′
δij eiki |x−x | + rij ei|ki x+kj x | , xx′ > 0

(13)

To obtain the corrections to the entries of the S matrix,
one has to substitute Eqs. (9) - (13) into (8) and then
to expand the resulting δΨα in transverse wave functions
i

3
φj (y). The presence of terms like (11) in the integrand
gives rise to terms that slowly decay as 1/|x| with the
distance from the obstacle and a logarithmic divergency
of the integral at E → EF . Meanwhile it is precisely
these values of energy that determine the conductance of
the wire. To cure this divergency, we have to introduce
a long-distance cutoﬀ Lc and a short-distance cutoﬀ d,
which is of the order of the characteristic scale of the
interaction potential V (r) (e.g of the distance from the
conducting channel to the screening gate). As a result,
one obtains the corrections to the reﬂection and transmission amplitudes in the form
δrij =

1
αij rij −
2

∗
∗
αkl (rik rkl rlj + tik rkl tlj ) ln
kl

Lc
,
d
(14)

δtij = −

1
2

∗
∗
αkl (rik rkl tlj + tik rkl rlj ) ln
kl

Lc
,
d

(15)

where diﬀerent interactions are characterized by dimensionless parameters
αij = (2 − δij )

Vijij (0) − gs Vijij (kiF + kjF )
, (16)
π (viF + vjF )

which are expressed in terms of Fourier transforms of the
interaction potential
Vklij (x − x′ ) =

dy

Then the procedure is repeated again and again until S
becomes independent of L. By treating the ratio L/d as
a continuous variable14,15 and using Eqs. (14) and (15),
one obtains a diﬀerential equation for the scattering matrix in the form
dS
= SF + S − F,
dl

dy ′ φ∗ (y) φ∗ (y ′ )
k
i

× V (x − x′ , y − y ′ ) φj (y ′ ) φl (y).

FIG. 1. Diagrams for the corrections to the reﬂection and
transmission amplitudes. The vertices with wavy lines show
scattering by Friedel oscillations, and the black dots show the
scattering at the obstacle

(17)

In what follows, we neglect the renormalization of the
interaction parameters by backward scattering obtained
in Ref. 1 for spinful electrons because it plays no role in
RG equations written to ﬁrst order in the interaction.
The three terms in Eq. (14) are illustrated by the diagrams in Fig. 1. The ﬁrst term is presented by diagram
a and corresponds simply to backward scattering from
channel i to channel j by Friedel oscillations. The second term is presented by diagram b and corresponds to
reﬂection from state j to state l with subsequent backward scattering by Friedel oscillations from channel l to
channel k and reﬂection from channel k to channel i. The
third term is presented by diagram c and corresponds to a
transmission from channel j to channel k, backscattering
by Friedel oscillations from l to k, and then transmission from k to i. Similar processes for δtij are shown by
diagrams d and e.
As the ﬁrst-order correction to the transmission and reﬂection amplitudes logarithmically diverges, one has to
sum up the perturbative series to inﬁnite order. The
summation of most divergent contributions may be performed using the renormalization procedure suggested by
Yue et al.14 . The idea is to choose the length Lc in such a
way that the corrections δrij and δtij are still small and
to treat the inhomogeneity together with Friedel oscillation within this distance as a new composite scatterer.

(18)

where l = ln(L/d) and the block-diagonal matrix
F =

f 0
0 f

(19)

is formed of two identical N × N matrices with elements
fij = −αij rij /2. This equation is similar to the RG
equation obtained by Lal et al.15 for a junction of several
weakly interacting wires, but F contains now both diagonal and oﬀ-diagonal elements. It is easily veriﬁed that
the RG ﬂow preserves the unitarity and symmetry of S.
In general, the symmetric S matrix has N 2 + N different complex entries, which are reduced to the same
number of independent real quantities because of the unitarity condition. Note that the form of Eq. (18) remains
unchanged by the transformation
S ′ = U SU,

(20)

where U is a diagonal 2N × 2N matrix with elements
Uij = δij exp(iϕj ) and the real phases ϕj are independent of l and obey the condition ϕj+N = ϕj . Therefore
the matrices related by Eq. (20) belong to the same family and evolve precisely in the same way. This suggests
that the phases of N elements of matrices t and r with
diﬀerent indices may be chosen arbitrarily when studying
the RG ﬂow.

