1: \documentstyle[aps,prl,multicol,epsf,epsfig]{revtex}
2: %\documentstyle[amssymb,twocolumn,prl,aps]{revtex}
3:
4:
5: %\documentstyle[amssymb,preprint,aps]{revtex}
6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
7: %TCIDATA{Created=Fri Jan 05 11:44:33 2001}
8: %TCIDATA{LastRevised=Mon Sep 10 20:46:24 2001}
9: %TCIDATA{Language=American English}
10:
11: \begin{document}
12: \title{Interaction between Kondo impurities in a quantum corral}
13: \author{G. Chiappe$^{a}$ and A. A. Aligia$^{b}$}
14: \address{$^{a}$ Departamento de F\'{\i }sica, FCEyN Universidad de Buenos Aires,\\
15: Pabell\'{o}n I, Ciudad Universitaria, (1428) Buenos Aires, Argentina.\\
16: $^{b}$ Centro At\'{o}mico Bariloche and Instituto Balseiro, Comisi\'{o}n
17: Nacional de Energ\'{\i }a At\'{o}mica, 8400 Bariloche, Argentina.}
18: \date{Received \today }
19: \maketitle
20:
21: \begin{abstract}
22: We calculate the spectral densities for two impurities inside an elliptical
23: quantum corral, using exact diagonalization in the relevant Hilbert subspace
24: and embedding into the rest of the system. Fore one impurity, the space and
25: energy dependence of the change in differential conductance $\Delta dI/dV$
26: observed in the quantum mirage experiment is reproduced. In presence of
27: another impurity, $\Delta dI/dV$ is very sensitive to the hybridization
28: between impurity and bulk. The impurities are correlated ferromagnetically
29: between them. A hopping $\gtrsim 0.15$ eV between impurities destroys the
30: Kondo resonance.
31: \end{abstract}
32:
33: \pacs{Pacs Numbers: 72.15.Qm, 68.37.Ef, 73.20.Fz}
34:
35: %\draft
36:
37: \twocolumn[
38: \hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
39: ]
40: \narrowtext
41: %\begin{multicols}{2}
42: In recent years, the manipulation of single atoms on top of a surface using
43: scanning tunneling microscopy (STM) was made possible,\cite{eig} and quantum
44: corrals have been assembled by depositing a closed line of atoms or
45: molecules on noble metal surfaces .\cite{cro,hel,man} The local conduction
46: spectral density of states $\rho _{c}(r,\omega )$, measured by differential
47: conductance $dI/dV$ reveals patterns that remind the wave functions of
48: two-dimensional noninteracting electrons under the corresponding confinement
49: potential. In a recent experiment, a Co atom has been placed at a focus of
50: an elliptic quantum corral, and the corresponding Kondo feature is
51: observed
52: not only at that position, but also at the other focus, where a ``mirage''
53: is formed as a consequence of the quantum interference.\cite{man} Several
54: variants of this experiment, some of them involving several impurities (Co
55: atoms) and eventually mirages
56: inside the corral, are being performed.\cite{man2} The main features
57: of the observed space and voltage dependence of $dI/dV$ have been reproduced
58: by several theories.\cite{aga,fie,por,wei,ali,ali2} Since in Refs. \cite
59: {aga,fie,por} the density of states per spin at the impurity $\rho
60: _{d}(\omega )$ is assumed rather than calculated, these theories cannot
61: account for the interaction between impurities. In Ref. \cite{wei} the Kondo
62: effect is absent, and perturbation theory in the Coulomb repulsion $U$ \cite
63: {ali,ali2} is restricted to small values of $U$.
64:
65: The aim of the present work is to present a theory of the quantum mirage
66: which is able to reproduce the experimental results for the case of one
67: impurity and give reliable predictions when more than one impurity is inside
68: the corral. We show that experiments with two impurities can elucidate the
69: role of the direct hybridization between the impurity and the bulk $V_{b}$.
70: Scattering theories \cite{hel,aga,fie} obtained an excellent agreement with
71: experiment assuming that the resonant level width due to hybridization with
72: bulk states $\delta _{b}$ is as large as that due to the surface $\delta
73: _{s} $. On the other hand, the larger density of $s$ and $p$ states at the
74: surface \cite{eu} and the rapid decay of the hybridization matrix elements
75: with distance suggest that $\delta _{b}$ is negligible, and the experiment
76: for one impurity can also be explained if $V_{b}=0$.\cite{ali,ali2} A calculation
77: of $\delta _{b}$ has not been made and experimentally the situation is still
78: unclear.
79: The role of $V_{b}$ is not only crucial for a correct theory of the mirage experiment,
80: but also for the general understanding of the interaction between metallic
81: surfaces and adsorbates.
82: Since actually $\delta _{b}$ was introduced as a phenomenological
83: parameter which takes into account the electrons lost in the scattering
84: process,\cite{hel,aga,fie} one expects that if $\delta _{b}=\delta _{s}$,
85: the interaction between impurities is roughly a fourth of that for $V_{b}=0$
86: if the same total width $\delta _{b}+\delta _{s}$ is kept.
87:
88: We obtain the ground state of the Anderson model in a cluster which contains
89: one or two impurities and the relevant conduction states inside a hard wall
90: ellipse using the Lanczos method. These states are then mixed with bulk
91: states using an embedding method.\cite{fer} This embedding is essential to
92: describe the low energy physics.\cite{hal} The average separation between
93: the relevant conduction states $d\sim 100$ meV \cite{por,ali} is much larger
94: than the Kondo temperature $T_{K}\sim 5$ meV.\cite{man} Under these
95: circumstances a Kondo peak at the Fermi level $\epsilon _{F}$ is absent in
96: the finite system \cite{ali,thi} and the experimental line shape for $dI/dV$
97: cannot be reproduced (unless an artificial Lorentzian broadening is
98: introduced). This is confirmed by our calculations.
99:
100: The Hamiltonian can be written as:
101: \begin{eqnarray}
102: H &=&\sum_{j\sigma }\varepsilon _{j}c_{j\sigma }^{\dagger }c_{j\sigma
103: }+E_{d}\sum_{i\sigma }d_{i\sigma }^{\dagger }d_{i\sigma
104: }+U\sum_{i}d_{i\uparrow }^{\dagger }d_{i\uparrow }d_{i\downarrow }^{\dagger
105: }d_{i\downarrow } \nonumber \\
106: &&+\sum_{ij\sigma }V[\varphi _{j}(R_{i})d_{i\sigma }^{\dagger }c_{j\sigma }+%
107: \text{H.c.}]+H^{\prime }. \label{ham}
108: \end{eqnarray}
109: Here $c_{j\sigma }^{\dagger }$ creates an electron on the $j^{th}$
110: conduction eigenstate of a hard wall elliptic corral with wave function $%
111: \varphi _{j}(r)$ \cite{note} and $d_{i\sigma }^{\dagger }$ is the
112: corresponding operator for the impurity at site $R_{i}$. The hybridization
113: of these states with bulk states of the same symmetry is described by $%
114: H^{\prime }$. We assume that each of the impurity and conduction states
115: mixes with a different continuum of bulk states:
116:
117: \begin{equation}
118: H^{\prime }\cong t\sum_{j\sigma }(c_{j\sigma }^{\dagger }b_{j\sigma }+\text{%
119: H.c.})+V_{b}\sum_{i\sigma }(d_{i\sigma }^{\dagger }b_{i\sigma }+\text{H.c.}).
120: \label{h2}
121: \end{equation}
122: The $b_{l\sigma }$ represent bulk states for which the unperturbed density
123: is 0.05 states/eV, similar to the density of bulk $s$ and $p$ states.\cite
124: {eu} Approximation (\ref{h2}) is justified by comparison of the
125: non-interacting Green functions for hard wall corrals and more realistic
126: boundary potentials.\cite{ali2,cor}
127:
128: The dressed matrix ${\bf G}$ describing the one-particle Green function is
129: calculated by solving the Dyson equation ${\bf G=g+g}H^{\prime }{\bf G}$,
130: where ${\bf g}$ is the corresponding matrix for $H^{\prime }=0$.\cite{fer}
131: This equation is exact for $U=0$ or $H^{\prime }=0$, and in the general case
132: represents an infinite sum of particular diagrams in perturbation theory in $%
133: H^{\prime }$ (the chain approximation\cite{cha}). The $\varphi _{j}(r)$ are
134: obtained as described elsewhere.\cite{ali} We choose the ellipse with
135: eccentricity $e=1/2$ and size such that the state $j=42$ lies at $\epsilon
136: _{F}$.\cite{man} The change in $dI/dV$ ($\Delta dI/dV$) after an impurity is
137: placed inside the corral is determined by the conduction states which lie
138: near $\epsilon _{F}$ and have a strong amplitude $\left| \varphi
139: _{j}(R_{i})\right| $ at the impurity position. For $R_{i}$ at one focus they
140: are $j=32$, 35, 42 and 51.\cite{ali} We have also included $j=24$ and 62,
141: although this inclusion leads to negligible changes in the results. We took
142: the impurity parameters $E_{d}=-1$ eV and $U=3$ eV.\cite{ujs} We consider
143: first the case $V_{b}=0$ and one impurity at the left focus ($R_{i}=(-0.5a,0)$).
144: The value $V=0.04$ eV was chosen to lead to the observed width of $\Delta
145: dI/dV$. The remaining parameter $t$ controls the amplitude of the mirage at
146: the right focus.
147:
148: In Fig. 1(a) we represent the resulting impurity spectral density $\rho
149: _{d}(\omega )$ for two values of $t$. A clear Kondo peak is obtained and for
150: $t\gtrsim 0.3$ eV its width is very weakly dependent on $t$. Instead, for $%
151: t\rightarrow 0$, the peak splits into two very narrow peaks out of $\epsilon
152: _{F}$. In contrast to $\rho _{d}(\omega )$, the magnitude of the change in
153: the conduction density $\Delta \rho _{c}(r,\omega )$ at the empty focus ($%
154: r=-R_{i}$) is quite sensitive to $t\gtrsim 0.3$ eV: as $t$ increases, the
155: width of the conduction states increases, the weight of the states 32, 35
156: and 51 (odd under the reflection through the minor axis of the ellipse $%
157: \sigma $) at $\epsilon _{F}$ increases, and the depression of $\rho
158: _{c}(-R_{i},\omega )$ decreases as a consequence of the negative
159: interference of these states with the even state 42.\cite{ali} The
160: differential conductance $dI/dV$ at zero temperature is proportional to the
161: density $\rho _{f}$ of the state\cite{schi}
162:
163: \begin{equation}
164: f_{\sigma }(r)=\sum_{j}\varphi _{j}(r)c_{j\sigma }+qd_{j\sigma }. \label{q}
165: \end{equation}
166: $q$ is related to Fano`s interference parameter and represents the effect of
167: a direct tunneling from the tip to the impurity. Therefore, it is relevant
168: only very near the impurity. For $q=0$, $\rho _{f}(r,\omega )=\rho
169: _{c}(r,\omega )$. In Fig. 1 (b) we represent the effect of adding the
170: impurity on $\rho _{f}(\pm R_{i},\omega )$ ($\Delta \rho _{f}\sim \Delta
171: dI/dV$). At the impurity site $R_{i}$, $\Delta \rho _{c}(r,\omega )$ is
172: asymmetric and smaller at the right of the valley. This is a consequence of
173: the asymmetry of the hybridization around $\epsilon _{F}$ ($\left| \varphi
174: _{51}(R_{i})\right| >\left| \varphi _{35}(R_{i})\right| $). A symmetric line
175: shape, as observed in the experimental $\Delta dI/dV$ is restored for $q\sim
176: 1$. The effect of this $q$ is consistent with the fact that on a clean
177: surface, $\Delta dI/dV$ is {\em larger} at the right of the peak. Another
178: nice fact is that the minimum of $\Delta \rho _{f}$ for $q=1$ lies at the
179: experimental position 1 meV. At the right focus ($r=-R_{i}$) we obtain a
180: similar valley, although slightly asymmetric and shifted to the left.
181: Increasing $t$ from 0.4 to 0.5, the magnitude of this valley is strongly
182: reduced (its minimum is shifted above -5/eV) but its shape and width is
183: retained. At the impurity position there are no significant changes.
184:
185: The space dependence of $\Delta \rho _{f}$ for $q=0$ is represented in Fig.
186: 2. As in the experimental $\Delta dI/dV$, the main features of $\left|
187: \varphi _{42}(r)\right| ^{2}$, attenuated at the right focus, are displayed.
188: Thus, the theory reproduces the space and energy dependence of $\Delta dI/dV$
189: observed in the experiment.\cite{man} All results so far agree
190: semiquantitatively
191: with perturbative calculations.\cite{ali,ali2} To see how the results
192: change if $\delta _{b}\cong \delta _{s}$ is assumed, we have reduced $V$ by
193: a factor $\sqrt{2}$. This should reduce $\delta _{s}$ by a factor 2.
194: Increasing $V_{b}$ from zero to 1.2 eV, the original width of $\rho _{d}$ is
195: restored. The intensity is reduced by a factor $\sim 2$ (due to the strong
196: energy dependence of $\delta _{s}$). $\Delta \rho _{c}$ turns out to be $%
197: \sim 4$ times smaller. The additional factor 2 can be understood from the
198: fact that the change in conduction electron Green function is proportional
199: to $V^{2}G_{d}(\omega )$, where $G_{d}(\omega )$ is the impurity Green
200: function.\cite{ali} Except for these factors, the results are surprisingly
201: similar to the previous ones. Some of them will be displayed in Fig. 4.
202:
203: We now turn to the case of two impurities, one at each focus, for $V_{b}=0$.
204: The spectral density for one of these impurities is represented in Fig. 3
205: (a). Comparison with the previous case (Fig. 1), shows that the peak around $%
206: \epsilon _{F}$ broadens (by a factor $\sim 1.5$), looses intensity and
207: shifts to lower energies. In addition, another very narrow peak appears $%
208: \sim 13$ meV below $\epsilon _{F}$. An analysis of the energy dependence of
209: the density of the individual conduction states shows that the broad peak
210: around $\epsilon _{F}$ is due to hybridization with even states (mainly 42),
211: while the narrow peak reflects the hybridization of the impurity states with
212: odd states (mainly 51). The difference in $dI/dV$ with respect to the empty
213: corral is however, not so different as in the previous case. This is due to
214: the effect of the unperturbed Green functions of the conduction states and
215: is also present in the one impurity case.\cite{ali} Nevertheless, a decrease
216: in the amplitude and a broadening of the depression should be observed in $%
217: \Delta dI/dV$ and seems in qualitative agreement with recent experiments.%
218: \cite{man2} The space dependence is similar to that for one impurity (Fig.
219: 2) but it is of course, symmetric under reflection through the minor axis $%
220: \sigma $, and not attenuated at the right focus.
221:
222: Qualitatively, the shape of $\rho _{d}$ can be understood looking at the
223: non-interacting case $U=0$, $E_{d}\sim $ $\epsilon _{F}$. In this case, for
224: one impurity, the Kondo peak is replaced by a Lorentzian near $\epsilon _{F}$%
225: . For two impurities, a change of basis of the $d$ orbitals to $e_{\sigma
226: }=(d_{1\sigma }+d_{2\sigma })/\sqrt{2}$, $o_{\sigma }=(d_{1\sigma
227: }-d_{2\sigma })/\sqrt{2}$, separates the problem into those corresponding to
228: even and odd states under $\sigma $. The even state hybridizes mainly with
229: conduction state 42, to form a resonance near $\epsilon _{F}$, roughly twice
230: wider than for one impurity due to the larger effective hybridization.
231: Instead, the odd state $o_{\sigma }$ is displaced towards lower energies due
232: to hybridization with state 51. The interactions should modify the
233: quantitative details of this picture. However, we expect that it remains
234: qualitatively valid, as suggested by the above mentioned energy distribution
235: of the different conduction states.
236:
237: To gain insight into the nature of the ground state, we have also calculated
238: spin-spin correlation functions for $t=0$. A reliable method to include $%
239: H^{\prime }$ in these calculations has not been developed yet. For one
240: impurity we obtain $\langle {\bf S}_{i}\cdot {\bf s}_{42}\rangle =-0.73$,
241: where ${\bf S}_{i}$ is the spin of the impurity $i$ and ${\bf s}_{j}$ is the
242: spin of the conduction state $j$. This value is close to the minimum
243: possible one -3/4. For $j\neq 42$, $\langle {\bf S}_{i}\cdot {\bf s}%
244: _{j}\rangle $ are very small, but this, and the large magnitude of $\langle
245: {\bf S}_{i}\cdot {\bf s}_{42}\rangle $, are affected to a certain degree by
246: the neglect of $H^{\prime }$ in this calculation. The space dependence of $%
247: \langle {\bf S}_{i}\cdot {\bf s}(r)\rangle $, where ${\bf s}(r)$ the
248: conduction spin at position $r$ follows closely $\left| \varphi
249: _{42}(r)\right| ^{2}$. For two impurities we find $\langle {\bf S}_{i}\cdot
250: {\bf s}_{42}\rangle =-0.47$ and$\ \langle {\bf S}_{1}\cdot {\bf S}%
251: _{2}\rangle =0.21$. In the limit of large $U$, one expects that the main
252: features of the spin dynamics for $V_{b}=0$ are described by the Hamiltonian
253: $H_{0}=J({\bf S}_{1}+{\bf S}_{2})\cdot {\bf s}_{42}$, where $J>0$ is the
254: Kondo coupling. The ground state of this Hamiltonian is a doublet in which
255: the impurity spins are correlated ferromagnetically between them ($\langle
256: {\bf S}_{1}\cdot {\bf S}_{2}\rangle =1/4$) and antiferromagnetically with
257: state 42 ($\langle {\bf S}_{i}\cdot {\bf s}_{42}\rangle =-1/2$). These
258: values are near to those we find. The effect of the hybridization of
259: state 42 with bulk states can be modelled by a tight binding Hamiltonian in
260: terms of Wilson's orbitals. $H_{0}$ is the strong coupling fixed point of
261: Wilson's renormalization group. An analysis of the stability of this fixed
262: point using perturbation theory as in Ref. \cite{allub} leads to the
263: conclusion that the ground state is a doublet for $V_{b}=0$. However, we
264: expect that as soon as $V_{b}\neq 0$, the doublet is screened at a very low
265: temperature.
266:
267: For the set of parameters corresponding to $\delta _{b}\cong \delta _{s}$, $%
268: \rho _{d}(\omega )$ is much more similar to the one impurity case, although
269: a structure reminiscent of a splitting is also present near its maximum. In
270: contrast to the case of $V_{b}=0$, when a second impurity is added, the
271: depression in $\Delta dI/dV$ at one impurity site $R_{i}$ increases and its
272: width is roughly the same (see Fig. 4). Comparison with results when $t$ is
273: increased from 0.4 eV to $t=0.5$ eV (not shown) suggests that when $\delta
274: _{b}\cong \delta _{s}$, $\Delta dI/dV$ at $\pm $ $R_{i}$ for two impurities
275: is roughly the sum of the results at $R_{i}$ and $-R_{i}$ for one impurity.
276: This is what one would expect if the interaction is very small.
277:
278: Coming back to the case $V_{b}=0$, we have also verified that qualitatively
279: similar features in $\Delta dI/dV$ are obtained at one focus, if one
280: impurity is placed there and the second impurity is put at another extremum
281: of $\varphi _{42}(r)$, like (0.22$a$,0) (instead of placing it at the other
282: focus). In this case, the spectral densities at (0.22$a$,0) have some
283: additional structure due to an important admixture of the state 41.\cite
284: {note2} In contrast, if both impurities are placed close to the same focus
285: and near each other, a moderate hopping $t\sim 0.15$ eV or larger between
286: them is sufficient to destroy the Kondo resonance. In particular $\Delta
287: dI/dV$ becomes flat and featureless near $\epsilon _{F}$.
288:
289: In summary, we have studied the spectral density for impurities inside a
290: quantum corral, using a many-body approach which treats exactly the
291: correlations in the impurities and their hybridization with the relevant
292: conduction states at the surface, and treats approximately the hybridization
293: with bulk states. We have been able to reproduce the main features of the
294: mirage experiment for one impurity inside the corral. The experiment for one
295: impurity cannot determine the relative importance of the direct
296: hybridization of the impurity with bulk states, unless the tunneling matrix
297: elements and other details are known accurately. Instead, for two impurities
298: inside the corral, the differential conductance is very sensitive to this
299: hybridization. For the parameters of the experiment, the spins of both
300: impurities are antiferromagnetically coupled with the conduction electrons,
301: and ferromagnetically correlated between them provided they are placed
302: sufficiently far apart, so that the hopping between them can be neglected.
303: If this hopping is larger than 0.15 eV, there is a tendency to form a
304: singlet state between both impurity spins and the Kondo resonance disappears.
305: To our knowledge, this is the first theory which is able to describe the
306: line shape of the differential conductance when more than one Kondo impurity
307: is inside the quantum corral.
308:
309: This work benefitted from PICT 03-00121-02153 of ANPCyT and PIP 4952/96 of
310: CONICET. We are partially supported by CONICET.
311:
312: \begin{references}
313: \bibitem{eig} D.M. Eigler and E.K. Schweizer, Nature{\ {\bf 344}, 524
314: (1990).}
315:
316: \bibitem{cro} M.F. Crommie, C.P. Lutz, and D.M. Eigler, Science {\bf 262}{,
317: 218 (1993).}
318:
319: \bibitem{hel} E.J. Heller {\it et al.}, Nature{\ {\bf 363}, 464 (1994).}
320:
321: \bibitem{man} H.C. Manoharan, C.P. Lutz, and D.M. Eigler,
322: Nature{\ {\bf 403}, 512 (2000).}
323:
324: \bibitem{man2} H.C. Manoharan, PASI Conference, {\it Physics and Technology
325: at the Nanometer Scale} (Costa Rica, June 24 - July 3, 2001).
326:
327: \bibitem{aga} O. Agam and A. Schiller, Phys. Rev. Lett. {\bf 86}, 484
328: (2001).
329:
330: \bibitem{fie} G.A. Fiete {\it et al.}, Phys. Rev. Lett. {\bf 86}, 2392
331: (2001).
332:
333: \bibitem{por} D. Porras, J. Fern\'{a}ndez-Rossier, and C. Tejedor, Phys.
334: Rev. B {\bf 63}, 155406 (2001).
335:
336: \bibitem{wei} M. Weissmann and H. Bonadeo, Physica E {\bf 10}, 44 (2001).
337:
338: \bibitem{ali} A.A. Aligia, Phys. Rev. B {\bf 64}, 121102(R) (2001).
339:
340: \bibitem{ali2} A.A. Aligia, cond-mat/0110081.
341:
342: \bibitem{eu} A. Euceda, D.M. Bylander, and L. Kleinman,
343: Phys. Rev. B {\bf 28}, 528 (1983).
344:
345: \bibitem{fer} V. Ferrari {\it et al.}, Phys. Rev. Lett. {\bf 82}, 5088
346: (1999); C.A. B\"{u}sser {\it et al.}, Phys. Rev. B {\bf 62,} 9907 (2000).
347:
348: \bibitem{hal} A numerical diagonalization without embedding has been
349: performed to study mirages on a spherical surface [K. Hallberg, A. Correa,
350: and C.A. Balseiro, cond-mat/0106082].
351:
352: \bibitem{thi} W.B. Thimm, J. Kroha, and J. von Delft, Phys. Rev. Lett. {\bf %
353: 82}, 2143 (1999).
354:
355: \bibitem{note} Here the wave functions are adimensional and normalized as
356: $\int dxdy\bar{\varphi}_{i}(r)\varphi _{j}(r)/(ab)=\delta _{ij}$, where
357: $a$ ($b$) is the semimajor (semiminor) axis of the ellipse.
358:
359: \bibitem{cor} A. Correa, K. Hallberg, and C.A. Balseiro, unpublished.
360:
361: \bibitem{cha} E.V. Anda, J. Phys. C {\bf 14}, L1037 (1981); W. Metzner,
362: Phys. Rev. B {\bf 43}, 8549 (1991)..
363:
364: \bibitem{ujs} O.\'{U}js\'{a}ghy {\it et al.}, Phys. Rev. Lett. {\bf 85},
365: 2557 (2000).
366:
367: \bibitem{schi} A. Schiller and S. Hershfield,
368: Phys. Rev. B {\bf 61}, 9036 (2000).
369:
370: \bibitem{allub} R. Allub and A.A. Aligia, Phys. Rev. B {\bf 52}, 7987
371: (1995).
372:
373: \bibitem{note2} This calculation was made replacing the states 24 and 62,
374: which practically do not change the results, with states 41 and 49, which
375: have an important hybridization with the impurity at (0.22$a$,0).
376: \end{references}
377:
378: \figure{\noindent Fig. 1: (a) Impurity spectral density as a function of
379: energy for two values of }$t${. (b) Change in the density of the mixed state
380: }$f_{\sigma }$ (Eq.(\ref{q})) at the impurity site (left focus) for two
381: values of $q$ and at the other focus for $t=0.4$ eV{.}
382:
383: \figure{\noindent Fig 2: Contour plot of $\Delta \rho _{c}(r,\omega )$ for
384: $t=0.4$ eV and {\ $\omega =10$ meV.}
385:
386: \figure{\noindent Fig 3: (a) Impurity spectral density for one impurity at
387: each focus and two values of }$t${. (b) Change in the density of the mixed
388: state }$f_{\sigma }$ after addition of both impurities (Eq.(\ref{q})), at
389: one impurity site for two values of $q$. Parameters are $V=0.04$eV, $V_{b}=0$
390: and $t=0.4$ eV{.}
391:
392: \figure{\noindent Fig 4: $\Delta \rho _{c}(r,\omega )$ as a function of
393: $\omega $ for the case of one impurity at the left focus (full and dashed
394: lines) or one impurity at each focus (dashed dot dot line). Parameters are
395: $V=0.04$eV/$\sqrt{2}$, $V_{b}=1.2$ eV and $t=0.4$ eV.}
396:
397: %\end{multicols}}
398: \end{document}
399: