1: \documentclass[prb,twocolumn,showpacs,floatfix,amsfonts]{revtex4}
2:
3: \usepackage{graphicx,graphics,color,epsfig}% Include figure files
4:
5: \usepackage{bm}
6:
7: \usepackage{amsmath}
8:
9: \usepackage{amssymb}
10:
11:
12:
13: \begin{document}
14:
15: \preprint{}
16:
17:
18:
19: \title{ Quantum interference in dirty $d$-wave
20: superconductors }
21: \author{Hong-Yi Chen$^*$, Lingyin Zhu$^\dagger$, and C.S. Ting$^*$}
22: \affiliation{$^*$Texas Center for Superconductivity and Department of
23: Physics, University of Houston, Houston, TX 77204 \\
24: $^\dagger$Department of Physics, State University of New York at Buffalo,
25: Buffalo, NY, 14260}
26:
27:
28:
29:
30:
31: \begin{abstract}
32: The local differential tunneling conductance on a Zn impurity in a
33: disordered $d$-wave superconductors is studied. Quantum
34: interference between many impurities leads to definitive
35: quasiparticle spectra. We suggest that an elaborate analysis on
36: impurity-induced spectra with quantum interference effect included
37: may be able to pin down the sign and strength of the scattering
38: potential of a Zn impurity in low density limit. Numerical
39: simulations calculated with appropriately determined impurity
40: parameters are in satisfactory agreement with the observations from
41: scanning tunneling microscopy (STM) experiments even in subtle
42: details.
43: \end{abstract}
44:
45:
46:
47: \pacs{74.25.Jb, 74.20.-z, 74.50.+r}
48:
49:
50:
51: \maketitle
52:
53: Impurity effect has served as a unique probe to the mechanism of
54: unconventional superconductivity. Recently, remarkable
55: improvements in high resolution STM experiments provided
56: unprecedentedly delicate images of electronic structure around
57: doped Zn in Bi$_2$Sr$_2$CaCu$_{2}$O$_{8+\delta}$ (BSCCO)
58: \cite{pan2000}. The defect states in differential conductance map
59: at low temperatures bear on almost identical spectroscopic and
60: spatial structures from one impurity to the other, with
61: characteristic on-site zero bias resonance peak, four-fold
62: symmetry pattern in local density of states (LDOS) and locally
63: suppressed superconductivity at immediate surroundings of Zn
64: atoms. These observations are in qualitative agreement with
65: preceding theoretical works which model Zn atoms as isolated
66: pointlike unitary centers embedded in a superconducting background
67: of $d$-wave like pairing symmetry \cite{balatsky1995,
68: balatsky0411}. However, although the single-impurity scenario (SI)
69: is approximately successful in interpreting the energy and
70: spatial symmetry of resonant states, it fails to reproduce the
71: subtle features in the spatial profile of LDOS extracted by STM
72: measurements. Attempts to reconcile this discrepancy include
73: postulations on the tunneling matrix of STM probes
74: \cite{zhujx2001, martin2002}, possible Kondo physics from
75: staggered moment \cite{sachdev} and etc.; nevertheless, the
76: determination of the attributes of Zn impurities, with which we
77: can substantiate reliable calculations to compare with
78: experiments, remains divergent itself: even assuming Zn impurities
79: are purely potential scatterers, one could still reach completely
80: opposite conclusions that they are either repulsive (as desired to
81: yield unitary limit with realistic cuprates band) or attractive
82: ({\it relative to the background Cu ions}) according to their
83: ionic configuration. Most recently, an {\it ab-initio} calculation
84: based on density functional theory concludes that Zn should be
85: repulsive centers to electrons \cite{wangll0505}.
86:
87:
88: In the meanwhile, the effect of {\it interference} between the
89: quasiparicle resonances cannot be neglected. Overlapping between
90: disorder-induced Fridel oscillations alters local landscape in
91: LDOS and populates the low energy excitations, which greatly
92: modify the spectroscopic and transport properties of cuprates.
93: There have been numerous theoretical studies on many-impurity
94: problem \cite{onishi1996, pepin2001, morr2002, andersen2003,
95: atkinson2003, zhuly2003}, with emphasis either on the effect of
96: impurity network on Fourier transformed power spectrum which can
97: be used to extract the kinematics of pure systems or on the
98: asymptotic behavior of Fermi level density of quasiparticle
99: states. Despite the creditable achievements of these endeavor in
100: interpreting STM data, the lingering unjustification of impurity
101: parameters could place their arguments on a shaky ground;
102: moreover, none of them have given full attention to the potential
103: of combining interference effect to discuss the impurity spectra
104: and identify the impurity characteristics itself. It is then the
105: purpose of this paper to argue that through elaborate
106: investigation on quantum interference effect in a fully disordered
107: system, we could nail the identities of Zn impurity such as its
108: sign and strength and hopefully close the debate on this issue to
109: our best.
110:
111:
112: We start by writing down the generic many-impurity Hamiltonian as,
113: \begin{eqnarray}
114: H_{Zn} = \sum_{{\bf i}\sigma} \phi(r_{\bf i}) c_{{\bf i}\sigma}^{\dagger}
115: c_{{\bf i}\sigma}^{} \;,
116: \end{eqnarray}
117: where $c_{{\bf i}\sigma}^{\dagger}$($c_{{\bf i}\sigma}$) creates
118: (annihilates) an electron with spin $\sigma$ at site $r_{\bf i}$
119: and $\phi(r_{\bf i})$ is the impurity strength on the perturbed
120: site. Randomness of impurity distribution makes counting relative
121: coordinates information between all the impurities a formidable
122: task. Therefore a systematic and reliable manipulation of the
123: interference, yet not losing the reality and generality, will be
124: helpful. We tacitly suggest to fix the first impurity at the
125: center ($r=0$) of the lattice and keep injecting others while
126: requiring that the inter-impurity distance of any pair, i.e.,
127: $R_{\bf ij}=|r_{\bf i}-r_{\bf j}|$ has to be greater than a preset
128: cutoff $d_0$. This constraint excludes the possibility of cluster
129: formation and further guarantees a uniform distribution of
130: disordered sites. The central impurity, which is the focus of our
131: discussion later, is believed to be the representative of a
132: arbitrary one surrounded by other disordered sites that are
133: randomly distributed in real samples. Therefore, a single
134: parameter, i.e., the concentration, will be the defining parameter
135: instead of the relative orientation and distance between
136: impurities. The effect of the quantum interference is then
137: parameterized as a function of the concentration $x$ (or the
138: number of impurities $N_{imp}$) in a fixed lattice size.
139:
140:
141: We study this problem using the conventional Bogolibuv-de Gennes' (BdG)
142: formalism:
143: \begin{eqnarray}
144: \sum_{\bf j}^N \left(
145: \begin{array}{cc}
146: {\cal H}_{\bf ij} & \Delta_{\bf ij}^{} \\
147: \Delta_{\bf ij}^* & -{\cal H}_{\bf ij}^*
148: \end{array} \right) \left(
149: \begin{array}{c}
150: u_{\bf j}^n \\
151: v_{\bf j}^n
152: \end{array} \right) = E_n \left(
153: \begin{array}{c}
154: u_{\bf i}^n \\
155: v_{\bf i}^n
156: \end{array} \right)\;,
157: \end{eqnarray}
158: where
159: ${\cal H}_{\bf ij} =
160: - t_{\bf ij}+\left(\phi(r_{\bf i})-\mu\right)\delta_{\bf ij}$
161: is the single particle Hamiltonian, $t_{\bf ij}$ is the hopping integral,
162: $\mu$ is
163: the chemical potential,
164: $\Delta_{\bf ij}^{} =
165: \frac{V}{2}\langle c_{{\bf i}\uparrow}^{} c_{{\bf j}\downarrow}^{}
166: - c_{{\bf i}\downarrow}^{} c_{{\bf j}\uparrow}^{} \rangle$
167: is the spin-singlet bond order parameter, $V$ is the
168: nearest-neighbor attractive potential), $u_{\bf i}^n$ and $v_{\bf
169: i}^n$ are the eigenfunctions of the BdG equations, and $E_n$ is the
170: eigen-energy. The self-consistent conditions are applied to solve
171: the BdG equations:
172: \begin{eqnarray}
173: & {} \langle n_{{\bf i}\uparrow} \rangle =
174: \displaystyle{\sum_{n=1}^{N}} \left| u_{\bf i}^n \right|^2 f(E_n)\;,
175: {}& \\
176: & {} \langle n_{{\bf i}\downarrow} \rangle =
177: \displaystyle{\sum_{n=1}^{N}} \left| v_{\bf i}^n \right|^2
178: [1-f(E_n)]\;, {}& \\
179: & {} \Delta_{\bf ij}^{} =
180: \displaystyle{\sum_{n=1}^{N}} \frac{V}{4} \left( u_{\bf i}^n v_{\bf
181: j}^{n*} + v_{\bf i}^{n*} u_{\bf j}^n \right) \tanh \left( \frac{\beta
182: E_n}{2} \right) \;, {}&
183: \end{eqnarray}
184: where $f(E)=1\slash(e^{\beta E}+1)$ is the Fermi distribution. We use
185: the following parameters throughout this paper: $\langle t_{\bf ij}
186: \rangle =t =150$ meV, $\langle t_{\bf ij} \rangle =t'=-0.3t$, $V=1t$.