4
G/G0 No |r11 |2 |r12 |2 |r22 |2 |t11 |2 |t12 |2 |t22 |2

λ1
1
(α11
2

1

I

0

1

0

0

0

0

II

0

1

0

1

0

0

0

III

0

0

1

1

0

0

1
(α12
2

+ α22 ) −

1
− 2 (α11 + α22 )

IV

1

0

0

0

0

1

1
(α12
2

V

0

0

0

0

1

0

1
(α11
2

VI

2

λ2
α12 1 (α11
2

0

0

0

1

0

1

λ4

λ5

λ6

+ α22 ) − α12 −α12 −α12 −α12 0

α12 − 1 (α11 + α22 ) −α11 −α22
2

− α22 )

1
(α12
2

− α22 )

− α11 )

1
(α12
2

− α11 )

+ α22 )

1
(α11
2

+ α22 )

α11

λ3

α22

α11 −α22
−α11 α22

0

0

0

0

0

0

α12

0

0

0

α12

0

0

0

TABLE I. Interaction-independent ﬁxed points of RG. The number of conductance quanta per spin direction, the number of
the ﬁxed point, the scattering amplitudes, and the eigenvalues describing the behaviour of small deviations from these points.

III.

FIXED POINTS

Now we study the ﬁxed points of RG ﬂow Eq. (18) by
considering the interaction parameters as formal quantities of arbitrary sign. The RG equation for the singlechannel wire with a scatterer has only two ﬁxed points,
which correspond to r = 1 and t = 1. The former is stable and the latter is unstable for the repulsive interaction
while for the attractive one, they just exchange places.
In the case of two channels, the analysis is much more
complicated because it involves six diﬀerent entries of S
matrix and three diﬀerent interaction constants α11 , α22 ,
and α12 . According to Eq. (18), the ﬁxed points of RG
should be determined from the condition
SF + = F S + ,

(21)

which results in four quadratic equations for rij and tij .
They should be supplemented by six equations that follow from the unitarity of S matrix. Fortunately, all the
ﬁxed points can be found analytically even for this case.
It is easily shown (see Appendix) that all the three reﬂection amplitudes at the ﬁxed points can be chosen as real
numbers without a loss of generality. Any ﬁxed point
can be obtained from one with real rij by means of the
transform (20), which involves two independent phases
ϕ1 and ϕ2 .
Our analysis revealed three types of ﬁxed points. First
of all, there are four isolated universal points that do
not depend on the interaction parameters and correspond
to an integer number of conductance quanta (points I IV in Table I). The second type includes several oneparametric families of ﬁxed points that correspond to
smooth transitions between the points of the ﬁrst type
(see Appendix, Table II). One of these families with
|r11 | = |r12 | = |r22 | = 0 and |t11 |2 = |t22 |2 = 1−|t12 |2 corresponds to a transition between the point with |t12 | = 1
and the point with |t11 | = |t22 | = 1 (points V and VI
in Table I) and exists for any values of the interaction
parameters. The rest of one-parameter families emerge
only for deﬁnite relations between αij . Note the asymmetry between the transmission and reﬂection amplitudes.
There is also a third type of ﬁxed points. These points

are isolated, but the corresponding scattering amplitudes
depend on the interaction parameters (see Appendix).
To investigate the stability of ﬁxed points, we linearize
Eq. (18) in δrij and δtij near them and eliminate the
extra increments of these quantities using the unitarity
condition. The evolution of the remaining six increments
is determined by a 6 × 6 matrix, and the ﬁxed point is
stable if all the corresponding eigenvalues λi are not positive. The eigenvalues describing the behaviour of small
deviations of S matrix from the isolated ﬁxed points are
listed in Table I. In general, this matrix has six eigenvalues, but some of them are zero because they correspond
to the rotations of S matrix (20), which involve two independent phases ϕ1 and ϕ2 and leave Eq. (18) invariant. As S at points II - IV has two nonzero entries, there
are two independent phase transformations and only four
nonzero λ’s. At point I, S has only one nonzero entry,
so there is only one independent phase rotation and ﬁve
nonzero eigenvalues. Points V and VI belong to the oneparametric continuous family of ﬁxed points with zero
reﬂectance and diﬀerent |t11 | = |t22 |, and the three zero
eigenvalues at any point of this family correspond to variations of |tij | and two phase rotations. The three nonzero
eigenvalues for this family are given by equations
1
λ1 = (α11 + α22 ) (1 − |t12 |) + α12 |t12 |,
(22a)
2
1
λ2,3 = α11 + α22 + 2α12 + (α11 + α22 − 2α12 ) |t12 |
4
± (α11 + α22 ) (1 + |t12 |) + 2α12 (1 − |t12 |)

2

− 8 α12 (α11 + α22 )
+ |t12 |(2α11 α22 − α11 α12 − α22 α12 )

1/2

. (22b)

It is easily seen that ﬁxed points of this family may be
stable only if α12 < 0. If one goes from larger to smaller
values of α11 and α22 , the ﬁrst stable point with |t12 | = 1
(point V in Table I) appears at α11 + α22 = 0. The
range of stability broadens as these quantities decrease
and ﬁnally includes the extreme point with |t11 | = |t22 | =
1 (point VI in Table I) as both the conditions α11 < 0
and α22 < 0 are fulﬁlled.

5

FIG. 2. The diagram of stability of isolated ﬁxed points from Table I for negative and positive α12 in the α11 - α22 plane.

The regions of stability of the isolated ﬁxed points from
Table I for diﬀerent values of αij are shown in Fig. 2. The
darker areas correspond to lower conductance and the
lighter areas, to higher conductance. It is easily seen that
the diagram of stability in the α11 - α22 plane essentially
depends on the sign of α12 . For α12 < 0, the conductance
per spin projection varies from 0 to 2e2 /h as α11 and α22
decrease from ∞ to −∞. If only one of these quantities
tends to −∞ and the other is positive, two stable ﬁxed
points are possible. One of them is point III or IV with
a conductance e2 /h, and the other is point V with a
conductance 2e2 /h. Hence the actual conductance, i. e.
the ﬁnal point of RG ﬂow in these regions depends on
where it starts, i. e. on the scattering properties of the
bare inhomogeneity.
The diagram in α11 and α22 looks very diﬀerent for
α12 > 0. Surprisingly, the conductance is zero not only if
α11 and α22 are both positive, but also if they are both
negative. However it may equal e2 /h if these quantities
have opposite signs, so changing the sign of one of the
interaction parameters from negative to positive may increase the conductance. There are also regions where
stable points with G = 0 and G = e2 /h coexist.
To conclude this section, the stationary points of RG
for the two-subband wire are essentially the same as for
two independent 1D quantum wires, but the diagram of
their stability is radically diﬀerent from the latter case,
especially for positive α12 .

IV.

INTERACTION PARAMETERS FOR
SPINFUL ELECTRONS

So far we considered the interaction parameters αij as
arbitrary independent quantities. For actual physical interactions, this is not the case and the three parameters
are related with each other. Consider a typical case of
a quantum wire patterned in a two-dimensional electron

FIG. 3. The stability regions of ﬁxed points II and IV in the
EF - d plane for the harmonic conﬁning potential. The Fermi
energy EF is normalized to the band separation ω, and the
distance to the gates d is normalized to the eﬀective width of
the wire aef f = 2 /mω.

gas by means of electrostatic gates. The Coulomb interaction between the electrons is screened by the gates and
has the standard form
V (r) =

e2
e2
,
− √
ǫr
ǫ r2 + 4d2

(23)

where ǫ is the dielectric constant of the material and d
is the distance between the wire and the gates, which
determines the characteristic interaction length. It is evident that for spinless electrons with gs = 1, all the three
interaction parameters (16) are positive because Vijij is
a monotonically decreasing function of momentum transfer. In this case, ﬁxed point II with zero conductance is
the only stable one.
The case of spinful electrons with gs = 2 appears to be
more interesting. If the Fermi level crosses the upper subband not far from its bottom, k2F may be much smaller

6
than 1/d, so V2222 (2k2F ) ≈ V2222 (0) and α22 may be negative while α11 is positive. This is just the case where
ﬁxed point IV is stable, so there is a full reﬂection of electrons in the lower subband and full transmission in the
upper subband at zero temperature. We determined the
region of stability of this ﬁxed point by calculating the
interaction parameters for a quantum wire with parabolic
conﬁnement and band separation ω. The curve separating the regions of stability of ﬁxed points IV and II in the
EF - d plane obtained by numerical calculations is shown
in Fig. 3. The Fermi energy is normalized to the separation between the energy levels in the transverse harmonic
potential ω, and d is normalized to the eﬀective width
of the quantum wire aef f = 2 /mω. At small d, the
separation curve follows the law EF / ω ∝ (aef f /d)2 .
The transition from ﬁxed point IV to II may be observed by changing the Fermi level in the system by
means of the gate voltage. If α11 is positive for EF
ω,
the transmission of electrons in the lower subband is fully
suppressed. Crossing the bottom of the upper subband
by the Fermi level will bring the wire to the ﬁxed point
IV with a conductance 2e2 /h, and further increase of EF
will result in crossing the separation line between points
IV and II and returning to zero conductance. Therefore
the electron - electron interaction may lead to a spike in
the dependence of conductance on EF .

V.

TEMPERATURE DEPENDENCE OF THE
CONDUCTANCE

The interaction eﬀects in two-channel quantum wires
may manifest themselves as a nontrivial temperature dependence of the conductance. At a ﬁnite temperature
the RG ﬂow (18) must be stopped when the cutoﬀ length
reaches the distance at which the Friedel oscillations are
smeared out by thermal broadening of the Fermi step. In
the case of a multichannel wire, there is a set of lengths
LT = (viF + vjF )/T at which the interaction parameij
ters αij turn into zero. As L increases, Eq. (18) must
be solved separately in each interval between the consecutive LT . The interaction parameters corresponding to
ij
smaller LT must be set equal to zero and the value of S
ij
in the end of previous interval must be used as the initial
condition for the RG equation. The procedure must be
T
T
stopped as L reaches LT
max ≡ max(Lij ), and S(Lmax )
gives the conductance at temperature T .
If the RG ﬂow tends to point IV, the conductance of
the lower subband tends to zero, and the conductance of
the upper subband tends to unity. In the vicinity of this
point, the deviations from these quantities decrease with
increasing cutoﬀ length L according to power laws with
exponents λi from Table I. As the exponents for lower
and upper subbands are diﬀerent, a nonmonotonic temperature dependence of the conductance may be observed
if the initial scattering amplitudes are chosen appropriately.
The results of numerical simulations for the parameters

FIG. 4. The temperature dependence of dimensionless conductance for a quantum wire with ω/EF = 0.98 and
d/aef f = 1.5. The initial transmission amplitudes are r12 =
t12 = 0, t11 = 0.65i, r22 = 0.6i, r11 =
1 − |t11 |2 , and
2.
t22 = 1 − |r22 |

typical of GaAs heterostructures are shown in Fig. 4. The
Fermi energy is EF = 14 meV, the subband separation
is ω = 0.98 EF , and the distance to the gate is d =
1.5 aef f . A substitution of these values into Eqs. (17)
and (16) gives the interaction parameters α11 = 0.106,
α12 = −0.01, and α22 = −0.28. Assume that the bare
inhomogeneity is described by the scattering matrix with
r12 = t12 = 0, t11 = 0.65i, r22 = 0.6i, r11 = 1 − |t11 |2 ,
and t22 = 1 − |r22 |2 . A solution of Eq. (18) with these
initial conditions leads to a maximum in the conductance
at T ∼ 0.4 K.
VI.

CONCLUSION

In summary, we have considered the eﬀects of electron - electron interaction on the conductance of a quantum wire with an obstacle. The obstacle gives rise to
Friedel oscillations of electron density, which cause backward scattering of electrons both within the transverse
subbands and from one subband to another. In the case
of two populated transverse subbands, the conductance
of the wire is determined by two intraband and one interband interaction parameter. Each subband is either in a
fully insulating or fully conducting state as in the case of
zero interaction between them, but which of these states
will be realized depends on all the three interaction parameters. The wire has zero conductance if all the three
parameters are positive and maximum possible conductance 2e2 /h per spin direction if all of them are negative.
However the conductance is zero if the interband parameter is positive and both intraband parameters are negative. Changing the sign of one of them from negative to
positive results in an increase in the conductance from 0
to e2 /h per spin direction.
For spinful electrons with Coulomb interaction, the interaction parameters may change their signs depending
on the ﬁlling of the subbands. This may result in a non-

7
monotonic dependence of the conductance on the position of the Fermi level in the wire. The temperature
dependence of the conductance may be also nonmonotonic if the intraband interaction parameters have opposite signs.
Our results diﬀer from those obtained in the Luttingerliquid model because the limiting transitions are made in
a diﬀerent sequence. In the Luttinger-liquid papers, the
length of the wire was assumed to be inﬁnite from the
very beginning, and therefore the gaps in the spectrum
developed even for a weak interaction, though they were
exponentially small in this case. Nevertheless these gaps
were essential because the low-energy limit was considered there. In contrast to this, we perform calculations
for a wire of a ﬁnite length in the limit of weak interactions and only then let this length to inﬁnity. This is
why the gaps are washed out by a ﬁnite dwell time of
an electron in the wire. Similar results can be obtained
by starting with an inﬁnite wire if the temperature is
assumed to be much higher than the gap width.

Hence the condition Eq. (21) for the existence of a ﬁxed
point leads to a system of another four independent equations
∗
∗
∗
∗
α11 r11 r12 + α12 r12 r22 = α12 r11 r12 + α22 r12 r22
∗
∗
α11 r11 t∗ + α12 r12 t∗ = α11 t11 r11 + α12 t12 r12
11
12
∗
∗
∗
∗
α11 r11 t12 + α12 r12 t22 = α12 t11 r12 + α22 t12 r22
∗
∗
α12 r12 t∗ + α22 r22 t∗ = α12 t12 r12 + α22 t22 r22
12
22

Equation (18) is invariant with respect to the transform
(20), where


0
0
eiϕ1 0


0 
 0 eiϕ2 0
U =
(A5)

 0
0 eiϕ1 0 
0
0
0 eiϕ2
The reﬂection and transmission amplitudes are changed
by this transform as follows:

Appendix A: Interaction-dependent ﬁxed points of
RG

If the obstacle is symmetric and the number of channels
is N = 2, the scattering matrix has the general form


r11 r12 t11 t12


t
t 
r r
(A1)
S =  12 22 12 22  .
 t11 t12 r11 r12 
t12 t22 r12 r22

The unitarity condition SS + = 1 leads to a system of six
independent equations
|r11 |2 + |r12 |2 + |t11 |2 + |t12 |2 = 1
2

2

2

2

|r12 | + |r22 | + |t12 | + |t22 | = 1
∗
∗
r11 r12 + r12 r22 + t11 t∗ + t12 t∗ = 0
12
22
∗
∗
∗
∗
r11 t11 + r12 t12 + t11 r11 + t12 r12 = 0
∗
∗
r11 t∗ + r12 t∗ + t11 r12 + t12 r22 = 0
12
22
∗
∗
∗
∗
r12 t12 + r22 t22 + t12 r12 + t22 r22 = 0.

