1: \documentclass[twocolumn,prb,aps,bm,amssymb,amsmath,floats,showpacs]{revtex4}
2: %\documentstyle[twocolumn,prb,aps,epsf,floats]{revtex}
3: \usepackage{epsf}
4: \begin{document}
5: %\twocolumn[ %+++++
6: \hsize\textwidth\columnwidth\hsize\csname @twocolumnfalse\endcsname %+++++
7: \title{Crossover from tunneling to incoherent (bulk) transport in
8: a correlated nanostructure}
9:
10: \author{ J.~K. Freericks }
11:
12: \affiliation{
13: Department of Physics, Georgetown University,
14: Washington, D.C. 20057-0995, U.S.A. }
15:
16: \begin{abstract}
17: {
18: We calculate the junction resistance for a metal-barrier-metal device with
19: the barrier tuned to lie just on the insulating side of the metal-insulator
20: transition. We find that the crossover from tunneling behavior in thin
21: barriers at low temperature to incoherent transport in thick barriers
22: at higher temperature is governed by a generalized Thouless energy. The
23: crossover temperature can be estimated from the low temperature resistance
24: of the device and the bulk density of states of the barrier.
25: }
26: \end{abstract}
27:
28: \pacs{73.63.-b, 71.30.+h, 71.27.+a}
29: %]
30:
31: \maketitle
32:
33: Many electronic devices employ quantum-mechanical tunneling in
34: determining their transport properties. Examples include Josephson
35: junctions~\cite{josephson} and magnetic tunnel junctions~\cite{magnetic}.
36: When designing a device
37: manufacturing process, or when optimizing the operational characteristics
38: of a device, it is important to have diagnostic tools that can determine
39: if the transport is via tunneling or via defects in the barrier (such as
40: pinholes). In superconductor-based devices, this is well understood, and was
41: described in detail by Rowell~\cite{rowell}
42: in the 1970's. However, the criteria relied
43: on testing the device in the superconducting state. Interest in this problem for
44: normal metals and for higher device operating temperatures has been driven
45: by recent activity in magnetic tunnel junctions~\cite{magnetic}. A number of
46: useful criteria for tunneling~\cite{schuller}
47: have emerged for these normal-metal-based devices: (i) the junction
48: resistance should increase with decreasing temperature; (ii) the fit of an
49: $IV$ characteristic to a Simmons model~\cite{simmons}
50: should have a barrier height that
51: does not decrease and a fitted thickness that does not increase as $T$
52: decreases; and (iii) the junction noise should not increase at finite
53: bias. It has also been well established that the naive criterion for
54: tunneling, that the
55: resistance increases exponentially with the barrier thickness is insufficient,
56: since a rough interface plus pinholes will also yield this exponential
57: dependence~\cite{pinholes}.
58:
59: In this contribution, we perform a theoretical analysis of tunneling through
60: a correlated barrier to investigate the crossover from a tunneling regime,
61: where transport is dominated by quantum processes that provide ``shorts''
62: across the barrier, to an incoherent bulk transport regime, where the
63: transport occurs via incoherent
64: thermal excitations of carriers in the barrier. In the
65: latter case, one expects the junction resistance to scale linearly with the
66: barrier thickness, with the slope proportional to the bulk resistivity of the
67: barrier (which has a strong temperature dependence in an insulator). As the
68: barrier is made thinner (or the temperature is decreased), the direct
69: quantum-mechanical coupling of the metallic leads through states localized
70: in the barrier begins to dominate the transport process, and the resistance
71: is reduced from that predicted by the incoherent transport mechanism to a
72: relatively temperature independent tunneling-based resistance. Since the
73: wavefunctions that connect the two metallic leads decay exponentially in the
74: barrier, the tunneling resistance depends exponentially on the barrier
75: thickness. Most commercial devices operate in this tunneling regime because the
76: junction resistance is low enough to generate reasonable current values for
77: low voltages and because the weak temperature dependence simplifies
78: variations of the device parameters with temperature.
79:
80: In conventional tunneling devices, which use an insulator with a large
81: energy gap (like AlO$_x$), one cannot see the crossover to the bulk
82: transport regime, because it occurs at too high a temperature, or for too
83: resistive junctions to be of interest. But there has been recent work in
84: examining barriers that are tuned to lie closer to the metal-insulator
85: transition~\cite{newman}
86: (like Ta$_x$N), and thereby have much smaller ``energy gaps''.
87: Barriers of this type may be easier to work with because they can be made
88: thicker and thereby be less susceptible to pinhole formation. They also
89: can be advantageous for different applications. As the energy gap of the
90: barrier
91: material is made smaller (or equivalently, if the barrier potential height is
92: reduced), it becomes possible to observe and study the
93: crossover from tunneling to bulk transport.
94:
95: We consider a device constructed out of stacks of infinite two-dimensional
96: planes stacked in registry on top of each other. This kind of inhomogeneous
97: layered device can be used to describe a wide range of different
98: multilayer-based structures. We couple a bulk ballistic semi-infinite
99: metal lead to thirty self-consistent ballistic metal planes; then we stack
100: 1 to 20 barrier planes and then top with another thirty self-consistent
101: ballistic metal planes followed by another bulk ballistic semi-infinite metal
102: lead. The ballistic metal is described by a simple hopping Hamiltonian
103: with no interactions. The barrier is described by a spin-one-half
104: Falicov-Kimball model~\cite{falicov_kimball}
105: with the same hopping parameters as the metal
106: plus strong scattering that yields correlations for the electron motion.
107: The Hamiltonian is
108: \begin{equation}
109: {\mathcal{H}}=-t\sum_{\langle i,j\rangle\sigma}c^\dagger_{i\sigma}c_{j\sigma}
110: +\sum_{i\sigma}U_i^{FK}w_i(n_{i\sigma}-\frac{1}{2}),
111: \label{eq: ham}
112: \end{equation}
113: where $c^\dagger_{i\sigma}$ ($c_{i\sigma}$) creates (destroys) a conduction
114: electron at site $i$ with spin $\sigma$ and $t$ is the hopping parameter.
115: The hopping is on a simple cubic lattice
116: constructed from the stacked two-dimensional planes; i.e., the hopping integral
117: is chosen to be the same within a plane and between two planes. $U_i^{FK}$ is
118: the Falicov-Kimball interaction and $w_i$ is a classical variable, equal to zero
119: or one, which denotes the presence of a scatterer at site $i$. Finally,
120: $n_{i\sigma}=c^\dagger_{i\sigma}c_{i\sigma}$ is the electron number operator.
121: The Falicov-Kimball interaction is nonzero only within the barrier, where
122: we set it equal to $6t$---large enough to create an insulator with a gap of
123: $0.4t$. The average concentration of scatterers is $\langle w_i\rangle=1/2$
124: and we choose half filling for the electrons as well (with our choice of
125: interaction, this corresponds to a vanishing chemical potential). In order to be
126: quantitative, we pick the hopping parameter to satisfy $t=0.25$~eV, which
127: yields a bandwidth of 3~eV for the metallic leads and a gap of 100~meV for
128: the correlated insulator (much smaller than a conventional oxide
129: insulator). We solve for the Green's functions using an
130: inhomogeneous dynamical mean field theory calculation described
131: elsewhere~\cite{potthoff_nolting,miller,jkf_scms,jkf_review}.
132: The resistance-area product for this device is calculated by a real-space
133: version of Kubo's formula. We take the lattice constant to be 0.3~nm.
134:
135: \begin{figure}[htbf]
136: \epsfxsize=3.0in
137: \centerline{\epsffile{fig1.eps}}
138: \caption{Ratio of the resistance of the junction at temperature $T$ to the
139: resistance at 30~K. The different curves correspond to different thicknesses
140: of the barrier, which are labeled with an integer denoting the number of
141: atomic planes in the barrier. As expected, the temperature dependence of the
142: resistance increases as the barrier is made thicker, because the barrier is
143: becoming more bulk-like. However, in this regime, all of the transport is
144: still dominated by tunneling.
145: \label{fig: rn_t}
146: }
147: \end{figure}
148:
149:
150: In Fig.~\ref{fig: rn_t}, we plot the ratio of the junction resistance
151: at temperature $T$ to the resistance at 30~K for junctions with a barrier
152: thickness ranging from 1 to 10 atomic planes. In all cases, the resistance
153: shows a weak temperature dependence with an insulator-like character. This
154: low-$T$ behavior is often used as a diagnostic to indicate that tunneling
155: is occuring in a junction~\cite{schuller,rudiger}, and that certainly is the
156: case here.
157: Note how the temperature dependence increases as the thickness increases.
158: This is because the thicker the barrier is, the more it looks like
159: a bulk material, and an insulating barrier has strong (exponentially activated)
160: temperature dependence in the bulk.
161:
162: \begin{figure}[htbf]
163: \epsfxsize=3.0in
164: \centerline{\epsffile{fig2.eps}}
165: \caption{Resistance-area product as a function of the barrier thickness $L$
166: for a number of different temperatures (the labels on the curves are
167: in K). Notice how the thin barriers have an
168: exponential dependence on thickness, which gives way to a linear dependence
169: as the junctions are made thick enough. This crossover region moves to thinner
170: barriers as the temperature is increased.
171: \label{fig: rn_l}
172: }
173: \end{figure}
174:
175:
176: Since our junctions are defect free, with atomically smooth interfaces, we
177: can analyze the resistance at fixed temperature as a function of the
178: barrier thickness to look for exponential dependence in the tunneling
179: regime, with a crossover to linear dependence in the incoherent (bulk)
180: transport regime. This is plotted in Fig.~\ref{fig: rn_l} for a number
181: of different temperatures, ranging from 30~K to 1000~K. Note how we see
182: a perfect exponential dependence on thickness for thin barriers, which then
183: gives way to a crossover to linear behavior as the junctions are made thicker
184: and the transport becomes incoherent and thermally activated. Because
185: of the thermal activation, this crossover moves to thinner barriers as the
186: temperature is increased. But it is interesting to note that there is no
187: simple relationship between the bulk gap (approximately 50~meV or 550~K when
188: measured from the $T=0$ chemical potential) and the location of the
189: crossover thickness as a function of temperature. Indeed, as $T$ is
190: increased, this crossover region is pushed to thinner and thinner
191: barriers. This type of behavior has been seen in Josephson junctions
192: made from high temperature superconductors using
193: molecular-beam-epitaxy~\cite{eckstein}.
194: When the barrier was increased from 1 to 3 to 5 to 7 atomic planes, the
195: junction resistance initially increased exponentially, and then started
196: to turn over to a more linear dependence on thickness. However, because the
197: high-temperature superconductor is a d-wave superconductor, there is
198: strong temperature dependence to the junction resistance, even in the
199: tunneling regime, so direct comparison with results given here is impossible.
200: This behavior has also been seen in some magnetic tunnel
201: junctions~\cite{covington} where an exponential increase as a function of
202: thickness gives way to an essentially constant dependence on thickness
203: for thicker Aluminum regions. What is less known about this data is how much
204: of the Aluminum is oxidized in the manufacturing process. Also, no
205: temperature scans at fixed thickness were reported.
206:
207: \begin{figure}[htbf]
208: \epsfxsize=3.0in
209: \centerline{\epsffile{fig3.eps}}
210: \caption{Resistance-area product as a function of temperature
211: for a number of different barrier thicknesses plotted on a log-log
212: plot. Notice how the thin barriers have a weak dependence on temperature,
213: and a constant step-size increase in the logarithm of the resistance as
214: the thickness increases,
215: indicating tunneling behavior, and how there is a crossover to incoherent
216: transport as $T$ is increased. The dashed line shows the boundary where
217: the generalized Thouless energy is equal to $k_BT$. This marks the
218: approximate crossover from tunneling (for $E_{\rm{Th}}(T)\gg k_BT$) and
219: incoherent transport (for $E_{\rm{Th}}(T)\ll k_BT$).
220: \label{fig: thouless}
221: }
222: \end{figure}
223:
224: In Fig.~\ref{fig: thouless}, we plot $R_n(T)A$ versus $T$ for a variety of
225: barrier thicknesses on a log-log plot. This figure clearly shows the
226: tunneling regime, where the resistance-area product is approximately
227: constant, and it shows the incoherent regime, where the resistance-area product
228: has a strong temperature dependence. The dashed line, that divides these two
229: regions is an approximate boundary that denotes the crossover region for
230: the two different types of transport. This crossover line is determined
231: by equating an energy scale extracted from the resistance with the temperature.
232: When this energy scale is larger than $k_BT$ we have tunneling, when it is lower
233: than $k_BT$ we have incoherent transport. The energy scale is a generalized
234: Thouless energy~\cite{thouless}, valid for a barrier that is described by an
235: insulator that does not have either ballistic or diffusive transport. The
236: generalized Thouless energy $E_{\rm{Th}}$ is the energy scale constructed
237: from the resistance at temperature $T$ via the expression
238: \begin{equation}
239: E_{\rm{Th}}(T)=\frac{\hbar}{R_n(T)\frac{2e^2}{\hbar}\int d\omega
240: \left (-\frac{df(\omega)}{d\omega}\right ) \rho_{int}(\omega)L}
241: \label{eq: thouless}
242: \end{equation}
243: where $f(\omega)=1/[1+\exp(\omega/k_BT)]$ is the Fermi-Dirac distribution,
244: $\rho_{int}(\omega)$ is the bulk density of states in the insulator, and
245: $L$ is the barrier thickness.
246: This definition of $E_{\rm{Th}}$ agrees with the conventional notion of
247: $\hbar/t_{dwell}$, relating the Thouless energy to the dwell time in
248: the barrier, when the transport in the barrier is described by either
249: a ballistic metal
250: (where the Thouless energy varies like $C/L$) or a diffusive metal
251: (where the Thouless energy varies like $C/L^2$),
252: but it can now be generalized for an insulating barrier as well (where the
253: Thouless energy now picks up a substantial temperature dependence).
254:
255: The notion of a Thouless energy can be employed as a diagnostic for tunneling
256: devices. Since $R_n(T)$ depends weakly on $T$ in the tunneling regime,
257: one can measure $R_n$ at low $T$, and estimate the crossover temperature,
258: by computing a simple integral over the bulk insulator density of states.
259: Then one evaluates $E_{\rm{Th}}(T)$ employing Eq.~(\ref{eq: thouless}) using
260: the low-temperature value of the resistance. The crossover temperature
261: is estimated by the point where $E_{\rm{Th}}(T)=k_BT$.
262: \textit{Note further that this crossover temperature is not proportional to the
263: gap of the bulk insulator, but rather is a complicated function of the
264: barrier thickness, and the strength of the correlations.}
265:
266: In summary, we have determined an energy scale extracted from the resistance of
267: a junction, that governs the crossover from tunneling to incoherent
268: transport. This energy scale approaches zero as the barrier thickness
269: becomes large, hence it could have applicability to any tunneling-based
270: device, but when we examine the common resistance-area products of actual
271: devices, it becomes clear that this concept will have the most applicability
272: to junctions with barriers tuned to lie close to the metal-insulator transition.
273: Since it is possible such devices will be used for devices of the future,
274: the concept of a generalized Thouless energy should become an important
275: diagnostic tool in evaluating the quality of devices, and allow one to engineer
276: the thickness and operating temperature range to guarantee tunneling with
277: the chosen barrier.
278:
279:
280: {\it Acknowledgments:}
281: We acknowledge support from the National Science Foundation under grant
282: number DMR-0210717 and from the Office of Naval Research under grant number
283: N00014-99-1-0328. High performance computer time was provided by the Arctic
284: Region Supercomputer Center and the U. S. Army Engineering Research and
285: Development Center. We also acknowledge useful discussions
286: with J. Eckstein, B. Jones, N. Newman, B. Nikoli\'c, S. Parkin, J. Rowell,
287: I. Schuller and S. Shafraniuk.
288:
289: \begin{thebibliography}{99}
290:
291: \bibitem{josephson} B. D. Josephson, Phys. Lett. {\bf 1}, 251 (1962).
292:
293: \bibitem{magnetic} J. S. Moodera, L. R. Kinder, T. M. Wong, and R. Meservey,
294: Phys. Rev. Lett. {\bf 74}, 3273 (1995); T. Miyazaki and N. Tezuka, J. Magn.
295: Magn. Mater. {\bf 139}, L231 (1995); S. Tehrani, J. M. Slaughter, E. Chen,
296: M. Durlam, J. Shi, and M. DeHerrera, IEEE Trans. Magn. {\bf 35}, 2814 (1999);
297: J. Zhang, Data Storage {\bf 5}, 31 (1998); S. A. Wolf and D. Treger, IEEE
298: Trans. Magn. {\bf 36}, 2748 (2000).
299:
300: \bibitem{rowell} J. M. Rowell, in \textit{Tunneling Phenomena in Solids},
301: edited by E. Burnstein and S. Lundqvist (Plenum, New York, 1969), p. 273.
302:
303: \bibitem{schuller} J. J. Akerman, J. M. Slaughter, R. W. Dave, and
304: I. K. Schuller, Appl. Phys. Lett. {\bf 79}, 3104 (2001).
305:
306: \bibitem{simmons} J. G. Simmons, J. Appl. Phys. {\bf 34}, 1793 (1963).
307:
308: \bibitem{pinholes} D. A. Rabson, B. J. J\" onsson-Akerman, A. H. Romero,
309: R. Escudero, C. Leighton, S. Kim, and I. K. Schuller, J. Appl. Phys. {\bf 89},
310: 2786 (2001); W. H. Rippard, A. C. Perrella, F. J. Albert, and R. A. Buhrman,
311: Phys. Rev. Lett. {\bf 88}, 046805 (2002); L. S. Dorneles, D. M. Schaefer,
312: M. Carara, and L. F. Schelp, Appl. Phys. Lett. {\bf 82}, 2832 (2003).
313:
314: \bibitem{newman} A.B. Kaul, S.R. Whitely, T. van Duzer, L. Yu, N.
315: Newman, and J.M. Rowell, Appl. Phys. Lett. {\bf 78}, 99 (2001).
316:
317: \bibitem{falicov_kimball} L.~M.~Falicov and J.~C.~Kimball, Phys. Rev.
318: Lett. {\bf 22}, 997 (1969).
319:
320: \bibitem{potthoff_nolting} M. Potthoff and W. Nolting, Phys. Rev.
321: B {\bf 59}, 2549 (1999).
322:
323: \bibitem{miller} P. Miller and J.~K.~Freericks, J. Phys.: Conden. Matter
324: {\bf 13}, 3187 (2001).
325:
326: \bibitem{jkf_scms} J.K. Freericks, B.K. Nikoli\'c, and P. Miller,
327: Phys. Rev. B {\bf 64}, 054511 (2001); Phys. Rev. B {\bf 68}, 099901 (E)
328: (2003).
329:
330: \bibitem{jkf_review} J.K. Freericks, B.K. Nikoli\'c, and P. Miller,
331: Int. J. Mod. Phys. B 16, 531--561 (2002).
332:
333: \bibitem{rudiger} U. R\" udiger, R. Calarco, U. May, K. Samm, J. Hauch,
334: H. Kittur, M. Sperlich, and G. G\"untherodt, J. Appl. Phys. {\bf 89}, 7573
335: (2001).
336:
337: \bibitem{eckstein} J. N. Eckstein and I. Bozovic, Ann. Rev. Mater. Sci.
338: {\bf 25}, 679 (1995).
339:
340: \bibitem{covington} M. Covington, J. Nowak, and D. Song, Appl. Phys. Lett.
341: {\bf 76}, 3965 (2000).
342:
343: \bibitem{thouless} J. T. Edwards and D. J. Thouless, J. Phys. C {\bf 5},
344: 807 (1972); D. J. Thouless, Phys. Rep. {\bf 13}, 93 (1974); A. Atland, Y. Gefen,
345: and G. Montambaux, Phys. Rev. Lett. {\bf 76}, 1130 (1996); M. Janssen,
346: \textit{Fluctuations and Localization in Mesoscopic Electron Systems},
347: (World Scientific, Singapore, 2001).
348:
349: \end{thebibliography}
350:
351:
352: \end{document}
353:
354:
355: