PHYSICAL REVIEW B, VOLUME 64, 085322

Mean-field theory of the Kondo effect in quantum dots with an even number of electrons
Mikio Eto1,2 and Yuli V. Nazarov1
1
Department of Applied Physics/DIMES, Delft University of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands
2
Faculty of Science and Technology, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan
共Received 9 January 2001; published 8 August 2001兲
We investigate the enhancement of the Kondo effect in quantum dots with an even number of electrons,
using a scaling method and a mean field theory. We evaluate the Kondo temperature T K as a function of the
energy difference between spin-singlet and -triplet states in the dot, ⌬, and the Zeeman splitting, E Z . If the
Zeeman splitting is small, E ZⰆT K , the competition between the singlet and triplet states enhances the Kondo
effect. T K reaches its maximum around ⌬⫽0 and decreases with ⌬ obeying a power law. If the Zeeman
splitting is strong, E ZⰇT K , the Kondo effect originates from the degeneracy between the singlet state and one
of the components of the triplet state at ⫺⌬⬃E Z . We show that T K exhibits another power-law dependence on
E Z . The mean field theory provides a unified picture to illustrate the crossover between these regimes. The
enhancement of the Kondo effect can be understood in terms of the overlap between the Kondo resonant states
created around the Fermi level. These resonant states provide the unitary limit of the conductance G⬃2e 2 /h.

DOI: 10.1103/PhysRevB.64.085322 PACS number共s兲: 73.23.Hk, 72.15.Qm, 85.35.Be

I. INTRODUCTION singlet and -triplet states are almost degenerate. We have

The Kondo effect observed in semiconductor quantum
dots has attracted a lot of interest.1–5 In a quantum dot, the
number of electrons N is fixed by the Coulomb blockade to
integer values and can be tuned by the gate voltage. Usually
the discrete spin-degenerate levels in the quantum dot are
consecutively occupied, and the total spin is zero or 1/2 for
an even and odd number of electrons, respectively. The
Kondo effect takes place only in the latter case. The spin 1/2
in the dot is coupled to the Fermi sea in external leads
through tunnel barriers, which results in the formation of the
Kondo resonant state at the Fermi level.6 – 8 The conductance
through the dot is enhanced to a value of the order of e 2 /h at
low temperatures of TⰆT K 共Kondo temperature兲.9–13 This is
called the unitary limit. When N is even, there is no localized
spin and thus the Kondo effect is not relevant.
Recently Sasaki et al. has found a large Kondo effect in
so-called ‘‘vertical’’ quantum dots with an even N.14 The
spacing of discrete levels in such dots is comparable with the
strength of electron-electron Coulomb interaction. Hence the
electronic states deviate from the simple picture mentioned
above.15,16 If two electrons are put into nearly degenerate
levels, the exchange interaction favors a spin triplet 共Hund’s
rule兲.15 This state is changed to a spin singlet by applying a
magnetic field perpendicularly to the dots, which increases
the level spacing. Hence the energy difference between the
singlet and triplet states, ⌬, can be controlled experimentally
by the magnetic field. The Kondo effect is significantly en-
hanced around the degeneracy point between the triplet and
singlet states, ⌬⫽0. Tuning of the energy difference between
the spin states is hardly possible in traditional Kondo sys-
tems of dilute magnetic impurities in metal and thus this
situation is quite unique to the quantum dot systems.
The Kondo effect in multilevel quantum dots has been
investigated theoretically by several groups.17–20 They have
shown that the contribution from multilevels enhances the
Kondo effect. In our previous paper,21 we have considered
the experimental situation by Sasaki et al. in which the spin-
calculated the Kondo temperature T K as a function of ⌬,
using the ‘‘poor man’s’’ scaling method.22–24 We have shown
that T K(⌬) is maximal around ⌬⫽0 and decreases with in-
creasing ⌬ obeying a power law, T K(⌬)⬀1/⌬ ␥ . The expo-
nent ␥ is not universal but depends on a ratio of the initial
coupling constants. Our results indicate that the Kondo effect
is enhanced by the competition between singlet and triplet
states, in agreement with the experimental findings.14
We have disregarded the Zeeman splitting of the spin-
triplet state, ⫺E ZM (M ⫽0,⫾1 is z component of the total
spin S⫽1), since this is a small energy scale in the experi-
mental situation, E ZⰆT K . 14 Pustilnik et al. have studied an-
other situation where the Zeeman effect is relevant, E Z
ⰇT K . 25 They have considered ‘‘lateral’’ quantum dots with
an even N, when the ground state is a spin singlet and the
first excited state is a triplet (⌬⬍0). By applying a quite
large magnetic field parallel to the dots, the Zeeman effect
reduces the energy of one component of the triplet state,
兩 SM 典 ⫽ 兩 11 典 , and finally makes it the ground state. At the
critical magnetic field of E Z⫽⫺⌬, the energy of the state
兩 11 典 is matched with that of the singlet state, 兩 00 典 . They
have found that a Kondo effect arises from the degeneracy
between the two states. This is contrast to the usual case with
spin 1/2, in which the Zeeman effect lifts off the degeneracy
of the spin states and, as a result, breaks the Kondo effect. A
similar idea has been proposed by Giuliano and
Tagliacozzo.26 Their mechanism might explain some experi-
mental results of the Kondo effect in quantum dots under
high magnetic fields.4,27 Indeed, this type of Kondo effect
has been reported in carbon nanotubes where the Zeeman
effect is stronger than in semiconductor heterostructures.28
The purpose of the present paper is to construct a general
theory for the enhancement of the Kondo effect in quantum
dots with an even number of electrons, with changing ⌬ and
E Z . Hence various experimental situations are analyzed in a
unified way. We adopt the poor man’s scaling method along
with the mean field theory. It is well known that the charac-
teristic energy scale of Kondo physics, the Kondo tempera-

0163-1829/2001/64共8兲/085322共11兲/$20.00 64 085322-1 ©2001 The American Physical Society

MIKIO ETO AND YULI V. NAZAROV
ture T K , is determined by all the energies from T K up to the
upper cutoff.7,8 By the scaling method, we can evaluate T K
共its exponential part at least兲 by taking all the energies
properly.22–24 When E Z is negligible, the energies from ⌬ to
the upper cutoff would feel fourfold degeneracy of the dot
states, 兩 1M 典 (M ⫽0,⫾1) and 兩 00 典 , which enhances the
Kondo temperature. With increasing ⌬, T K decreases by a
power law.21 We extend our previous calculations to the case
of E Z⫽⫺⌬ⰇT K which has been discussed by Pustilnik
et al.25 and Giuliano and Tagliacozzo.26 We take into account
the energies not only from T K to E Z , where only two degen-
erate states 兩 11 典 and 兩 00 典 are relevant, but also from E Z to
the upper cutoff, where the dot states seem fourfold degen-
erate. The latter energy region has been neglected in Refs. 25
and 26. In consequence we find a power-law dependence of
T K on E Z again.
The mean field theory of the Kondo effect was pioneered
by Yoshimori and Sakurai29 and is commonly used for the
Kondo lattice model.30 It is useful to capture main qualitative
features of the Kondo effect; renormalizability at the scale of
T K , resonances at the Fermi level, and resonant transmis-
sion. The simplicity and universality of the mean field theory
have driven us to apply it to the problem in question. Gen-
erally the Kondo effect gives rise to a many-body ground
state which consists of the dot states 兩 SM 典 ⫽ f †SM 兩 0 典 and the
conduction electrons ⌸c k†␴ 兩 0 典 . The total spin of this ground
state is less than the original spin S localized in the dot. The
binding energy is of the order of the Kondo temperature T K .
We take into account the spin couplings between the dot
states and conduction electrons, 具 f †SM c k ␴ 典 , by the mean field,
neglecting their fluctuations.31 These spin couplings give rise
to resonant states around the Fermi level ␮ with the width of
the order of T K . The conduction electrons can be transported
through the resonant levels, which yields the unitary limit of
the conductance G⬃2e 2 /h. For our study, the mean field
calculations have the following advantages. 共i兲 The enhance-
ment of T K by the competition between the singlet and triplet
states can be directly understood in terms of the overlap be-
tween their Kondo resonant states. 共ii兲 The power-law depen-
dence of T K on ⌬ or E Z is obtained, which is in accordance
with the calculated results by the scaling method. 共iii兲 The
mean field calculations are applicable to the intermediate re-
gions where two of T K , ⌬, and E Z , are of the same order.
The poor man’s scaling method hardly gives any results in
these regions. Hence we can examine the whole parameter
region of ⌬ and E Z by the mean field theory. The disadvan-
tage of the mean field calculations is that they only give
qualitative answers.31 Hence the mean field theory and scal-
ing method are complementary to each other for understand-
ing the Kondo effect.
We shall discuss the relation of our approach to the renor-
malization group analysis of the multilevel Kondo effect.23,24
Our model effectively reduces to the one with two channels
in the leads and spin-triplet 共and -singlet兲 state in the dot
when E ZⰆT K . The ground state of this model would be
believed to be a spin singlet, which corresponds to the full
screening of the dot spin. The poor man’s scaling approach
and our mean field theory, however, show a tendency to the
085322-2
PHYSICAL REVIEW B 64 085322
formation of the underscreened Kondo ground state with
spin 1/2. We should mention that the exact ground state can-
not be determined within the limits of the applicability of
these approaches. Pustilnik and Glazman have recently pro-
posed a different model for the ‘‘triplet-singlet Kondo
effect.’’32 In our notation, they set C 1 ⫽ 冑2,C 2 ⫽0 in Eq. 共4兲
for the singlet state. Their model can be directly mapped
onto a special case of the two-impurity Kondo model,33 for
which the ground state is a spin singlet. We are concerned
about the case of C 1 ⬇C 2 , and we find that the difference
between C 1 and C 2 reduces as a result of the
renormalization.21 This suggests that the case considered in
Ref. 32 is by no means a generic one.
This paper is organized as follows. Our model is pre-
sented in the next section. In Sec. III, we rederive T K(⌬)
when the Zeeman splitting is irrelevant, using the poor man’s
scaling method, in a simpler form than our previous work.21
Then we extend our calculations to the case of E Z⫽⫺⌬
ⰇT K . Section IV is devoted to the mean field theory for the
Kondo effect in quantum dots. First we explain this theory
for the usual Kondo effect in a quantum dot with S⫽1/2.
Then we apply the mean field scheme to our model with an
even number of electrons in the dot. The conclusions and
discussion are given in the last section.
II. MODEL
We are interested in the competition between the spin-
singlet and -triplet states in a quantum dot. To model the
situation, it is sufficient to consider two extra electrons in a
quantum dot at the background of a singlet state of all other
N⫺2 electrons, which we will regard as the vacuum 兩 0 典 .
These two extra electrons occupy two levels of different or-
bital symmetry.34 The energies of the levels are ␧ 1 and ␧ 2 .
Possible two-electron states are 共i兲 the spin-triplet state, 共ii兲
the spin-singlet state of the same orbital symmetry as the
triplet state, 1/冑2(d 1↑ d 2↓ ⫺d 1↓
† † † †
d 2↑ ) 兩 0 典 , and 共iii兲 two singlet
states of different orbital symmetry, d 1↑ d 1↓ 兩 0 典 , d 2↑
† † † †
d 2↓ 兩 0 典 .
Among the singlet states, we only consider a state of the
lowest energy, which belongs to group 共iii兲. Thus we restrict
our attention to four states, 兩 SM 典 :
兩 11 典 ⫽d 1↑
† †
d 2↑ 兩 0 典 , 共1兲
1
兩 10 典 ⫽ † †
共 d 1↑ d 2↓ ⫹d 1↓
† †
d 2↑ 兲 兩 0 典 , 共2兲
冑2
兩 1⫺1 典 ⫽d 1↓
† †
d 2↓ 兩 0 典 , 共3兲
1
兩 00 典 ⫽ † †
共 C 1 d 1↑ d 1↓ ⫺C 2 d 2↑
† †
d 2↓ 兲 兩 0 典 , 共4兲
冑2
where d i†␴ creates an electron with spin ␴ in level i. The
coefficients in the singlet state, C 1 and C 2 ( 兩 C 1 兩 2 ⫹ 兩 C 2 兩 2
⫽2), are determined by the electron-electron interaction and
one-electron level spacing ␦ ⫽␧ 2 ⫺␧ 1 . We set C 1 ⫽C 2 ⫽1.
This is the case for ␦ ⫽0.35 Although C 1 ⫽C 2 in general, the
scaling analysis shows that the Kondo temperature is the
MEAN-FIELD THEORY OF THE KONDO EFFECT IN . . . PHYSICAL REVIEW B 64 085322

⫺E(N)⫿␮, where ␮ is the Fermi energy in the leads, are

FIG. 1. 共a兲 The energy diagram for the spin states 兩 SM 典 consid-
ered in our model. ⌬⫽E 00⫺E S⫽1 and E Z is the Zeeman splitting.
共b兲 Spin-flip processes between the spin states. The exchange cou-
plings J (i) involving the spin-triplet state only are accompanied by
the scattering of conduction electrons of channel i. Those involving
both the spin-triplet and -singlet states, J̃, are accompanied by the
interchannel scattering of conduction electrons.
same as that in the case of C 1 ⫽C 2 ⫽1, apart from a
prefactor.21 The energies of the triplet state are given by
positive. We are interested in the case where E ⫾ Ⰷ 兩 ⌬ 兩 , ␦ and
also exceed the level broadening ⌫ i ⫽ ␲␯ V 2i and temperature
T 共Coulomb blockade region兲. In this case we can integrate
out the states with one or three extra electrons. This is
equivalent to Schrieffer-Wolff transformation that is used to
obtain the conventional Kondo model.7,8 We obtain the fol-
lowing effective low-energy Hamiltonian
⬘ .
H eff⫽H leads⫹H dot⫹H S⫽1 ⫹H S⫽1↔0 ⫹H eff 共9兲
The Hamiltonian of the dot H dot reads
H dot⫽ 兺 E SM f †SM f SM ,
S,M
共10兲
using pseudofermion operators f †SM ( f SM ) which create 共an-
nihilate兲 the state 兩 SM 典 . The condition of

E S⫽1,M ⫽E S⫽1 ⫺E ZM , 共5兲 兺

SM
f †SM f SM ⫽1 共11兲
and the energy of the singlet state is denoted by E 00 . We
define ⌬ by should be fulfilled. The third term H S⫽1 represents the spin-
flip processes among three components of the spin-triplet
⌬⫽E 00⫺E S⫽1 . 共6兲 state. This resembles a conventional Kondo Hamiltonian for
The energy diagram for the spin states is indicated in Fig. S⫽1 in terms of the spin operator Ŝ:
1共a兲.
The dot is connected to two external leads L, R with free
electrons being described by H S⫽1 ⫽ 兺 兺 (i)† (i)
J (i) 关 Ŝ ⫹ c k ⬘ ↓ c k↑
(i)† (i)
⫹Ŝ ⫺ c k ⬘ ↑ c k↓
(i)† (i)
⫹Ŝ z 共 c k ⬘ ↑ c k↑
kk ⬘
i⫽1,2

(i)†
⫺c k ⬘ ↓ c k↓
兺 兺 ␧ (i)k c ␣(i)†,k ␴ c ␣(i),k ␴ ,
(i)
H leads⫽ 兲兴
␣ ⫽L,R k ␴ i

,k ␴ (c ␣ ,k ␴ ) is the creation 共annihilation兲 operator of

where c ␣(i)† (i) ⫽ 兺 兺 J (i) 关 冑2 共 f 11
†
f 10⫹ f 10
† (i)† (i)
f 1⫺1 兲 c k ⬘ ↓ c k↑
an electron in lead ␣ with momentum k, spin ␴ , and orbital kk ⬘
i⫽1,2

symmetry i(⫽1,2). The density of states ␯ in the leads re- ⫹ 冑2 共 f 10

(i)†
†
f 11⫹ f 1⫺1
† (i)
f 10兲 c k ⬘ ↑ c k↓ ⫹ 共 f 11
†
f 11⫺ f 1⫺1
†
f 1⫺1 兲
mains constant in the energy band of 关 ⫺D,D 兴 . The tunnel-
(i)† (i)†
ing between the dot and the leads is written as ⫻共 c k ⬘ ↑ c k↑
(i)
⫺c k ⬘ ↓ c k↓
(i)
兲兴 . 共12兲

The exchange coupling J (i) is accompanied by the scattering

H T⫽ 兺 兺 共 V ␣ ,i c ␣(i)†,k ␴ d i ␴ ⫹H.c.兲 .
␣ ⫽L,R k ␴ i of conduction electrons of channel i. The fourth term
H S⫽1↔0 in H eff describes the conversion between the spin-
We assume that the orbital symmetry is conserved in the triplet and -singlet states accompanied by the interchannel
tunneling processes.34 To avoid the complication due to the scattering of conduction electrons
fact that there are two leads ␣ ⫽L,R, we perform a unitary
transformation for electron modes in the leads along the
lines of Ref. 9; c k(i)␴ ⫽(V L,i (i)
* c L,k ␴ ⫹V R,i
(i)
* c R,k (i)
␴ )/V i , c̄ k ␴
H S⫽1↔0 ⫽ 兺 J̃ 关 冑2 共 f 11
†
f 00⫺ f 00
† (1)† (2)
f 1⫺1 兲 c k ⬘ ↓ c k↑ ⫹ 冑2 共 f 00
†
f 11
⬘
⫽(⫺V R,i c L,k ␴ ⫹V L,i c R,k ␴ )/V i , with V i ⫽ 冑兩 V L,i 兩 ⫹ 兩 V R,i 兩 2 .
(i) (i) 2 kk
(1)† (1)†
The modes c̄ k(i)␴ are not coupled to the quantum dot and shall ⫺ f 1⫺1
† (2)
f 00兲 c k ⬘ ↑ c k↓ ⫺ 共 f 10
†
f 00⫹ f 00
† (2)
f 10兲共 c k ⬘ ↑ c k↑
be disregarded hereafter. Then H leads and H T are rewritten as (1)†
⫺c k ⬘ ↓ c k↓
(2)
兲 ⫹ 共 1↔2 兲兴 . 共13兲

H leads⫽ 兺
k␴i
␧ (i) (i)† (i)
k c k␴ c k␴ , 共7兲 The coupling constants are given by

V 2i
J (i) ⫽ , 共14兲
H T⫽ 兺
k␴i
␴ d i ␴ ⫹H.c. 兲 .
V i 共 c k(i)† 共8兲 2E c

We assume that the state of the dot with N electrons is V 1V 2

J̃⫽ , 共15兲
stable, so that addition/extraction energies, E ⫾ ⬅E(N⫾1) 2E c

085322-3
MIKIO ETO AND YULI V. NAZAROV PHYSICAL REVIEW B 64 085322

where 1/E c⫽1/E ⫹ ⫹1/E ⫺ . Note that J̃ 2 ⫽J (1) J (2) . The last
⬘ represents the scattering processes of conduction
term H eff
electrons without any change of the dot state and is not rel-
evant for the current discussion. The spin-flip processes in-
cluded in our model are shown in Fig. 1共b兲.

III. SCALING CALCULATIONS

In this section we calculate the Kondo temperature T K

using the poor man’s scaling technique.22–24 By changing the
energy scale 共bandwidth of the conduction electrons兲 from D
to D⫺ 兩 dD 兩 , we obtain the scaling equations using the
second-order perturbation calculations with respect to the ex-
change couplings, J (1) , J (2) , and J̃. With decreasing D, the
exchange couplings are renormalized. The Kondo tempera-
ture is determined as the energy scale at which the exchange
couplings become so large that the perturbation breaks down.

A. In the absence of Zeeman effect

When the Zeeman splitting is small and irrelevant, E Z
ⰆT K , we obtain a closed form of the scaling equations for
J (1) , J (2) , and J̃ in two limits.21 共i兲 When the energy scale D FIG. 2. The scaling calculations of the Kondo temperature T K as
is much larger than the energy difference 兩 ⌬ 兩 , H dot can be a function of ⌬, 共a兲 when the Zeeman splitting is irrelevant, E Z
safely disregarded in H eff . The scaling equations can be writ- ⰆT K , and 共b兲 in a case of E Z⫽⫺⌬. D 0 is the bandwidth in the
ten as leads. In both the figures, curve a, ␪ / ␲ ⫽0.25; curve b, 0.15; and
curve c, 0.10, where tan2 ␪ ⫽J (2) /J (1) .
d
d ln D J̃ 冉
J (1) J̃
J (2)
冊 冉
⫽⫺2 ␯
J (1)
J̃
J̃
J (2)
冊 2

. 共16兲 In the intermediate region of T K(0)Ⰶ⌬ⰆD 0 , the ex-

change couplings develop by Eq. 共16兲 for DⰇ⌬. Around D
共ii兲 For DⰆ⌬, the ground state of the dot is a spin triplet and ⫽⌬, J̃ saturates while J (1) and J (2) continue to grow with
the singlet state can be disregarded. Then J (1) and J (2) decreasing D, following Eq. 共17兲 for DⰆ⌬. We match the
evolve independently, solutions of these scaling equations at D⯝⌬ and obtain a
power law of T K(⌬)
d 2
J (i) ⫽⫺2 ␯ J (i)2 , 共17兲 T K共 ⌬ 兲 ⫽T K共 0 兲关 T K共 0 兲 /⌬ 兴 tan ␪ , 共21兲
d ln D
with
whereas J̃ does not change.
In the case of 兩 ⌬ 兩 ⰆT K , the scaling equations 共16兲 remain tan ␪ ⫽J̃/ 关 冑共 J (1) ⫺J (2) 兲 2 /4⫹J̃ 2 ⫹ 共 J (1) ⫺J (2) 兲 /2兴 ⫽ 冑J (2) /J (1)
valid till the scaling ends. The matrix in Eq. 共16兲 has eigen- 共22兲
values of
for J (1) ⭓J (2) . Here (cos ␪,sin ␪)T is the eigenfunction of the
J ⫾ ⫽ 共 J (1) ⫹J (2) 兲 /2⫾ 冑共 J (1) ⫺J (2) 兲 2 /4⫹J̃ 2 ⫽J (1) ⫹J (2) ,0. matrix in Eq. 共16兲 corresponding to J ⫹ . ␪ ⬃0 for J (1)
共18兲 ⰇJ (2) and ␪ ⫽ ␲ /4 for J (1) ⫽J (2) . In general, 0⬍ ␪ ⭐ ␲ /4 and
thus 0⬍tan2 ␪ ⭐1.
The larger one, J ⫹ , diverges upon decreasing the bandwidth Finally, for ⌬⬍0, all the coupling constants saturate and
D and determines T K : no Kondo effect is expected, provided 兩 ⌬ 兩 ⰇT K(0). Thus T K
quickly decreases to zero at ⌬⬃⫺T K(0). The Kondo tem-
T K共 0 兲 ⫽D 0 exp关 ⫺1/2␯ J ⫹ 兴 ⫽D 0 exp关 ⫺1/2␯ 共 J (1) ⫹J (2) 兲兴 . perature as a function of ⌬ is schematically shown in Fig.
共19兲 2共a兲.
Here D 0 is the initial bandwidth, which is given by
冑E ⫹ E ⫺ .36 B. Case of E ZÄÀ⌬
When ⌬⬎D 0 , the scaling equations 共17兲 work in the When E Z⫽⫺⌬, the energies of states 兩 00 典 and 兩 11 典 are
whole scaling region. This yields degenerate. Then the Kondo effect is expected even when
兩 ⌬ 兩 ⰇT K(0).25,26 In this subsection we evaluate T K in the
T K共 ⬁ 兲 ⫽D 0 exp关 ⫺1/2␯ J (1) 兴 共20兲
special case of E Z⫽⫺⌬ by the poor man’s scaling method.
for J ⭓J . This is the Kondo temperature for spin-triplet
(1) (2) 共i兲 For the energy scale of DⰇ 兩 ⌬ 兩 ⫽E Z , H dot can be dis-
localized spins.37 regarded in H eff . The exchange couplings, J (1) , J (2) , and J̃,

085322-4
MEAN-FIELD THEORY OF THE KONDO EFFECT IN . . . PHYSICAL REVIEW B 64 085322

evolve following Eq. 共16兲. 共ii兲 In another limit of DⰆ 兩 ⌬ 兩

⫽E Z , only the states 兩 00 典 and 兩 11 典 are relevant. In H eff , H⫽ 兺
k␴
␧ k c k†␴ c k ␴ ⫹ 兺 E ␴ f ␴† f ␴ ⫹J 兺 兺
␴
†
f ␴† f ␴ ⬘ c k ⬘ ␴ ⬘ c k ␴ ,
kk ␴ , ␴⬘ ⬘
共28兲
00 典 , 兩 11 典
H 兩eff ⫽ 兺 兺 关 1
2 J s(i) 共 f 11
†
f 11⫺ f 00
† (i)† (i)
f 00兲共 c k ⬘ ↑ c k↑ with the constraint of
⬘
kk i⫽1,2
(i)†
⫺c k ⬘ ↓ c k↓ 兲 ⫹ 21 J (i)
(i)† f †↑ f ↑ ⫹ f †↓ f ↓ ⫽1. 共29兲
c 共 f 11 f 11⫹ f 00 f 00 兲共 c k ⬘ ↑ c k↑
(i) † † (i)

For electrons in leads L, R, we have performed a unitary

(i)† (i)
⫺c k ⬘ ↓ c k↓ 兲兴 ⫹ 兺 冑2J̃ 关 f 11
† (1)† (2)
f 00c k ⬘ ↓ c k↑ transformation of c k ␴ ⫽(V L* c L,k ␴ ⫹V R* c R,k ␴ )/ 冑兩 V L 兩 2 ⫹ 兩 V R 兩 2
kk ⬘ where V ␣ is the tunneling coupling to lead ␣ .9 The last term
(1)† (2) in Eq. 共28兲 represents the exchange coupling between S
⫹ f 00
†
f 11c k ⬘ ↑ c k↓ ⫹ 共 1↔2 兲兴 . 共23兲
⫽1/2 in the dot and conduction electrons 共see Appendix A兲.
J s(i) ⫽J (i) In the mean field theory, we introduce the order parameter
c ⫽J
(i)
initially. The scaling procedure yields
1
d
J (i) ⫽⫺4 ␯ J̃ 2 ,
d ln D s
具⌶典⫽
冑2 兺k 共 具 f †↑ c k↑ 典 ⫹ 具 f †↓ c k↓ 典 兲 共30兲

to describe the spin couplings between the dot states and

d
J̃⫽⫺ ␯ 共 J s(1) ⫹J s(2) 兲 J̃, 共24兲 conduction electrons. The mean field Hamiltonian reads
d ln D

and J (i)
c do not change. These scaling equations are nearly H MF⫽ 兺
k␴
␧ k c k†␴ c k ␴ ⫹ 兺 E ␴ f ␴† f ␴ ⫺ 兺 共 冑2J 具 ⌶ 典 c k†␴ f ␴
␴ k, ␴
equivalent to those of the anisotropic Kondo model with S
⫽1/2,22 as pointed out in Refs. 25,26.
When 兩 ⌬ 兩 ⫽E Z⬎D 0 , the scaling equations 共24兲 remain
valid in the whole scaling region. This yields the Kondo
⫹H.c.兲 ⫹2J 兩 具 ⌶ 典 兩 2 ⫹␭ 冉兺 ␴
f ␴† f ␴ ⫺1 . 冊 共31兲

temperature The constraint, Eq. 共29兲, is taken into account by the last
term with a Lagrange multiplier ␭. By minimizing the ex-
T K共 ⬁ 兲 ⫽D 0 exp关 ⫺A 共 ␪ 兲 /2␯ 共 J (1) ⫹J (2) 兲兴 共25兲 pectation value of H MF , 具 ⌶ 典 is determined self-consistently
共Appendix A兲.
with

再
In the absence of the Zeeman effect, E ↑ ⫽E ↓ ⫽E 0 . The

A共 ␪ 兲⫽
1
␭
ln 冉 冊
1⫹␭
1⫺␭
共 0⬍ ␪ ⭐ ␲ /8兲
共26兲
mean field Hamiltonian, H MF , represents a resonant tunnel-
ing through an ‘‘energy level,’’ Ẽ 0 ⫽E 0 ⫹␭, with ‘‘tunneling
coupling,’’ Ṽ⫽⫺ 冑2J 具 ⌶ 典 . Ṽ provides a finite width of the
2 ⫺1 ˜ 0 ⫽ ␲␯ 兩 Ṽ 兩 2 , with ␯ being the density of states in
tan ␭ 共 ␲ /8⬍ ␪ ⭐ ␲ /4兲 , resonance, ⌬
␭ the leads. The constraint, Eq. 共29兲, requires that the states for
where ␭⫽ 冑兩 cos 4␪兩.38 A( ␪ ) decreases monotonically with the pseudofermions are half-filled, that is, Ẽ 0 ⫽ ␮ . Hence the
increasing ␪ . A( ␪ )→⬁ as ␪ →0. A( ␲ /8)⫽2 and A( ␲ /4) Kondo resonant state appears just at the Fermi level ␮ , as
⫽ ␲ /2. When J (1) ⫹J (2) is fixed, T K(⬁) is the largest at indicated in the inset 共A兲 in Fig. 3共a兲. The self-consistent
J (1) ⫽J (2) ( ␪ ⫽ ␲ /4) and becomes smaller with decreasing calculations give us the resonant width
J (2) /J (1) (⫽tan2 ␪ ).
In the intermediate region, T K(0)Ⰶ 兩 ⌬ 兩 ⫽E ZⰆD 0 , we ˜ 0 ⫽ ␲␯ 兩 冑2J 具 ⌶ 典 兩 2 ⫽D 0 exp关 ⫺1/2␯ J 兴 .
⌬ 共32兲
match the solutions of Eqs. 共16兲 and 共24兲 at D⯝ 兩 ⌬ 兩 . We This is identical to the Kondo temperature T K .
obtain a power law In the presence of the Zeeman splitting, E ↑ ⫽E 0 ⫺E Z and
E ↓ ⫽E 0 ⫹E Z . Hence the resonant level is split for spin-up
T K共 ⌬ 兲 ⫽T K共 0 兲关 T K共 0 兲 / 兩 ⌬ 兩 兴 A( ␪ )⫺1 . 共27兲
and -down electrons, Ẽ ↑/↓ ⫽E ↑/↓ ⫹␭. The constraint, Eq.
Figure 2共b兲 shows the behaviors of T K(⌬) in the case of 共29兲, yields E 0 ⫹␭⫽ ␮ 关see inset 共B兲 in Fig. 3共a兲兴. The reso-
E Z⫽⫺⌬. nant width ⌬ ˜ is determined as

IV. MEAN FIELD CALCULATIONS ˜ 2 ⫹E Z2⫽⌬

⌬ ˜ 20 , 共33兲
A. Kondo resonance for spin SÄ1Õ2 where ⌬˜ 0 is given by Eq. 共32兲. The Kondo temperature is
To illustrate the mean field theory for the Kondo effect in evaluated by this width, T K(E Z)⫽⌬˜ . T K decreases with in-
quantum dots, we begin with the usual case of S⫽1/2. We creasing E Z and disappears at E Z⫽T K(0), as shown in Fig.
assume that one level (E 0 ) in a quantum dot is occupied by 3共a兲.
an electron with spin either up or down ( ␴ ⫽↑,↓). The ef- The conductance G through the dot is expressed, using
fective low-energy Hamiltonian is ⌫ ␣ ⫽ ␲␯ 兩 V ␣ 兩 2 , as

085322-5
MIKIO ETO AND YULI V. NAZAROV PHYSICAL REVIEW B 64 085322

ជ †典 ⌶
H MF⫽H lead⫹H dot⫺J MF关 具 ⌶ ជ ⫹⌶
ជ †具 ⌶
ជ 典⫺兩具⌶
ជ 典兩2兴

⫹␭ 冉兺
SM
f †SM f SM ⫺1 ,冊 共36兲

where

J MF⫽J (1) ⫹ 冑J (1)2 ⫹3J̃ 2 共37兲

and

tan ␸ ⫽ 冑3J̃/J MF . 共38兲

The last term in H MF considers the restriction of Eq. 共11兲.

The expectation value of H MF is minimized with respect to
ជ 兩 2 . The Kondo temperature can be estimated by
兩⌶

ជ 典兩2,
T K⫽ ␲␯ 兩 J MF具 ⌶ 共39兲

using 具 ⌶ ជ 典 determined by the self-consistent calculations

FIG. 3. The mean field calculations for the Kondo effect in a
quantum dot with S⫽1/2. 共a兲 The Kondo temperature T K and 共b兲
conductance through the dot, G, as functions of the Zeeman split-
ting E Z . T K and E Z are in units of D 0 exp(⫺1/2␯ J) and G is in units
of (2e 2 /h)4⌫ L ⌫ R /(⌫ L ⫹⌫ R ) 2 . Inset in 共a兲: The Kondo resonant
states created around the Fermi level ␮ in the leads, 共A兲 in the
absence and 共B兲 presence of the Zeeman splitting. The resonant
width is given by T K .
共Appendix B兲.
First let us consider the case in the absence of the Zeeman
effect, E 1M ⫽E S⫽1 and E 00⫽E S⫽1 ⫹⌬. The resonant level
for the triplet state is threefold degenerate at Ẽ S⫽1 ⫽E S⫽1
⫹␭, whereas the resonant level for the singlet state is at
Ẽ 0 ⫽E 00⫹␭. These levels are separated by the energy ⌬.
The Lagrange multiplier ␭ is determined to fulfill Eq. 共11兲.
Figure 4共a兲 shows the calculated results of T K as a function
of ⌬. Both of T K and ⌬ are in units of D 0 exp(⫺1/␯ J MF). We

冋 冉 冊册
2e 2 4⌫ L ⌫ R EZ 2
G⫽ 1⫺
h 共 ⌫ L ⫹⌫ R 兲 2 T K共 0 兲
This is the conductance in the unitary limit for E Z⫽0. Figure
3共b兲 presents the E Z dependence of the conductance. With
increasing E Z , the splitting between the resonant levels for
spin up and down becomes larger. In consequence the am-
plitude of the Kondo resonance decreases at ␮ , which re-
duces the conductance.
B. Kondo resonance in the present model
Now we apply the mean field theory to our model which
has the spin-triplet and -singlet states in a quantum dot. The
spin states of the coupling to a conduction electron are (S
⫽1) 丢 (S⫽1/2)⫽(S⫽3/2) 丣 (S⫽1/2) for the former, and
(S⫽0) 丢 (S⫽1/2)⫽(S⫽1/2) for the latter 共Appendix B兲. To
represent the competition between the triplet and single
states, therefore, the order parameter should be a spinor of
S⫽1/2. It is 具 ⌶ជ 典 where
find that 共i兲 T K(⌬) reaches its maximum at ⌬⫽0, 共ii兲 for
. 共34兲 ⌬ⰇT K(0), T K(⌬) obeys a power law
2
T K共 ⌬ 兲 ⌬ tan ␸ ⫽const., 共40兲
and 共iii兲 for ⌬⬍0, T K decreases rapidly with increasing 兩 ⌬ 兩
and disappears at ⌬⫽⌬ c⬃⫺T K(0):
2
⌬ c⫽⫺D 0 exp共 ⫺1/␯ J MF兲共 1⫹tan2 ␸ 兲共 tan2 ␸ 兲 ⫺sin ␸ .
共41兲
These features are in agreement with the results of the scal-
ing calculations.
The behaviors of T K(⌬) can be understood as follows.
The inset of Fig. 4共a兲 schematically shows the Kondo reso-
nant states. The resonance of the triplet state is denoted by
solid lines, whereas that of the singlet state is by dotted lines.
共A兲 When ⌬ⰇT K(0), the triplet resonance appears around
␮ , whereas the singlet resonance is far above ␮ . 共B兲 With a
decrease in ⌬, the two resonant states are more overlapped at

冉 冊
␮ , which raises T K gradually. This results in a power law of
cos ␸ 共 冑2 f 11 c k↑ ⫹ f 10
† (1)
c k↓ 兲 / 冑3⫹ 共 sin ␸ 兲 f 00
† (1) † (2)
c k↓ T K(⌬), Eq. 共40兲. The largest overlap yields the maximum of
ជ⫽
⌶ 兺k cos ␸ 共 冑2 f 1⫺1
† (1)
c k↓ ⫹ f 10 c k↑ 兲 / 冑3⫺ 共 sin ␸ 兲 f 00
† (1) † (2)
c k↑ T K at ⌬⫽0. 共C兲 When ⌬⬍0, the singlet and triplet reso-
共35兲 nances are located below and above ␮ , respectively, being
sharper and farther from each other with increasing 兩 ⌬ 兩 . Fi-
for J (1) ⬎J (2) . A mode of the largest coupling is taken into nally the Kondo resonance disappears at ⌬⫽⌬ c .
account in this approximation. The Hamiltonian reads The conductance through the dot is given by

085322-6
MEAN-FIELD THEORY OF THE KONDO EFFECT IN . . . PHYSICAL REVIEW B 64 085322

FIG. 4. The mean field calculations for the Kondo effect in the
present model. The Zeeman splitting is disregarded (E ZⰆT K). 共a兲
The Kondo temperature T K and 共b兲 conductance through the dot, G,
as functions of ⌬⫽E 00⫺E S⫽1 . T K and ⌬ are in units of
D 0 exp(⫺1/␯ J MF). G, in units of 2e 2 /h, is evaluated in a symmetric
case of ⌫ Li ⫽⌫ Ri (i⫽1,2). tan ␸ ⫽ 冑3J̃/J MF where curve a, ␸ / ␲
⫽0.25; curve b, 0.15; and curve c, 0.10. Note that ␸ / ␲ ⭐1/6 in this
approximation 共case a is only for reference兲. Inset in 共a兲: The
Kondo resonant states for S⫽1 共solid line兲 and for S⫽0 共dotted
line兲 when 共A兲 ⌬ⰇT K(0), 共B兲 ⌬⬃T K(0), and 共C兲 ⌬⬍0.
FIG. 5. The mean field calculations for the Kondo temperature
T K in the present model. ␸ / ␲ ⫽0.15 where tan ␸ ⫽ 冑3J̃/J MF . All of
E Z , ⌬, and T K are in units of D 0 exp(⫺1/␯ J MF). 共a兲 T K is plotted in
E Z-⌬ plane, by contour lines drawn every 0.25. The lighter shade
indicates the larger values of T K . 共b兲 T K as a function of ⌬ when
E Z is fixed at curve a, 0; curve b, 1; curve c, 2; curve d, 5; and
curve e, 10. The broken line indicates T K in the case of ⫺⌬

冉
⫽E Z .
4⌫ L1 ⌫ R1 ⌬ 2
˜ 11
G/ 共 e 2 /h 兲 ⫽
共 ⌫ L1 ⫹⌫ R1 兲 2 共 ␧⫺Ẽ 11兲 2 ⫹⌬ 2
˜ 11 of ␸ ⫽0.15␲ . Figure 5共b兲 presents T K as a function of ⌬ for

冊
several values of E Z . When E Z is large enough, the Kondo
⌬ 2
˜ 10
effect takes place only when the resonant state of 兩 11 典 is
⫹
共 ␧⫺Ẽ 10兲 2 ⫹⌬ 2
˜ 10 overlapped with that of 兩 00 典 . Then T K is the largest at ⌬

冏
⫽⫺E Z and decreases with ⌬ being away from this value. At
4⌫ L2 ⌫ R2 ⌬ 2
˜ 00 ⌬⫽⫺E Z , T K obeys a power law
⫹ , 共42兲
共 ⌫ L2 ⫹⌫ R2 兲 2 共 ␧⫺Ẽ 00兲 2 ⫹⌬ 2
˜ 00
␧⫽ ␮
2␸)
T K共 ⌬ 兲 兩 ⌬ 兩 1/(2⫹3 tan ⫽const., 共43兲
where ⌫ ␣i ⫽ ␲␯ 兩 V ␣ ,i 兩 2 . The resonant widths are ⌬ ˜ 11 /⌬
˜0
⫽(2 cos2␸)/3, ⌬ ˜ 10 /⌬˜ 0 ⫽(cos2␸)/3, and ⌬ ˜ 0 ⫽sin2␸ with
˜ 00 /⌬ which is indicated by a broken line in Fig. 5共b兲. This is
˜ ជ
⌬ 0 ⫽ ␲␯ 兩 J MF具 ⌶ 典 兩 . The conductance G as a function of ⌬ is
2
qualitatively in agreement with the calculated results by the
shown in Fig. 4共b兲, in a symmetric case of ⌫ Li ⫽⌫ Ri (i scaling method.
⫽1,2). G⫽2e 2 /h for ⌬⬎0, whereas G goes to zero sud- Figure 6 indicates the conductance G in E Z-⌬ plane,
denly for ⌬⬍0. Around ⌬⫽0, G is larger than the value in when ␸ ⫽0.15␲ and ⌫ Li ⫽⌫ Ri (i⫽1,2). G takes the value of
the unitary limit, 2e 2 /h, which is attributable to nonuniversal 2e 2 /h around E Z⫽0 and ⌬⬎0, and also along the line of
contribution from the multichannel nature of our model.21 E Z⫽⫺⌬. (G⬎2e 2 /h in the neighborhood of E Z⫽⌬⫽0, as
In the presence of the Zeeman splitting, E 1M ⫽E S⫽1 discussed above.兲 For sufficiently large E Z , our model is
⫺E ZM , the resonant level of the triplet state is split into nearly equivalent to the anisotropic Kondo model with S
three. With increasing E Z , the Kondo effect is rapidly ⫽1/2.25,26 Hence G⫽2e 2 /h at ⌬⫽⫺E Z and reduces to zero
weaken except in the region of ⌬⬃⫺E Z . In Fig. 5共a兲, we as ⌬ deviates from this value, in the same way as in Fig. 3共b兲
show the Kondo temperature T K in E Z-⌬ plane, in the case for the case of S⫽1/2.

085322-7
MIKIO ETO AND YULI V. NAZAROV
FIG. 6. The mean field calculations for the conductance G in the
present model. G is plotted in E Z-⌬ plane, by contour lines drawn
every 0.2(2e 2 /h). The lighter shade indicates the larger values of
G. E Z and ⌬ are in units of D 0 exp(⫺1/␯ J MF). ␸ / ␲ ⫽0.15 where
tan ␸ ⫽ 冑3J̃/J MF , and ⌫ Li ⫽⌫ Ri (i⫽1,2).
V. CONCLUSIONS AND DISCUSSION
The Kondo effect in quantum dots with an even number
of electrons has been investigated theoretically. The Kondo
temperature T K has been calculated as a function of the en-
ergy difference ⌬⫽E 00⫺E S⫽1 and the Zeeman splitting E Z ,
using the poor man’s scaling method and mean field theory.
The scaling calculations have indicated that the competition
between the spin-triplet and -singlet states significantly en-
hances the Kondo effect. When the Zeeman effect is irrel-
evant, E ZⰆT K , T K is maximal around ⌬⫽0 and decreases
with ⌬ obeying a power law. In a case of ⫺⌬⫽E Z , the
Kondo effect takes place from the degeneracy between two
states, 兩 00典 and 兩 11典 . Even in this case, the contribution from
the other states of higher energy, 兩 10典 and 兩 1 ⫺1 典 , plays an
important role in the enhancement of T K . As a result, T K is
maximal around E Z⫽0 and depends on E Z by a power law
again.
The mean field theory yields a clear-cut view for the
Kondo effect in quantum dots. Considering the spin cou-
plings between the dot states and conduction electrons as a
mean field, 具 f †SM c k, ␴ 典 , we find that the resonant states are
(i)
created around the Fermi level ␮ . The resonant width is
given by the Kondo temperature T K . The unitary limit of the
conductance, G⬃2e 2 /h, can be easily understood in terms of
the tunneling through these resonant states. In our model, the
overlap between the resonant states of S⫽1 and S⫽0 in the
dot enhances the Kondo effect. The mean field calculations
have led to a power-law dependence of T K on ⌬ and on E Z ,
in accordance with the scaling calculations.
The mean field theory is not quantitatively accurate for
the evaluation of T K . 31 共In the case of S⫽1/2, the exact
value of T K is obtained accidentally.兲 In our model, the scal-
ing calculations indicate that all the exchange couplings,
J (1) , J (2) , and J̃, are renormalized altogether following Eq.
共16兲 when 兩 ⌬ 兩 and E Z are much smaller than the energy scale
D. In consequence two channels in the leads are coupled
085322-8
PHYSICAL REVIEW B 64 085322
effectively for an increase in T K . In the mean field calcula-
tions, the interchannel couplings are taken into account in
Eq. 共37兲 only partly. In fact, conduction electrons of channel
1 and 2 independently take part in the conductance, Eq. 共42兲.
By the perturbation calculations with respect to the exchange
couplings, we find that mixing terms between the channels
appear in the logarithmic corrections to the conductance.21
We could improve the mean field calculations by adopting
another form of the order parameter than Eq. 共35兲.
Our calculated results being expressed in terms of ⌬, E Z ,
and T K are applicable to any experimental realization, either
at small or large magnetic field B. At small B, our results
explain the experimental findings by Sasaki et al.:14 The
Kondo effect is largely enhanced around ⌬⫽0 when ⌬ is
tuned by the orbital effect of the magnetic field. ⌬⫽0 at B
⬇0.2 T, where the Zeeman effect can be safely disregarded.
At large B, the Zeeman effect splits three components of the
spin-triplet state. The Kondo effect has been observed in car-
bon nanotubes at B⬇1 T (E Z⫽g ␮ BB with g⫽2.0) where
the energy of one component of the triplet state coincides
with that of the singlet state.28 The Kondo effect induced by
the Zeeman effect may be observed in quantum dots on
semiconductor heterostructures with a smaller g factor (g
⬇0.4), under higher magnetic fields.25,26 In such experi-
ments, the value of the Zeeman splitting E Z can be controlled
by applying a large magnetic field parallel to the dot while
⌬⫽E 00⫺E S⫽1 by a small magnetic field perpendicularly to
the dot.
In this paper, we have restricted our considerations to the
space of C 1 ⫽C 2 in Eq. 共4兲. This is a good starting point in
the vicinity of ⌬⫽0.35 Investigation in the space of C 1
⫽ 冑2 and C 2 ⫽0, however, has shown different fixed points
of the renormalization flow in cases of E Z⫽0 32 and E Z
⫽⫺⌬ⰇT K . 25 In possible realizations, C 2 ⬇C 1 for small 兩 ⌬ 兩
and C 2 ⰆC 1 for large 兩 ⌬ 兩 ,38 and our calculations could be
incomplete in the latter region. The crossover between the
regions requires further studies. A generalized renormaliza-
tion analysis is in progress for this problem.
ACKNOWLEDGMENTS
The authors are indebted to L. P. Kouwenhoven, S. De
Franceschi, J. M. Elzerman, K. Maijala, S. Sasaki, W. G. van
der Wiel, Y. Tokura, L. I. Glazman, M. Pustilnik, and G. E.
W. Bauer for valuable discussions. The authors acknowledge
financial support from the ‘‘Netherlandse Organisatie voor
Wetenschappelijk Onderzoek’’ 共NWO兲. M. E. is also grateful
for financial support from the Japan Society for the Promo-
tion of Science for his stay at Delft University of Technol-
ogy.
APPENDIX A: MEAN FIELD CALCULATIONS FOR SÄ1Õ2
The original Hamiltonian for a quantum dot with one en-
ergy level reads
H⫽ 兺 兺 ␧ k c ␣† ,k ␴ c ␣ ,k ␴ ⫹ ␣ ⫽L,R
␣ ⫽L,R k ␴
兺 兺 k␴
共 V ␣ c ␣† ,k ␴ d ␴ ⫹H.c.兲
⫹H dot 共A1兲
MEAN-FIELD THEORY OF THE KONDO EFFECT IN . . . PHYSICAL REVIEW B 64 085322

with This yields E 0 ⫹␭⫽0. The minimization of E MF with respect

˜ 共or 兩 具 ⌶ 典 兩 2 ) determines ⌬
to ⌬ ˜
H dot⫽ 兺␴ E 0 d ␴† d ␴ ⫹Ud †↑ d ↑ d †↓ d ↓ . 共A2兲
⳵ E MF 1 Ẽ ␴2 ⫹⌬
˜2 1
For the state of one electron in the dot, the addition and ⳵⌬˜
⫽
2␲ 兺␴ ln
D 20
⫹
␲␯ J
⫽0. 共A7兲

extraction energies are given by E ⫹ ⫽E 0 ⫹U⫺ ␮ and E ⫺

For E Z⫽0, we find
⫽ ␮ ⫺E 0 , respectively. The parameters, E 0 and U, in Eq.
共A2兲 should be determined to fit these energies to experimen- ˜ ⫽D 0 exp关 ⫺1/2␯ J 兴 ⬅⌬
⌬ ˜ 0. 共A8兲
tal data. For conduction electrons in leads L, R, we perform a
unitary transformation, c k ␴ ⫽(V L* c L,k ␴ ⫹V R* c R,k ␴ )/V, c̄ k ␴ This is equal to the Kondo temperature, T K . For E Z⫽0, Eq.
⫽(⫺V R,i c L,k ␴ ⫹V L,i c R,k ␴ )/V, with V⫽ 冑兩 V L 兩 2 ⫹ 兩 V R 兩 2 , 共A7兲 yields
along the lines of Ref. 9. We disregard the modes c̄ k ␴ that are ˜ 2 ⫹E Z2⫽⌬
⌬ ˜ 20 . 共A9兲
uncoupled to the quantum dot.
We consider the Coulomb blockade region for one elec-
tron, where both E ⫹ and E ⫺ are much larger than the level Using the T matrix, T̂, the conductance through the dot,
broadening ⌫⫽ ␲␯ V 2 ( ␯ being the density of states in the G, is given by

冏
leads兲 and temperature. Integrating out the dot states with e2
zero or two electrons by the Schrieffer-Wolff G⫽
h
共 2 ␲␯ 兲 2 兺␴ 兩 具 R,k ⬘ ␴ 兩 T̂ 兩 L,k ␴ 典 兩 2
transformation,7,8 we obtain the effective low-energy Hamil- ␧ k ⫽␧ k ⬘ ⫽ ␮
tonian
e2 兩 V L兩 2兩 V R兩 2
⫽ 共 2 ␲␯ 兲 2
H⫽ 兺
k␴
␧ k c k†␴ c k ␴ ⫹ 兺 E ␴ f ␴† f ␴ ⫹J 兺 关 Ŝ ⫹ c k ⬘ ↓ c k↑
␴
† h 共 兩 V L兩 2⫹ 兩 V R兩 2 兲 2

冏
kk ⬘
† † †
⫹Ŝ ⫺ c k ⬘ ↑ c k↓ ⫹Ŝ z 共 c k ⬘ ↑ c k↑ ⫺c k ⬘ ↓ c k↓ 兲兴 共A3兲 ⫻ 兺␴ 兩 具 ␺ k ⬘␴兩 T̂ 兩 ␺ k ␴ 典 兩 2
␧ k ⫽␧ k ⬘ ⫽ ␮
under a constraint of Eq. 共29兲. In the second term we have
included the Zeeman effect, E ↑,↓ ⫽E 0 ⫾E Z . The third term
represents the exchange coupling between the dot spin and
conduction electrons with J⫽V 2 /E c , where 1/E c⫽1/E ⫹
⫽
e 2 4⌫ L ⌫ R
h 共 ⌫ L ⫹⌫ R 兲 2 兺␴ 共 ␧⫺Ẽ
⌬
˜2

␴兲
2
⫹⌬
˜2 冏 ␧⫽ ␮
, 共A10兲

⫹1/E ⫺ . By expressing the spin operator Ŝ as Ŝ ⫹ ⫽ f †↑ f ↓ ,
Ŝ ⫺ ⫽ f †↓ f ↑ , Ŝ z ⫽( f †↑ f ↑ ⫺ f †↓ f ↓ )/2, one finds that Eq. 共A3兲 is
identical to Eq. 共28兲.
The mean field Hamiltonian, Eq. 共31兲, includes ‘‘energy
levels’’ for pseudofermions, Ẽ ␴ ⫽E ␴ ⫹␭, which are coupled
to the leads with ‘‘tunneling amplitude,’’ Ṽ⫽⫺ 冑2J 具 ⌶ 典 . The
Green function for the pseudofermions is
where ⌫ ␣ ⫽ ␲␯ 兩 V ␣ 兩 2 . This yields Eq. 共34兲 in the text. On the
second line in Eq. 共A10兲, 兩 ␺ k ␴ 典 ⫽c k†␴ 兩 0 典 ⫽(V L 兩 L,k ␴ 典
⫹V R 兩 R,k ␴ 典 )/V, and the T matrix is evaluated in terms of
the Green function, Eq. 共A4兲, 兩 Ṽ 兩 2 G ␴ (␧⫽␧ k ).
APPENDIX B: MEAN FIELD CALCULATIONS IN THE
PRESENT MODEL

For the spin states of the coupling between the spin triplet
1 S⫽1 in the dot and a conduction electron, we introduce
G ␴共 ␧ 兲 ⫽ , 共A4兲
␧⫺Ẽ ␴ ⫹i⌬
˜ spinors of S⫽1/2 and 3/2. Using the Clebsch-Gordan coef-
ficients, they are given by
where ⌬˜ ⫽ ␲␯ 兩 Ṽ 兩 2 . This represents the resonant tunneling
with the resonant width ⌬ ˜.
The expectation value of the Hamiltonian, Eq. 共31兲, is
ជ 1/2
⍀ (i)
⫽ 兺k 冉 共 冑2 f 11
共 冑2 f 1⫺1
†
c k↑ ⫹ f 10
† (i)

(i)
c k↓ ⫹ f 10
c k↓ 兲 / 冑3
† (i)

c k↑ 兲 / 冑3
† (i) 冊 , 共B1兲

冉 冊
written as

冋 册
† (i)
˜ Ẽ ␴2 ⫹⌬ f 11 c k↓
⌬
˜ Ẽ ␴ ⌬
˜ ⌬ ˜2 ⌬
˜
E MF⫽ 兺␴ ⫺ ⫹ tan⫺1
␲ ␲
⫹
Ẽ ␴ 2 ␲
ln ⫺␭⫹
␲␯ J
, 共 ⫺ f 11 c k↑ ⫹ 冑2 f 10
† (i)
c k↓ 兲 / 冑3
† (i)

兺k
2
D0 ជ 3/2
⍀ (i)
⫽ . 共B2兲
共A5兲 †
共 f 1⫺1 (i)
c k↓ ⫺ 冑2 f 10 c k↑ 兲 / 冑3
† (i)

where D 0 is the bandwidth in the leads.7 We set ␮ ⫽0 in this ⫺ f 1⫺1

† (i)
c k↑
appendix. The constraint of Eq. 共29兲 is equivalent to the con-
The exchange couplings between the triplet state and con-
dition
duction electrons, Eq. 共12兲, can be rewritten as
⳵ E MF 1 ⌬
˜
⳵␭
⫽
␲ 兺␴ tan⫺1 Ẽ ⫺1⫽0. 共A6兲 H S⫽1 ⫽ 兺 ជ 1/2
J (i) 关 ⫺2⍀ (i)† ជ (i)
⍀ 1/2⫹⍀ជ 3/2
(i)† ជ (i)
⍀ 3/2兴 . 共B3兲
␴ i⫽1,2

085322-9
MIKIO ETO AND YULI V. NAZAROV PHYSICAL REVIEW B 64 085322

In the same way we define the spinors of S⫽1/2 to represent where the resonant widths are ⌬ ˜ 0 ⫽(2 cos2␸)/3,
˜ 11 /⌬
the spin couplings between the singlet state S⫽0 and a con- ⌬ 10 /⌬ 0 ⫽(cos ␸)/3, and ⌬ 00 /⌬ 0 ⫽sin2␸ with ⌬
˜ ˜ 2 ˜ ˜ ˜0
duction electron ជ 典 兩 . We set ␮ ⫽0 here. Minimizing E MF with
⫽ ␲␯ 兩 J MF具 ⌶ 2

ជ (i) ⫽
⌿ 兺k 冉 † ( ī )
f 00
⫺ f 00
c k↓
† ( ī )
c k↑
冊 , 共B4兲
respect to ⌬ 0 , we obtain

2
˜

共 cos ␸ 兲 ln
2
2
Ẽ 11⫹⌬ 2
˜ 11 1
⫹ 共 cos ␸ 兲 ln
2
2
Ẽ 10 ⫹⌬ 2
˜ 10
3 D 20 3 D 20
where ī ⫽2 and 1 for i⫽1 and 2, respectively. The conver-
sion between the triplet and singlet states, Eq. 共13兲, is rewrit- 2
Ẽ 00 ⫹⌬ 2
˜ 00 2
ten as ⫹ 共 sin2 ␸ 兲 ln ⫹ ⫽0. 共B8兲
D 20 ␯J
H S⫽1↔0 ⫽⫺ 冑3J̃ 兺
i⫽1,2
ជ (i)† ⍀
关⌿ ជ 1/2
(i)
⫹H.c.兴 . 共B5兲 Equations 共B7兲 and 共B8兲 determine ␭ and ⌬ ជ 典 兩 2 ).
˜ 0 共or 兩 具 ⌶
The conductance through the dot is given by

冏
In H S⫽1 ⫹H S⫽1↔0 , a mode of the largest coupling with S
e2
⫽1/2 is given by G⫽
h
共 2 ␲␯ 兲 2 兺
i, j, ␴ , ␴ ⬘
兩 具 R,k ⬘ ␴ ⬘ , j 兩 T̂ 兩 L,k ␴ ,i 典 兩 2
␧ k ⫽␧ k ⬘ ⫽ ␮
ជ ⫽ 共 cos ␸ 兲 ⍀
⌶ ជ 1/2
(1) ជ (1)
⫹ 共 sin ␸ 兲 ⌿ 共B6兲
for J (1) ⭓J (2) , which is Eq. 共35兲 in the text. The correspond- e2 ⌫ Rj ⌫ Li
ing eigenvalue is given by Eq. 共37兲 and ␸ is determined as in
⫽ 共 2 ␲␯ 兲
h
2
兺
i, j, ␴ , ␴ ⬘ ⌫ L ⫹⌫ R ⌫ L ⫹⌫ R
j j i i
Eq. 共38兲.
The mean field Hamiltonian, Eq. 共36兲, represents the reso-
nant tunneling through the energy levels for the pseudo-
fermions, Ẽ SM ⫽E SM ⫹␭. The expectation value of Eq. 共36兲,
( j)
冏
⫻ 兩 具 ␺ k ⬘ ␴ ⬘ 兩 T̂ 兩 ␺ k(i)␴ 典 兩 2
␧ k ⫽␧ k ⬘ ⫽ ␮
, 共B9兲

E MF , is evaluated in the same way as in Appendix A. where ⌫ ␣i ⫽ ␲␯ 兩 V ␣ ,i 兩 2

and 兩 ␺ k(i)␴ 典 ⫽(V L,i 兩 L,k ␴ ,i 典
⳵ E MF / ⳵ ␭⫽0 yields ⫹V R,i 兩 R,k ␴ ,i 典 )/V i . The T matrix can be evaluated, using
the Green function for the pseudofermions, G SM (␧)⫽ 关 ␧
⌬
˜ 11 ⌬
˜ 10 ⌬
˜ 00
tan⫺1 ⫹tan⫺1 ⫹tan⫺1 ⫽␲, 共B7兲 ⫺Ẽ SM ⫹i⌬ ˜ SM 兴 ⫺1 , as in Appendix A. This yields Eq. 共42兲 in
Ẽ 11 Ẽ 10 Ẽ 00 the text.

1 14
D. Goldhaber-Gordon, H. Shtrikman, D. Mahalu, D. Abusch- S. Sasaki, S. De Franceschi, J.M. Elzerman, W.G. van der Wiel,
Magder, U. Meirav, and M.A. Kastner, Nature 共London兲 391, M. Eto, S. Tarucha, and L.P. Kouwenhoven, Nature 共London兲
156 共1998兲; D. Goldhaber-Gordon, J. Göres, M.A. Kastner, H. 405, 764 共2000兲.
Shtrikman, D. Mahalu, and U. Meirav, Phys. Rev. Lett. 81, 5225 15
S. Tarucha, D.G. Austing, T. Honda, R.J. van der Hage, and L.P.
共1998兲. Kouwenhoven, Phys. Rev. Lett. 77, 3613 共1996兲.
2
S.M. Cronenwett, T.H. Oosterkamp, and L.P. Kouwenhoven, Sci- 16
L.P. Kouwenhoven, T.H. Oosterkamp, M.W.S. Danoesastro, M.
ence 281, 540 共1998兲.
3 Eto, D.G. Austing, T. Honda, and S. Tarucha, Science 278, 1788
F. Simmel, R.H. Blick, J.P. Kotthaus, W. Wegscheider, and M.
共1997兲.
Bichler, Phys. Rev. Lett. 83, 804 共1999兲. 17
4 T. Inoshita, A. Shimizu, Y. Kuramoto, and H. Sakaki, Phys. Rev.
J. Schmid, J. Weis, K. Eberl, and K.v. Klitzing, Phys. Rev. Lett.
B 48, R14725 共1993兲; T. Inoshita, Y. Kuramoto, and H. Sakaki,
84, 5824 共2000兲.
5 Superlattices Microstruct. 22, 75 共1997兲.
W.G. van der Wiel, S. De Franceschi, T. Fujisawa, J.M. Elzerman, 18
S. Tarucha, and L.P. Kouwenhoven, Science 289, 2105 共2000兲. T. Pohjola, J. König, M.M. Salomaa, J. Schmid, H. Schoeller, and
6
J. Kondo, Prog. Theor. Phys. 32, 37 共1964兲. G. Schön, Europhys. Lett. 40, 189 共1997兲.
19
7
A. C. Hewson, The Kondo Problem to Heavy Fermions 共Cam- W. Izumida, O. Sakai, and Y. Shimizu, J. Phys. Soc. Jpn. 67, 2444
bridge, Cambridge, England, 1993兲. 共1998兲.
20
8
K. Yosida, Theory of Magnetism 共Springer, New York, 1996兲. A. Levy Yeyati, F. Flores, and A. Martı́n-Rodero, Phys. Rev. Lett.
9
L.I. Glazman and M.É. Raı̆kh, Pis’ma Zh. Éksp. Teor. Fiz. 47, 83, 600 共1999兲.
378 共1988兲 关 JETP Lett. 47, 452 共1988兲兴.
21
M. Eto and Yu.V. Nazarov, Phys. Rev. Lett. 85, 1306 共2000兲.
10
T.K. Ng and P.A. Lee, Phys. Rev. Lett. 61, 1768 共1988兲. 22
P.W. Anderson, J. Phys. C 3, 2436 共1970兲.
11
A. Kawabata, J. Phys. Soc. Jpn. 60, 3222 共1991兲. 23
P. Nozières and A. Blandin, J. Phys. 共Paris兲 41, 193 共1980兲.
12
S. Hershfield, J.H. Davies, and J.W. Wilkins, Phys. Rev. Lett. 67, 24
D.L. Cox and A. Zawadowski, Adv. Phys. 47, 599 共1998兲.
25
3720 共1991兲; Phys. Rev. B 46, 7046 共1992兲. M. Pustilnik, Y. Avishai, and K. Kikoin, Phys. Rev. Lett. 84, 1756
13
Y. Meir, N.S. Wingreen, and P.A. Lee, Phys. Rev. Lett. 70, 2601 共2000兲.
共1993兲. 26
D. Giuliano and A. Tagliacozzo, Phys. Rev. Lett. 84, 4677 共2000兲.

085322-10
MEAN-FIELD THEORY OF THE KONDO EFFECT IN . . .
27
L. P. Kouwenhoven 共private communications兲.
28
J. Nygård, D.H. Cobden, and P.E. Lindelof, Nature 共London兲 408,
342 共2000兲.
29
A. Yoshimori and A. Sakurai, Suppl. Prog. Theor. Phys. 46, 162
共1970兲.
30
C. Lacroix and M. Cyrot, Phys. Rev. B 20, 1969 共1979兲.
31
This mean field theory is equivalent to the mean field theory
using the slave bosons for U⫽⬁ Anderson model in the Kondo
region 共Ref. 7兲. This method is exact in the large-N limit when
the dot state is N-fold degenerate.
32
M. Pustilnik and L.I. Glazman, Phys. Rev. Lett. 85, 2993 共2000兲.
33
I. Affleck, A.W.W. Ludwig, and B.A. Jones, Phys. Rev. B 52,
9528 共1995兲.
34
The different symmetry means different orbital quantum number
for the one-electron states in a quantum dot. We consider the
situation for vertical quantum dots where the orbital quantum
085322-11
PHYSICAL REVIEW B 64 085322
number i is conserved in the tunneling processes between the dot
and leads.
35
One finds that C 1 ⫽C 2 ⫽1 in the case of ␦ ⫽0, considering the
matrix element of the Coulomb interaction between d †1↑ d †1↓ 兩 0 典
and d †2↑ d †2↓ 兩 0 典 , 具 22兩 e 2 /r 兩 11典 ⬅K. In rectangular dots 共Ref. 14兲,
the matrix element K is not zero and of the same order as the
exchange interaction, 具 12兩 e 2 /r 兩 21典 ⬅J, and smaller than the
Hartree term, 具 11兩 e 2 /r 兩 11典 ⬃ 共charging energy兲, typically by one
order. In a case of 具 11兩 e 2 /r 兩 11典 ⫽ 具 22兩 e 2 /r 兩 22典 , C 1 /C 2 ⫽( ␦
⫹ 冑␦ 2 ⫹K 2 )/K and thus C 1 ⬇C 2 for ␦ ⬃K 共when the singlet and
triplet states are degenerate, ␦ ⬃J).
36
F.D.M. Haldane, J. Phys. C 11, 5015 共1978兲.
37
I. Okada and K. Yosida, Prog. Theor. Phys. 49, 1483 共1973兲.
38
This result is different from that in Ref. 25, which treats the
situation of C 1 ⫽ 冑2,C 2 ⫽0 in Eq. 共4兲. Their calculations might
be better when 兩 ⌬ 兩 ⫽E Z is so large that ␦ ⰇK 共Ref. 35兲 ( 兩 ⌬ 兩
increases with ␦ ). See discussion in Sec. V.