187: This prepares us a hole-like Fermi surface with the hole doping
188: $1 - n_f =
189: 1 - \sum_{{\bf i} \sigma} \langle c_{{\bf i}\sigma}^\dagger c_{{\bf
190: i}\sigma}^{} \rangle / N_xN_y = 0.15$,
191: i.e., optimally doped region and a maximum gap magnitude about 45
192: meV. The local density of states takes the following expression:
193: \begin{eqnarray}
194: \rho_{\bf i}(E) &=& -\frac{2}{M_x M_y} \sum_{n{\bf k}}^{N} \biggl[
195: \left| u_{\bf i}^{n{\bf
196: k}} \right|^2 f'(E_{n,{\bf k}}-E) \nonumber \\
197: && +\left| v_{\bf i}^{n{\bf k}} \right|^2 f'(E_{n,{\bf k}} + E)
198: \biggr],
199: \end{eqnarray}
200: where the factor of 2 arises from spin degeneracy. The summation in
201: $\rho_{\bf i}(E)$ is averaged over $M_x \times M_y$ (in our paper 20
202: $\times$ 20) wavevectors in the first Brillouin zone.
203:
204:
205: {\it Single impurity} We start our discussion with a brief review on
206: the single impurity model and plot the resulting spectra in Fig.1
207: with the ``blocking effect''(which is addressed below) included. We
208: unbiasedly picked up a repulsive potential $\phi=3.0$eV. Because of
209: the Bogoliubov symmetry of quasiparticles, the on-site LDTC is
210: expected to display two resonance peaks at $\Omega^{\pm}_0=\pm
211: 0.015t$ \cite{balatsky1995}[see Fig.1(a),(b)]. While the spatial
212: scale and symmetry of impurity states observed in STM fit well with
213: the prediction of SI theory, the relative spectral weight
214: distribution is completely reverted \cite{balatsky0411, zhuly2003}:
215: experimental images at $\Omega$ = -1.5 meV \cite{pan2000} displays
216: an on-site maximum spectral intensity, a local minimum on its
217: nearest neighbor sites and a second maximum on the next nearest
218: sites, as illustrated schematically in Fig.1(d).
219: This inconsistency gave birth to the conjectures
220: regarding the indirect tunneling through the insulating BiO layer:
221: the ``filter effect'' due to the coherent tunneling
222: \cite{martin2002} and an alternative ``blocking effect'' due to the
223: incoherent tunneling \cite{zhujx2000}. In our paper, we will stick
224: to the argument of the latter, in which the ``local differential
225: tunneling conductance (LDTC)" fetched by STM probes is believed to
226: measure acutally the averaged LDOS over the four nearest neighbor
227: sites, i.e.,
228: \begin{eqnarray}
229: G_{\bf i}(E) = \frac{1}{4} \sum_{\bf\hat e} \rho_{\bf i+\bf\hat e}(E)\;,
230: \end{eqnarray}
231: where $\bf\hat e$ is $\pm \hat{x}$ or $\pm \hat{y}$. Fig
232: 1.(a),(b),(c) are then plotted with the ``blocking effect" included
233: intentionally. In Fig.1(c), the spatial distribution of spectral
234: intensity mimics experimental results under this manipulation.
235: However, corresponding on-site spectra [Fig 1(a)]deviates from the
236: result of Pan {\it et al.} remarkably: the former shows a sharp peak with
237: excess spectral background at negative bias and a second resonance
238: at positive side; the latter displays a zero bias conductance peak
239: and a second small peak on negative bias; furthermore, Pan {\it et al.}
240: showed that the Friedel oscillation along the nodal direction decays
241: surprisingly faster than that of the antinodal direction. This is
242: rather counterintuitive since the subgap resonances in $d$-wave
243: superconductors overlap with the vanishing continuum and hence are
244: virtually bounded, with supposedly extended tails along the nodal
245: direction due to the vanishing order parameter. We argue that with
246: the inclusion of quantum interference effect, all this discrepancy
247: will be reconciled provided that the impurity parameter is properly
248: settled.
249:
250:
251: \begin{figure}[!t]
252: \centerline{\epsfxsize=8.5cm\epsfbox{fig1.eps}} \caption[*]{(a) LDTC
253: on the impurity-site. Dashed green: pure system; solid black: LDTC
254: with the present of single impurity; (b) LDTC on the nearest
255: neighbor site. Dashed green: pure system; solid black: LDTC with the
256: present of single impurity; (c) calculation on spatial distribution
257: of LDTC at $\Omega_0=-0.015t$; impurity site is placed at the center
258: of the squared window. (d) Schematic plot of the STM image of one
259: impurity \cite{pan2000}.}
260: \end{figure}
261:
262:
263:
264:
265: \begin{figure}[!t]
266: \centerline{\epsfxsize=8.5cm\epsfbox{fig2.eps}} \caption[*] {The
267: local differential tunneling conductance (LDTC) on the central Zn
268: impurity with 10 impurities embedded in the lattice. Left panel:
269: negative potentials (a) -3.0 eV (c) -1.2 eV (e) -0.3 eV. Right
270: panel: positive potentials (b) 3.0 eV (d) 1.2 eV (f) 0.3 eV. Green
271: curves in (e) and (f) are on-site LDTC of single-impurity case.}
272: \end{figure}
273:
274:
275:
276: {\it Many impurity} We investigate the quantum interference effect
277: by embedding 10 impurities in an otherwise clean lattice, which
278: yields $x=0.41\%$, close to laboratory terms. The lattice size is of
279: $N_x \times N_y = 49 \times 49$. Fig. 2 enumerates the on-site LDTC
280: for different values of impurity strength systematically. Generally,
281: quantum interference splits the single impurity resonances into
282: multiple peaks, as illustrated in Fig. 2 and extracting distinctive
283: information from those peaks directly is not easy.
284: However, the zero energy residual LDOS $\rho(\omega=0)$ is
285: particularly useful as it evolves systematically with respect to the
286: impurity potential. When $\phi=-3.0$eV, $\rho(\omega=0)$ is
287: considerably large; as $\phi$ increases but remains negative ({\it
288: attractive}), $\rho(\omega=0)$ decreases until it is completely
289: depleted. This tendency can be understood qualitatively in the sense
290: that the sample recovers its clean spectra which has vanishing
291: $\rho(\omega=0)$ when less contaminated. When $\phi$ becomes
292: positive and increases ({\it repulsive}), $\rho(\omega=0)$
293: aggregates remarkably and multiple peaks merge into a ``single
294: resonance peak'' when $\phi$ is around the value with which unitary
295: limit is obtained in single impurity analysis. Fig. 2(d) shows such
296: a strong resonance peak at negative bias slightly below zero, whose
297: width and height are in fair agreement with the result of Pan {\it
298: et al.}. The likeness between Fig. 1(a) and Fig. 2(d) will plausibly
299: suggests the single-impurity physics as dominating mechanism and the
300: fact that Zn impurities in the experimental conductance maps appear to be
301: isolated entities with nearly undisturbed four-fold symmetry in
302: LDTC seemingly supports this viewpoint. However, it is worthwhile
303: to point out that exact width of this zero-bias resonance in Fig.
304: 2(d) and STM experiments \cite{pan2000} is $\sim 10$ meV, an order
305: of magnitude bigger than what single-impurity theory expects but
306: agrees well with the impurity bandwidth in unitarity low density
307: limit, i.e., $\gamma = \sqrt{n_i\Delta_0E_F}$ \cite{sctma}. Hence
308: the figurative resemblance between the on-site LDTC's of
309: many-impurity and single-impurity must be the consequence of
310: quantum interference between dilute impurity states and the
311: ``homogeneous broadening'' effect which was proposed by Atkinson
312: {\it et al.} \cite{atkinson2003} and is not negligible in
313: laboratory terms. The possibility of $|\phi|$ being weak (less
314: than 0.4 eV) is excluded, as it will not introduce any prominent
315: resonant behavior and observable interference effect requires
316: inter-impurity distance as close as $3a$, corresponding to an
317: unreasonably large population of disordered sites. This is
318: demonstrated in Fig. 2(e) and 2(f): when $\phi=\pm0.3$ eV, the
319: on-site LDTC of dirty samples are essentially coincident with that
320: of single-impurity case. Collection of all the facts above and
321: comparison with experiments lead us to believe that Zn atoms in
322: BSCCO-2212 be {\it repulsive } potentials at least in the zeroth
323: order.
324:
325:
326: \begin{figure}[!t]
327: \centerline{\epsfxsize=8.5cm\epsfbox{fig3.eps}} \caption[*] {(a)
328: The LDTC $G_{\bf i}(E)$ on a Zn impurity at $r=0$ for $x=0.42\%$.
329: The dashed line represents the LDTC far away from Zn impurities;
330: (b) The spatial profile of LDTC maps at
331: $\Omega=-2.4$ meV;(c) The LDTC (normalized to the peak value)
332: versus distance from the Zn impurity; (d)(e)(f) The LDTC $G_{\bf
333: i}(E)$ on the Zn impurity at $r=0$ for $x=0.75\%,1.0\%,1.83\%$.}
334: \end{figure}
335:
336:
337: The exact magnitude of Zinc impurity is still open to
338: determination, but favorably it should not differ too much from
339: the value which approximates single-impurity unitary limit (which
340: is $\phi=1.2$eV here) with the cuprates band structure since the
341: unitary scattering due to doped Zn is verified by the scattering
342: phase shift of $0.49\pi$ in experiments \cite{pan2000}. We then
343: adopted $\phi=1.2$eV as the input of calculations and the results
344: are plotted in Fig. 3. The spectrum in channel (a) shows a
345: striking agreement with the result of Pan {\it et al.} with a sharp
346: resonant peak forming at $\Omega=0$ and a second smaller peak on
347: negative bias. Apparently, single-impurity scattering fails to
348: interpret both the asymmetry of the peaks positions and the
349: broadened peak widths. While the staggered magnetic interaction
350: may also address the two-peak structure, we would rather believe
351: that the negative peak should arise collectively from
352: inter-impurity correlations and our numerical inspection confirms
353: that it actually persists in a fairly wide region of positive
354: $\phi$'s; in Fig. 3(b), the spatial distribution of the resonance
355: around a Zn impurity shows the local minima of the four
356: nearest-neighbor sites, and the local maxima of the eight
357: 2nd-nearest- and 3rd-nearest-neighbor sites clearly forming a
358: ``box'' around this Zn atom; in Fig. 3(c), the normalized LDTC
359: $G({\bf r})$ at resonance frequency is plotted as a function of
360: distance away from the impurity along the nodal and antinodal
361: directions. The strength of the normalized LDTC in the nodal
362: direction is found to decay faster than that along the antinodal
363: direction, reinforcing what is observed in STM experiments
364: \cite{pan2000}. All this above agree with STM observations
365: remarkably and are confirmed to be robust against the change of
366: $\phi$ within a wide range centered at $1.2$eV; When impurity
367: concentration increases, the low energy excitations are further
368: populated and a subgap impurity band is established gradually,
369: which will suppress superconductivity eventually and is referred
370: to as ``impurity band", in analogy to similar phenomena in
371: semiconductors. This is illustrated schematically in Fig.
372: 3(d)(e)(f), where the zero bias impurity resonance is further
373: broadened and finally into multiple-peak structure with excessive
374: spectral weight filling up the gap region when impurity
375: concentration is several times bigger than the experimental value
376: ($0.2\%-0.5\%$).
377:
378:
379: {\it Conclusion}: We discuss how quantum interference between many
380: impurities can produce qualitatively different spectral features
381: when the impurity variables change. With a numerical
382: study on the disorder induced local spectra (on impurity site), we
383: would like to close the debate over the identity of Zn impurity by
384: concluding that Zn atoms in Bi$_2$Sr$_2$CaCu$_2$O$_{8+\delta}$ are
385: {\it repulsive} to electrons in nature with a strength close to
386: unitary limit provided that other internal degrees of freedom was
387: not considered. The discrepancy between STM experiments and the
388: results of single-impurity analysis can be reconciled
389: satisfactorily by taking the quantum interference and insulating
390: layer blocking effect into consideration simultaneously. Our
391: numerical calculations with impurity parameter determined in this
392: paper match STM experiments up to subtle details. We then would
393: emphasize that the results obtained in laboratories should
394: actually be interpreted within the frame of collective quantum
395: interference processes rather than the single impurity physics.
396:
397:
398: ${\bf Acknowledgements}$: We thank S.H. Pan, J.X. Zhu, J. O$'$Neal, Ang
399: Li and P.J. Hirschfeld for stimulating comments and suggestions. This
400: work is supported by the Texas Center for Superconductivity at the
401: University of Houston, and by a grant from the Robert A. Welch Foundation
402: under No. E-1146.
403:
404:
405: \begin{thebibliography}{99}
406:
407: \bibitem{pan2000} S.H. Pan, E.W. Hudson, K.M. Lang, H. Eisaki, S. Uchida,
408: and J.C. Davis, Nature {\bf 403}, 746 (2000).
409: \bibitem{balatsky1995} A.V. Balatsky, M.I. Salkola, and A. Rosengren,
410: Phys. Rev. B {\bf 51}, 15547 (1995).
411: \bibitem{balatsky0411} A.V. Balatsky, I. Vekhter, and Jian-Xin Zhu,
412: cond-mat/0411318.
413: \bibitem{zhujx2001} J.X. Zhu and C.S. Ting, Phys. Rev. B {\bf 64},
414: 060501(R) (2001).
415: \bibitem{martin2002} I. Martin, A.V. Balatsky, and J. Zaanen, Phys. Rev.
416: Lett. {\bf 88}, 097003 (2002).
417: \bibitem{sachdev} A. Polkovnikov, S. Sachdev, and M. Vojta, Phys. Rev.
418: Lett. {\bf 86}, 296-299 (2001).
419: \bibitem{wangll0505} L.-L. Wang, P.J. Hirschfeld and Haiping Chen,
420: cond-mat/0505014.
421: \bibitem{zhuly2003} Lingyin Zhu, W.A. Atkinson, and P.J. Hirschfeld,
422: Phys. Rev. B {\bf 67}, 094508 (2003).
423: \bibitem{onishi1996} Y. Onishi, Y. Ohashi, Y. Shingaki, and K. Miyake, J.
424: Phy. Soc. Jpn. {\bf 65}, 675 (1996).
425: \bibitem{pepin2001} C. Pepin and P.A. Lee, Phys. Rev. B {\bf 63} 054502
426: (2001).
427: \bibitem{morr2002} D.K. Morr and N.A. Stavropoulos, Phys. Rev. B {\bf 66},
428: 140508(R) (2002).
429: \bibitem{andersen2003} Brian M{\o}ller Andersen and Per Hedeg{\aa}rd,
430: Phys. Rev. B {\bf 67}, 172505 (2003)
431:
432: \bibitem{atkinson2003} W.A. Atkinson, P.J. Hirschfeld and Lingyin Zhu,
433: Phys. Rev. B, {\bf 68}, 054501 (2003).
434: \bibitem{zhujx2000} J.X. Zhu, C.S. Ting, and C.R. Hu, Phys. Rev. B {\bf
435: 62}, 6027 (2000).
436: \bibitem{sctma}The unitary low density limit is obtained by
437: {\it neglecting} quantum interference effect in self-consistent
438: $T$-matrix approximation. However, it gives the impurity band
439: width with the right order of magnitude.
440: \end{thebibliography}
441:
442: \end{document}
443: