1: \documentclass[prb,twocolumn,showpacs,superscriptaddress]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{subfigure}
4:
5: \begin{document}
6:
7: \title{New Numerical Results Indicate a Half-Filling SU(4) Kondo State in Carbon Nanotubes}
8:
9: \author{C. A. B\"usser}
10: \affiliation{Condensed Matter Sciences Division, Oak Ridge National Laboratory, Oak Ridge,
11: Tennessee 37831}
12: \affiliation{Department of Physics and Astronomy, The University of Tennessee, Knoxville,
13: Tennessee 37996}
14: \author{G. B. Martins}
15: \email[corresponding author: ]{martins@oakland.edu}
16: \affiliation{Department of Physics, Oakland University, Rochester, MI 48309}
17:
18: \begin{abstract}
19: Numerical calculations simulate transport experiments in carbon nanotube
20: quantum dots (P. Jarillo-Herrero {\it et. al}, Nature {\bf 434}, 484 (2005)), where
21: a strongly enhanced Kondo temperature ${\rm T_K} \approx 8.0 ~{\rm K}$ was associated with the
22: SU(4) symmetry of the Hamiltonian at quarter-filling
23: for an orbitally double-degenerate single-occupied electronic shell.
24: Our results clearly suggest that the Kondo conductance measured for an adjacent shell
25: with ${\rm T_K} \approx 16.0 ~{\rm K}$, interpreted as a singlet-triplet
26: Kondo effect, can be associated instead to an SU(4) Kondo effect at {\it half-filling}. Besides presenting
27: spin-charge Kondo screening similar to the quarter-filling SU(4), the half-filling SU(4) has been
28: recently associated to very rich physical behavior, including
29: a non-Fermi-liquid state (M. R. Galpin {\it et al.}, Phys. Rev. Lett. {\bf 94}, 186406 (2005)).
30: \end{abstract}
31:
32: \pacs{71.27.+a,72.15.Qm,73.63.-b,73.63.Kv}
33: \maketitle
34:
35: \section{Introduction}
36:
37: The synthesis of nanostructures such as quantum dots (QDs) has attained a
38: high level of sophistication, allowing control over systems displaying complex many-body
39: properties. Recently, the observation of the Kondo effect in
40: orbitally degenerate carbon nanotube (CNT) QDs by Jarillo-Herrero {\it et al.}
41: \cite{jarillo} has renewed interest in the so-called
42: SU(4) Kondo effect.
43: Early measurements of orbital Kondo effect in double QDs can be found
44: in work by U. Wilhelm {\it et al.} \cite{wilhelm}, while more recent results
45: are reported in A. W. Holleitner {\it et al.} \cite{holleitner}.
46: However, no conclusive evidence for an SU(4) Kondo effect in double QDs has been
47: established. To date, besides the results in CNT QDs \cite{jarillo}, clear evidence
48: of SU(4) Kondo has been reported in vertical QDs \cite{sasaki1}.
49: Early theoretical work can be found in T. Pohjola {\it et al.} \cite{pohjola} and
50: in L. Borda {\it et al.} \cite{borda}, while a
51: review of SU(4) Kondo in nanostructures was written by G. Zarand \cite{zarand}.
52: Recently, a flurry of theoretical results exploring more detailed
53: aspects of the SU(4) Kondo effect have been presented for a diversity of setups \cite{lehur}.
54: Quite recent transport measurements in ambipolar semiconducting CNT QDs \cite{gleb}
55: report conductance results for a large sequence of electronic shells in the QD, with
56: clear indication of SU(4) states.
57:
58: Besides the fact that the Kondo temperature of
59: an SU(4) Kondo state is in general at least one order of magnitude higher than the
60: traditional SU(2) Kondo temperatures \cite{note1}, there is also great interest in
61: studying mesoscopic
62: systems with two or more interacting SU(4) Kondo impurities, since this could
63: shed light into the puzzling behavior of some bulk systems displaying the orbitally
64: degenerate Kondo effect. For instance, $\rm Ce_x La_{1-x} B_6$, a well-known Kondo system
65: with orbitally degenerate impurities, presents a magnetic phase diagram which
66: still defies theoretical description \cite{iroi}.
67: Another intriguing aspect recently discussed is the possibility of
68: orbitally degenerate QDs being Jahn-Teller active \cite{toonen}. In addition, the simultaneous
69: Kondo screening of charge and spin, resulting in a many-body entangled state
70: for these two degrees of freedom \cite{goldhaber}, points to the exciting
71: possibility of observing new many-body states.
72:
73: The interpretation of the CNT experimental results \cite{jarillo} through numerical
74: calculations has concentrated specifically
75: on quarter-filling (QF) (1 electron occupying the topmost electronic shell in the QD)
76: \cite{aguado1, aguado2}. In reality, most of the theoretical research on the SU(4)
77: Kondo effect in QDs has concentrated on the QF regime, with the sole exception of
78: the work by M. R. Galpin {\it et al.} \cite{galpin1, galpin2}, where NRG
79: calculations analyzed the properties of the SU(4) Kondo effect at half-filling (HF), i.e.,
80: with 2 electrons in the topmost electronic shell.
81:
82: In this paper, motivated by these interesting new possibilities regarding the
83: SU(4) Kondo effect in QDs, the authors will use a recently developed numerical
84: method, called Embedded Cluster Approximation (ECA) \cite{method}, to reanalyze
85: the conductance measurements performed in a CNT QD by
86: Jarillo-Herrero {\it et al.} \cite{jarillo} and also to extend
87: the already mentioned QF \cite{aguado1,aguado2} and HF \cite{galpin1,galpin2}
88: results to all fillings, paying special attention to
89: the robustness of the SU(4) state in respect to the tunneling properties
90: of the QD.
91: The rest of the paper is organized as follows: The model used is presented
92: in section II, where the ECA method will be briefly described.
93: As an illustration of the capabilities of the ECA method, in section III the
94: authors qualitatively reproduce the NRG results presented by Galpin {\it et al.}
95: \cite{galpin1}.
96: In section IV it will be shown that the state where the orbital degree
97: of freedom is not conserved upon tunneling (see Fig. 1), also called Two-Level SU(2) (2LSU(2)),
98: is qualitatively different from the SU(4) state (where the orbital degree of freedom
99: is conserved) even at zero magnetic field. Also in section IV,
100: the authors will analyze the transition between SU(4) and 2LSU(2) states,
101: comparing our results to previously published results \cite{aguado2}.
102: In section V, by realizing that there {\it is} spin-charge entanglement also at
103: HF, the authors will suggest that the conductance of one of the electronic
104: shells observed in the experiments (the third shell) can be associated to an SU(4)
105: Kondo effect, offering an alternative to the single-triplet effect \cite{sasaki} interpretation
106: suggested by Jarillo-Herrero {\it et al.} \cite{jarillo}.
107: The authors also note that recent experimental results by Makarovski {\it et al.}
108: \cite{gleb} n CNT QDs give support to our HF SU(4) Kondo interpretation.
109: This new interpretation of the CNT results \cite{jarillo} implies that the rich
110: physics unveiled by the NRG results
111: of Galpin {\it et al.} \cite{galpin1}, also confirmed by our results (see Fig. 2), could,
112: at least in principle, be probed in CNT QDs.
113: To further support our interpretation, in section V results in agreement with the experiments will
114: be presented for a magnetic field applied along the axis of the CNT \cite{jarillo,jarillo2}.
115: In section VI the conclusions are presented.
116:
117: \section{Model}
118:
119: \begin{figure}[h]
120: \centering
121: \includegraphics[height=4.0cm]{fig1.eps}
122: \caption{ Schematic representation of the system being analyzed. Top:
123: Hopping matrix elements (in red) conserving the orbital degree of freedom
124: (indicated by the blue arrows) lead to a Hamiltonian with SU(4) symmetry.
125: Bottom: The orbital degree of freedom is not conserved
126: upon tunneling (notice the green arrows), leading to a so-called 2LSU(2) state
127: (see reference \onlinecite{aguado2} for a detailed discussion of this
128: transition at QF).
129: }
130: \end{figure}
131:
132: The CNT QD will be modeled by an orbitally degenerate Anderson impurity
133: coupled to leads with two conduction channels:
134: \begin{eqnarray}
135: H_{\rm d}=\sum_{\sigma;\lambda=\alpha,\beta}
136: \left[ {U \over 2} n_{\lambda \sigma} n_{\lambda \bar{\sigma}} + V_g n_{\lambda \sigma}\right]
137: + U^{\prime} \sum_{\sigma \sigma^{\prime}} n_{\alpha \sigma} n_{\beta \sigma'} ,
138: \end{eqnarray}
139: \begin{eqnarray}
140: H_{\rm leads} &=& t \sum_{l=R,L} \sum_{\sigma;\lambda=\alpha,\beta;i}
141: \left[ c_{l_{\lambda} i\sigma}^{\dagger}
142: c_{l_{\lambda} i+1\sigma} +\mbox{h.c.} \right], \\
143: H_{\rm int} &=& \sum_{l=R,L} \sum_{\lambda;\lambda^{\prime};\sigma} t_{\lambda \lambda^{\prime}}
144: \left[ d_{\lambda \sigma}^{\dagger} c_{l_{\lambda^{\prime}} 0\sigma} + \mbox{h.c.} \right] ,
145: \end{eqnarray}
146: where $H_{\rm d}$ describes the orbitally degenerate Anderson impurity, subjected to a gate potential
147: $V_g$, and the second and third equations describe the leads and their interaction with the
148: CNT QD, respectively. More specifically, $\lambda=\alpha, \beta$ are
149: two degenerate orbitals associated to the wrapping mode (clockwise or counterclockwise)
150: of the electron propagation along the axial direction of the CNT \cite{minot},
151: while $d_{\alpha \sigma}$ ($d_{\beta \sigma}$) annihilates an electron with spin $\sigma$ in the
152: $\alpha$ ($\beta$) orbital in the CNT and $c_{l_{\alpha}i\sigma}$ ($c_{l_{\beta}i\sigma}$)
153: annihilates an electron with spin $\sigma$ in the i-th site of the
154: $\alpha$ ($\beta$) channel in the $l=R, L$ (right or left) lead \cite{channels}.
155: We introduce intra- and inter-orbital Coulomb repulsions $U$ and $U^{\prime}$, respectively.
156: To decrease the number of parameters in the model, the hopping matrix elements connecting
157: the CNT QD to the leads (eq. 3) are the same at left and right,
158: and assumed to follow the equalities $t_{\alpha \alpha}=t_{\beta \beta}=t^{\prime}$ (red
159: arrows in Fig. 1)
160: and $t_{\alpha \beta}=t_{\beta \alpha}=t^{\prime \prime}$ (green arrows in Fig. 1). As discussed in
161: references \onlinecite{aguado1} and \onlinecite{ aguado2},
162: when $t^{\prime}$ is finite and $t^{\prime \prime}=0$,
163: one has an SU(4) Kondo state. On the other hand, when $t^{\prime}=t^{\prime \prime}$,
164: one has the so-called 2LSU(2) Kondo state.
165: As discussed in the Supplementary Information in reference
166: \onlinecite{jarillo}, there are fundamental differences between these two states.
167: The 2LSU(2) Kondo effect is more akin to the singlet-triplet Kondo effect \cite{sasaki},
168: where spin and orbital degrees of freedom have different roles: only the spin degree of freedom
169: is screened by the conduction electrons, while the degenerate orbital levels just
170: contribute to the increase in the possible number of co-tunneling processes,
171: leading to an enhanced Kondo temperature. In this case, the orbital degree
172: of freedom is {\bf not} screened by the conduction electrons. On the other
173: hand, in the SU(4) state, spin and orbital degrees of freedom
174: participate in the same footing. Both of them are screened by the conduction
175: electrons, and if the screening of one of the degrees of freedom is somehow
176: suppressed, the system is then left in an SU(2) Kondo effect stemming from
177: the other degree of freedom.
178:
179: To calculate the conductance $G$, using the Keldysh formalism \cite{Meir-cnd},
180: a cluster containing the orbitally degenerate Anderson impurity plus a few sites of
181: the leads is solved exactly, the Green
182: functions are calculated, and the leads are then incorporated through a Dyson Equation
183: embedding procedure \cite{method}. All the results shown were obtained for
184: (in units of $t$) $U=0.5$, $t^{\prime}=0.2$,
185: zero-bias, and zero temperature. The value of $t^{\prime \prime}$ varies between zero and
186: $t^{\prime}$ and most of the results shown are for $U^{\prime}=U$.
187: When a magnetic field is applied along the CNT axis,
188: besides the Zeeman splitting coming from the spin degree of freedom, the orbital
189: levels behave as a pseudo-spin $1/2$ and are also split \cite{minot}:
190: $H_{\rm Zeeman} = B\left[\mu_{orb} \sum_{\sigma} \left(n_{\beta\sigma}-n_{\alpha\sigma}\right)
191: + \mu_{sp} \sum_{\lambda} \left(n_{\lambda+}-n_{\lambda-}\right)\right]$.
192: As reported by Jarillo-Herrero {\it et al.} \cite{jarillo}, the
193: orbital magnetic moment $\mu_{orb}$ experimentally measured is such that the orbital
194: splitting is one order of magnitude larger than the spin splitting.
195: For the actual simulation of the experimental results (Figs. 6 and 7), the degeneracy
196: of the orbital levels will be raised by introducing a small energy
197: splitting $\delta E$ \cite{splitting}.
198:
199: \section{Comparison with NRG results at Half-Filling}
200:
201: Before presenting the main results in this work, the authors
202: will show that ECA can qualitatively reproduce the NRG results of Galpin
203: {\it et al.} (Fig. 2). Besides showing that ECA captures correctly the
204: physics of the model, this qualitative agreement shows that the NRG
205: low-energy results in ref. \onlinecite{galpin1} are robust, suggesting
206: that they could be observable if one can find an experimental realization
207: of the SU(4) state at HF (more on that below).
208:
209: \begin{figure}[h]
210: \centering
211: \includegraphics[angle=90,height=6.5cm]{fig2.eps}
212: \caption{
213: (a) Blue: Conductance $G$ (solid) and average occupancy $\left< n \right>$
214: per spin orientation, per orbital (dashed) obtained when the two orbital levels are
215: not correlated ($U^{\prime}=0.0$). Red: Same as blue curves, but
216: now the orbital levels are correlated ($U^{\prime}=0.5$). In both cases,
217: $U=0.5$, $t^{\prime}=0.2$, and $t^{\prime \prime}=0.0$.
218: (b) Variation with $U^{\prime}$ of the LDOS at HF, from $U^{\prime}=0.0$ (SU(2)$\times$SU(2),
219: bottom curve) to $U^{\prime}=U=0.5$ (SU(4), top curve).
220: (c) Same as in (b), but now for $U^{\prime}\geq U$ ($U^{\prime}=U$ top curve).
221: (d) Variation with $U^{\prime}$ of the width of the peaks ($\Gamma$) in (b) and (c), which is proportional
222: to $T_K$. The qualitative behavior matches the results presented in reference \onlinecite{galpin1}.
223: }
224: \end{figure}
225:
226: Figure 2 displays the influence of $U^{\prime}$ (Coulomb repulsion
227: between the two orbital levels) over the conductance in the case
228: where the two channels are independent ($t^{\prime \prime}=0.0$),
229: i. e., a transmitted electron that tunnels into the CNT QD coming from the
230: $\alpha$ ($\beta$)-channel in the left lead, can only tunnel out through
231: the $\alpha$ ($\beta$)-channel in the right lead. The blue curves
232: ($U^{\prime}=0$) in Fig. 2a are representative of two {\it independent} spin SU(2) Kondo
233: effects (associated to each channel) which are simply added together.
234: In this case, each channel magnetically screens the spin situated
235: on the level to which it is connected ($\alpha$ or $\beta$). This
236: situation changes if the orbital levels are
237: correlated with each other (finite $U^{\prime}$, red curves), i. e., in the
238: SU(4) state. As can be seen in Fig. 2b (at HF, i.e., $\left< n \right> =0.5$),
239: the width of the Kondo resonance
240: in the local density of states (LDOS) becomes larger as $U^{\prime}$ increases, indicating
241: an enhancement of $T_K$: in the SU(4) state at HF, both
242: degrees of freedom (spin and orbital) are participating in the Kondo effect
243: and are being simultaneously screened (magnetically and electrostatically) by
244: the conduction electrons \cite{galpin1}. Similarly to QF (as described in Fig. 1
245: of Jarillo-Herrero {\it et al.} \cite{jarillo}), the increase in the number
246: of possible co-tunneling processes now available
247: for electron transport through the CNT QD results in an enhanced Kondo effect \cite{range}.
248: Fig. 2c shows the abrupt suppression of the Kondo resonance for $U^{\prime} > U$.
249: The difference between the
250: regions above and below the SU(4) point is more clearly seen in Fig. 2d, where the width
251: of the LDOS peaks in 2b and 2c (which is proportional to $T_K$) is plotted.
252: This result is in qualitative agreement with ref. \onlinecite{galpin1}.
253:
254: \begin{figure}[h]
255: \centering
256: \includegraphics[angle=90,height=3.5cm]{fig3a.eps}
257: \includegraphics[angle=90,height=3.5cm]{fig3b.eps}
258: \caption{ Illustration of the difference between SU(4) and
259: 2LSU(2) Kondo states. (a) By applying an external magnetic field
260: and making $\mu_{orb}=0.0$ and $\mu_{sp}B=2.0$, one suppresses
261: the spin Kondo effect. In this case, the SU(4) Kondo peak seen
262: in Fig. 2 (red curve) splits into two orbital SU(2) Kondo peaks (red
263: curve), where the degree of freedom being screened is the
264: orbital one. However, in the case of the 2LSU(2) (green curve), as the spin
265: Kondo effect has been suppressed by the field and since
266: there is no orbital Kondo effect in the 2LSU(2) state, one
267: can see the two sets of CB peaks separated by $U^{\prime}$ (green curve).
268: (b) On the other hand, when $\mu_{orb}B=2.0$ and $\mu_{sp}=0.0$, one is left with
269: a spin SU(2) Kondo effect for each orbital level in both states (SU(4) and 2LSU(2)),
270: as is clearly shown by the two Kondo peaks separated by a distance proportional to the
271: applied field.
272: }
273: \end{figure}
274:
275: \section{Comparison of SU(4) with 2LSU(2) and transition between both states}
276:
277: To illustrate the entanglement of the charge and spin degrees
278: of freedom in the SU(4) state, we compare results for $t^{\prime \prime}=0$ and
279: $t^{\prime \prime}=t^{\prime}$, i.e., for the SU(4) and 2LSU(2)
280: states (see Fig. 1), respectively. In Figs. 3a and 3b it is shown a comparison between them when
281: a magnetic field acts only on the spin degree of freedom (Fig. 3a) or
282: only on the orbital degree of freedom (3b). By making $\mu_{orb}=0.0$ and
283: $\mu_{sp}B=2.0$ (3a), the spin Kondo effect is suppressed, since the spin levels are
284: split by an energy larger than the Kondo temperature, while the orbital levels are
285: unaffected. Therefore, the SU(4) Kondo peak splits into two
286: {\it orbital} SU(2) Kondo peaks separated by a distance proportional to the field
287: (red curve) \cite{lipinski}.
288: On the other hand, since in the 2LSU(2) state there is no orbital
289: Kondo effect (since the orbital quantum number is not conserved upon tunneling),
290: by suppressing the spin Kondo with the magnetic field one is left with just
291: two sets of Coulomb Blockade (CB) peaks (split by $U^{\prime}$)
292: separated by a splitting proportional to the field (green curve). In contrast, when only the
293: orbital levels are split by the magnetic field ($\mu_{orb}B=2.0$ and $\mu_{sp}=0.0$),
294: one is left with two spin SU(2) Kondo peaks (one for each
295: orbital level) for both states, as can be seen in Fig. 3b.
296:
297: At zero field, it is then not really surprising that the SU(4) state depicted
298: by the red curve in Fig. 2a will change once the channels are allowed to
299: `talk' to each other (finite $t^{\prime \prime}$, see green arrows in Fig. 1), i. e., if
300: an electron that tunnels into the CNT QD through one channel has a
301: finite probability of tunneling out through the other channel.
302: Figure 4 shows results comparing the SU(4) state ($t^{\prime}=0.2$,
303: $t^{\prime \prime}=0.0$) with the 2LSU(2) state ($t^{\prime}=t^{\prime \prime}=0.1$)
304: \cite{note2}. Notice that the conductances for the two states, although being both equal
305: to $2e^2/h$ at QF \cite{friedel},
306: are {\it qualitatively} very different: the 2LSU(2) state reaches a maximum of
307: $2e^2/h$, half of the maximum in the SU(4) state, and it resembles more the
308: results for a single-channel system \cite{discontinuity}.
309:
310: \begin{figure}[h]
311: \centering
312: \includegraphics[angle=90,height=5.0cm]{fig4.eps}
313: \caption{ Comparing SU(4) and 2LSU(2) conductances: The red
314: curves show the conductance $G$ (solid) and average occupancy $\left< n \right>$ (dashed)
315: for the SU(4) state (same as the red curves in Fig. 2), while the
316: green curves show results for 2LSU(2).
317: It is interesting to note that
318: for both states, when the CNT is occupied by one electron ($V_g=-0.25$
319: for 2LSU(2), and $V_g=-0.385$ for SU(4)), the conductance
320: reaches the unitary limit ($G=2e^2/h$), as discussed in ref. \onlinecite{aguado1},
321: however there is no question that the conductances for both states are qualitatively different.
322: }
323: \end{figure}
324:
325: It is reasonable to assume that in a realistic experimental situation
326: the conservation of the orbital quantum number lies somewhere between the two schemes
327: represented in fig. 1, therefore a careful analysis of the
328: robustness of the SU(4) state (in the presence of some channel
329: mixing) is needed if one wants to correlate any of the
330: experimental observations to the results obtained by numerical modeling. Recent calculations
331: \cite{aguado1, aguado2} have analyzed the transition between these two states
332: (SU(4) and 2LSU(2)) only at QF. In figure 5, we present
333: complementary calculations for all fillings. Our results at QF confirm (as discussed below) that
334: SU(4) and 2LSU(2) at QF are experimentally indistinguishable, reinforcing then the need to extend
335: the analysis to other fillings, especially between QF and HF.
336:
337: Figure 5 shows how the conductance evolves from the SU(4) to the
338: 2LSU(2) state, for $t^{\prime \prime}$ varying from $0$ to $t^{\prime}$. Notice that,
339: as the value of $t^{\prime \prime}$ increases from zero (solid black curve),
340: the conductance peak at $4e^2/h$ (characteristic of the SU(4) state)
341: becomes gradually narrower, until (for $t^{\prime \prime}>0.175$) the
342: central peak splits into three very narrow peaks (not shown). An indication
343: of their presence can be seen already in the curve for
344: $t^{\prime \prime}=0.175$ (cyan), as indicated by the arrows.
345: As $t^{\prime \prime}$ approaches $t^{\prime}$, these three peaks continue to narrow, until they
346: vanish. On the other hand, for values of conductance around $2e^2/h$, the
347: curve develops shoulders which become broader as $t^{\prime \prime}$
348: increases. We want to stress the qualitative agreement of our QF results
349: to those obtained by J.-S. Lim {\it et al.} \cite{aguado2}.
350: In fig. 5, all curves cross at $V_g \approx -0.3$ (where $\langle n_d \rangle \approx 1$, QF)
351: where they have approximately unitary conductance $G_0$. Therefore, as stressed in
352: references \onlinecite{aguado1} and \onlinecite{aguado2}, SU(4) and 2LSU(2) are experimentally
353: indistinguishable {\it at} QF and zero magnetic field.
354: In addition, it is also interesting to note, as described above, that our results for $t^{\prime \prime}
355: \approx t^{\prime}$ change discontinuously to the 2LSU(2) result (dashed curve).
356: A similar discontinuity is seen in the Slave Boson Mean Field results at QF in reference
357: \onlinecite{aguado2} (please, check their fig. 14).
358:
359: Finally, the green solid curve in fig. 4 has a discontinuity in the
360: conductance for $V_g = -0.5$. We have seen this kind
361: of behavior in other multi-orbital systems \cite{discontinuity},
362: and in some cases associated it to the crossing through
363: the Fermi energy of a very narrow level. This causes an
364: abrupt charging of the QD (clearly visible
365: in the green dashed curve for $\langle n_d \rangle$ vs. $V_g$),
366: with the consequent abrupt change of the conductance. In this specific case, this level
367: can be identified to the $\epsilon_-$ energy level in the upper
368: panel of fig. 11 in reference \onlinecite{aguado2}, which,
369: as described there, has a vanishingly narrow width.
370:
371: \begin{figure}[h]
372: \centering
373: \includegraphics[angle=90,height=5.0cm]{fig5.eps}
374: \caption{ Transition from the SU(4) to the 2LSU(2) state:
375: Results showing the variation of the conductance for $t^{\prime \prime}$
376: varying from 0.0 (black curve) to 0.2 (dashed), going through the values
377: 0.05 (blue), 0.1 (green), 0.15 (red) and 0.175 (cyan). $U = U^{\prime} = 0.5$ and
378: $t^{\prime}=0.2$.
379: }
380: \end{figure}
381:
382: \section{Experimental observation of SU(4) Kondo at Half-Filling}
383:
384: Figure 6a reproduces Fig. SI2 in the Supplementary Information of
385: ref. \onlinecite{jarillo}. In it, the temperature variation of the
386: conductance for three shells in the CNT QD is reported (thick red line
387: at higher temperature and black thin line for the lowest temperature).
388: Notice that the coupling to the leads increases from right to left \cite{jarillo,jarillo2}
389: (that is why the conductance of shell 1 is the lowest). In reference
390: \onlinecite{jarillo}, the conductance in regions I and III in the second
391: shell was associated to a QF SU(4) state and the conductance at HF in
392: shell $n=3$ was associated to a singlet-triplet effect \cite{sasaki,yeyati}. However,
393: in Fig. 6b, our results indicate an alternative interpretation: by breaking
394: the degeneracy of the orbital levels (by introducing a small energy
395: splitting $\delta E = 0.032$) \cite{splitting} and by increasing the coupling of the
396: third shell to the leads ($t^{\prime}=0.2$) in relation to the second
397: shell ($t^{\prime}=0.11$), the experimental results can be qualitatively reproduced.
398: Note that since our calculations are done at zero-temperature, the curve in Fig. 6b should
399: be compared to the highest conductance curve in Fig. 6a (thin black line).
400: It is clear that the qualitative agreement is quite good.
401:
402: \begin{figure}[h]
403: \centering
404: \includegraphics[height=3.3cm]{fig6a.eps}
405: \includegraphics[angle=90,height=4.5cm]{fig6b.eps}
406: \caption{ (a) Adapted from Supplementary Information in ref. \onlinecite{jarillo} (Fig. SI2):
407: temperature dependence of the conductance for the three shells observed in
408: the CNT QD. Thick red curve at higher temperature and thin black curve at lowest temperature.
409: Note that the coupling to the leads increases from right to left (from shell
410: $n=1$ to $n=3$) \cite{jarillo,jarillo2}.
411: In ref. \onlinecite{jarillo}, the regions indicated as I and III in the second shell were
412: interpreted as indicative of the presence of a QF SU(4) state and
413: the conductance of the third shell (leftmost) was associated to
414: a singlet-triplet state \cite{sasaki}, however,
415: our simulations in (b) show that the conductance for both shells can be interpreted
416: as indication of a QF SU(4) state for the second shell and
417: of {\it both} QF and HF SU(4) states for the third shell. These results are obtained
418: by adding a small energy separation between the orbital levels
419: ($\delta E=0.032$) and by increasing the coupling of the third shell
420: ($t^{\prime}=0.2$) in relation to the second shell ($t^{\prime}=0.11$).
421: }
422: \end{figure}
423:
424: To further test our simulations against the experimental results,
425: Fig. 7a shows a color-scale plot with numerical results for
426: the variation of the conductance (at zero bias) with applied magnetic field
427: (along the CNT axis) for the third shell (same parameters as the ones used
428: in Fig. 6b).
429: Since the orbital moment is much larger than
430: the spin one ($\mu_{orb}=0.2$ and $\mu_{sp}=0.04$), at lower fields
431: one sees first the splitting of the SU(4) conductance peak into two spin SU(2) Kondo
432: peaks, which at higher field values will each further split into two CB peaks.
433: Figure 7b shows a figure adapted from ref. \onlinecite{jarillo2} containing
434: field-dependent conductance results (third shell in Fig. 6a)
435: which are clearly in qualitative agreement with the numerical results in Fig. 7a.
436: The combined results presented in Figs. 6 and 7 present compelling evidence that
437: the conductance results of the third shell can be interpreted as a manifestation
438: of the HF SU(4) state. It is interesting to note the asymmetry in the
439: conductance in the experimental results in Fig. 7b, i.e., the CB regime
440: is reached at lower values of field for lower values of gate voltage.
441: The same kind of asymmetry was observed by Makarovski {\it et al.} \cite{gleb}
442: in all shells and always with the higher conductance for 3 electrons in the shell.
443: It is not clear yet the reason for this higher Kondo temperature for 3 electrons
444: when compared to QF (1 electron inside the shell).
445:
446: \begin{figure}[h]
447: \centering
448: \includegraphics[angle=90,height=3.5cm]{fig7a.eps}
449: \includegraphics[height=2.5cm]{fig7b.eps}
450: \caption{ Splitting of the Kondo SU(4) peak caused by an external magnetic
451: field applied along the CNT axis. (a) Color-scale plot of the conductance
452: showing the progressive splitting of the zero-field SU(4) peak into
453: two spin SU(2) peaks and then into 4 CB peaks
454: (dark red $\approx 4e^2/h$ and blue $=0$). (b) Experimental results
455: adapted from ref. \onlinecite{jarillo2} (same shell as the leftmost one
456: in Fig. 6a) indicating that the numerical results in part (a) qualitatively
457: reproduce the experiments.
458: }
459: \end{figure}
460:
461: \section{Conclusions}
462:
463: In summary, using a recently developed numerical method (ECA) \cite{method},
464: the authors offer a reinterpretation of recent
465: transport measurements in CNT QDs. In these experiments,
466: the conductance of one of the electronic shells of a CNT QD was
467: associated to the SU(4) Kondo effect at QF, and
468: the conductance at HF of an adjacent shell was interpreted as
469: resulting from a singlet-triplet effect \cite{jarillo}.
470: Our results clearly show that the conductance of {\it both}
471: shells can be interpreted instead as resulting from an SU(4) state\cite{note-st}:
472: this is achieved by introducing a small energy splitting between the orbital levels
473: and increasing the coupling to the leads of one of the shells in relation to the other
474: (see Fig. 6).
475: Furthermore, simulations of conductance at finite magnetic field give support
476: to our interpretation. These results open the possibility that the SU(4) state at
477: HF could be analyzed in detail in CNT QDs. The fact that recent NRG results
478: by Galpin {\it et al.} \cite{galpin1}
479: have associated this state to rich physical behavior, including a non-Fermi-liquid phase, adds
480: to the importance of our results. In addition, simulations presented in
481: Fig. 2 are in qualitative agreement with Galpin {\it et al.}'s NRG results.
482: This suggests that the low-energy physics associated to their NRG results is quite
483: robust and should in principle be observed experimentally.
484:
485: One last point the authors would like to stress is related to the implications
486: of the SU(4) `spin-orbital entanglement' to the structure of the Kondo cloud.
487: A qualitative description of the screening mechanism in the Kondo effect involves the existence
488: of the so-called `Kondo cloud': associated to the Kondo temperature $T_K$
489: (a universal energy scale which emerges naturally from Renormalization Group arguments)
490: there is a universal length scale $\xi_K = \hbar v_F/T_K$, which can be interpreted
491: as the size of the many-body wave function containing the conduction electron
492: that forms a singlet with the impurity spin. Even before the first measurement of the Kondo effect in
493: semiconducting QDs \cite{goldhaber2}, there was great interest in experimentally detecting the Kondo cloud,
494: with theoretical \cite{bergmann} and experimental \cite{giordano}
495: efforts being made to evaluate and measure its size and dependence on dimensionality.
496: The failure to actually observe the extent of the Kondo cloud (or even ascertain its existence)
497: underscores the difficulties involved.
498: This has lead more recently to theoretical efforts to analyze setups where the Fermi
499: sea is effectively `confined' {\it inside} the nanostructure, like for example
500: in the so-called `Kondo Box' \cite{thimm} or in QDs embedded in Aharonov-Bohm (AB) rings \cite{affleck}.
501: In such setups, interesting new effects are expected to occur, hopefully leading to a
502: better understanding of the screening effect and a possible direct
503: or indirect breakthrough measurement of the Kondo cloud. The difficulty of fabricating
504: the proposed devices and performing the necessary measurements may explain the fact that most of
505: the work in this area is theoretical.
506: Referring back to fig. 1, the many-body wave function formed by conduction electrons
507: that screen the localized moment has to have quite different properties
508: in the SU(4) Kondo state when compared to the 2LSU(2) state.
509: Indeed, the recognition that orbital Kondo correlations can only form if the orbital QN
510: is conserved upon tunneling, and that electron states in the metallic leads do
511: not have a defined `wrapping mode', results in the natural conclusion that in the SU(4) Kondo
512: state the `Fermi sea' is in reality formed by
513: electrons that have a well defined orbital QN, and therefore should reside
514: primarily in the regions of the CNT contained between the tunnel
515: barrier and the metallic contacts \cite{jarillo, aguado1,aguado2}. These regions, as
516: for example in the proposed setups involving QDs embedded in AB rings \cite{affleck},
517: naturally constrain the extent of the Kondo cloud. Obviously, for CNT devices where
518: the tunneling barrier is exactly at the interface between the CNT and the metallic
519: contacts, there is no conservation of the orbital QN, leading to
520: a 2LSU(2) state, where there is no entanglement of
521: the spin and orbital degrees of freedom and therefore its associated Kondo cloud
522: is free to spread inside the metallic contacts.
523: One of the problems with the experimental realization of setups suggested as
524: possible probes of the Kondo cloud properties is that, once leads are attached
525: to the nanostructure to perform actual measurements, the Kondo cloud spreads into them,
526: making the measurements of its properties difficult.
527: What we suggest here is that the SU(4) state in carbon nanotubes naturally provides
528: a system where the Kondo cloud should be constrained inside the nanostructure
529: itself, with the advantage that the leads needed for conductance
530: measurements, at least in principle, do not `accept' the Kondo cloud.
531: In that case, careful analysis of the change in transport properties as the system
532: transitions from the SU(4) to the 2LSU(2) state should provide valuable information about the screening
533: process and at least some indirect information about the Kondo cloud.
534:
535: The authors acknowledge useful discussions with E. Anda, E. Dagotto, G. Finkelstein,
536: D. Goldhaber-Gordon, P. Jarillo-Herrero and J. Riera.
537: G. B. M. acknowledges support from Research Corporation;
538: C. A. B. acknowledges support from UT, Knoxville.
539:
540: \begin{thebibliography}{}
541:
542: \bibitem{jarillo} P. Jarillo-Herrero {\it et al.}, Nature {\bf 434}, 484 (2005).
543:
544: \bibitem{wilhelm} U. Wilhelm {\it et al.}, Physica E {\bf 14}, 385 (2002).
545:
546: \bibitem{holleitner} A. W. Holleitner {\it et al.}, Phys. Rev. B {\bf 70}, 075204 (2004).
547:
548: \bibitem{sasaki1} S. Sasaki {\it et al.}, Phys. Rev. Lett. {\bf 93}, 017205 (2004).
549:
550: \bibitem{pohjola} T. Pohjola {\it et al.}, Europh. Lett. {\bf 54}, 241 (2001).
551:
552: \bibitem{borda} L. Borda {\it et al.}, Phys. Rev. Lett. {\bf 90}, 026602 (2003).
553:
554: \bibitem{zarand} G. Zarand, Philos. Mag. {\bf 86}, 2043 (2006).
555:
556: \bibitem{lehur} P.G. Silvestrov and Y. Imry, cond-mat/0609355 (2006) and
557: K. Le Hur {\it et al.}, cond-mat/0609298 (2006).
558:
559: \bibitem{gleb} A. Makarovski {\it et al.}, cond-mat/0608573.
560:
561: \bibitem{note1} The Kondo temperature observed in the orbitally degenerate CNT QD
562: was as high as $T_{\rm K}=16.0 {\rm K}$ \cite{jarillo}.
563:
564: \bibitem{iroi} M. Iroi {\it et al.}, Phys. Rev. B {\bf 55}, 8339 (1997).
565:
566: \bibitem{toonen} R. C. Toonen {\it et al.}, cond-mat/0512235 (2005).
567:
568: \bibitem{goldhaber} R. M. Potok and D. Goldhaber-Gordon, Nature {\bf 434}, 451 (2005).
569:
570: \bibitem{aguado1} M.-S. Choi {\it et al.}, Phys. Rev. Lett. {\bf 95}, 067204 (2005).
571:
572: \bibitem{aguado2} J.-S. Lim {\it et al.}, cond-mat/0608110. In this manuscript,
573: Renormalization Group, Numerical Renormalization Group, Slave Boson Mean Field, and
574: Non Crossing Approximation where employed to analyze the transport properties of CNT QDs.
575:
576: \bibitem{galpin1} M. R. Galpin {\it et al.}, Phys. Rev. Lett. {\bf 94}, 186406 (2005),
577: where the properties of a Quantum Critical Point at HF were studied.
578: The fact that the model for a double-orbital QD coupled to double channel leads displays
579: a Kondo effect with entanglement of spin and charge degrees of freedom not only at QF, but also
580: at HF, has been mostly overlooked in the analysis of experimental results \cite{jarillo,aguado1}.
581:
582: \bibitem{galpin2} M. R. Galpin {\it et al.}, cond-mat/0608169 and cond-mat/0608186.
583:
584: \bibitem{method} V. Ferrari {\it et al.} Phys. Rev. Lett. {\bf 82}, 5088 (1999)
585:
586: \bibitem{sasaki} S. Sasaki {\it et al.}, Nature {\bf 405}, 764 (2000).
587:
588: \bibitem{jarillo2} P. Jarillo-Herrero {\it et al.}, Phys. Rev. Lett. {\bf 94}, 156802 (2005).
589:
590: \bibitem{minot} E. D. Minot {\it et. al}, Nature {\bf 428}, 536 (2004).
591:
592: \bibitem{channels}
593: For a discussion about the
594: origin and properties of these two channels in the devices used in
595: the experiments, please see last paragraph in ref. \onlinecite{jarillo}.
596:
597: \bibitem{Meir-cnd} Y. Meir {\it et al.}, Phys. Rev. Lett. {\bf 66}, 3048 (1991).
598:
599: \bibitem{splitting} M. R. Buitelaar {\it et al.}, Phys. Rev. Lett. {\bf 88}, 156801 (2002).
600:
601: \bibitem{range} As noted in reference \onlinecite{galpin1}, one expects that the
602: SU(4) state on each side of the $U=U^{\prime}$ point will survive for an interval
603: $| U^{\prime}-U | \approx T_{K}^{SU(4)}$.
604:
605: \bibitem{lipinski} A similar effect, taking advantage of the existence of an
606: SU(4) state in double-QDs (where $\mu_{orb}=0.0$), is suggested by S. Lipinski and
607: D. Krychowski (cond-mat/0512726 (2005)) as a way to perform spin filtering of
608: the transmitted electrons.
609:
610: \bibitem{note2} A smaller value is used for the hoppings in the 2LSU(2) state
611: to compensate for the fact that the number of hoppings between the QD and
612: the leads in 2LSU(2) is double that in SU(4) (see Fig. 1).
613:
614: \bibitem{friedel} This can be obtained in a very simple way by invoking the
615: Friedel sum rule (check reference \onlinecite{aguado1}).
616:
617: \bibitem{discontinuity} Single-channel results for multi-orbital systems obtained
618: with the ECA method were reported in C. A. Busser {\it et al.},
619: Phys. Rev. B {\bf 70}, 245303 (2004) for 1 QD,
620: and in G. B. Martins {\it et al.}, Phys. Rev. Lett. {\bf 94}, 026804 (2005) for 2 QDs.
621: The discontinuities seen in $G$ and $\left< n \right>$
622: for the 2LSU(2) state (green curves) have also been obtained in similar systems
623: by applying a variety of different techniques (see, for example, P. G. Silvestrov
624: and Y. Imry, Phys. Rev. B {\bf 65}, 035309 (2001)
625: and M. Sindel {\it et al.}, Phys. Rev. B {\bf 72}, 125316 (2005)).
626:
627: \bibitem{yeyati} Notice that the temperature variation obtained for the
628: SU(4) HF effect (A. L. Yeyati {\it et al.}, Phys. Rev. Lett. {\bf 83}, 600 (1999))
629: describes the data for the third shell better than the singlet-triplet
630: effect (W. Izumida {\it et. al}, Phys. Rev. Lett. {\bf 87}, 216803 (2001)).
631:
632: \bibitem{note-st} This claim can be tested by measuring the variation of $T_K$
633: away from the orbital degeneracy point: it should be asymmetric for
634: ST and symmetric for SU(4), as indicated in reference \onlinecite{sasaki1}.
635:
636: \bibitem{goldhaber2} D. Goldhaber-Gordon {\it et al.}, Nature {\bf 391}, 156 (1998).
637:
638: \bibitem{bergmann} G. Bergmann, Phys. Rev. Lett. {\bf 67}, 2545 (1991).
639:
640: \bibitem{giordano} M. A. Blachly and N. Girodano, Phys. Rev. B {\bf 51}, 12537 (1995).
641:
642: \bibitem{thimm} W. B. Thimm {\it et al.}, Phys. Rev. Lett. {\bf 82}, 2143 (1999).
643:
644: \bibitem{affleck} I. Affleck and P. Simon, Phys. Rev. Lett. {\bf 86}, 2854 (2001).
645:
646: \end{thebibliography}
647:
648: \end{document}
649: