CFT approach to multi-channel SU(N) Kondo effect Sho Ozaki (Keio Univ.) Seminar @ Chiba Institute of Technology, 2017 July 8 In collaboration with Taro Kimura (Keio Univ.)
CFT approach to multi-channel SU(N) Kondo effect
Sho Ozaki (Keio Univ.)
Seminar @ Chiba Institute of Technology, 2017 July 8
In collaboration with
Taro Kimura (Keio Univ.)
Contents
Introduction
CFT approach to multi-channel SU(N) Kondo effect
Summary
I)
II)
III) Application to high energy physics: QCD Kondo effect
IV)
T. Kimura and S. O., in preparation
T. Kimura and S.O., arXiv: 1611.07284 to be published in JPSJ
Kondo effect
近藤効果: 磁性不純物の入った金の電気抵抗の低温での振る舞い
近藤効果出典: フリー百科事典『ウィキペディア(Wikipedia)』
近藤効果(こんどうこうか、Kondo effect)とは、磁性を持った極微量な不純物(普通磁性のある鉄原子など)がある金属では、温度を下げていくとある温度以下で電気抵抗が上昇に転じる現象である。これは通常の金属の、温度を下げていくとその電気抵抗も減少していくという一般的な性質とは異なっている。現象そのものは電気抵抗極小現象とよばれ、1930年頃から知られていたが、その物理的機構は1964年に日本の近藤淳が初めて理論的に解明した[1]。近藤はこの仕事により1973年に日本学士院恩賜賞を受章した。
目次1 現象2 理論3 理論の拡張と応用4 脚注5 参考文献6 関連項目7 外部リンク
現象金属は電圧を加えると、金属内の伝導電子が加速され電流が流れる。これを電気伝導という。
一方で、この伝導電子には電気抵抗がはたらく。金属の電気抵抗の主な要因は、金属内に含まれる不純物などによる格子欠陥と、原子の熱振動の2つである。不純物による抵抗は温度に依存せず一定である。熱振動による抵抗は、温度を下げると小さくなり、低温では抵抗は温度Tの5乗に比例する。そのため、金属の電気抵抗は通常、温度を下げると減少し、絶対零度で、一定値(=不純物による抵抗値)に落ち着く。
しかし、金属によっては、ある温度までは温度が下がると電気抵抗も減少するが、さらに温度を下げると電気抵抗は逆に増大するという、通常では起こりえないふるまいを見せる。この現象は、1933年、ド・ハース、ド・ブール、ファン・デン・バーグが、金の電気抵抗を測定したときに初めて観測された[2]。
その後の研究により、この現象は金(Au)、銀(Ag)、銅(Cu)などに鉄(Fe)、マンガン(Mn)、クロム(Cr)などの磁性不純物を微量に加えた金属で起こることが明らかになった。
T 2 (Classical)logT
(Quantum)
By “infrared divergence”
Kondo effect is firstly observed in experiment as an enhancement of electrical resistivity of impure metals.
Jun Kondo(1930-)
J. Kondo has explained the phenomenon based on the second order perturbation of interaction betweenconduction electron and impurity.
Conditions for the appearance of Kondo effect
0) Localized (Heavy) impurity
i) Fermi surface
ii) Quantum fluctuation (loop effect)
iii) Non-Abelian property of interaction
(spin-flip int.)
s-d model (Kondo model)
Born term
q q’
P P’
J
|k "i = c†k"|FSi
Scattering amplitude T (k "! k0 ")
k " k0 "
M M
Hsd =X
k,�
✏kc†k,�ck,� � J
X
k,k0
~sk0k · ~S (J < 0)
T (1) = hk0 " |� J~sk0k · ~S|k "i = �JSz
Second order perturbation theory
q q’k
P P+q-k P’
(a)
q q’
kP
P-q’ +k
P’
(b)
- only spin flip processes -
T (2)b = hk0 " |
X
k00
⇣J2S
+c†k00#ck"⌘⇣
J2S
�c†k0"ck00#
⌘
✏k00 � ✏� i⌘|k "iT (2)
a = hk0 " |X
k00
⇣J2S
�c†k0"ck00#
⌘⇣J2S
+c†k00#ck"⌘
✏� ✏k00 + i⌘|k "i
✏ = ✏k0 = ✏k
S+ S� S� S+
⇢F : density of state on the Fermi surface, D : Bandwidth
k " k "k0 " k0 "
M M + 1 M � 1M MM
k00 #
k00 #
particle hole
T (2)=
J2
4
⇥S+, S�⇤ ⇢F log
✓D
T
◆
T = T (1) + T (2) + · · ·
At low temperature (IR) regions: T ⌧ D
The perturbative expansion breaks down.
Total amplitude
We need a some non-perturbative method to analyze the Kondo effect at IR regions.
T ' TK@
' T (1)
✓1 +
J
2
⇢F log
✓T
D
◆◆
1 ' J
2
⇢F log
✓T
D
◆(J < 0)
Non-perturbative approach for Kondo effect
Numerical renormalization group
Bethe ansatz
1+1 dim. conformal field theory (CFT) approach
[Wilson]
[Andrei] [Wiegmann] ….
[Affleck-Ludwig]
k(multi)-channel SU(2) Kondo
Non-perturbative approach for Kondo effect
Numerical renormalization group
1+1 dim. conformal field theory (CFT) approach
[Wilson]
[Affleck-Ludwig]
k(multi)-channel SU(2) Kondo
k-channel SU(N) Kondo with k >= N
Bethe ansatz [Andrei] [Wiegmann] ….
Non-perturbative approach for Kondo effect
Numerical renormalization group
1+1 dim. conformal field theory (CFT) approach
[Wilson]
[Affleck-Ludwig]
k(multi)-channel SU(2) Kondo
k-channel SU(N) Kondo with k >= N
k-channel SU(N) Kondo including N > k >1T. Kimura and S. O, arXiv:1611.07284
Bethe ansatz [Andrei] [Wiegmann] ….
Boundary CFT
H = i
†(x)@ (x)
@x
+ �K †(x)ta (x)Sa
�(x)
Assuming that the impurity is sufficiently dilute, and the interaction is a contact-type one, the s-wave approx. is valid, which leads to the following one-dim. effective theory:
Boundary CFT
Currents
color
flavor
charge
H = i
†(x)@ (x)
@x
+ �K †(x)ta (x)Sa
�(x)
J
a(x) =: †(x)ta (x) :
J
A(x) =: †(x)TA (x) :
J(x) =: †(x) (x) :
: OO(x) := lim✏!0
{O(x)O(x+ ✏)� hO(x)O(x+ ✏)i}
Normal order product
Assuming that the impurity is sufficiently dilute, and the interaction is a contact-type one, the s-wave approx. is valid, which leads to the following one-dim. effective theory:
Boundary CFT
Currents
color
flavor
charge
The Sugawara form of the Hamiltonian density
H = i
†(x)@ (x)
@x
+ �K †(x)ta (x)Sa
�(x)
J
a(x) =: †(x)ta (x) :
J
A(x) =: †(x)TA (x) :
J(x) =: †(x) (x) :
: OO(x) := lim✏!0
{O(x)O(x+ ✏)� hO(x)O(x+ ✏)i}
Normal order product
H =1
N + k
J
aJ
a +1
k +N
J
AJ
A +1
2kNJJ + �KJ
aS
a�(x)
Assuming that the impurity is sufficiently dilute, and the interaction is a contact-type one, the s-wave approx. is valid, which leads to the following one-dim. effective theory:
x
Impurity effect
Boundary of the theory
x = 0
⌧
Boundary CFT
H =1
N + k
J
aJ
a +1
k +N
J
AJ
A +1
2kNJJ + �KJ
aS
a�(x)
J a = J
a +�K
2(N + k)Sa
�(x)
H =1
N + kJ aJ a +
1
k +NJAJA +
1
2kNJJ
g-factor and Impurity entropy
Partition function
gRimp ⇥ e⇡cL6�Z(L,�)
bulkboundary
Free energy
Entropy
Impurity entropy at IR fixed point (T=0)
gRimp =SRimp0
S00
: central charge
: g-factor
Smn : modular S-matrix
,
( )
S(T ) = �@F
@T
L ! 1
F = � 1
�logZ
Simp = S(T )� Sbulk(T )|T=0 = log (gRimp)
universal quantities
(impurity)
c = Nk
g-factor and Impurity entropy
g-factor provides the impurity entropy:
Simp = log(gRimp)
Rimp : fundamental representation
ex) SU(2), k=1 (standard Kondo effect)
Simp
log2
0IR UV
T
Simp = log(2s+ 1)
s = 1/2
Simp ! log2
In UV,
In IR, s ! 0 (Kondo singlet)
Simp ! 0
,
Overscreening Kondo effect in multi-channel SU(2) Kondo model
k = 1
Fermi liquid at IR fixed point
: integerg
(single channel)
k = 2
non-Fermi liquid at IR fixed point
: non-integerg
(two channel)
Standard Kondo effect
Overscreening Kondo effect
N k = 1 k = 2 k = 3 k = 4 k = ∞
2 1 1.4142... 1.6180... 1.7320... 2
3 1 1.6180... 2 2.2469... 3
4 1 1.7320... 2.2469... 2.6131... 4
∞ 1 2 3 4 ∞
TABLE II. Numerical values of the g-factor for the (anti)fundamental representation. It approaches
to an integer in the large (N, k) limit.
where ρ is the Weyl vector (A4), and the q-number [x]q is defined by
[x]q =qx/2 − q−x/2
q1/2 − q−1/2. (25)
This q-number is reduced to the ordinary number [x]q → x in the limit q → 1, so that
dimq R → dimR. Therefore, in this limit, the quantum dimension becomes
dimq Nq→1−→ N. (26)
We remark the anti-fundamental representation N gives the same quantum dimension.
Let us consider the large N behavior of the g-factor. Expanding the expression (24) with
respect to the large N at a fixed k, we obtain
g = k −k(k2 − 1)
N2
π2
6+O(N−3) . (27)
In the large N limit, the g-factor is approximated to g = k, and the correction starts with
O(N−2). This implies that the SU(N)k Kondo effect is described as the Fermi liquid in
the large N limit, and thus the low-temperature scaling of the specific heat and so on is
expected to exhibit the Fermi liquid behavior.
Table II shows the numerical values of the g-factor for the (anti)fundamental representa-
tion. Although there is an accidental case giving an integer value for three-channel SU(3)
system SU(3)k=3, we obtain irrational values for k > 1 in general cases. This is a signature
of the non-Fermi liquid behavior at the IR fixed point of the multi-channel Kondo system,
corresponding to zero temperature (ground state). We remark that the coincidence of g-
factor for SU(N)k and SU(k)N reflects the level-rank duality of the Kac–Moody algebra.
The SU(3)3 system is self-dual in this sense (N, k) = (3, 3).
9
g-factor in SU(N) Kondo effect @ IR fixed point(zero temperature)
For k=1, g-factor is always unity, and thus the system becomes the Fermi liquid.
In general with k > 1, the g-factor becomes non-integer, which indicates that the system is described by non-Fermi liquid.
1+1 dim. (boundary) CFT approach
Correlation functions are exactly determined by
Conformal symmetry in 1+1 dim.
Kac-Moody algebra
From the correlation functions, one can evaluate T-dep. of several observables of k-channel SU(N) Kondo effect in IR regions.
hO1(x)O2(y)i =C1,2
|x� y|�
C1,2
� determined by conformal symmetry
determined by KM algebra
⇥J a(x),J b(y)
⇤= if
abcJ c(x)�(x� y) +i
2⇡
✓k
2
◆�
ab @
@x
�(x� y)
Specific heat, susceptibility and the Wilson ratio
Bulk contributions to C �&
Cbulk =⇡
3Nk T
�bulk =k
2⇡
These are well known properties of free Nk (bulk) fermions in 1+1 dim.
Impurity contributions to C �&
i) k=1 & arbitrary N [Affleck 1990]
Leading irrelevant operator
From the perturbation w.r.t.
with k = 1
�H1 = �1J aJ a(x)�(x)
�1 ⇠ 1/TK
�H1
Cimp = ��1k(N2 � 1)
3⇡2T
�imp = ��1k(N + k)
2
Typical Fermi liquid behaviors
Leading irrelevant operator
:adjoint operator, appearing when k >1 scaling dimension is
:Fourier mode of
.
From the perturbation with respect to the leading irrelevant operator, we can evaluate .Cimp �imp,
�a
�H = �J a�1�
a(x)�(x)
J an J a(x)
� ⇠ 1/T�K
ii) k > 1, Overscreening case
� < 1
� =N
N + k
Z = e��F (T,�,h)
=
ZD D ¯
e
� Rd
2xH
exp
(+
Z�/2
��/2d⌧
"�J a
�1�a
(⌧, 0) +
h
2⇡
ZL
�L
dx J 3(⌧, x)
#)
= Z0
*exp
(+
Z �/2
��/2d⌧
"�J a
�1�a(⌧, 0) +
h
2⇡
Z L
�Ldx J 3
(⌧, x)
#)+
F = Lfbulk + fimp
C = �T@2F
@T 2� = �@2F
@h2
Free energy can be divided in to bulk and impurity parts
which is expressed in terms of the correlation functions of the leading irrelevant operators.
,
Observables
h = 0
L ! 1
D D †
�H = �J a�1�
a(x)�(x)�H1 = �1J aJ a(x)�(x)
Cimp =
8>>>>>>>><
>>>>>>>>:
�2
2
⇡1+2�(2�)
2(N2 � 1)(N + k/2)
✓1� 2�
2
◆�(1/2��)�(1/2)
�(1��)
T 2�(k > N)
�2⇡1+2�(N2 � 1)(N + k/2)(2�)
2 T log
✓TK
T
◆(k = N)
��1k
3
(N2 � 1)⇡2T + 2�2⇡2(N2 � 1)(N + k/2)
2�
1 + 2�
��2�+1K
2�� 1
!T (N > k > 1)
Specific heat of k-channel SU(N) Kondo effect
Specific heat of k-channel SU(N) Kondo effect
Cimp =
8>>>>>>>><
>>>>>>>>:
�2
2
⇡1+2�(2�)
2(N2 � 1)(N + k/2)
✓1� 2�
2
◆�(1/2��)�(1/2)
�(1��)
T 2�(k > N)
�2⇡1+2�(N2 � 1)(N + k/2)(2�)
2 T log
✓TK
T
◆(k = N)
��1k
3
(N2 � 1)⇡2T + 2�2⇡2(N2 � 1)(N + k/2)
2�
1 + 2�
��2�+1K
2�� 1
!T (N > k > 1)
Low T scaling
Cimp /
8><
>:
T 2�(k > N)
T log(TK/T ) (k = N)
T (N > k > 1)
For N > k >1, although the g-factor (at IR fixed point) exhibits non-Fermi liquid signature, T-dep. of Cimp shows Fermi liquid behavior. Fermi/non-Fermi mixing [T. Kimura and S. O, arXiv:1611.07284]
Non-Fermi
Non-FermiFermi
�imp =
8>>>>>>>><
>>>>>>>>:
�2
2
⇡2��1(N + k/2)2(1� 2�)
�(1/2��)�(1/2)
�(1��)
T 2��1(k > N)
2�2(N + k/2)2 log
✓TK
T
◆(k = N)
��1k(N + k)
2
+ 2�2(N + k/2)2
��2�+1K
2�� 1
!(N > k > 1)
Low T scaling
Susceptibility of k-channel SU(N) Kondo effect
�imp =
8><
>:
T 2��1(2k > N)
log(TK/T ) (2k = N)
const. (N > k > 1)
Non-Fermi
Non-FermiFermi
The Wilson ratio of QCD Kondo effect
For N >= k, the Wilson ratio is no longer universal, which depends on the detail of the system, such as
RW =
✓�imp
Cimp
◆,✓�bulk
Cbulk
◆
Unknown parameters are canceled, and thus the Wilson ratio is universal.
� = 4�2
�1T 2��1Kwith
�, TK
=(N + k/2)(N + k)2
3N(N2 � 1)(k � N)
RW =(N + k/2)(N + k/3)
N2 � 1
� � k(N + k)
(N + k/2)2
� � k(N + k/3)
N(N + k/2)
(N > k > 1)
k >= N N > k > 1
g-factor (IR fixed point) non-Fermi non-Fermi
Low T scaling non-Fermi Fermi
Wilson ratio universal non-universal
T. Kimura and S. O, arXiv:1611.07284
IR behaviors of k-channel SU(N) Kondo effect
Fermi/non-Fermi mixing
Application to high energy physics: QCD Kondo effect
Strong interaction
QCD Lagrangian
LQCD = �1
4F aµ⌫F
aµ⌫ + q (i�µDµ �Mq) q
Dµ = @µ � igAaµt
a
F aµ⌫ = @µA
a⌫ � @⌫A
aµ + gfabcAb
µAc⌫
Quarks
ColorRed Green Blue
Flav
or up
down
strage
charm
bottom
top
u
d d d
s
c
u u
s s
c c
b b b
t t t
2.3MeV
4.8MeV
95MeV
⇤QCD ⇠ 200MeV
1200MeV
4200MeV
173⇥ 103MeV
Electron
e�
⇠ 0.5MeV
ColorRed Green Blue
Flav
or up
down
strage
charm
bottom
top
u
d d d
s
c
u u
s s
c c
b b b
t t t
2.3MeV
4.8MeV
95MeV
⇤QCD ⇠ 200MeV
1200MeV
4200MeV
173⇥ 103MeV
Impurity effectby heavy flavors
Kondo effect induced by color d.o.f.
Electron
e�
⇠ 0.5MeV
Quarks
Asymptotic freedom in Kondo effect and QCD
⇤Fermi
Surface
0 ⇤K
G(⇤)
0 ⇤
Kondo effect Running coupling of QCD
q q’
P P’
q’-q
FIG. 2: The tree diagram. Solid and double solid lines are massless and heavy quarks, respectively.
Here ⇧(p2k/m2) has a rather complicated form (see Refs. [11–15] for the explicit expression).
But in our analysis with the massless QCD, it is su�cient to know that ⇧(p2k/m2) ! 1 in
the massless limit m ! 0. It should be emphasized that there is a strong screening e↵ect
in the first term of the gluon propagator (4) owing to the gluon mass. The propagator (4)
is an analog of that employed in analyses of the Schwinger model, namely, 1+1 dimensional
QED [16].
A. Tree amplitude
Now we compute the amplitude for scattering between a light (massless) quark near the
Fermi surface and a heavy quark impurity. Under the strong magnetic field, the light quark
moves only in the direction parallel to the magnetic field. In the LLL with e
q
> 0, the spin
of the light quark is fixed to the magnetic field direction. We set the momentum of the
initial quark as positive direction of the z-axis: qz
> 0. Then, the leading order amplitude
as shown in Fig. 2 is given by
�iM0
= (ig)2⇥u
LLL
(q0)�µ(TA)a
0a
u
LLL
(q)⇤DAB
µ⌫
(q0 � q|eq
B)⇥U(P 0)�⌫(TB)
b
0b
U(P )⇤, (7)
where the color indices of quarks can take a, a0, b, b0 = 1, 2, · · ·Nc
. The spinors are defined by
u
LLL
(q) = Nq
⇣�",
�zqz
|qz | �"
⌘t
with �
z
�" = +�", and U(P ) = NQ
(⇠�
, 0)t with �
z
⇠± = ±⇠±.
Nq
and NQ
are normalization constants. By using these spinors, we find u
LLL
�
µ
u
LLL
=
u
LLL
�
µ
u
LLL
where µ = 0, 3, and U�
⌫
U = U�
0
U . Then, in the gluon propagator (4), only
the first term proportional to g
k00
contributes to the amplitude. Furthermore, as we will
see soon, only forward scattering is allowed in the massless limit of the light quark. Then,
7
q
q
g
g
Asymptotic freedom in Kondo effect and QCD
⇤Fermi
Surface
0 ⇤K
G(⇤)
0 ⇤
Kondosinglet
Colorsinglet
(Hadron)
q q’
P P’
q’-q
FIG. 2: The tree diagram. Solid and double solid lines are massless and heavy quarks, respectively.
Here ⇧(p2k/m2) has a rather complicated form (see Refs. [11–15] for the explicit expression).
But in our analysis with the massless QCD, it is su�cient to know that ⇧(p2k/m2) ! 1 in
the massless limit m ! 0. It should be emphasized that there is a strong screening e↵ect
in the first term of the gluon propagator (4) owing to the gluon mass. The propagator (4)
is an analog of that employed in analyses of the Schwinger model, namely, 1+1 dimensional
QED [16].
A. Tree amplitude
Now we compute the amplitude for scattering between a light (massless) quark near the
Fermi surface and a heavy quark impurity. Under the strong magnetic field, the light quark
moves only in the direction parallel to the magnetic field. In the LLL with e
q
> 0, the spin
of the light quark is fixed to the magnetic field direction. We set the momentum of the
initial quark as positive direction of the z-axis: qz
> 0. Then, the leading order amplitude
as shown in Fig. 2 is given by
�iM0
= (ig)2⇥u
LLL
(q0)�µ(TA)a
0a
u
LLL
(q)⇤DAB
µ⌫
(q0 � q|eq
B)⇥U(P 0)�⌫(TB)
b
0b
U(P )⇤, (7)
where the color indices of quarks can take a, a0, b, b0 = 1, 2, · · ·Nc
. The spinors are defined by
u
LLL
(q) = Nq
⇣�",
�zqz
|qz | �"
⌘t
with �
z
�" = +�", and U(P ) = NQ
(⇠�
, 0)t with �
z
⇠± = ±⇠±.
Nq
and NQ
are normalization constants. By using these spinors, we find u
LLL
�
µ
u
LLL
=
u
LLL
�
µ
u
LLL
where µ = 0, 3, and U�
⌫
U = U�
0
U . Then, in the gluon propagator (4), only
the first term proportional to g
k00
contributes to the amplitude. Furthermore, as we will
see soon, only forward scattering is allowed in the massless limit of the light quark. Then,
7
q
q
g
g
Asymptotic freedom in Kondo effect and QCD
⇤Fermi
Surface
0 ⇤K
G(⇤)
0 ⇤
Kondosinglet
Colorsinglet
(Hadron)
QCD Kondo
⇤K
G(⇤)
Conditions for the appearance of Kondo effect
0) Heavy impurity
i) Fermi surface
ii) Quantum fluctuation (loop effect)
iii) Non-Abelian property of interaction
(spin-flip int.)
Conditions for the appearance of QCD Kondo effect
0) Heavy quark impurity
i) Fermi surface of light quarks
ii) Quantum fluctuation (loop effect)
iii) Color exchange interaction in QCD
QCD Kondo effect
K. Hattori, K. Itakura, S. O. and S. Yasui, PRD92 (2015) 065003
Heavy quark impurity
(light) quark matter with
charm or bottom quark
µ � ⇤QCD
Q
(light) quark matter with µ � ⇤QCD
Q
q q
q q’
P P’
q’-q
FIG. 2: The tree diagram. Solid and double solid lines are massless and heavy quarks, respectively.
Here ⇧(p2k/m2) has a rather complicated form (see Refs. [11–15] for the explicit expression).
But in our analysis with the massless QCD, it is su�cient to know that ⇧(p2k/m2) ! 1 in
the massless limit m ! 0. It should be emphasized that there is a strong screening e↵ect
in the first term of the gluon propagator (4) owing to the gluon mass. The propagator (4)
is an analog of that employed in analyses of the Schwinger model, namely, 1+1 dimensional
QED [16].
A. Tree amplitude
Now we compute the amplitude for scattering between a light (massless) quark near the
Fermi surface and a heavy quark impurity. Under the strong magnetic field, the light quark
moves only in the direction parallel to the magnetic field. In the LLL with e
q
> 0, the spin
of the light quark is fixed to the magnetic field direction. We set the momentum of the
initial quark as positive direction of the z-axis: qz
> 0. Then, the leading order amplitude
as shown in Fig. 2 is given by
�iM0
= (ig)2⇥u
LLL
(q0)�µ(TA)a
0a
u
LLL
(q)⇤DAB
µ⌫
(q0 � q|eq
B)⇥U(P 0)�⌫(TB)
b
0b
U(P )⇤, (7)
where the color indices of quarks can take a, a0, b, b0 = 1, 2, · · ·Nc
. The spinors are defined by
u
LLL
(q) = Nq
⇣�",
�zqz
|qz | �"
⌘t
with �
z
�" = +�", and U(P ) = NQ
(⇠�
, 0)t with �
z
⇠± = ±⇠±.
Nq
and NQ
are normalization constants. By using these spinors, we find u
LLL
�
µ
u
LLL
=
u
LLL
�
µ
u
LLL
where µ = 0, 3, and U�
⌫
U = U�
0
U . Then, in the gluon propagator (4), only
the first term proportional to g
k00
contributes to the amplitude. Furthermore, as we will
see soon, only forward scattering is allowed in the massless limit of the light quark. Then,
7
q k q’
P P+q-k P’
(a)
q-k q’-k
q k q’
P P-q’+k P’
q-k q’-k
(b)
FIG. 3: One-loop diagrams.
the momentum transfer q
0 � q carried by a gluon equals to zero. Accordingly, the gluon
propagator contributing to the leading order amplitude simplifies to
D
AB
00
=ig
00
m
2
g
�
AB
, (8)
and the other componets are vanishing. Since in this study we concentrate on the quarks
near the Fermi surface, we set q3 = q
03 = k
F
. Then, the four momentum vectors of the initial
and final state quarks are given by q
0µ = (q00, 0, 0, q03) = q
µ = (q0, 0, 0, q3) = (✏F
, 0, 0, kF
) with
✏
F
= k
F
. [need to mention about transverse momentum] The leading scattering amplitude,
thus, reads
�iM0
= �iG
⇥u
LLL
(q0)�0(TA)a
0a
u
LLL
(q)⇤ ⇥
U(P 0)�0(TA)b
0b
U(P )⇤
= �iG (TA)a
0a
(TA)b
0b
N 2
q
⇣1 + sgn(q0
z
)⌘�
†"�" N 2
Q
⇠
†�
0⇠�
, (9)
where we have introduced a dimensionful coupling G as
G =g
2
m
2
g
. (10)
In the large mass limit of the heavy quark impurity: M ! 1, the heavy-quark spin is frozen
as ⇠†�
0⇠�
= �
�
0�
(thus does not play a role in the QCD Kondo e↵ect). The tree amplitude (9)
is proportional to the factor (1 + sgn(q0z
)), and thus only the forward scattering is allowed.
This is due to the helicity conservation of the massless quarks.
8
q k q’
P P+q-k P’
(a)
q-k q’-k
q k q’
P P-q’+k P’
q-k q’-k
(b)
FIG. 3: One-loop diagrams.
the momentum transfer q
0 � q carried by a gluon equals to zero. Accordingly, the gluon
propagator contributing to the leading order amplitude simplifies to
D
AB
00
=ig
00
m
2
g
�
AB
, (8)
and the other componets are vanishing. Since in this study we concentrate on the quarks
near the Fermi surface, we set q3 = q
03 = k
F
. Then, the four momentum vectors of the initial
and final state quarks are given by q
0µ = (q00, 0, 0, q03) = q
µ = (q0, 0, 0, q3) = (✏F
, 0, 0, kF
) with
✏
F
= k
F
. [need to mention about transverse momentum] The leading scattering amplitude,
thus, reads
�iM0
= �iG
⇥u
LLL
(q0)�0(TA)a
0a
u
LLL
(q)⇤ ⇥
U(P 0)�0(TA)b
0b
U(P )⇤
= �iG (TA)a
0a
(TA)b
0b
N 2
q
⇣1 + sgn(q0
z
)⌘�
†"�" N 2
Q
⇠
†�
0⇠�
, (9)
where we have introduced a dimensionful coupling G as
G =g
2
m
2
g
. (10)
In the large mass limit of the heavy quark impurity: M ! 1, the heavy-quark spin is frozen
as ⇠†�
0⇠�
= �
�
0�
(thus does not play a role in the QCD Kondo e↵ect). The tree amplitude (9)
is proportional to the factor (1 + sgn(q0z
)), and thus only the forward scattering is allowed.
This is due to the helicity conservation of the massless quarks.
8
+ +
MQ ! largeHeavy quark:
�iM =
q
Q
ta
ta
tb
ta
ta
tb ta
tatb
tb
Renormalization group equation of scattering amplitude
q q’
P P’
q q’
P P’
q q’k
P P+q-k P’
(a)
q q’
kP
P-q’ +k
P’
(b)
Q Q Q Q
Q
Q
G(⇤� d⇤) G(⇤)
=
+ +
G(⇤) G(⇤)
⇤ ⇠ ⇤� d⇤
⇤ ⇠ ⇤� d⇤
G(⇤) G(⇤)
0 ⇤⇤� d⇤ ⇤0
· · · · · ·
⇤0 = ⇤UV ' kF
Initial scale
~poor man’s scaling~
Renormalization group equation of scattering amplitude
⇤dG(⇤)
d⇤= �Nc
2⇢FG
2(⇤)
SolutionG(⇤) =
G(⇤0)
1 +
Nc2 ⇢FG(⇤0)log(⇤/⇤0)
Kondo scale (from the Landau pole)
⇤K ' kF exp
✓� 8⇡
Nc↵slog(⇡/↵s)
◆
⇤0 = ⇤UV ' kF
Initial scale
⇤
FermiSurface
⇤00 ⇤K
q q
Q
QCD Kondo effect
The strength of the q-Q interaction increases as the energy scale decreases, and the system becomes non-perturbative one below the Kondo scale.
This indicates a change of mobility of light quarks.
Several transport coefficients will be largely affected by QCD Konde effect.
' kF exp
✓� 8⇡
Nc↵slog(⇡/↵s)
◆
G(⇤)
Magnetically induced QCD Kondo effect
S. O., K. Itakura and Y. Kuramoto, PRD94 (2016) 074013
⇢F
⇢LLL
3+1 D
3+1 D
1+1 D
1+1 D
S-wave projection(Partial wave decomposition)
LLL projection
Degenerate fermions on the fermi surface
Degenerate fermions in the Landau levels
pz
|~p|
kF
eB
|~p|
Super conductivity
Kondo effect
Magnetic catalysis
BB(Dimensional reduction)
Conditions for the appearance of QCD Kondo effect
0) Heavy quark impurity
i) Fermi surface of light quarks
ii) Quantum fluctuation (loop effect)
iii) Color exchange interaction in QCD
0) Heavy quark impurity
i) Strong magnetic field
ii) Quantum fluctuation (loop effect)
iii) Color exchange interaction in QCD
The magnetic field does not affect color degrees of freedom.
“Magnetically induced QCD Kondo effect”Conditions for the appearance of
Renormalization group equation
Kondo scale (from the Landau pole)
⇤dG(⇤)
d⇤= �Nc
2⇢LLLG
2(⇤)
solution
⇤K 'peqB↵1/2
s exp
(� 2⇡
Nc↵slog(⇡/↵s)+ log
✓⇡
↵s
◆1/6)
'peqB↵1/3
s exp
⇢� 2⇡
Nc↵slog(⇡/↵s)
�
G(⇤) =
G(⇤0)
1 +
Nc2 ⇢LLLG(⇤0)log(⇤/⇤0)
QCD Kondo effect from CFT
T. Kimura and S. O, in preparation
Non-perturbativeregion
Λ
G(Λ)
ΛK ΛUV
Perturbative region
Fermi surface
IR fixedpoint
μ
Figure 1: Schematic picture of the flow of the e↵ective interaction G(⇤). The black solid
line is a perturbative flow of G(⇤), while the blue dashed line is a non-perturbative flow. ⇤K
stands for the Kondo scale. fig:scale_dependence
related to the resistivity
⇢(T ) ⇠ ⇢u
1� ReS(1)
2+ C�T�
!(2.25) eq:rho_NFL
where ⇢u
is the resistivity at the unitary limit, and C is a dimensionless constant which can
be specified by explicit computation of the first order perturbation theory. This factor is
|S(1)| < 1 for the non-Fermi liquid case, |S(1)| = 1 for the Fermi liquid case, and S(1) ! 1
in the large k limit. Although we cannot determine the sign of the temperature dependence
in this formalism, which is given by the sign of the coupling constant �, it is expected that
the correction would be positive for strong coupling and negative for weak coupling region.
We also remark that the resistivity for the Fermi liquid case is given by
⇢(T ) ⇠ ⇢u
⇥1� C 0�2T 2
⇤(2.26) eq:rho_FL
where C 0 is again a dimensionless constant. In this case the finite temperature correction is
negative, and second order in the coupling constant �, while it is first order for the non-Fermi
liquid case.
3 Application to QCD Kondo e↵ect
sec:QCD-Kondo
The QCD Kondo e↵ect is the Kondo e↵ect induced by the color exchange interaction between
light quarks near the Fermi surface and a heavy quark impurity. One of the present authors
8
⇡⇡
⇤QCD
QCD Kondo effect
In order to investigate QCD Kondo effect in IR region below Kondo scale, we have to rely on non-perturbative method.
High density QCD in the presence of the heavy quark
1+1 dim.
This is nothing but k-channel SU(N) Kondo modelin 1+1 dim., where
s-wave
k = 2Nf , N = Nc
is light quark fields with 2Nf components of flavor
.
and Nc colors. The 2 comes from spin d.o.f. in 4 dim.
(Dimensional reduction)
with G = ↵s log4µ2
m2g
= ↵s log4⇡
↵s⌧ 1
[E. Shuster & D. T. Son, and T. Kojo et al.]
S
1+1eff =
Zd
2x [i�µ@µ] �G †
t
a Q†t
aQ
Effective 1+1 dim. theory at high density
g-factor in QCD Kondo effect @ IR fixed point
In general Nc and Nf, the g-factor is non-integer, and thus QCD Kondo effect has non-Fermi liquid IR fixed point.
In large Nc limit:
Nc = 3
g =1 +
p5
2
g = 2.24598...
g = 2.53209...
(Nf = 1)
(Nf = 2)
(Nf = 3)
Nc ! 1Fermi liquid at IR fixed point
Nf, : fixed
k = 2Nf
(zero temperature)
g ! k = 2Nf
Specific heat of QCD Kondo effect
Cimp =
8>>>>>>>><
>>>>>>>>:
�2
2
⇡1+2�(2�)
2(N2
c � 1)(Nc +Nf )
✓1� 2�
2
◆�(1/2��)�(1/2)
�(1��)
T 2�(2Nf > Nc)
�2⇡1+2�(N2 � 1)(Nc +Nf )(2�)
2 T log
✓TK
T
◆(2Nf = Nc)
��12
3
(N2c � 1)⇡2T + 2�2⇡2
(N2c � 1)(Nc +Nf )
2�
1 + 2�
��2�+1K
2�� 1
!T (Nc > 2Nf )
Cimp /
8><
>:
T 2�(2Nf > Nc)
T log(TK/T ) (2Nf = Nc)
T (Nc > 2Nf )
Non-FermiNon-FermiFermi
For Nc > 2Nf, QCD Kondo effect shows Fermi/non-Fermi mixing.
Low T scaling
Susceptibility of QCD Kondo effect
�imp =
8>>>>>>>><
>>>>>>>>:
�2
2
⇡2��1(Nc +Nf )
2(1� 2�)
�(1/2��)�(1/2)
�(1��)
T 2��1(2Nf > Nc)
2�2(Nc +Nf )
2log
✓TK
T
◆(2Nf = Nc)
��1Nf (Nc + 2Nf ) + 2�2(Nc +Nf )
2
��2�+1K
2�� 1
!(Nc > 2Nf )
�imp =
8><
>:
T 2��1(2Nf > Nc)
log(TK/T ) (2Nf = Nc)
const. (Nc > 2Nf )
Non-Fermi
Non-FermiFermi
Low T scaling
The Wilson ratio of QCD Kondo effect
For Nc >= 2Nf, the Wilson ratio is no longer universal, which depends on the detail of the system, such as
RW =
✓�imp
Cimp
◆,✓�bulk
Cbulk
◆
=(Nc +NF )(Nc + 2Nf )2
3Nc(N2c � 1)
(2Nf � Nc)
RW =(Nc +Nf )(Nc + 2Nf/3)
N2c � 1
� � 2Nf (Nc + 2Nf )
(Nc +Nf )2
� � 2Nf (Nc + 2Nf/3)
Nc(Nc +Nf )
Unknown parameters are canceled, and thus the Wilson ratio of QCD Kondo effect is universal for 2Nf >= Nc.
(Nc > 2Nf )
� = 4�2
�1T 2��1Kwith
�, TK
2Nf >= Nc Nc > 2Nf
g-factor (IR fixed point) non-Fermi non-Fermi
Low T scaling non-Fermi Fermi
Wilson ratio universal non-universal
IR behaviors of QCD Kondo effect
(k >= N) (N > k >1)
Fermi/non-Fermi mixing
Summary
We apply CFT approach to QCD Kondo effect and determineits IR behaviors below the Kondo scale.
In the vicinity of IR fixed point, the Kondo system shows Fermi/non-Fermi mixing for N > k > 1, while it shoes non-Fermi liquid behaviors for k >= N.
We develop the CFT approach to general k-channel SU(N)Kondo effect and investigate its IR behaviors.
Our CFT analysis for k-channel SU(N) Kondo effect can be alsoapplied to SU(3) Kondo effect in cold atom and SU(4) Kondo effect in Quantum dot systems with multi-channels.