(A2a)
(A2b)
(A2c)
(A2d)
(A2e)
(A2f)

Matrix F in Eq. (18) is of the form


−α11 r11 −α12 r12
0
0

1 −α r −α22 r22
0
0


F =  12 12
.
 (A3)
2
0
0
−α11 r11 −α12 r12 
0
0
−α12 r12 −α22 r22

t′ = t11 e2iϕ1
11
t′ = t22 e2iϕ2
22

′
r12

We are grateful to S. T. Carr and B. N. Narozhny for
a useful discussion. This work was supported by Russian
Foundation for Basic Research, grant 13-02-01238-a, and
by the program of Russian Academy of Sciences.

′
r11 = r11 e2iϕ1 ,
′
r22 = r22 e2iϕ2 ,

ACKNOWLEDGMENTS

(A4a)
(A4b)
(A4c)
(A4d)

t′
12

= r12 e

iϕ1 +iϕ2

,

= t12 e

iϕ1 +iϕ2

(A6)
.

This allows us to make r11 and r22 real by appropriately
choosing ϕ1 and ϕ2 . Therefore Eq. (A4a) is brought to
the form
∗
(α11 − α12 ) r11 r12 + (α12 − α22 ) r22 r12 = 0,

(A7)

which suggests that r12 is either purely real or purely
imaginary. We choose it to be purely real because in the
opposite case, it can be made such just by adding π either
to ϕ1 or ϕ2 in the transform (20).
If one or more of the scattering amplitudes is zero, Eqs.
(A2) and (A4) are easily solved, and the corresponding
interaction-independent solutions are listed in Table I. In
addition to them, there are ﬁxed points that exist only
for the speciﬁc combinations of interaction parameters,
which are not considered here. Assume now that none of
the scattering amplitudes are zero. Express r22 in terms
of r11 by means of Eq. (A7), and the real and imaginary parts of t12 and t11 by means of Eqs. (A2d), (A2f),
(A4b), and (A4d). Substituting these quantities into the
imaginary part of Eq. (A2c) results in the equation
α22
α11 − α12
α22 (α11 − α12 )
1−
− 1−
α12
α22 − α12
α11 (α22 − α12 )
× Im t12 Re t12 = 0. (A8)
Hence for arbitrary interaction parameters, t22 has either
zero real or imaginary part. In the case of real t22 , the
solution is
(α11 − α12 )(α22 − α12 )
α11 + α22 − 2α12
α22 − α12
r11 = ±t22 =
α11 + α22 − 2α12
α11 − α12
,
t11 = ±r22 =
α11 + α22 − 2α12

r12 = ∓t12 = ±

(A9a)
(A9b)
(A9c)

8
Path

Condition

|r11 |2

|r12 |2

|r22 |2

5↔6
none
0
0
0
2↔4
α11 = 0
0÷1
0
1
3↔6
α11 = 0
0÷1
0
0
1↔5
α12 = 0
0
0÷1
0
1↔6
α12 = 0
0
0÷1
0
2↔3
α22 = 0
1
0
0÷1
4↔6
α22 = 0
0
0
0÷1
2 ↔ 5 α11 + α22 = 0 1 − |t12 |2 0 1 − |t12 |2
1 ↔ 2 α11 + α22 = 2α12 1 − |r12 |2 0 ÷ 1 1 − |r12 |2

|t11 |2

|t12 |2

1 − |t12 |2 0 ÷ 1
1 − |r11 |2
0
1 − |r11 |2
0
0
1 − |r12 |2
2
1 − |r12 |
0
0
0
1
0
0
1÷0
0
0

|t22 |2
1 − |t12 |2
0
1
0
1 − |r12 |2
1 − |r22 |2
1 − |r22 |2
0
0

TABLE II. One-parametric families of ﬁxed points of RG. The ﬁrst column indicates the extreme points of the family according
to Table I, and the second column shows the condition for the existence of the family.

t11 = i(−1)n
If t22 is imaginary, the scattering amplitudes are given
by equations

r11 =

α12 (α22 − α12 )
,
α2 − α11 α22
12

×

r22 =

α12 (α11 − α12 )
,
α2 − α11 α22
12

α12 (α22 + α11 − 2α12 )
, (A10d)
α11 (α22 − α12 ) + α22 (α11 − α12 )

(A10a)
t22 = i(−1)n
×

α22 (α11 − α12 )
α2 − α11 α22
12

α11 (α22 − α12 )
α2 − α11 α22
12

α12 (α22 + α11 − 2α12 )
, (A10e)
α11 (α22 − α12 ) + α22 (α11 − α12 )

(A10b)
t12 = i(−1)n+m
×

m

r12 = (−1)

×

1
2

3
4

5
6

7

8

9

α11 α22 (α11 − α12 )(α22 − α12 )
,
α2 − α11 α22
12

(A10c)

J. Solyom, Adv. Phys. 28, 201 (1978).
D. L. Maslov and M. Stone, Phys. Rev. B 52, R5539
(1995).
I. Saﬁ and H. J. Schulz, Phys. Rev. B 52, R17040 (1995).
C. L. Kane and M. P. A. Fisher, Phys. Rev. Lett. 68, 1220
(1992).
A. Furusaki and N. Nagaosa, Phys. Rev. B 47, 4631 (1993).
K. A. Matveev, D. Yue, and L. I. Glazman, Phys. Rev.
Lett. 71, 3351 (1993).
S. Tarucha, T. Honda, and T. Saku, Solid State Commun.
94, 413 (1995).
A. M. Chang, L. N. Pfeiﬀer, and K. W. West, Phys. Rev.
Lett. 77, 2538 (1996).
O.A. Starykh, D.L. Maslov, W. Husler, L.I. Glazman, Interactions and Transport Properties of Lower Dimensional
Systems, in: T. Brandes (Ed.), Lecture Notes in Physics,

α11 α22 (α11 − α12 )(α22 − α12 )
α2 − α11 α22
12

α12 (α22 + α11 − 2α12 )
, (A10f)
α11 (α22 − α12 ) + α22 (α11 − α12 )

where n, m = 1, 2. Fixed points (A9) and (A10) are
unstable for any interaction parameters.

10

11

12

13

14

15
16

17

vol. 544 (Springer, Berlin, 2000), p. 37.
J. S. Meyer, K. A. Matveev, and A. I. Larkin, Phys. Rev.
Lett. 98, 126404 (2007).
J. S. Meyer and K. A. Matveev, J. Phys.: Condens. Matter
21, 023203 (2009).
S. T. Carr, B. N. Narozhny, and A. A. Nersesyan, Phys.
Rev. Lett. 106, 126805 (2011); Annals of Physics 339, 22
(2013).
T. Giamarchi, Quantum Physics in One Dimension (Oxford University Press, Oxford, 2004).
D. Yue, L. I. Glazman, and K. A. Matveev, Phys. Rev. B
49, 1966 (1994).
S. Lal, S. Rao, and D. Sen, Phys. Rev. B 66, 165327 (2002).
D. G. Polyakov and I. V. Gornyi, Phys. Rev. B 68, 035421
(2003).
M. Titov, M. M¨ller, and W. Belzig, Phys. Rev. Lett. 97,
u
237006 (2006).