1: %\documentclass[twocolumn]{jpsj2}
2: \documentclass[preprint]{jpsj2}
3: %\usepackage[mtbold]{mathtime}
4: \usepackage{graphicx}
5: \renewcommand{\baselinestretch}{0.833}
6:
7: \def\runtitle{Enhancement of Kondo effect in multilevel quantum dots}
8: \def\runauthor{M.\ {\sc Eto}}
9:
10: \title{%
11: Enhancement of Kondo effect in multilevel quantum dots
12: }
13:
14: \author{%
15: Mikio {\sc Eto}\thanks{E-mail address: eto@rk.phys.keio.ac.jp}
16: }
17:
18: \inst{%
19: Faculty of Science and Technology, Keio University,
20: 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan
21: }
22:
23: \recdate{July 17, 2004}
24:
25: \abst{%
26: We theoretically study enhancement mechanisms of the Kondo effect in
27: multilevel quantum dots. In quantum dots fabricated
28: on semiconductors, the energy difference between
29: discrete levels $\Delta$ is tunable by applying a magnetic field.
30: With two orbitals and spin $1/2$ in the quantum dots, we evaluate the Kondo
31: temperature $T_{\rm K}$ as a function of $\Delta$,
32: using the scaling method. $T_{\rm K}$ is maximal around $\Delta=0$
33: and decreases with increasing $|\Delta|$, following a power law,
34: $T_{\rm K}(\Delta)=T_{\rm K}(0)\cdot ( T_{\rm K}(0)/|\Delta| )^{\gamma}$,
35: which is understood as a crossover from SU(4) to SU(2) Kondo effect.
36: The exponents on both sides of a level crossing, $\gamma_{\rm L}$
37: and $\gamma_{\rm R}$, satisfy a relation of
38: $\gamma_{\rm L}\cdot\gamma_{\rm R}=1$.
39: We compare this enhanced Kondo effect with that by
40: spin-singlet-triplet degeneracy for an even number of electrons,
41: to explain recent experimental results using vertical quantum dots.
42: }
43: \kword{quantum dot, Kondo effect, multilevel, scaling method}
44:
45:
46: \begin{document}
47: \sloppy
48: \maketitle
49:
50:
51: \section{Introduction}
52:
53: Kondo effect is one of the most important many-body problems in solid-state
54: physics.\cite{Kondo}
55: When a localized spin ($S=1/2$) is brought in contact with electron Fermi sea,
56: the Kondo effect gives rise to a new many-body ground state of a spin singlet
57: and markedly influences the transport properties.\cite{Hewson,Yosida}
58: The Kondo effect was originally discovered in metals including dilute
59: magnetic impurities.
60: The Kondo effect also takes place in semiconductor quantum dots connected to
61: external leads through tunnel barriers, as predicted by several theoretical
62: groups.\cite{Glazman,Ng,Kawabata,Hershfield,Hershfield2,Meir}
63: In this case, a localized spin $S$ is formed by electrons confined
64: in the quantum dots due to the Coulomb blockade.\cite{Leo}
65: Usually, the spin is zero or $1/2$ in the Coulomb blockade region with
66: an even and odd number of electrons, respectively, since discrete
67: spin-degenerate levels are consecutively occupied in the quantum dots.
68: Hence the Kondo effect is observed in the latter case
69: only.\cite{GGNat,Cronen,GGPRL}
70:
71: The semiconductor quantum dots offer the following merits for the
72: investigation of the Kondo physics.
73: First, we can examine the properties of single ``magnetic impurity.''
74: Second, the transport of conduction electrons is observed through the magnetic
75: impurity in the case of quantum dots, whereas the scattering by the magnetic
76: impurities is observed in the traditional metallic systems. In the former,
77: the resonant nature of the Kondo effect is directly detected:
78: The conductance is enhanced to a universal value of $2e^2/h$ at
79: temperatures much below the Kondo temperature, $T_{\rm K}$,\cite{Wilfred}
80: which is ascribable to the resonant tunneling through Kondo singlet state.
81: In the latter, the electric resistance is anomalously enhanced at low
82: temperatures since the scattering probability is amplified by the resonance.
83: This explains a
84: ``resistance minimum'' as a function of temperature ($T$), together with the
85: resistivity due to the phonons which increases with $T$.\cite{Kondo}
86: (In the Coulomb blockade region of quantum dots, the normal conductance
87: increases with $T$ by the thermal activation. Thus the Kondo effect results
88: in a ``conductance minimum.'')
89: Third, various parameters are tunable in quantum dots using gate electrodes,
90: {\it e.g}., number of electrons, strength of tunnel coupling.
91: Nonequilibrium situations are also realized by applying a finite bias voltage.
92:
93: Besides the conventional Kondo effect with spin $1/2$,
94: unconventional Kondo effects have been discovered which stem from
95: the interplay between spin and orbital degrees of freedom in quantum dots.
96: An example is the enhanced Kondo effect at the energy degeneracy
97: between spin-singlet and -triplet states for an even number of
98: electrons (S-T Kondo effect).\cite{Sasaki}
99: In vertically fabricated quantum dots, the spacing between discrete
100: levels can be tuned by applying a magnetic field perpendicularly to
101: the quantum dots.
102: Around a level degeneracy, high-spin states appear by the exchange
103: interaction (Hund's rule).\cite{Tarucha}
104: Hence the energy difference $\Delta$ between spin-singlet and -triplet
105: states is tunable by the magnetic field, as schematically shown in Fig.\ 1(b).
106: The Kondo effect is significantly enhanced around $\Delta=0$.
107: Note that the Zeeman effect is irrelevant in the experiment
108: since the Zeeman energy is much
109: smaller than $T_{\rm K}$, owing to a small $g$ factor in semiconductors.
110: This S-T Kondo effect has been explained theoretically by the author and
111: Nazarov\cite{Eto,Eto2,Eto3} and
112: by Pustilnik and Glazman.\cite{Pustilnik,Pustilnik2}
113: (The Zeeman effect is relevant to the S-T Kondo effect in carbon
114: nanotubes.\cite{nanotube} We refer to ref.\ 23 % \cite{Pustilnik0}
115: for the theoretical study.)
116:
117: Recently another large Kondo effect has been found around the
118: two-orbital degeneracy for an odd number of electrons with spin $1/2$
119: [doublet-doublet (D-D) Kondo effect].\cite{Sasaki2}
120: When the energy difference between the two orbitals $\Delta$ is tuned
121: [Fig.\ 1(a)], the conductance is largely enhanced around $\Delta = 0$ by
122: the Kondo effect. Though the Kondo effect is expected with single level
123: and spin $1/2$, the Kondo temperature is probably smaller than actual
124: temperature $T$ so that no conductance increase is seen except the vicinity of
125: the level crossing point in this experiment. Around $\Delta = 0$,
126: the estimated $T_{\rm K}$ is similar to that for the S-T Kondo effect
127: observed in the same samples, indicating that a total of fourfold spin and
128: orbital degeneracy contributes to enhance $T_{\rm K}$ in both the cases.
129:
130: The purpose of the present paper is to study the D-D Kondo effect
131: theoretically and to elucidate the differences between D-D and
132: S-T Kondo effects. It is known that the orbital degeneracy plays an important
133: role in the Kondo effect of magnetic impurities with $f$ electrons.
134: When the total degeneracy factor is $N_{\rm d}=2j+1$ ($j$ is the total
135: angular momentum consisting of orbital angular momentum and spin),
136: the Kondo effect is described by the Coqblin-Schrieffer model of
137: SU($N_{\rm d}$) symmetry.\cite{Coqblin,Hewson}
138: This model yields the Kondo temperature
139: of $T_{\rm K} \sim D_0 e^{-1/N_{\rm d} \nu J}$ with exchange coupling $J$,
140: where $D_0$ and $\nu$ are the bandwidth and density of states of
141: conduction electrons, respectively.
142: Hence the degeneracy factor $N_{\rm d}$ significantly
143: enhances the Kondo temperature.
144: We examine the Kondo effect in quantum dots with two orbitals and spin
145: $1/2$ and evaluate the Kondo temperature as a function of $\Delta$.
146: We find that $T_{\rm K}$ is maximal around $\Delta=0$ and decreases with
147: increasing $|\Delta|$ by a power law,
148: $T_{\rm K}(\Delta)=T_{\rm K}(0)\cdot ( T_{\rm K}(0)/|\Delta| )^{\gamma}$,
149: with exponent $\gamma=1$. This behavior of $T_{\rm K}$ is
150: understood as a crossover from SU(4) to SU(2) Kondo effect when
151: the strength of the tunnel coupling is equivalent for the two orbitals.
152: Since the spin-orbit interaction can be disregarded in the
153: quantum dots, the degenerate factor is $N_{\rm d}=4$ ($N_{\rm d}=2$) for
154: orbital and spin degrees of freedom when $\Delta=0$ ($\Delta \ne 0$).
155: In general situations, the exponent of the power law is not a universal
156: value since the fixed point of SU(4) Kondo effect is marginal. We find a
157: relation of $\gamma_{\rm L}\cdot\gamma_{\rm R}=1$, where $\gamma_{\rm L}$
158: and $\gamma_{\rm R}$ are the exponents on both sides of a level crossing.
159: The $\Delta$ dependence of $T_{\rm K}$ was not studied in details by
160: previous theoretical works on the Kondo effect in
161: multilevel quantum dots.\cite{Inoshita,Inoshita2,Pohjola,Izumida,Yeyati}
162:
163: The behavior of $T_{\rm K}(\Delta)$ in the D-D Kondo effect is similar to
164: that in the S-T Kondo effect.\cite{Eto,Eto2,Eto3,Pustilnik,Pustilnik2}
165: We clarify the differences between them to explain the above-mentioned
166: experimental results.\cite{Sasaki2}.
167: Our discussion is restricted to the case of
168: quantum dots vertically fabricated from double-barrier
169: heterostructures.\cite{Sasaki,Sasaki2,Tarucha,Leo}
170: In transport processes between a quantum dot and leads through
171: heterostructure tunnel junctions, the orbital quantum numbers are conserved
172: for the motion of electrons in the transverse direction.
173: Therefore there are two channels of conduction electrons in the leads
174: when two orbitals are relevant in the quantum dot; each channel couples to
175: only one of the two orbitals.
176: The situation is different in quantum dots of lateral geometry, in which a
177: single channel in each lead couples to both the orbitals through a tunnel
178: junction.\cite{Leo}
179: We refer to refs.\ 31 % \cite{Wilfred2}
180: and 32--34 %\cite{Pustilnik3,Hofstetter,Pustilnik4}
181: for experimental and theoretical studies on the S-T Kondo effect
182: in lateral quantum dots.
183:
184: The organization of the present paper is as follows.
185: In the next section (\S 2), we model a quantum dot with
186: two orbitals and spin $1/2$ for the D-D Kondo effect.
187: In \S 3, the Kondo temperature is evaluated as a function of energy
188: difference $\Delta$ between the two orbitals, using the scaling method.
189: In \S 4, the scaling analysis is given for the S-T Kondo effect, according
190: to our previous works.\cite{Eto,Eto2,Eto3} The final section (\S 5) is
191: devoted to conclusions and discussion. We compare the D-D and S-T Kondo
192: effects to explain the experimental results.\cite{Sasaki2}
193:
194: %-------------
195: \begin{figure}[t]
196: \begin{center}
197: \includegraphics[width=8cm]{Eto-fig1.eps}
198: \end{center}
199: \caption{In a quantum dot with two orbitals, the energy difference $\Delta$
200: can be tuned using a magnetic field $B$.
201: (a) $\Delta=\varepsilon_2-\varepsilon_1$, level spacing between the
202: two orbitals for an electron in the quantum dot, and
203: (b) $\Delta=E_{S=0}-E_{S=1}$, the energy difference between spin-singlet
204: and -triplet states for two electrons.}
205: \end{figure}
206: %-------------
207:
208:
209: \section{Model}
210:
211: As a model for the D-D Kondo effect, we consider a quantum dot with an
212: electron ($S=1/2$) and two orbitals ($i=1,2$). The energy difference between
213: the orbitals is denoted by $\Delta=\varepsilon_{2}-\varepsilon_{1}$.
214: The Hamiltonian in the quantum dot reads
215: \begin{equation}
216: H_{\rm{dot}} = \sum_{\mu}\varepsilon_i d_{\mu}^{\dagger} d_{\mu}
217: +E_{\rm c}(\hat{n}-{\cal N}_{\rm{gate}})^2,
218: \label{eq:Hdot}
219: \end{equation}
220: where $d_{\mu}^{\dagger}$ and $d_{\mu}$ are creation and annihilation
221: operators of the dot state $\mu=(i,S_z)$, respectively.
222: $\hat{n}=\sum_{\mu} d_{\mu}^{\dagger}d_{\mu}$ is the number operator of
223: electrons in the quantum dot. $E_{\rm c}$ is the charging energy,
224: whereas ${\cal N}_{\rm{gate}}$ is tuned by the gate voltage.
225: We set ${\cal N}_{\rm{gate}} = 1$ for the Coulomb blockade region with
226: an electron in the quantum dot.
227: The exchange interaction between electrons in the
228: quantum dot is neglected in sections 2 and 3 since it is irrelevant to the
229: D-D Kondo effect.
230:
231: The quantum dot is connected to two external leads, $L$ and $R$,
232: through tunnel barriers.
233: We assume that the orbital symmetry is conserved in the tunnel
234: processes, which is the case of vertical quantum dots.
235: Hence the leads have two channels;
236: channel $i$ ($=1,2$) in lead $L$ ($R$) couples to orbital $i$ by $V_{L,i}$
237: ($V_{R,i}$). We perform a unitary transformation for conduction
238: electrons in the two leads, along the lines of ref.\ 4; % \cite{Glazman};
239: \begin{eqnarray*}
240: c_{k \mu} & = & (V_{L,i}^*c_{L,k \mu}+V_{R,i}^*c_{R,k \mu})/V_i,
241: \\
242: \bar{c}_{k \mu} & = & (-V_{R,i}c_{L,k \mu}+V_{L,i}c_{R,k \mu})/V_i,
243: \end{eqnarray*}
244: with $V_i=\sqrt{|V_{L,i}|^2+|V_{R,i}|^2}$, where $c_{\alpha,k\mu}$
245: is the annihilation operator of an electron in lead $\alpha$,
246: with momentum $k$ and $\mu=(i,S_z)$.
247: The mode $c_{k \mu}$ is coupled to the quantum dot with $V_i$, whereas the mode
248: $\bar{c}_{k \mu}$ is decoupled and thus shall be disregarded.
249: The Hamiltonian of the leads and that of tunnel processes are written as
250: \begin{eqnarray}
251: H_{\rm{leads}} & = & \sum_{k} \sum_{\mu}
252: \varepsilon_{k} c_{k\mu}^{\dagger} c_{k\mu}, \\
253: H_{\rm{T}} & = & \sum_{k} \sum_{\mu}
254: V_i (c_{k\mu}^{\dagger} d_{\mu} + {\rm h.c.}),
255: \label{eq:tunnel}
256: \end{eqnarray}
257: respectively.
258:
259: We define addition and extraction energies as the energy cost to add or
260: subtract an electron on/from the quantum dot,
261: $E^+=\varepsilon_1 + E_{\rm c}-E_{\rm F}$, $E^-=E_{\rm F}-\varepsilon_1$,
262: where $E_{\rm F}$ is the Fermi level in the leads.
263: We assume that $E^{\pm} \gg |\Delta|$.
264: The level broadening of orbital $i$ is given by $\Gamma_i=\pi \nu V_i^2$,
265: where $\nu$ is the density of states in the leads.
266: The level broadening and temperature $T$ should be much smaller than
267: $E^{\pm}$ for the Coulomb blockade.
268:
269: %-------------
270: \begin{figure}[t]
271: \begin{center}
272: \includegraphics[width=4.5cm]{Eto-fig2r.eps}
273: \end{center}
274: \caption{Four states in a quantum dot with an electron ($S=1/2$) and
275: two orbitals ($i=1,2$). There are exchange processes among them ($J_1$,
276: $J_2$, $\tilde{T}$ and $T$) due to the second-order tunnel couplings with
277: conduction electrons in the leads.}
278: \end{figure}
279: %-------------
280:
281: Considering the second-order tunnel processes, $H_{\rm{T}}$,
282: we obtain the effective Hamiltonian $H_{\rm eff}$ in a subspace with
283: one electron in the quantum dot. We introduce the pseudofermion operators
284: $f^{\dagger}_{\mu}$ ($f_{\mu}$) that create (annihilate) the dot state
285: $\mu=(i,S_z)$. The constraint of
286: \begin{equation}
287: \sum_{\mu} f^{\dagger}_{\mu} f_{\mu}=1
288: \end{equation}
289: is required. The Hamiltonian is written as
290: \begin{eqnarray}
291: H_{\rm{eff}} & = & \sum_{\mu}\varepsilon_i f_{\mu}^{\dagger} f_{\mu}
292: + H_{\rm{lead}} \nonumber \\
293: & + &
294: 2J_1 {\mbi s}^{(1)} \cdot {\mbi S}^{(1)}
295: +2J_2 {\mbi s}^{(2)} \cdot {\mbi S}^{(2)} \nonumber \\
296: & + & 2\tilde{J} \left[ {\mbi s}^{(12)} \cdot {\mbi S}^{(21)}+
297: {\mbi s}^{(21)} \cdot {\mbi S}^{(12)} \right] \nonumber \\
298: & + & 2\tilde{T} \left[ \left(\sum_{kk'}\sum_{s}
299: c_{k',1,s}^{\dagger} c_{k,2,s} \right)
300: \left( \sum_{s} f_{2,s}^{\dagger}f_{1,s} \right)
301: +{\rm h.c.} \right] \nonumber \\
302: & + & 2T \left[ \sum_{kk'}\sum_{s} (c_{k',1,s}^{\dagger} c_{k,1,s}-
303: c_{k',2,s}^{\dagger} c_{k,2,s}) \right]
304: \left[ \sum_{s} (f_{1,s}^{\dagger}f_{1,s}- f_{2,s}^{\dagger}f_{2,s})
305: \right],
306: \label{eq:Hmltn0}
307: \end{eqnarray}
308: where
309: \begin{eqnarray*}
310: {\mbi s}^{(i)} & = & \sum_{kk'}\sum_{ss'}
311: c_{k',i,s'}^{\dagger} ({\mbi \sigma}_{s's}/2) c_{k,i,s}, \\
312: {\mbi S}^{(i)} & = & \sum_{ss'}
313: f_{i,s'}^{\dagger} ({\mbi \sigma}_{s's}/2) f_{i,s}, \\
314: {\mbi s}^{(ij)} & = & \sum_{kk'}\sum_{ss'}
315: c_{k',i,s'}^{\dagger} ({\mbi \sigma}_{s's}/2) c_{k,j,s}, \\
316: {\mbi S}^{(ij)} & = & \sum_{ss'}
317: f_{i,s'}^{\dagger} ({\mbi \sigma}_{s's}/2) f_{j,s}.
318: \end{eqnarray*}
319: The third (fourth) term in $H_{\rm{eff}}$ represents the exchange coupling
320: between conduction electrons of channel 1 (2) and dot spin $1/2$ at orbital
321: 1 (2). The terms with $\tilde{J}$ and $\tilde{T}$ describe the interchannel
322: scattering of conduction electrons by the dot electron
323: with or without spin flip. The coupling constants are given by
324: $J_1=V_1^2/\tilde{E}_{\rm C}$, $J_2=V_2^2/\tilde{E}_{\rm C}$
325: and $\tilde{J}=\tilde{T}=V_1V_2/\tilde{E}_{\rm C}$, where
326: $1/\tilde{E}_{\rm C}=1/E^++1/E^-$.
327: These processes are schematically shown in Fig.\ 2.
328: A potential scattering term with
329: $T=(V_1^2+V_2^2)/(2\tilde{E}_{\rm C})$ is also relevant to the Kondo
330: effect, whereas irrelevant potential scatterings are
331: disregarded in $H_{\rm{eff}}$.
332:
333: In the case of $V_1=V_2$, all the coupling constants are identical to
334: one another, $J_1=J_2=\tilde{J}=\tilde{T}=T \equiv J$. Then
335: the effective Hamiltonian is reduced to the Coqblin-Schrieffer model of
336: SU(4) symmetry if $\Delta=0$.
337: Indeed, using a pseudospin $T_z= \pm 1/2$ to represent orbitals $i=1,2$,
338: $H_{\rm eff}$ is rewritten as
339: \begin{equation}
340: H_{\rm eff}=H_{\rm{lead}}
341: % \sum_{k,\mu} \varepsilon_k c_{k,\mu}^{\dagger}c_{k,\mu}
342: +J {\mbi S} \cdot {\mbi s} - \tilde{\mbi B} \cdot {\mbi S},
343: \label{eq:Hmltn}
344: \end{equation}
345: where
346: % $\displaystyle
347: \begin{eqnarray*}
348: {\mbi S} & = & \frac{1}{2}\sum_{\mu \nu} f_{\mu}^{\dagger}
349: \hat{\mbi \Sigma}_{\mu \nu} f_{\nu}, \\
350: {\mbi s} & = & \frac{1}{2}
351: \sum_{k k'}\sum_{\mu \nu} c_{k',\mu}^{\dagger}
352: \hat{\mbi \Sigma}_{\mu \nu} c_{k,\nu}.
353: \end{eqnarray*}
354: $\hat{\mbi \Sigma}$ involves 15 components which are
355: generators of the Lie algebra su(4),
356: $\{ (\tau_x, \tau_y, \tau_z, I) \otimes
357: (\sigma_x, \sigma_y, \sigma_z, I) \} - \{ I \otimes I \}$, where $\sigma_i$
358: ($\tau_i$) are the Pauli matrices in the spin (pseudospin) space
359: and $I$ is the unit matrix.
360: The energy difference $\Delta$ is expressed by the fictitious magnetic
361: field $\tilde{\mbi B}$, the $\tau_z \otimes I$ component of which is
362: $\Delta$ and the other components are 0:
363: \[
364: \tilde{\mbi B} \cdot {\mbi S}=\frac{\Delta}{2}\sum_{s}
365: (f_{1,s}^{\dagger}f_{1,s}-f_{2,s}^{\dagger}f_{2,s}).
366: \]
367: When $\tilde{\mbi B}=0$, the Hamiltonian (\ref{eq:Hmltn})
368: is invariant with respect to the rotation in a product space of spin and
369: pseudospin spaces [SU(4) symmetry].
370:
371:
372: \section{Scaling calculations}
373:
374: To evaluate the Kondo temperature, the effective Hamiltonian,
375: (\ref{eq:Hmltn0}) or (\ref{eq:Hmltn}), is analyzed using ``poor man's''
376: scaling method.\cite{Anderson,Nozieres,Cox}
377: We assume that the density of states of conduction
378: electrons in the leads is constant $\nu$ in the energy range $[-D,D]$.
379: With changing the bandwidth from $D$ to $D-|d D|$, we renormalize the
380: exchange coupling constants ($J_1$, $J_2$, $\tilde{J}$, $\tilde{T}$, $T$)
381: not to change the low-energy physics, within the second-order perturbation with
382: respect to the exchange couplings. This procedure yields the scaling equations
383: in two limits, $D \gg |\Delta|$ and $D \ll |\Delta|$.
384: The Kondo temperature $T_{\rm K}$ is determined as the
385: energy scale $D$ at which the coupling constants become so large that
386: the perturbation breaks down.
387:
388:
389: \subsection{SU(4) Kondo effect}
390:
391: We begin with the case of $V_1=V_2$, where $J_1=J_2=\tilde{J}=\tilde{T}=T
392: \equiv J$. Using the Hamiltonian (\ref{eq:Hmltn}),
393: we obtain the scaling equations for a single coupling constant $J$.
394:
395: When the energy scale $D$ is much larger than the energy difference
396: $|\Delta|$, the scaling equation is given by
397: \begin{equation}
398: dJ/d \ln D= -4\nu J^2.
399: \label{eq:scalDDA}
400: \end{equation}
401: The exchange coupling $J$ develops rapidly with decreasing $D$,
402: reflecting a fourfold degeneracy of spin and pseudospin.
403: This represents the SU(4) Kondo effect.
404:
405: When $D \ll |\Delta|$, the equation is
406: \begin{equation}
407: dJ/d \ln D= -2\nu J^2.
408: \label{eq:scalDDB}
409: \end{equation}
410: The evolution of $J$ is slower since one component of $T_z$
411: (orbital of the upper level) is irrelevant.
412: This is the conventional Kondo effect of SU(2) symmetry.
413:
414: From these equations, we evaluate the Kondo temperature $T_{\rm K}$ as a
415: function of $\Delta$. (i) $T_{\rm K}$ is maximal around $\Delta = 0$
416: [$|\Delta| \ll T_{\rm K}(\Delta=0)$].
417: In this case, the scaling equation (\ref{eq:scalDDA}) remains valid till
418: the scaling ends. This yields
419: \begin{equation}
420: T_{\rm K}(\Delta=0)=D_0 \exp[-1/4\nu J],
421: \end{equation}
422: where $D_0$ is the initial bandwidth and given by
423: $\displaystyle D_0=\sqrt{E_+E_-}$.\cite{Haldane}
424:
425: (ii) $T_{\rm K}$ is minimal when $|\Delta| \gg D_0$. Then eq.\
426: (\ref{eq:scalDDB}) is valid in the whole scaling region. Solving the
427: equation, we obtain
428: \begin{equation}
429: T_{\rm K}(\Delta=\infty)=D_0 \exp[-1/2\nu J].
430: \end{equation}
431:
432: (iii) The intermediate region of $T_{\rm K}(0) \ll |\Delta| \ll D_0$ is
433: examined as follows. With decreasing the energy scale $D$,
434: the renormalization flow goes toward the fixed point of SU(4) Kondo effect
435: first [eq.\ (\ref{eq:scalDDA}) for $D \gg |\Delta|$],
436: and then goes to the fixed point of SU(2) Kondo effect
437: [eq.\ (\ref{eq:scalDDB}) for $D \ll |\Delta|$].
438: To consider this crossover behavior, we match the solutions of
439: eqs.\ (\ref{eq:scalDDA}) and (\ref{eq:scalDDB}) at
440: $D \approx |\Delta|$. We find that $T_{\rm K}(\Delta)$
441: decreases with increasing $|\Delta|$, following a power law,
442: \begin{equation}
443: T_{\rm K}(\Delta)=T_{\rm K}(0)\cdot
444: ( T_{\rm K}(0)/|\Delta| )^{\gamma},
445: \label{eq:powerDD}
446: \end{equation}
447: with exponent $\gamma=1$.
448: Our results of (i), (ii) and (iii) clearly indicate an enhancement of the Kondo
449: effect around $\Delta=0$ [see Fig.\ 6(a) for the schematic drawing of
450: $T_{\rm K}(\Delta)$], in accordance with
451: the experimental observation.\cite{Sasaki2}
452:
453: The expression of $T_{\rm K}(\Delta)$, eq.\ (\ref{eq:powerDD}),
454: has been originally derived by Yamada {\it et al}.\ for the Kondo effect in
455: metals with magnetic impurities in the presence of a crystal
456: field.\cite{Yamada} The exact solution of this Kondo effect has also been
457: obtained.\cite{Schlottmann,Kawakami}
458: Lately Borda {\it et al}.\ has examined the SU(4) Kondo effect in a double
459: quantum dot system with interdot capacitive coupling.\cite{Borda}
460: Their model is equivalent to ours when the tunnel coupling is absent
461: between the double dots. [The magnetic field in ref.\ 42 % \cite{Borda}
462: corresponds to $\tilde{{\mbi B}}$ in eq.\ (\ref{eq:Hmltn}).]
463: Using numerical renormalization group calculations,
464: they have estimated $T_{\rm K}$ as a function of magnetic field
465: and shown its enhancement around the fourfold degeneracy.
466: This is in qualitative agreement with our scaling results.
467:
468:
469: \subsection{General case of $V_1 \ne V_2$}
470:
471: Now we examine general situations of $V_1 \ne V_2$. We set
472: $\Delta=\varepsilon_{2}-\varepsilon_{1} \ge 0$ in this subsection.
473: We derive a set of scaling equations for $J_1$, $J_2$, $\tilde{J}$,
474: $\tilde{T}$ and $T$ in Hamiltonian (\ref{eq:Hmltn0}), using the poor man's
475: scaling method.
476: For $D \gg \Delta$,
477: \begin{equation}
478: \left\{
479: \begin{array}{rcl}
480: dJ_1/\nu d \ln D &=& -2J_1^2-\tilde{J}(\tilde{J}+\tilde{T}), \\
481: dJ_2/\nu d \ln D &=& -2J_2^2-\tilde{J}(\tilde{J}+\tilde{T}), \\
482: d\tilde{J}/\nu d \ln D &=& -\tilde{J}(J_1+J_2+T)-\tilde{T}(J_1+J_2)/2, \\
483: d\tilde{T}/\nu d \ln D &=& -3\tilde{J}(J_1+J_2)/2-\tilde{T}T, \\
484: dT/\nu d \ln D &=& -3\tilde{J}^2-\tilde{T}^2.
485: \end{array}
486: \right.
487: \label{eq:DDscale}
488: \end{equation}
489: For $D \ll \Delta$,
490: \begin{equation}
491: dJ_1/ d \ln D= -2 \nu J_1^2,
492: \label{eq:DDscale2}
493: \end{equation}
494: whereas the other coupling constants do not evolve.
495: % [$J_1$ should be replaced by $J_2$ in eq.\ (\ref{eq:DDscale2})
496: % when $\Delta<0$.]
497:
498: Analyzing eqs.\ (\ref{eq:DDscale}), we find that the fixed point of SU(4)
499: Kondo effect, $J_1=J_2=\tilde{J}=\tilde{T}=T=\infty$, is marginal (Appendix A).
500: With a decrease in $D$, a marginal variable, $(J_1-J_2)/(J_1+J_2)$, is
501: almost kept at the initial value while the other variables are renormalized
502: toward the values at the fixed point.
503: In consequence, $T_{\rm K}(\Delta)$ is not a universal function although
504: $T_{\rm K}$ is always maximal around $\Delta=0$.
505:
506: %-------------
507: \begin{figure}[t]
508: \begin{center}
509: \includegraphics[width=6cm]{Eto-fig3.eps}
510: \end{center}
511: \caption{
512: Calculated results of the Kondo temperature $T_{\rm K}$ for the D-D Kondo
513: effect, as a function of energy difference $\Delta$ between the two
514: orbitals, on a log-log scale.
515: Both $T_{\rm K}$ and $\Delta$ are normalized by the Kondo temperature at
516: $\Delta=0$, $T_{\rm K}(0)$.
517: (a) $(V_2/V_1)^2=1$, (b) 3/4, (c) 1/2, (d) 4/3 and (e) 2.
518: The scaling equations, (\ref{eq:DDscale}) for $D>\Delta$ and
519: (\ref{eq:DDscale2}) for $D<\Delta$, are solved numerically
520: with the initial condition of $\nu(J_1+J_2) = 0.02$.}
521: \end{figure}
522: %-------------
523:
524: To examine $T_{\rm K}(\Delta)$, we numerically solve the scaling equations,
525: eqs.\ (\ref{eq:DDscale}) for $D>\Delta$ and
526: eq.\ (\ref{eq:DDscale2}) for $D<\Delta$.
527: We put $\nu (J_1+J_2) = 0.02$ in the initial condition and determine
528: the Kondo temperature $T_{\rm K}$ as the energy scale $D$ at which
529: $\nu J_1$ becomes as large as unity.
530: The calculated results of $T_{\rm K}(\Delta)$ are shown
531: in Fig.\ 3 on a log-log scale, with (a) $(V_2/V_1)^2=1$,
532: (b) 3/4, (c) 1/2, (d) 4/3 and (e) 2. $T_{\rm K}(\Delta)$ seems to follow
533: a power law
534: \begin{equation}
535: T_{\rm K}(\Delta)=T_{\rm K}(0)\cdot
536: ( T_{\rm K}(0)/\Delta )^{\gamma},
537: \label{eq:powerDD2}
538: \end{equation}
539: but the exponent depends on model parameters. It is approximately
540: given by
541: \begin{equation}
542: \gamma=\left( \frac{V_2}{V_1} \right)^2=\frac{\Gamma_2}{\Gamma_1},
543: \label{eq:DDgamma}
544: \end{equation}
545: as long as $\Gamma_1 \sim \Gamma_2$ (Appendix A).
546: If $\Gamma_2/\Gamma_1 \gg 1$ or $\ll 1$,
547: $T_{\rm K}(\Delta)$ deviates from the power law. Then the renormalization
548: of $\Delta$,\cite{Kuzmenko} that is not considered in our calculations,
549: might have to be taken into account to determine $T_{\rm K}$, besides the
550: renormalization of exchange couplings.
551:
552: If we denote the exponents on the left and right sides of a level crossing
553: in Fig.\ 1(a) by $\gamma_{\rm L}$ and $\gamma_{\rm R}$, respectively,
554: we obtain a general relation of
555: \[
556: \gamma_{\rm L} \cdot \gamma_{\rm R}=
557: \frac{\Gamma_2}{\Gamma_1} \cdot \frac{\Gamma_1}{\Gamma_2}=1,
558: \]
559: since the roles of orbitals 1 and 2 are interchanged at the level crossing.
560:
561:
562: \section{Singlet-triplet Kondo effect}
563:
564: We examine the Kondo effect around the singlet-triplet degeneracy for an
565: even number of electrons in this section.
566: In our previous papers,\cite{Eto,Eto2,Eto3}
567: the scaling analysis has been presented for this S-T Kondo effect in vertical
568: quantum dots in rectangular shape.\cite{Sasaki} Here, the analysis is
569: modified for circular quantum dots used in the experiment of ref.\ 24.
570: % \cite{Sasaki2}.
571:
572: We consider a quantum dot with two orbitals and two electrons.
573: The exchange interaction which favors parallel spins,
574: \[
575: %-E_{\rm ex}\sum_{s=\pm 1/2} d_{s,1}^{\dagger}d_{s,2}^{\dagger}d_{s,2}d_{s,1},
576: -E_{\rm ex}\sum_{s_1s_2s_3s_4}
577: ({\mbi \sigma}/2)_{s_1s_2} \cdot ({\mbi \sigma}/2)_{s_3s_4}
578: d_{1,s_1}^{\dagger} d_{1,s_2} d_{2,s_3}^{\dagger} d_{2,s_4}
579: \]
580: with ${\mbi \sigma}=(\sigma_x,\sigma_y,\sigma_z)$,
581: is important in this problem. We add this term to the dot Hamiltonian
582: (\ref{eq:Hdot}) and fix ${\cal N}_{\rm{gate}} = 2$ for the
583: Coulomb blockade with two electrons in the quantum dot.
584: The possible ground state is a spin singlet ($S=0$) or triplet
585: ($S=1$, $S_z \equiv M=1,0,-1$). We denote the states by $|S M \rangle$,
586: \begin{eqnarray*}
587: |0 0 \rangle & = &
588: d_{1 \uparrow}^{\dagger} d_{1 \downarrow}^{\dagger} |0 \rangle, \\
589: |1 1 \rangle & = &
590: d_{1 \uparrow}^{\dagger} d_{2 \uparrow}^{\dagger} |0 \rangle, \\
591: |1 0 \rangle & = & \frac{1}{\sqrt{2}}
592: (d_{1 \uparrow}^{\dagger} d_{2 \downarrow}^{\dagger}
593: +d_{1 \downarrow}^{\dagger} d_{2 \uparrow}^{\dagger}) |0 \rangle, \\
594: |1 -1 \rangle & = &
595: d_{1 \downarrow}^{\dagger} d_{2 \downarrow}^{\dagger} |0 \rangle,
596: \end{eqnarray*}
597: where $|0 \rangle$ represents the vacuum, assuming
598: $\varepsilon_1<\varepsilon_2$.\cite{com1}
599: The energy difference between them is $E_{S=0}-E_{S=1}=E_{\rm ex}
600: -(\varepsilon_2-\varepsilon_1)$, which is denoted by $\Delta$ in this
601: section. We disregard two other singlet states,
602: $d_{2 \uparrow}^{\dagger} d_{2 \downarrow}^{\dagger} |0 \rangle$ and
603: $(1/\sqrt{2})(d_{1 \uparrow}^{\dagger} d_{2 \downarrow}^{\dagger}
604: -d_{1 \downarrow}^{\dagger} d_{2 \uparrow}^{\dagger}) |0 \rangle$,
605: since they have larger energies than state $|0 0 \rangle$.
606: We assume that addition and extraction energies, $E^{\pm}$, are much
607: larger than $|\Delta|$ and level broadenings, $\Gamma_1$, $\Gamma_2$.
608:
609: As in \S 2, we make an effective Hamiltonian $H_{\rm eff}$ in a
610: space of two-electron states, $|0 0 \rangle$, $|1 1 \rangle$,
611: $|1 0 \rangle$ and $|1 -1 \rangle$, taking account of the second-order
612: tunnel processes, $H_{\rm{T}}$. The Hamiltonian reads
613: \begin{equation}
614: H_{\rm{eff}}=H_{\rm{dot}}+H_{\rm{leads}}+
615: H^{S=1}+H^{S=1 \leftrightarrow 0}+H_{\rm{pot}}.
616: \end{equation}
617: The first term represents the dot state,
618: \begin{equation}
619: H_{\rm{dot}}=\sum_{S,M} E_{S} f_{SM}^{\dagger}f_{SM},
620: \end{equation}
621: using pseudofermion operators $f_{SM}^{\dagger}$ ($f_{SM}$) which
622: create (annihilate) the state $|SM \rangle$. It is required that
623: \begin{equation}
624: \sum_{SM} f_{SM}^{\dagger} f_{SM} =1.
625: \end{equation}
626: $H^{S=1}$ describes the spin-flip processes among three components of
627: the triplet state, $|1 1 \rangle$, $|1 0 \rangle$ and $|1 -1 \rangle$,
628: with a scattering of conduction electrons of channel 1 or 2.
629: The coupling constants are
630: given by $J_1=V_1^2/(2\tilde{E}_{\rm C})$ and $J_2=V_2^2/(2\tilde{E}_{\rm C})$,
631: respectively, where $1/\tilde{E}_{\rm C}=1/E^++1/E^-$.
632: %
633: $H^{S=1 \leftrightarrow 0}$ represents the conversion between the triplet
634: and singlet states with an interchannel scattering of conduction electrons.
635: The coupling constant is $\tilde{J}=V_1V_2/(\sqrt{2}\tilde{E}_{\rm C})$.
636: These spin-flip processes are schematically shown in Fig.\ 4.
637: A potential scattering $H_{\rm{pot}}$ with
638: $T=(V_1^2+V_2^2)/(2\tilde{E}_{\rm C})$ is also relevant to the S-T Kondo
639: effect. (See ref.\ 19 % \cite{Eto3}
640: for explicit expressions of the Hamiltonian.)
641:
642: %-------------
643: \begin{figure}[t]
644: \begin{center}
645: \includegraphics[width=6cm]{Eto-fig4.eps}
646: \end{center}
647: \caption{Four states that we consider in a quantum dot with two orbitals
648: and two electrons, $| S,M \rangle$,
649: and exchange processes among them.
650: The exchange process denoted by $J_1$ ($J_2$) is accompanied by a spin-flip
651: scattering of conduction electrons of channel 1 (2), whereas that by
652: $\tilde{J}$ is accompanied by an interchannel scattering of conduction
653: electrons.}
654: \end{figure}
655: %-------------
656:
657: We obtain the scaling equations for the coupling constants,
658: $J_1$, $J_2$, $\tilde{J}$, $T$, using the poor man's scaling method.
659: When $D \gg |\Delta|$,
660: \begin{equation}
661: \left\{
662: \begin{array}{rcl}
663: d J_1 / \nu d \ln D & = & -2 J_1^2 - \tilde{J}^2, \\
664: d J_2 / \nu d \ln D & = & -2 J_2^2 - \tilde{J}^2, \\
665: d \tilde{J}/ \nu d \ln D & = & -2 (J_1+J_2) \tilde{J} - T \tilde{J}, \\
666: d T/ \nu d \ln D & = & -8 \tilde{J}^2.
667: \end{array}
668: \right.
669: \label{eq:scalA}
670: \end{equation}
671: These equations have a stable fixed point of
672: $J_1=J_2=\infty$,
673: $\tilde{J}/J_1=\sqrt{2(2+\sqrt{5})}$ and $T/J_1=2(1+\sqrt{5})$
674: (Appendix B).
675: The renormalization flow of the exchange couplings goes toward this fixed
676: point, as the energy scale $D$ decreases.
677:
678: When $D \ll \Delta$, the singlet state can be safely disregarded.
679: The scaling equations are given by
680: \begin{equation}
681: \left\{
682: \begin{array}{rcl}
683: d J_1 / d \ln D & = & -2 \nu J_1^2, \\
684: d J_2 / d \ln D & = & -2 \nu J_2^2,
685: \end{array}
686: \right.
687: \label{eq:scalB}
688: \end{equation}
689: indicating that $J_1$ and $J_2$ evolve independently. The other coupling
690: constants do not change.
691: On the singlet side ($\Delta<0$), all the exchange couplings
692: saturate, provided $D \ll |\Delta|$.
693:
694: The Kondo temperature $T_{\rm K}$ is evaluated as a function of $\Delta$,
695: in the same way as in \S 3. (i) When $|\Delta| \ll T_{\rm K}(0)$,
696: the scaling equations (\ref{eq:scalA}) are valid till the end of the scaling.
697: Then $T_{\rm K}$ is maximal although the analytical expression of
698: $T_{\rm K}(0)$ cannot be obtained.\cite{com2}
699:
700: (ii) When $\Delta > D_0$, the ground state is a spin triplet, whereas the
701: singlet state is not relevant to the Kondo effect.
702: Then the scaling equations (\ref{eq:scalB}) work in the whole scaling region.
703: This yields
704: \begin{equation}
705: T_{\rm K}(\infty) = D_0 \exp [-1/2\nu J_1],
706: \label{eq:TK0b}
707: \end{equation}
708: when $J_1 \ge J_2$ ($J_1$ should be replaced by $J_2$ when $J_1 < J_2$).
709: This is the Kondo temperature of a localized
710: spin with $S=1$.\cite{Okada}
711:
712: (iii) In the intermediate region of $T_{\rm K}(0) \ll \Delta \ll D_0$,
713: all the coupling constants develop with decreasing $D$, following
714: eq.\ (\ref{eq:scalA}) when $D \gg \Delta$.
715: $\tilde{J}$ and $T$ saturate at $D \approx \Delta$,
716: whereas $J_1$ and $J_2$ continue to grow by eq.\ (\ref{eq:scalB}) when
717: $D \ll \Delta$. To see the $\Delta$ dependence of $T_{\rm K}$,
718: we perform numerical calculations.\cite{Eto3}
719: We solve eqs.\ (\ref{eq:scalA}) and (\ref{eq:scalB}) for $D>\Delta$ and
720: $D<\Delta$, respectively, with the initial condition of
721: $\nu (J_1+J_2) = 0.01$. The Kondo temperature $T_{\rm K}$ is determined
722: as the energy scale $D$ at which $\nu J_1 = 1$ or $\nu J_2 = 1$.
723: The calculated results of
724: $T_{\rm K}(\Delta)$ are shown in Fig.\ 5 on a log-log scale, with
725: (a) $J_2/J_1=(V_2/V_1)^2=1$, (b) $3/4$ or $4/3$, and (c) $1/2$ or $2$.
726: $T_{\rm K}(\Delta)$ shows the power law of eq.\ (\ref{eq:powerDD2})
727: with an exponent of $\gamma=2+\sqrt{5}$ if $\Delta/T_{\rm K}(0)$ is
728: not too large (Appendix B).\cite{Pustilnik,Pustilnik2}
729: With increasing $\Delta$, the exponent gradually deviates from the universal
730: value to a nonuniversal one,
731: Min$\{ \Gamma_2/\Gamma_1, \Gamma_1/\Gamma_2 \}$.\cite{Eto,Eto2,Eto3}
732:
733: (iv) On the singlet side ($\Delta<0$), no Kondo effect is expected when
734: $|\Delta| \gg T_{\rm K}(0)$. $T_{\rm K}(\Delta)$ drops to zero
735: suddenly at $|\Delta| \sim T_{\rm K}(0)$.
736: The behavior of $T_{\rm K}(\Delta)$ is schematically shown in Fig.\ 6(b).
737:
738:
739: %-------------
740: \begin{figure}[t]
741: \begin{center}
742: %\includegraphics[width=6cm]{Eto-fig5r.eps}
743: \includegraphics[width=6cm]{JPSJ-ST.eps}
744: \end{center}
745: \caption{
746: Calculated results of the Kondo temperature $T_{\rm K}$ for the S-T Kondo
747: effect, as a function of the energy difference $\Delta$ between spin-singlet
748: and -triplet states, on a log-log scale.
749: Both $T_{\rm K}$ and $\Delta$ are normalized by the Kondo temperature at
750: $\Delta=0$, $T_{\rm K}(0)$.
751: (a) $(V_2/V_1)^2=1$, (b) $3/4$ or $4/3$, and (c) $1/2$ or $2$.
752: The scaling equations, (\ref{eq:scalA}) for $D>\Delta$ and
753: (\ref{eq:scalB}) for $D<\Delta$, are solved numerically
754: with the initial condition of $\nu(J_1+J_2) = 0.01$.
755: % A thin solid line
756: A straight broken line
757: indicates a slope of $\gamma=2+\sqrt{5} \approx 4.2$.}
758: \end{figure}
759: %-------------
760:
761:
762: \section{Conclusions}
763:
764: We have theoretically examined D-D and S-T Kondo effects in multilevel
765: quantum dots. For the D-D Kondo effect, we have considered
766: a quantum dot with two orbitals and spin $1/2$.
767: Using the poor man's scaling method, the Kondo
768: temperature $T_{\rm K}$ has been evaluated as a function of energy separation
769: $\Delta$ between the two orbitals.
770: When the tunnel couplings are equivalent for the two orbitals,
771: $V_1=V_2$ in eq.\ (\ref{eq:tunnel}),
772: $T_{\rm K}(\Delta)$ is maximal around $\Delta = 0$, and
773: decreases with increasing $|\Delta|$ obeying a power law,
774: eq.\ (\ref{eq:powerDD}), with exponent $\gamma=1$. This behavior of
775: $T_{\rm K}(\Delta)$ is understood as a crossover
776: from SU(4) to SU(2) Kondo effect. In general case with different tunnel
777: couplings for the two orbitals, $V_1 \ne V_2$,
778: we observe a power law with nonuniversal
779: exponents, reflecting the marginality of the fixed point of SU(4) Kondo effect.
780: We find a relation of $\gamma_{\rm L} \cdot \gamma_{\rm R}=1$,
781: where $\gamma_{\rm L}$ and $\gamma_{\rm R}$ are the exponents on both sides
782: of a level crossing, as long as $V_1 \sim V_2$.
783:
784:
785: %-------------
786: \begin{figure}[t]
787: \begin{center}
788: \includegraphics[width=6cm]{Eto-fig6r.eps}
789: \end{center}
790: \caption{Schematical drawing of the Kondo temperature $T_{\rm K}$, as
791: a function of $\Delta$, for (a) D-D Kondo effect and (b) S-T Kondo effect.
792: $\Delta=\varepsilon_2-\varepsilon_1$ is the level spacing between
793: two orbitals in (a), whereas
794: $\Delta=E_{S=0}-E_{S=1}$ is the energy difference between spin-singlet
795: and -triplet states in (b).
796: The tunnel couplings are equivalent for two orbitals; $V_1 = V_2$
797: in eq.\ (\ref{eq:tunnel}). $T_{\rm K}$ is normalized by
798: the Kondo temperature at $\Delta=0$, $T_{\rm K}(0)$. $D_0$ is the bandwidth
799: of conduction electrons in the leads.}
800: \end{figure}
801: %-------------
802:
803:
804: For the S-T Kondo effect, we have examined a quantum dot with two orbitals and
805: two electrons. The Kondo temperature is maximal around $\Delta = 0$,
806: where $\Delta$ is the energy difference between spin-singlet and -triplet
807: states. $T_{\rm K}(\Delta)$ decreases with increasing
808: $|\Delta|$ in a different way on the triplet side ($\Delta>0$) or
809: on the singlet side ($\Delta<0$): $T_{\rm K}(\Delta)$ approximately
810: shows a power law with $\gamma=2+\sqrt{5}$ when $\Delta>0$,
811: whereas $T_{\rm K}$ drops to zero suddenly at $|\Delta| \sim T_{\rm K}$
812: when $\Delta<0$.
813:
814:
815: In both D-D and S-T Kondo effects, the Kondo temperature is maximal
816: around the degeneracy point ($\Delta=0$), indicating that a total of fourfold
817: spin and orbital degeneracy contributes to the enhancement of the Kondo
818: effect. Now we compare $T_{\rm K}(\Delta)$ of these Kondo effects in details to
819: explain recent experimental results.\cite{Sasaki2}
820: In the experimental situation of $V_1 \approx V_2$,
821: there are two major differences between them (see Fig.\ 6).
822:
823: First, when the energy difference $\Delta$ is tuned by a magnetic field,
824: the behavior of $T_{\rm K}(\Delta)$ is almost symmetric in D-D case and
825: largely asymmetric in S-T case, with respect to the degeneracy point
826: ($\Delta=0$) in Fig.\ 1. In the former, $T_{\rm K}$ decreases gradually with
827: increasing $|\Delta|$ obeying a power law of eq.\ (\ref{eq:powerDD}) on
828: both sides of $\Delta>0$ and $\Delta<0$.
829: In the latter, $T_{\rm K}(\Delta)$ approximately obeys a power law
830: on the triplet side ($\Delta>0$), whereas it drops to zero abruptly on
831: the singlet side ($\Delta<0$).
832:
833: Second, the exponent of eq.\ (\ref{eq:powerDD}) is about unity in D-D Kondo
834: effect when $V_1 \approx V_2$. The value is smaller than $\gamma=2+\sqrt{5}$,
835: the exponent in S-T Kondo effect on the triplet side. Therefore, $T_{\rm K}$
836: decreases with increasing $|\Delta|$ more slowly in D-D case.
837:
838: In ref.\ 24, % \cite{Sasaki2},
839: both D-D and S-T Kondo effects
840: have been observed in the same quantum dot with different gate voltage
841: and magnetic field. The enhancement of conductance due to the Kondo effect,
842: $\Delta G$, has been examined in the vicinity of D-D and S-T degeneracy
843: points. As the magnetic field is changed from the degeneracy point,
844: $\Delta G$ drops asymmetrically in the S-T case, whereas $\Delta G$ decreases
845: more slowly and symmetrically in the D-D case. These behaviors of $\Delta G$
846: are in accordance with our theoretical results of $T_{\rm K}(\Delta)$
847: although the direct measurement of $T_{\rm K}(\Delta)$ has not been reported.
848:
849:
850: Finally, we comment on the electronic states at $T \ll T_{\rm K}$.
851: In the D-D Kondo effect, the dot state is represented by a pair of orbital
852: index ($i=1,2$) and spin ($S_z=\pm 1/2$). They are fully screened by the
853: electrons in the leads with two conduction channels if
854: $|\Delta| \ll T_{\rm K}$. In consequence an usual Fermi liquid state of SU(4)
855: symmetry appears in which the orbital and spin degrees of freedom are
856: totally entangled, as discussed by Borda {\it et al}.\cite{Borda}
857: The Fermi liquid properties are not changed by the marginality of
858: the SU(4) Kondo effect.
859: If $|\Delta| \gg T_{\rm K}$, the upper orbital is not relevant to the
860: Kondo state. A single channel screens an electron spin in the lower
861: orbital, whereas the other channel is decoupled. Then we should observe
862: a usual Fermi liquid state. Despite the different properties,
863: the conductance through the quantum dot is
864: $G=2e^2/h$ [apart from an asymmetric factor of two tunnel junctions]
865: in both cases.\cite{Sakano}
866: (i) When $|\Delta| \ll T_{\rm K}$, two conduction channels contribute to the
867: transport, but each channel has a phase shift of $\pi/4$ in the SU(4) Kondo
868: effect.\cite{Borda,Sakano} Hence the total conductance is $2e^2/h$.
869: (ii) When $|\Delta| \gg T_{\rm K}$, a single channel takes part in the
870: transport. The phase shift of $\pi/2$ in the conventional SU(2) Kondo effect
871: (unitary limit) results in $G=2e^2/h$.
872:
873: In the S-T Kondo effect, the spin-triplet state is fully screened by
874: two conduction channels. The phase shift of each channel is given by
875: $\pi/2$ in this case. Thus we observe $G=4e^2/h$ everywhere on the triplet
876: side at $T=0$.
877: The $T$ dependence of the conductance has been discussed in a special
878: case of $V_1=V_2$ by Pustilnik {\it et al}.\cite{Pustilnik,Pustilnik2}
879:
880:
881: \acknowledgements
882:
883: The author gratefully acknowledges discussions with Yu.\ V.\ Nazarov,
884: L.\ I.\ Glazman, H.\ Hyuga, N.\ Kawakami, S.\ Sasaki and S.\ Tarucha.
885:
886:
887: \appendix
888: \section{Fixed point of D-D Kondo effect}
889:
890: For the D-D Kondo effect, the scaling equations for the exchange couplings,
891: $J_1$, $J_2$, $\tilde{J}$, $\tilde{T}$ and $T$,
892: are given by eq.\ (\ref{eq:DDscale}) for $D \gg \Delta$.
893: To find and analyze the fixed point, it is convenient to consider the
894: equations for $j_{\rm A}=J_{\rm A}/J_{\rm S}$,
895: $\tilde{j}=\tilde{J}/J_{\rm S}$,
896: $\tilde{t}=\tilde{T}/J_{\rm S}$ and $t=T/J_{\rm S}$, where
897: $J_{\rm S}=J_1+J_2$ and $J_{\rm A}=J_1-J_2$. They are given by
898: \begin{eqnarray}
899: \frac{d j_{\rm A}}{d x} & = &
900: \frac{2j_{\rm A}}{1+j_{\rm A}^2+2\tilde{j}(\tilde{j}+\tilde{t})}-j_{\rm A},
901: \label{eq:DDeq1} \\
902: \frac{d \tilde{j}}{d x} & = &
903: \frac{\tilde{j}+\tilde{t}/2+\tilde{j}t}
904: {1+j_{\rm A}^2+2\tilde{j}(\tilde{j}+\tilde{t})}-\tilde{j},
905: \label{eq:DDeq2} \\
906: \frac{d \tilde{t}}{d x} & = &
907: \frac{3\tilde{j}/2+\tilde{t}/2+\tilde{j}t}
908: {1+j_{\rm A}^2+2\tilde{j}(\tilde{j}+\tilde{t})}-\tilde{t},
909: \label{eq:DDeq3} \\
910: \frac{d t}{d x} & = &
911: \frac{3\tilde{j}^2+\tilde{t}^2}
912: {1+j_{\rm A}^2+2\tilde{j}(\tilde{j}+\tilde{t})}-t,
913: \label{eq:DDeq4}
914: \end{eqnarray}
915: with $x=\ln (\nu J_{\rm S})$ (we assume that $J_{\rm S}$ goes to infinity
916: at the fixed point). $J_{\rm S}$ obeys the scaling equation
917: \begin{equation}
918: \frac{d}{d {\cal L}}(1/\nu J_{\rm S})=
919: 1+j_{\rm A}^2+2\tilde{j}(\tilde{j}+\tilde{t}),
920: \label{eq:DDeq0}
921: \end{equation}
922: where ${\cal L}=\ln D/D_0$.
923:
924: From scaling equations, (\ref{eq:DDeq1})--(\ref{eq:DDeq4}),
925: we find a fixed point of
926: \begin{equation}
927: j_{\rm A}=0, \ \tilde{j}=\tilde{t}=t=1/2,
928: \end{equation}
929: with $J_{\rm S}=\infty$, which corresponds to the fixed point of
930: SU(4) Kondo effect.
931: By the linearization of the scaling equations around the fixed point,
932: we find that the fixed point is marginal. A marginal valuable is $j_{\rm A}$.
933:
934: We derive the power law of $T_{\rm K}(\Delta)$, eqs.\ (\ref{eq:powerDD2}) and
935: (\ref{eq:DDgamma}), approximately in the following.
936: First, we expand $1/\nu J_{\rm S}(D)$ around $D=T_{\rm K}(0)$
937: (${\cal L}=\ln T_{\rm K}(0)/D_0 \equiv {\cal L}_{\rm K}$) at which
938: $1/\nu J_{\rm S}=0$. We assume that $\tilde{j}=\tilde{t}=t=1/2$
939: (values at the fixed point), whereas the marginal variable $j_{\rm A}$ is
940: fixed at the initial value, $(V_1^2-V_2^2)/(V_1^2+V_2^2)$, at $D=T_{\rm K}(0)$.
941: To the first order in ${\cal L}-{\cal L}_{\rm K}=\ln D/T_{\rm K}(0)$,
942: \begin{eqnarray}
943: 1/\nu J_{\rm S}(D) & \approx &
944: \left. \frac{d}{d {\cal L}}(1/\nu J_{\rm S}) \right|_{{\cal L}_{\rm K}}
945: ({\cal L}-{\cal L}_{\rm K}) \nonumber \\
946: & = & (2+j_{\rm A}^2)({\cal L}-{\cal L}_{\rm K}).
947: \label{eq:DDtmp}
948: \end{eqnarray}
949: For $T_{\rm K}(0) \ll |\Delta| \ll D_0$, $T_{\rm K}(\Delta)$ is determined
950: as the energy scale at which $1/\nu J_1=0$ in the solution of
951: eq.\ (\ref{eq:DDscale2}),
952: \begin{equation}
953: 1/\nu J_1(D)=-2 \ln T_{\rm K}(\Delta)/D,
954: \label{eq:DDtmp2}
955: \end{equation}
956: Connecting eq.\ (\ref{eq:DDtmp}) [$J_1(D)=J_{\rm S}(D)(1+j_{\rm A})/2$] and
957: eq.\ (\ref{eq:DDtmp2}) at $D=\Delta$, we obtain
958: the power law of eq.\ (\ref{eq:powerDD2}) with
959: \begin{equation}
960: \gamma=\frac{1-j_{\rm A}+j_{\rm A}^2}{1+j_{\rm A}}
961: =\left( V_2/V_1 \right)^2+O(j_{\rm A}^2).
962: \end{equation}
963: This is identical to eq.\ (\ref{eq:DDgamma}) if $j_{\rm A}^2$ is neglected.
964:
965: Apart from the fixed point (with large $\Delta$),
966: another approximation is taken to explain the power law of $T_{\rm K}(\Delta)$.
967: We expand $1/\nu J_{\rm S}$ around the initial point ($D=D_0$, ${\cal L}=0$).
968: By the substitution of the initial values,
969: $j_{\rm A}=(V_1^2-V_2^2)/(V_1^2+V_2^2)$,
970: $\tilde{j}=V_1V_2/(V_1^2+V_2^2)$ and $\tilde{t}=1/2$,
971: into the right-hand side of eq.\ (\ref{eq:DDeq0}), we obtain
972: \begin{equation}
973: 1/\nu J_{\rm S}(D) - 1/\nu J_{\rm S}(D_0) \approx 2 {\cal L}.
974: \label{eq:DDtmp3}
975: \end{equation}
976: If we connect eqs.\ (\ref{eq:DDtmp3}) and (\ref{eq:DDtmp2}) at $D=\Delta$,
977: we find $T_{\rm K}(\Delta) \propto \Delta^{-\gamma}$ with exponent of
978: eq.\ (\ref{eq:DDgamma}).
979: This implies that the power law of eqs.\ (\ref{eq:powerDD2}) and
980: (\ref{eq:DDgamma}) holds in a wide range of $\Delta$, which is
981: consistent with numerical results shown in Fig.\ 3.
982:
983:
984: \section{Fixed point of S-T Kondo effect}
985:
986: For the S-T Kondo effect, the scaling equations (\ref{eq:scalA}) for
987: $D \gg \Delta$ are analyzed in the same way as in Appendix A.
988: The scaling equations for $j_{\rm A}=J_{\rm A}/J_{\rm S}$,
989: $\tilde{j}=\tilde{J}/J_{\rm S}$ and $t=T/J_{\rm S}$,
990: where $J_{\rm S/A}=J_1 \pm J_2$, have a fixed point of
991: \begin{equation}
992: j_{\rm A}=0, \ \tilde{j}=\sqrt{(2+\sqrt{5})/2}, \ t=1+\sqrt{5},
993: \label{eq:STfixp}
994: \end{equation}
995: with $J_{\rm S}=\infty$. This is a stable fixed point.
996:
997: The scaling equation for $J_{\rm S}$ is
998: \begin{equation}
999: \frac{d}{d {\cal L}}(1/\nu J_{\rm S})=1+j_{\rm A}^2+2\tilde{j}^2,
1000: \label{eq:eqST0}
1001: \end{equation}
1002: where ${\cal L}=\ln D/D_0$. We expand the solution of eq.\ (\ref{eq:eqST0})
1003: around the fixed point of eq.\ (\ref{eq:STfixp}) at which
1004: $D=T_{\rm K}(0)$ (${\cal L}={\cal L}_{\rm K}$).
1005: To the first order in ${\cal L}-{\cal L}_{\rm K}=\ln D/T_{\rm K}(0)$,
1006: \begin{equation}
1007: 1/\nu J_{\rm S}(D) \approx (3+\sqrt{5})({\cal L}-{\cal L}_{\rm K}).
1008: \label{eq:STtmp}
1009: \end{equation}
1010: For $T_{\rm K}(0) \ll |\Delta| \ll D_0$,
1011: we match eq.\ (\ref{eq:STtmp}) [$J_1(D)=J_2(D)=J_{\rm S}(D)/2$] and
1012: the solution of eqs.\ (\ref{eq:scalB}) at $D=\Delta$,\cite{Pustilnik}
1013: and find a power law, eq.\ (\ref{eq:powerDD2}), with
1014: \begin{equation}
1015: \gamma=2+\sqrt{5}.
1016: \end{equation}
1017:
1018: The universal exponent of the power law is observed in a limited range of
1019: $\Delta$ in Fig.\ 5. Far away from the fixed point, we expand $1/\nu J_{\rm S}$
1020: around the initial point of $D=D_0$ (${\cal L}=0$).
1021: Substituting the initial values, $j_{\rm A}=(V_1^2-V_2^2)/(V_1^2+V_2^2)$
1022: and $\tilde{j}=\sqrt{2}V_1V_2/(V_1^2+V_2^2)$,
1023: into the right-hand side of eq.\ (\ref{eq:eqST0}), we obtain
1024: \begin{equation}
1025: 1/\nu J_{\rm S}(D) - 1/\nu J_{\rm S}(D_0) \approx 2 {\cal L}.
1026: \label{eq:STtmp2}
1027: \end{equation}
1028: We match eq.\ (\ref{eq:STtmp2}) [$J_{1,2}(D)=J_{\rm S}(D)(1 \pm j_{\rm A})/2$]
1029: and the solution of eqs.\ (\ref{eq:scalB}) at $D=\Delta$, we find
1030: a power law, $T_{\rm K}(\Delta) \propto \Delta^{-\gamma}$, with exponent of
1031: \begin{equation}
1032: \gamma=\left( \frac{V_2}{V_1} \right)^2=\frac{\Gamma_2}{\Gamma_1},
1033: \label{eq:STgamma}
1034: \end{equation}
1035: if $\Gamma_1 \ge \Gamma_2$ ($J_1 \ge J_2$ and hence $J_1$ diverges faster
1036: than $J_2$), whereas
1037: \begin{equation}
1038: \gamma=\left( \frac{V_1}{V_2} \right)^2=\frac{\Gamma_1}{\Gamma_2},
1039: \label{eq:STgamma2}
1040: \end{equation}
1041: if $\Gamma_1 < \Gamma_2$. Note that $0<\gamma \le 1$ in this case.
1042: In consequence,
1043: the exponent of the power law of $T_{\rm K}(\Delta)$ changes gradually
1044: from the universal value, $2+\sqrt{5}$, to
1045: a nonuniversal one given by eq.\ (\ref{eq:STgamma}) or (\ref{eq:STgamma2}),
1046: with increasing $\Delta$, as shown in Fig.\ 5.\cite{Eto3}
1047: This is in contrast to the case of Fig.\ 3 for the D-D Kondo effect in
1048: which a power law with fixed exponent holds in a wide range of $\Delta$.
1049:
1050:
1051:
1052: \begin{thebibliography}{99}
1053: \bibitem{Kondo}
1054: J.\ Kondo: Prog.\ Theor.\ Phys.\ {\bf 32} (1964) 37.
1055:
1056: \bibitem{Hewson}
1057: A.\ C.\ Hewson: {\it The Kondo problem to heavy fermions}
1058: (Cambridge University Press, Cambridge, 1993).
1059:
1060: \bibitem{Yosida}
1061: K.\ Yosida: {\it Theory of Magnetism} (Springer, New York, 1996).
1062:
1063: \bibitem{Glazman}
1064: L.\ I.\ Glazman and M.\ {\'E}.\ Ra{\u\i}kh:
1065: Pis'ma Zh.\ Eksp.\ Teor.\ Fiz.\ {\bf 47} (1988) 378
1066: [JETP Lett.\ {\bf 47} (1988) 452].
1067: \bibitem{Ng} T.\ K.\ Ng and P.\ A.\ Lee:
1068: Phys.\ Rev.\ Lett.\ {\bf 61} (1988) 1768.
1069: \bibitem{Kawabata}
1070: A.\ Kawabata: J.\ Phys.\ Soc.\ Jpn.\ {\bf 60} (1991) 3222.
1071: \bibitem{Hershfield}
1072: S.\ Hershfield, J.\ H.\ Davies and J.\ W.\ Wilkins:
1073: Phys.\ Rev.\ Lett.\ {\bf 67} (1991) 3720.
1074: \bibitem{Hershfield2}
1075: S.\ Hershfield, J.\ H.\ Davies and J.\ W.\ Wilkins:
1076: Phys.\ Rev.\ B {\bf 46} (1992) 7046.
1077: \bibitem{Meir}
1078: Y.\ Meir, N.\ S.\ Wingreen and P.\ A.\ Lee:
1079: Phys.\ Rev.\ Lett.\ {\bf 70} (1993) 2601.
1080:
1081: \bibitem{Leo}
1082: L.\ P.\ Kouwenhoven, C.\ M.\ Marcus, P.\ L.\ McEuen, S.\ Tarucha,
1083: R.\ M.\ Westervelt and N.\ S.\ Wingreen: in
1084: {\it Mesoscopic Electron Transport}, NATO ASI Series E {\bf 345},
1085: eds.\ L.\ Y.\ Sohn, L.\ P.\ Kouwenhoven and G.\ Sch\"{o}n (Kluwer, 1997),
1086: p.\ 105.
1087:
1088: \bibitem{GGNat}
1089: D.\ Goldhaber-Gordon, H.\ Shtrikman, D.\ Mahalu, D.\ Abusch-Magder,
1090: U.\ Meirav and M.\ A.\ Kastner:
1091: Nature {\bf 391} (1998) 156.
1092: \bibitem{Cronen}
1093: S.\ M.\ Cronenwett, T.\ H.\ Oosterkamp and L.\ P.\ Kouwenhoven:
1094: Science {\bf 281} (1998) 540.
1095: \bibitem{GGPRL}
1096: D.\ Goldhaber-Gordon, J.\ G\"ores, M.\ A.\ Kastner, H.\ Shtrikman,
1097: D.\ Mahalu and U.\ Meirav:
1098: Phys.\ Rev.\ Lett.\ {\bf 81} (1998) 5225.
1099:
1100: \bibitem{Wilfred}
1101: W.\ G.\ van der Wiel, S.\ De Franceschi, T.\ Fujisawa,
1102: J.\ M.\ Elzerman, S.\ Tarucha and L.\ P.\ Kouwenhoven:
1103: Science {\bf 289} (2000) 2105.
1104:
1105: \bibitem{Sasaki}
1106: S.\ Sasaki, S.\ De Franceschi, J.\ M.\ Elzerman, W.\ G.\ van der Wiel,
1107: M.\ Eto, S.\ Tarucha and L.\ P.\ Kouwenhoven:
1108: Nature {\bf 405} (2000) 764.
1109:
1110: \bibitem{Tarucha}
1111: S.\ Tarucha, D.\ G.\ Austing, T.\ Honda, R.\ J.\ van der Hage and
1112: L.\ P.\ Kouwenhoven: Phys.\ Rev.\ Lett.\ {\bf 77} (1996) 3613.
1113:
1114: \bibitem{Eto}
1115: M.\ Eto and Yu.\ V.\ Nazarov: Phys.\ Rev.\ Lett.\ {\bf 85} (2000) 1306.
1116:
1117: \bibitem{Eto2}
1118: M.\ Eto and Yu.\ V.\ Nazarov: Phys.\ Rev.\ B {\bf 64} (2001) 85322.
1119:
1120: \bibitem{Eto3}
1121: M.\ Eto and Yu.\ V.\ Nazarov: Phys.\ Rev.\ B {\bf 66} (2002) 153319.
1122:
1123: \bibitem{Pustilnik}
1124: M.\ Pustilnik and L.\ I.\ Glazman: Phys.\ Rev.\ Lett.\ {\bf 85} (2000) 2993.
1125:
1126: \bibitem{Pustilnik2}
1127: M.\ Pustilnik and L.\ I.\ Glazman: Phys.\ Rev.\ B {\bf 64} (2001) 45328.
1128:
1129: \bibitem{nanotube}
1130: J.\ Nyg{\aa}rd, D.\ H.\ Cobden and P.\ E.\ Lindelof:
1131: Nature {\bf 408} (2000) 342.
1132: \bibitem{Pustilnik0}
1133: M.\ Pustilnik, Y.\ Avishai, and K.\ Kikoin:
1134: Phys.\ Rev.\ Lett.\ {\bf 84} (2000) 1756.
1135:
1136: \bibitem{Sasaki2}
1137: S.\ Sasaki, S.\ Amaha, N.\ Asakawa, M.\ Eto and S.\ Tarucha:
1138: Phys.\ Rev.\ Lett.\ {\bf 93} (2004) 17205.
1139:
1140: \bibitem{Coqblin}
1141: B.\ Coqblin and J.\ R.\ Schrieffer: Phys.\ Rev.\ {\bf 185} (1969) 847.
1142:
1143: \bibitem{Inoshita}
1144: T.\ Inoshita, A.\ Shimizu, Y.\ Kuramoto and H.\ Sakaki:
1145: Phys.\ Rev.\ B {\bf 48} (1993) R14725.
1146: \bibitem{Inoshita2}
1147: T.\ Inoshita, Y.\ Kuramoto and H.\ Sakaki:
1148: Superlattices Microstruct.\ {\bf 22} (1997) 75.
1149: \bibitem{Pohjola}
1150: T.\ Pohjola, J.\ K\"onig, M.\ M.\ Salomaa, J.\ Schmid, H.\ Schoeller,
1151: and G.\ Sch\"on: Europhys.\ Lett.\ {\bf 40} (1997) 189.
1152: \bibitem{Izumida}
1153: W.\ Izumida, O.\ Sakai and Y.\ Shimizu:
1154: J.\ Phys.\ Soc.\ Jpn.\ {\bf 67} (1998) 2444.
1155: \bibitem{Yeyati}
1156: A.\ Levy Yeyati, F.\ Flores and A.\ Mart\'in-Rodero:
1157: Phys.\ Rev.\ Lett.\ {\bf 83} (1999) 600.
1158:
1159: \bibitem{Wilfred2}
1160: W.\ G.\ van der Wiel, S.\ De Franceschi, J.\ M.\ Elzerman, S.\ Tarucha,
1161: L.\ P.\ Kouwenhoven, J.\ Motohisa, F.\ Nakajima and T.\ Fukui:
1162: Phys.\ Rev.\ Lett.\ {\bf 88} (2002) 126803.
1163: \bibitem{Pustilnik3}
1164: M.\ Pustilnik and L.\ I.\ Glazman: Phys.\ Rev.\ Lett.\ {\bf 87} (2001) 216601.
1165: \bibitem{Hofstetter}
1166: W.\ Hofstetter and H.\ Schoeller: Phys.\ Rev.\ Lett.\ {\bf 88} (2002) 16803.
1167: \bibitem{Pustilnik4}
1168: M.\ Pustilnik, L.\ I.\ Glazman and W.\ Hofstetter:
1169: Phys.\ Rev.\ B {\bf 68} (2003) 161303(R).
1170:
1171: \bibitem{Anderson}
1172: P.\ W.\ Anderson: J.\ Phys.\ C {\bf 3} (1970) 2436.
1173: \bibitem{Nozieres}
1174: P.\ Nozi\`eres and A.\ Blandin: J.\ Phys.\ (Paris) {\bf 41} (1980) 193.
1175: \bibitem{Cox}
1176: D.\ L.\ Cox and A.\ Zawadowski: Adv.\ Phys.\ {\bf 47} (1998) 599.
1177:
1178: \bibitem{Haldane}
1179: F.\ D.\ M.\ Haldane: J.\ Phys.\ C {\bf 11} (1978) 5015.
1180:
1181: \bibitem{Yamada}
1182: K.\ Yamada, K.\ Yosida and K.\ Hanzawa: Prog.\ Theor.\ Phys.\ {\bf 71} (1984)
1183: 450.
1184:
1185: \bibitem{Schlottmann}
1186: P.\ Schlottmann: Phys.\ Rev.\ B {\bf 30} (1984) 1454.
1187:
1188: \bibitem{Kawakami}
1189: N.\ Kawakami and A.\ Okiji: J.\ Phys.\ Soc.\ Jpn.\ {\bf 54} (1985) 685.
1190:
1191: \bibitem{Borda}
1192: L.\ Borda, G.\ Zar\'and, W.\ Hofstetter, B.\ I.\ Halperin and J.\ von Delft:
1193: Phys.\ Rev.\ Lett.\ {\bf 90} (2003) 26602.
1194:
1195: \bibitem{Kuzmenko}
1196: T.\ Kuzmenko, K.\ Kikoin and Y.\ Avishai:
1197: Phys.\ Rev.\ Lett.\ {\bf 89} (2002) 156602.
1198:
1199: \bibitem{com1}
1200: In our previous papers,\cite{Eto,Eto2,Eto3}
1201: we have considered a singlet state of
1202: $|0 0 \rangle =
1203: (C_1 d_{1 \uparrow}^{\dagger} d_{1 \downarrow}^{\dagger}
1204: -C_2 d_{2 \uparrow}^{\dagger} d_{2 \downarrow}^{\dagger}) |0 \rangle$,
1205: where $|C_1|^2+|C_2|^2=1$. In general, $C_2 \ne 0$ owing to
1206: the matrix element of the Coulomb interaction between
1207: $d_{1 \uparrow}^{\dagger} d_{1 \downarrow}^{\dagger} |0 \rangle$
1208: and $d_{2 \uparrow}^{\dagger} d_{2 \downarrow}^{\dagger} |0 \rangle$,
1209: $\langle 22|e^2/r|11 \rangle$.
1210: In this paper, we fix $C_1=1$ and $C_2=0$ to examine the
1211: experimental situation of circular dots,\cite{Sasaki2} in which
1212: the total angular momentum conserves and hence the matrix element is absent
1213: if orbitals 1 and 2 have different angular momentums.
1214:
1215: \bibitem{com2}
1216: A lower limit of $T_{\rm K}(0)$ is given by
1217: $D_0 \exp [-1/2\nu (J_1+J_2)]$.\cite{Eto3}
1218:
1219: \bibitem{Okada}
1220: I.\ Okada and K.\ Yosida: Prog.\ Theor.\ Phys.\ {\bf 49} (1973) 1483.
1221:
1222: \bibitem{Sakano}
1223: R.\ Sakano and N.\ Kawakami: private communications.
1224:
1225: \end{thebibliography}
1226:
1227: \end{document}
1228: