1: \documentclass[aps,twocolumn,showpacs,superscriptaddress]{revtex4}
2:
3: \usepackage[dvips]{graphicx} \usepackage{amsmath} \usepackage{amsfonts}
4: \usepackage{amssymb}
5:
6: \newcommand{\Tr}{\mathop{\mathrm{Tr}}} % matrix trace
7: %\newcommand{\iu}{{\mathrm{i}}} % imaginary unit
8: %\newcommand{\upd}{{\mathrm{d}}} % differential operator
9: \newcommand{\iu}{i} % imaginary unit
10: \newcommand{\upd}{d} % differential operator
11: \newcommand{\omat}{\boldsymbol} % orbital matrix
12:
13: \begin{document}
14:
15: \title{Role of electronic structure in photoassisted transport
16: through atomic-sized contacts}
17:
18: \author{J. K. Viljas}
19: \affiliation{Institut f\"ur Theoretische Festk\"orperphysik,
20: Universit\"at Karlsruhe, D-76128 Karlsruhe, Germany}
21: \affiliation{Forschungszentrum Karlsruhe,
22: Institut f\"ur Nanotechnologie, D-76021 Karlsruhe, Germany }
23:
24: \author{J. C. Cuevas}
25: \affiliation{Departamento de F\'{\i}sica Te\'orica de la Materia
26: Condensada, Universidad Aut\'onoma de Madrid, E-28049 Madrid, Spain}
27: \affiliation{Institut f\"ur Theoretische Festk\"orperphysik,
28: Universit\"at Karlsruhe, D-76128 Karlsruhe, Germany}
29: \affiliation{Forschungszentrum Karlsruhe,
30: Institut f\"ur Nanotechnologie, D-76021 Karlsruhe, Germany }
31:
32: \date{\today}
33:
34: \begin{abstract}
35: We study theoretically quantum transport through laser-irradiated
36: metallic atomic-sized contacts. The radiation field is treated
37: classically, assuming its effect to be the generation of an ac
38: voltage over the contact. We derive an expression for the dc current
39: and compute the linear conductance in one-atom thick contacts as a
40: function of the ac frequency, concentrating on the role played by
41: electronic structure. In particular, we present results for three
42: materials (Al, Pt, and Au) with very different electronic
43: structures. It is shown that, depending on the frequency and the
44: metal, the radiation can either enhance or diminish the conductance.
45: This can be intuitively understood in terms of the energy dependence
46: of the transmission of the contacts in the absence of radiation.
47: \end{abstract}
48:
49: \pacs{73.63.-b, 73.50.Pz, 73.63.Rt, 73.40.Jn}
50:
51: \maketitle
52:
53: \section{Introduction}
54: The study of electronic transport in microscopic and nanoscale
55: electrical contacts under the influence of time-dependent external
56: fields has a long history. Perhaps as the most famous example,
57: superconducting tunnel junctions subjected to microwave radiation
58: exhibit a step-like structure in their current-voltage ($I$-$V$)
59: characteristics \cite{Tucker85}. This can be understood in terms of
60: inelastic (``photoassisted'') transport of electrons across the
61: junction. In the early theoretical work of Tien and Gordon (TG)
62: \cite{Tien63}, this phenomenon was described by a harmonic voltage at
63: the radiation frequency $\omega$ applied to one of the leads, giving
64: rise to photo-``sidebands'' associated with the ``absorption'' or
65: ``emission'' of an integer multiple of the photon energy
66: $\hbar\omega$. This theory has also been extended to describe
67: superconducting atomic point contacts \cite{Cuevas02} and the
68: predictions have been confirmed experimentally using microfabricated
69: Al break-junctions under microwave irradiation \cite{Chauvin06}. In
70: addition, a similar approach has been used to describe
71: laser-irradiated junctions in scanning tunneling microscopes (STM)
72: \cite{Grafstrom02}. In these, as a result of inherent asymmetries in
73: the geometry and the materials of the junction, laser irradiation can
74: cause dc (rectification) currents even in the absence of a dc bias
75: voltage \cite{Voelcker91,Levy91}.
76:
77: Over the years, several types of model calculations have been employed
78: also to describe the ac-response of semiconductor heterostructures
79: \cite{Jauho94,Platero04} and other mesoscopic systems
80: \cite{Buttiker96,Pedersen98,Wang99}, as well as atomic and molecular
81: contacts \cite{Zheng00,Buker02,Kohler04,Galperin05,Wu05}. Among
82: these, the TG-like theories have been quite successful in gaining a
83: qualitative understanding of light-induced currents
84: \cite{Tucker85,Platero04,Kohler04}.
85:
86: \begin{figure}[!b]
87: \includegraphics[width=0.95\linewidth,clip=]{fig1_lcr.eps}
88: \caption{(Color online) (a) Two infinite fcc [001] surfaces contacted
89: with pyramids of 10 atoms in each, forming a ``dimer'' contact. The
90: tip-to-tip distance is denoted with $D$, all other interatomic
91: distances are the same as in bulk. (b) Two model ac voltage profiles
92: $U_i^{ac}$, where $i$ labels atomic sites and orbitals in the $C$
93: region: a double step (A) and a linear ramp (B) between the lead
94: values $U_{L}^{ac}$ and $U_{R}^{ac}$.}
95: \label{f.lcr}
96: \end{figure}
97:
98: Metallic atomic-sized wires fabricated with STM or break-junction
99: techniques have turned out to be ideal systems for investigating
100: electronic transport at the nanoscale \cite{Agrait03}. The bulk of the
101: research in this field so far has concentrated on stationary transport
102: properties, while systems being driven by time-varying external fields
103: (such as laser light) have received less attention. Very recent
104: experiments on laser-irradiated gold contacts support the idea that
105: photoassisted processes may play an important role in their transport
106: properties \cite{Guhr06}. To take the first steps towards a
107: microscopic description of experiments of this type, we address in
108: this paper the role of electronic structure in photoassisted
109: transport through atomic-sized junctions. This problem is not only
110: relevant for the field of atomic contacts, but also for molecular
111: electronics, where the role of the metallic contacts on photoassisted
112: transport through molecular junctions remains to be understood.
113:
114: In different types of metals, the nature and number of conduction
115: channels in one-atom contacts reflect the valence of the metal,
116: \emph{i.e.}, what type of atomic orbitals are available at the Fermi
117: energy \cite{Scheer98}. How is this difference between the metals seen
118: in their response to irradiation? To investigate this question, we use
119: a tight-binding model and explore mainly one-atom thick contact
120: geometries like the one shown in Fig.\ \ref{f.lcr}(a). We are
121: interested in the linear conductance $G_{dc}=\partial I/\partial
122: V|_{V=0}$ as a function of the radiation frequency $\omega$, which we
123: call \emph{photoconductance}. Following the TG ideology, we model the
124: effect of radiation with a time-periodic voltage, but we also take
125: into account the effect of the voltage profile across the contact [see
126: Fig.\ \ref{f.lcr}(b)]. While many previous model calculations with
127: periodic driving fields are based on Floquet theory
128: \cite{Platero04,Kohler04}, our method is based on non-equilibrium
129: Green functions \cite{Levy91,Jauho94,Brandes97,Datta92}. Our approach
130: allows for a realistic description of the photoconductance of atomic
131: contacts, with a single free parameter describing the local intensity
132: of the radiation.
133:
134: As examples of metals with very different electronic structures we
135: choose Al, Pt, and Au. We study one-atom thick junctions of these
136: materials mainly in the so-called contact regime, but also in the
137: tunneling regime. In the case of the contact regime we find the
138: following results. For Pt the effect of the ac voltage is almost
139: always to decrease $G_{dc}$ from its value in the absence of
140: radiation. This is because the Fermi energy $\epsilon_F$ lies at the
141: edge of the $d$ band, and exciting electrons above $\epsilon_F$
142: necessarily decreases their transmission probability, since less
143: transmission channels are available there. For Al, where $\epsilon_F$
144: is at the beginning of the $p$ band, also an enhancement of the
145: conductance is possible. The magnitude of these changes depends on the
146: intensity of the radiation, but can be up to $10\%$ or more at visible
147: frequencies. For Au, $\epsilon_F$ is in the $s$ band, with a single
148: completely open transmission channel for a wide range of energies.
149: Thus, low-frequency radiation has no effect on the conductance, while
150: in the visible range both an increase and a decrease are possible.
151: These conclusions are based on detailed numerical simulations, but
152: they can be understood in an appealing way in terms of the energy
153: dependence of the transmission in the absence of the ac drive.
154:
155: The rest of the paper is organized as follows. In Sec.\ \ref{s.model}
156: we introduce the theoretical model used to describe the electronic
157: structure of atomic-sized contacts and the effect of irradiation.
158: Section \ref{s.formula} is devoted to the derivation of a general
159: formula for the dc current through atomic contacts subjected to an ac
160: voltage. In Sec.\ \ref{s.results} we describe examples of numerical
161: results of the frequency dependence of the linear conductance for
162: one-atom thick contacts of Al, Pt, and Au. Finally, in Sec.\
163: \ref{s.discussion} we discuss some of the assumptions and restrictions
164: of our model and present the main conclusions.
165:
166:
167: \section{Model}\label{s.model}
168: We describe the laser-irradiated point contact with an
169: $spd$-tight-binding Hamiltonian of the form
170: %
171: \begin{equation}
172: \begin{split} \label{e.hamilton}
173: H(t) & = H_{0} + H_{1}(t) \\
174: \hspace*{-0.5cm} H_{0} =\sum_{ij}d_i^\dagger H_{ij}d_j , &
175: \hspace*{0.5cm} H_{1}(t) = \sum_{ij}d_i^\dagger W_{ij}(t)d_j,
176: \end{split}
177: \end{equation}
178: %
179: where the indices $i,j$ run over the different atoms and orbitals,
180: including the two spin directions. The creation and annihilation
181: operators ($d^\dagger_i$ and $d_j$) satisfy $\{d_i,d_j\}=0$ and
182: $\{d_i,d^\dagger_j\}=[\omat{S}^{-1}]_{ij}$, where $\omat{S}$ is the
183: overlap matrix of the non-orthogonal basis \cite{Viljas05}. The
184: Hamiltonian $H_{0}$ is for the system without radiation and dc
185: voltage, while $H_{1}(t)$ includes them both. The matrix elements of
186: the Hamiltonian and the overlap matrix ($H_{ij}$ and $S_{ij}$) are
187: taken from the parametrization of Ref.\ \onlinecite{Papa}. All
188: matrices in the spin-orbital basis (such as $\omat{H}$ and $\omat{S}$)
189: are denoted with a boldface symbol.
190:
191: We consider ideal symmetric geometries of the type shown in Fig.\
192: \ref{f.lcr}(a), with a single-atom thick constriction. This type of
193: ``dimer'' structure is suggested by molecular dynamics simulations as
194: the most common one in the last conductance plateau \cite{Dreher05}.
195: When the distance $D$ between the tip atoms corresponds to the bulk
196: interatomic distance, the junction is said to be in the contact
197: regime. We also study larger $D$ values, where the junction enters the
198: tunneling regime. For the calculation of transport we shall, in the
199: usual way, divide the system into left lead ($L$), central ($C$), and
200: right lead ($R$) regions. The leads are modeled with infinite
201: surfaces, where the fcc [001] axis coincides with the transport
202: direction. We index the regions with the label $\Omega=L,C,R$. As a
203: matrix index, this label also indicates collectively all the orbital
204: indices of the respective region \cite{Viljas05}.
205:
206: We assume that the $L$ and $R$ lead potentials are spatially constant
207: and harmonic with angular frequency $\omega$, such that
208: $T_\omega=2\pi/\omega$ is the oscillation period. Thus
209: $\omat{W}_{\Omega\Omega}(t)=U_\Omega(t)\omat{S}_{\Omega\Omega}$, where
210: $U_{\Omega}(t)=U^{dc}_{\Omega}+U^{ac}_{\Omega}\cos(\omega t)$, and
211: $\Omega=L,R$. The applied dc voltage $V=(U_L^{dc}-U_R^{dc})/e$, where
212: $-e$ is the electron charge, is assumed to be infinitesimal. For the
213: lead-center hoppings we also assume
214: $\omat{W}_{C\Omega}(t)=U_{\Omega}(t)\omat{S}_{C\Omega}$, etc. The
215: central potential is assumed to be of the form
216: $\omat{W}_{CC}(t)=\omat{W}_{CC}^{dc}+\omat{W}_{CC}^{ac}\cos(\omega
217: t)$, or $[\omat{W}_{CC}(t)]_{ij} = [\omat{S}_{CC}]_{ij}
218: (U_i(t)+U_j(t))/2$ \cite{Note1}. Here,
219: $U_{i}(t)=U_i^{dc}+U_i^{ac}\cos(\omega t)$ is the same for all
220: orbitals $i$ on the same atom in region $C$. The actual shape of
221: $U_{i}(t)$ within $C$ should in principle be obtained
222: self-consistently through the solution of a Poisson equation, such
223: that screening and local field-enhancement effects would be properly
224: accounted for. In this case the appearance of higher harmonics of
225: $\omega$ and phase shifts in $\omat{W}_{CC}(t)$ would be possible. As
226: this would be computationally very costly, we shall do the following:
227: the dc part $U_i^{dc}$ is fixed by a requirement of local charge
228: neutrality in equilibrium \cite{Viljas05,Note1} and incorporated into
229: $H_{CC}$, while for $U_i^{ac}$ we just assume some simple forms.
230: Below we consider two model profiles [see Fig.\ \ref{f.lcr}(b)]:
231: $U_i^{ac}=(U_L^{ac}+U_R^{ac})/2$ for all orbitals $i$ in region $C$
232: (profile A), and one which linearly interpolates between $U_L^{ac}$
233: and $U_R^{ac}$ (profile B). For the symmetric junctions that we are
234: considering, symmetric profiles of this type are the most reasonable.
235:
236:
237: \section{Current formula}\label{s.formula}
238: We only consider the time-averaged current, and can thus neglect
239: displacement contributions without losing current conservation
240: \cite{Pedersen98,Buttiker96,Wang99}. The particle current is given in
241: terms of non-equilibrium Green functions by \cite{Jauho94}
242: %
243: \begin{equation}
244: \begin{split}\label{e.curr1}
245: I &= \frac{e}{\hbar}\int_0^{T_\omega}\frac{\upd t}{T_\omega}
246: \Tr[(\omat{G}^{<}_{CL}\circ \omat{t}_{LC}
247: -\omat{t}_{CL}\circ \omat{G}^{<}_{LC})(t,t)] . \\
248: \end{split}
249: \end{equation}
250: %
251: Here, in the case of the non-orthogonal basis \cite{Viljas05}
252: %
253: \begin{equation}
254: \begin{split} \label{e.hoppings}
255: \omat{t}_{CL}(t,t')&=[(\omat{H}_{CL}+\omat{W}_{CL}(t))
256: -\omat{S}_{CL}\iu\hbar\partial_t]\hbar\delta(t-t') , \nonumber
257: \end{split}
258: \end{equation}
259: %
260: and using the ``Langreth rules'', $\omat{G}^{<}_{CL} =
261: \omat{G}^{r}_{CC}\circ \omat{t}_{CL}\circ \omat{g}^{<}_{LL}
262: +\omat{G}^{<}_{CC}\circ \omat{t}_{CL}\circ \omat{g}^{a}_{LL}$
263: \cite{Jauho94}. The product $\circ$ is defined by $(\omat{A}\circ
264: \omat{B})(t,t')=\int(\upd s/\hbar)\omat{A}(t,s)\omat{B}(s,t')$, where
265: $\omat{A}$ and $\omat{B}$ are matrices in the spin-orbital basis, over
266: which the trace $\Tr$ acts. The Green functions
267: $\omat{G}^x_{\Omega\Omega'}$ with $x=r,a,\gtrless$ and
268: $\Omega,\Omega'=L,C,R$ are defined as usual \cite{Jauho94,Viljas05},
269: and $\omat{g}^{x}_{\Omega\Omega}$ are the functions for uncoupled
270: leads. The $CC$ component of the full retarded function satisfies
271: %
272: \begin{equation}
273: \begin{split} \label{e.gcc}
274: [\omat{S}_{CC}\iu\hbar \frac{\partial}{\partial t} -\omat{H}_{CC}-\omat{W}_{CC}(t)]
275: \omat{G}^{r}_{CC}(t,t') =\hbar \delta(t-t') \\
276: +(\omat{\Sigma}^r_{L}\circ \omat{G}^{r}_{CC})(t,t') +
277: (\omat{\Sigma}^r_{R}\circ \omat{G}^{r}_{CC})(t,t') ,
278: \end{split}
279: \end{equation}
280: %
281: whereas
282: $\omat{G}^{\gtrless}_{CC}=\omat{G}^r_{CC}\circ(\omat{\Sigma}^{\gtrless}_L
283: +\omat{\Sigma}^{\gtrless}_R)\circ \omat{G}^a_{CC}$. Here we defined the
284: ``lead self-energies'' $\omat{\Sigma}^x_{\Omega}
285: = \omat{t}_{C\Omega}\circ \omat{g}^x_{\Omega\Omega}\circ \omat{t}_{\Omega C}$,
286: with $\Omega=L,R$. The solutions of
287: %
288: \begin{equation}
289: \begin{split}
290: [\omat{S}_{\Omega\Omega}\iu\hbar\frac{\partial}{\partial t}
291: -\omat{H}_{\Omega\Omega}-\omat{W}_{\Omega\Omega}(t)]
292: \omat{g}^r_{\Omega\Omega}(t,t') & = \hbar\delta(t-t')
293: \end{split}
294: \end{equation}
295: %
296: lead to self-energies of the form
297: %
298: \begin{equation}\begin{split}
299: \omat{\Sigma}^x_\Omega(t,t') =
300: e^{-\iu\int_{t'}^t\frac{\upd s}{\hbar} U_\Omega(s)}
301: \int\frac{\upd\epsilon}{2\pi} e^{-\iu\epsilon(t-t')/\hbar}
302: \omat{\Sigma}^{x,eq}_\Omega(\epsilon) .
303: \end{split}\end{equation}
304: %
305: The equilibrium self-energies are given by
306: $\omat{\Sigma}^{x,eq}_\Omega(\epsilon)=\omat{t}_{C\Omega}(\epsilon)
307: \omat{g}^{x,eq}_{\Omega\Omega}(\epsilon)\omat{t}_{\Omega C}(\epsilon)$,
308: where the Green functions
309: $\omat{g}^{x,eq}_{\Omega\Omega}(\epsilon)$ of the infinite surfaces
310: are obtained by a decimation method \cite{Viljas05}, while
311: $\omat{t}_{C\Omega}(\epsilon)=\omat{H}_{C\Omega}-\epsilon
312: \omat{S}_{C\Omega}$ and $\omat{t}_{\Omega C}=[\omat{t}_{C\Omega}]^\dagger$.
313: We also have
314: $\omat{g}^{<,eq}_{\Omega\Omega}(\epsilon)=-f(\epsilon)$
315: $[\omat{g}^{r,eq}_{\Omega\Omega}(\epsilon)-\omat{g}^{a,eq}_{\Omega\Omega}(\epsilon)]$
316: where $f(\epsilon)$ is the Fermi function.
317: %$f(\epsilon)=1/[\exp(\beta(\epsilon))+1]$ is the Fermi function,
318: %$\beta$ being the inverse temperature.
319:
320: Since the Green functions and self-energies $\omat{A}(t,t')$ are all
321: periodic in the time $(t+t')/2$, we can simplify the analysis by
322: working in Fourier coordinates. We define the harmonic matrices
323: $\hat{A}(\epsilon)$ with components
324: $[\hat A]_{m,n}(\epsilon)=\omat{A}_{m-n}(\epsilon+(m+n)\hbar\omega/2)$,
325: where $m,n$ are integers, and
326: %
327: \begin{equation}\begin{split}
328: \omat{A}_n(\epsilon) &= \int_0^{T_\omega}
329: \frac{\upd T}{T_\omega} e^{\iu n\omega T}
330: \int\frac{\upd \tau}{\hbar}e^{\iu\epsilon \tau/\hbar}
331: \omat{A}(T+\tau/2,T-\tau/2) . \nonumber
332: \end{split}\end{equation}
333: %
334: In this way we find
335: $\hat{G}_{}^{\gtrless}=\hat{G}_{}^r(\hat{\Sigma}^{\gtrless}_L
336: +\hat{\Sigma}^{\gtrless}_R)\hat{G}_{}^a$,
337: while
338: %
339: \begin{equation}\begin{split}
340: [\hat{G}_{}^{r,a}]^{-1}
341: &=[\hat\epsilon S_{CC}-H_{CC} ]\hat{1}
342: -\hat W_{} -\hat{\Sigma}^{r,a}_L-\hat{\Sigma}^{r,a}_R .
343: \end{split}\end{equation}
344: %
345: Here $[\hat\epsilon]_{m,n}=(\epsilon+m\hbar\omega)\delta_{m,n}$,
346: $[\hat W_{}]_{m,n}=\omat{W}_{CC}^{ac}(\delta_{m-1,n}+\delta_{m+1,n})/2$,
347: and we have dropped the indices $CC$ from the harmonic matrices.
348: The self-energies as well as the scattering rates $\hat{\Gamma}_{\Omega}=
349: \iu(\hat{\Sigma}^r_\Omega-\hat{\Sigma}^a_\Omega)$
350: are related to the corresponding equilibrium quantities by
351: %
352: \begin{equation}
353: \begin{split}
354: [\hat\Sigma^x_\Omega]_{m,n}
355: &=\sum_{l} [\hat\Sigma^{x(l)}_\Omega]_{m,n}, \quad
356: [\hat\Gamma_\Omega]_{m,n}
357: =\sum_{l} [\hat\Gamma^{(l)}_\Omega]_{m,n} ,
358: \end{split}
359: \end{equation}
360: %
361: where we define the components
362: \begin{equation}\begin{split}
363: [\hat\Gamma^{(l)}_\Omega]_{m,n}(\epsilon) &=
364: J_{m+l}\left(\alpha_\Omega\right)
365: J_{n+l}\left(\alpha_\Omega\right)
366: \omat{\Gamma}_\Omega^{eq}(\epsilon-U_\Omega^{dc}-l\hbar\omega) , \nonumber
367: \end{split}\end{equation}
368: with a similar equation for $\hat{\Sigma}^{x(l)}_\Omega(\epsilon)$.
369: %
370: Here
371: $\omat{\Gamma}_\Omega^{eq}(\epsilon)
372: =\iu[\omat{\Sigma}^{r,eq}(\epsilon)-\omat{\Sigma}^{a,eq}(\epsilon)]$,
373: $J_l$ are Bessel functions of the first kind,
374: and $\alpha_\Omega=U_\Omega^{ac}/\hbar\omega$.
375: All of the harmonic matrices satisfy the symmetry
376: $[\hat A^{(l)}]_{m+k,n+k}(\epsilon) =
377: [\hat A^{(l+k)}]_{m,n}(\epsilon+k\hbar\omega)$.
378: Finally, we may note that
379: $\int(\upd t/T_\omega)\Tr[(\omat{A}\circ \omat{B})(t,t)]
380: =\int_0^{\hbar\omega}(\upd\epsilon/2\pi)\Tr_{\omega}
381: [\hat{A}(\epsilon)\hat{B}(\epsilon)]$,
382: where ${\Tr}_{\omega}[\hat{A}(\epsilon)]=\sum_m\Tr [\hat{A}]_{m,m}(\epsilon)$.
383: For numerical calculations, the matrices $\hat A(\epsilon)$ must be
384: truncated to a few lowest indices $m,n$, but the results converge
385: rapidly with the cutoff.
386:
387: Using the above definitions and general symmetries like
388: $\hat{G}^r-\hat{G}^a=\hat{G}^>-\hat{G}^<$,
389: Eq.\ (\ref{e.curr1}) yields the dc current
390: %
391: \begin{equation}\begin{split}\label{e.elcurr2}
392: I &= \frac{e}{\hbar}
393: \int_0^{\hbar\omega}\frac{\upd\epsilon}{2\pi}
394: \sum_{k,l}
395: {\Tr}_\omega[\hat{G}_{}^r\hat{\Gamma}_R^{(k)}
396: \hat{G}_{}^a\hat{\Gamma}_L^{(l)}]
397: (f_L^{(l)}-f_R^{(k)}) ,
398: \end{split}\end{equation}
399: %
400: where
401: $f_{\Omega}^{(k)}(\epsilon) = f(\epsilon-U_\Omega^{dc}-k\hbar\omega)$.
402: In the absence of an ac field this reduces to the standard Landauer-type
403: formula \cite{Datta95}. Although we are not computing $U_i(t)$
404: self-consistently, Eq.\ (\ref{e.elcurr2}) is still gauge invariant in
405: the sense that a spatially constant potential added everywhere has no
406: effect. Thus, the results only depend on $U_L(t)-U_R(t)$ \cite{Pedersen98}.
407:
408:
409: \section{Results}\label{s.results}
410: The experimentally accessible quantity which we calculate
411: is the linear conductance $G_{dc}(\omega)=\partial I/\partial V|_{V=0}$.
412: In what follows we shall assume zero temperature. If the
413: lead self-energies $\Sigma^{r,eq}_{L,R}$
414: are furthermore assumed to be energy-independent (``wide-band''
415: approximation), then the full result for potential profile A simplifies to
416: %
417: \cite{Pedersen98}
418: \begin{equation}\begin{split} \label{e.simpleg}
419: G_{dc}(\omega) =
420: G_0 \sum_l [J_l(\alpha/2)]^2 T^{eq}(\epsilon_F + l\hbar\omega) ,
421: \end{split}\end{equation}
422: %
423: where $G_0=2e^2/h$ and $T^{eq}(\epsilon)$ is the equilibrium transmission
424: function \cite{Note2}.
425: We have also defined the parameter
426: $\alpha=\alpha_L-\alpha_R=(U^{ac}_L-U^{ac}_R)/\hbar\omega$,
427: which measures the local intensity of the radiation \cite{Tien63}.
428: Equation (\ref{e.simpleg}) describes electrons incident from the
429: different sidebands (after having ``absorbed'' or ``emitted'' $l$ photons)
430: being transmitted elastically through the constriction, which mostly
431: determines $T^{eq}(\epsilon)$.
432: In reality, the electric field is nonzero only in the constriction
433: and thus the actual physical transitions must occur there.
434: Note that Eq.\ (\ref{e.simpleg}) should reproduce the
435: results for profile A only in the limit of small $\omega$
436: and $\alpha$. Still, it works surprisingly well for all the
437: cases presented below.
438:
439: \begin{figure}[!tb]
440: \includegraphics[width=0.95\linewidth,clip=]{fig2_al.eps}
441: \caption{(Color online) Top panel: Equilibrium transmission $T^{eq}$
442: and its decomposition into conduction channels $T_{1,2,3,4}$ for an Al dimer
443: contact. The position of $\epsilon_F$ is indicated by a vertical dotted line.
444: Lower panels: Zero-temperature photoconductance for several values of
445: $\alpha$ as a function of frequency $\omega$ using the voltage
446: profile A (a), profile B (b) and Eq.\ (\ref{e.simpleg}) (c).
447: In (b) the wavelengths $\lambda$ with a tick spacing of 400 nm are shown.
448: The range of visible light is indicated by vertical dotted lines.
449: }
450: \label{f.aluminum}
451: \end{figure}
452: %
453: Let us first discuss the results for the contact regime, which
454: is the subject of our main interest in this paper.
455: In Fig.\ \ref{f.aluminum} we show results for a dimer Al
456: contact, where $\epsilon_F$ lies in the $3p$ band.
457: The upper panel shows the transmission function $T^{eq}(\epsilon)$
458: and the lower panels show $G_{dc}(\omega)$ for the two voltage profiles
459: as well Eq.\ (\ref{e.simpleg}) for several values of $\alpha$.
460: In the absence of an ac voltage ($\omega=0$) the conductance
461: is close to $G_0$, and is dominated by three conduction channels
462: due to the contribution of $3s$ and $3p$ orbitals \cite{Scheer98}.
463: At finite $\omega$, the relative change
464: $\delta G_{dc}(\omega)=[G_{dc}(\omega)-G_{dc}(\omega=0)]/G_{dc}(\omega=0)$
465: is initially \emph{negative}, but can then rise to \emph{positive} values of
466: $\approx 10\%$ towards visible frequencies. This behavior is similar
467: for both profiles A and B, as well as for Eq.\ (\ref{e.simpleg}). From
468: the latter result, we can interpret our findings in the following
469: appealing way. For $\alpha<1$ only the first sidebands ($l=0,\pm 1$)
470: contribute to the transport. In this limit, according to Eq.\
471: (\ref{e.simpleg}), $\delta G_{dc}(\omega)$ measures, roughly speaking,
472: the ``second derivative'' of $T^{eq}(\epsilon)$ on the scale of
473: $\hbar\omega$ around $\epsilon_F$. Thus, for instance, the conductance
474: enhancement in the visible range follows from the transmission increase
475: for electrons promoted above $\epsilon_F$ (due to ``absorption'')
476: overcoming the corresponding decrease for electrons moved below
477: $\epsilon_F$ (due to ``emission'').
478:
479: Figure \ref{f.platinum} shows the corresponding results for Pt.
480: In the absence of radiation the conductance is close to $2.1G_0$
481: due to the contributions of mainly three conduction channels, which
482: originate from the $6s$ and $5d$ orbitals.
483: In this case, and in the contact regime in general for Pt, the
484: effect of the radiation is almost always a significant reduction
485: in conductance. This is understandable, since $\epsilon_F$ lies at
486: the edge of the $d$ band, and photon absorption leads to an energy
487: region where fewer open transmission channels are available and $T^{eq}$
488: is smaller. Note that for low $\omega$, the full results are again
489: well described by Eq.\ (\ref{e.simpleg}). Let us remark that we only
490: compute $G_{dc}(\omega)$ for low enough $\omega$ and $\alpha$, so that the
491: electric fields of the radiation remain reasonable
492: ($\lesssim 3\cdot 10^9$ V/m).
493:
494: \begin{figure}[!t]
495: \includegraphics[width=0.95\linewidth,clip=]{fig3_pt.eps}
496: \caption{ (Color online)
497: Same as Fig.\ \ref{f.aluminum}, but for Pt.
498: }
499: \label{f.platinum}
500: \end{figure}
501: %
502:
503: The results for Au are shown in Fig.\ \ref{f.gold}. As can be seen
504: in the upper panel, the conductance for $\omega=0$ is equal to
505: $1G_0$ with a single open channel arising from the contribution
506: of the $6s$ orbitals \cite{Scheer98}. Moreover, notice that
507: the transmission around $\epsilon_F$ is very flat. Due to this flatness,
508: for frequencies up to $\hbar\omega\approx 1$ eV the effect of radiation
509: is practically negligible. In the red part of the visible
510: range ($\hbar\omega\lesssim 2$ eV) we find that $\delta G_{dc}(\omega)>0$
511: up to a few percent, although this depends rather strongly on the
512: choice of the voltage profile. This increase in the conductance
513: is due to a contribution of the $5d$ bands located $2$ eV below
514: $\epsilon_F$, where the number of open transmission channels is
515: higher than at $\epsilon_F$. At higher frequencies
516: $\delta G_{dc}(\omega)<0$, as for Pt. We have also studied
517: Au contacts with atomic chains of varying length and the results
518: remain qualitatively similar, although in the case of profile B
519: the amplitude $\delta G_{dc}(\omega)<0$ becomes smaller as
520: the number of chain atoms increases.
521:
522: \begin{figure}[!t]
523: \includegraphics[width=0.95\linewidth,clip=]{fig4_au.eps}
524: \caption{ (Color online)
525: Same as Fig.\ \ref{f.aluminum}, but for Au.
526: }
527: \label{f.gold}
528: \end{figure}
529:
530: \begin{figure}[!t]
531: \includegraphics[width=0.95\linewidth,clip=]{fig5_al_t.eps}
532: \caption{ (Color online)
533: Tunneling limit for Al contacts.
534: Top left panel: zero-frequency conductance $G_{dc}(\omega=0)$
535: as a function of tip distance $D$. The vertical lines indicate
536: the distances where the examples with corresponding linestyles
537: in the other panels are computed. Top right panel: transmission
538: $T^{eq}(\epsilon)$ for the example distances. Lower panels:
539: the left-hand (a,d), central (b,e),
540: and right-hand (c,d) panels are for profile A, profile B, and
541: Eq.\ (\ref{e.simpleg}), respectively. The panels (a)-(c) are for
542: $\alpha=0.5$ and (d)-(f) for $\alpha=1.0$.
543: }
544: \label{f.altunnel}
545: \end{figure}
546:
547: When the distance $D$ between the tip atoms of the dimer contact
548: (Fig.\ \ref{f.lcr}) is increased, the tunneling regime is approached.
549: Here $G_{dc}(\omega)$ decreases exponentially with increasing $D$, but
550: $\delta G_{dc}(\omega)$ has a tendency to saturate. This is easy to
551: understand, because it may be shown that for very large $D$ the
552: magnitude of the conductance is approximately determined by the square
553: of the slowest-decaying hopping integral between the tip atoms, which
554: enters the conductance formula as a prefactor. The form of $\delta
555: G_{dc}(\omega)$ can, however, be very different from the contact
556: regime.
557:
558: An example of the tunneling regime results for Al is presented in
559: Fig.\ \ref{f.altunnel}. The top panels illustrate the exponential
560: decay of $G_{dc}(\omega=0)$ and $T^{eq}(\epsilon)$ with $D$, while in
561: the lower panels the quantity $G_{dc}(\omega)/G_{dc}(\omega=0)$ is
562: shown for two values of $\alpha$ and for several distances $D$. The
563: results again look quite similar for the two voltage profiles as well
564: as for Eq.\ (\ref{e.simpleg}). The quantity $\delta G_{dc}$ can obtain
565: both positive and negative values, and its saturation with $D$ is
566: clearly visible. For Pt (see Fig.\ \ref{f.pttunnel}) we find that
567: $G_{dc}(\omega)$ is otherwise flat, but there is a sharp resonance at
568: $\hbar\omega\approx1$ eV where $\delta G_{dc}(\omega)$ can take on
569: \emph{positive} values of up to a few hundred percent. This is due to
570: resonant transmission through a level formed by the $d$ orbitals of
571: the tip atoms, which can be seen also in the $T^{eq}(\epsilon)$
572: curves. In the case of Au there exists a rather similar, but broader
573: positive peak covering the visible range (see Fig.\
574: \ref{f.autunnel}). For each metal, we only consider large enough $D$
575: to see the saturation of $\delta G_{dc}(\omega)$. Indeed, the
576: tight-binding parametrization we are using is based on bulk
577: calculations \cite{Papa} and is probably not good for very large
578: interatomic distances. Furthermore, the charge neutrality shifts
579: mentioned in Sec.\ \ref{s.model} are strongest for the tip atoms.
580: Therefore the peaks in $G_{dc}(\omega)$ observed for Pt and Au, for
581: example, should be taken with some reservations. For very large $D$,
582: we would also not expect to see such a good agreement between the
583: results for profiles A and B, because the tunneling conductance for
584: profile B depends on the local densities of states of the tip atoms in
585: a way that cannot be written in the form (\ref{e.simpleg}).
586: Nevertheless, our results serve as good illustrations of the different
587: phenomena that may potentially arise in the tunneling regime.
588:
589: \begin{figure}[!t]
590: \includegraphics[width=0.95\linewidth,clip=]{fig6_pt_t.eps}
591: \caption{ (Color online)
592: Same as Fig.\ \ref{f.altunnel}, but for Pt.}
593: \label{f.pttunnel}
594: \end{figure}
595:
596: \begin{figure}[!t]
597: \includegraphics[width=0.95\linewidth,clip=]{fig7_au_t.eps}
598: \caption{ (Color online)
599: Same as Fig.\ \ref{f.altunnel}, but for Au.}
600: \label{f.autunnel}
601: \end{figure}
602:
603: Above we have only shown some example results for very idealized
604: symmetric geometries. In general, the signs and magnitudes of $\delta
605: G_{dc}(\omega)$ depend sensitively on details of the atomic structure,
606: as does $T^{eq}(\epsilon)$. For a more complete analysis one should
607: therefore study also larger contacts and carry out a statistical
608: exploration of geometrical variations along the lines of Ref.
609: \onlinecite{Dreher05}. Based on Eq. (\ref{e.simpleg}), we can still
610: expect that in the limit of several-atoms-wide contacts the relative
611: effect of the ac voltage gradually becomes smaller as
612: $T^{eq}(\epsilon)$ becomes smoother. A direct comparison with the
613: ongoing experiments \cite{Guhr06} will be postponed for later.
614:
615: \section{Discussion and conclusions}\label{s.discussion}
616: Some of the assumptions of our model and the effects not taken into
617: account are worth discussing. First, we assume a flat potential in the
618: leads, which requires a complete screening of the electric field. This
619: is well satisfied in metals for frequencies $\omega$ much below the
620: plasma frequency but, as $\omega$ begins to approach the visible
621: range, the screening is weakened. On the other hand, as we have
622: mentioned, the screening in the central region is not treated
623: self-consistently. One of the main concerns here is that local
624: surface-plasmon modes in small geometries tend to have their
625: frequencies close to the visible range, and their excitation can lead
626: to huge field-enhancement effects \cite{Grafstrom02}. Although this is
627: not a problem for our model, one should bear in mind that $\alpha$ is
628: not simply related to the field intensity away from the contact and,
629: following the spirit of the experiments in superconducting contacts
630: \cite{Chauvin06}, it must be understood as an adjustable parameter.
631: Also, at visible frequencies ``multiphoton'' processes can already
632: cause a photoelectric effect, \emph{i.e.}, excite electrons above the
633: vacuum level, which typically lies $4$--$6$ eV above $\epsilon_F$.
634:
635: Heating accompanies any possible effect arising from absorption of
636: light, which in metals is specially pronounced in the optical range
637: due to the onset of interband transitions. While we may expect that
638: the effect of temperature is just to broaden our results, in practice
639: thermal expansion can play an important role. This is well documented
640: in the STM context \cite{Grafstrom02}, where expansion typically
641: brings the tip closer to the sample, thus reducing the tunneling gap
642: width. This results in a strong enhancement of the tunneling current.
643: In the case of atomic wires (\emph{i.e.}, in the contact regime), it
644: is not obvious in which sense and to what degree thermal expansion
645: affects the conductance. Assuming that the expansion simply mimics a
646: mechanical closing process in the STM or break-junction experiments,
647: it can lead to either an increase or a decrease of the conductance,
648: depending on the metal. For instance, for Al contacts, which exhibit
649: raising plateaus upon stretching \cite{Cuevas1998}, one would expect a
650: decrease of the conductance due to thermal expansion, as opposed to
651: the effect of the electronic structure in the visible range (see Fig.\
652: \ref{f.aluminum}). In this sense, our predictions can be valuable for
653: distinguishing in an experiment (such as Ref.\ \onlinecite{Guhr06})
654: between the contributions of the different effects to the
655: photoconductance.
656:
657: In conclusion, we have modeled electronic transport in atomic point
658: contacts subjected to external electromagnetic radiation. The
659: radiation has been described by an ac voltage over the contact.
660: Within a non-equilibrium Green function method, we have derived a
661: formula for the dc current in the presence of such an ac drive. Using
662: a tight-binding model, we have applied the method for describing
663: atomic-sized contacts of Al, Pt, and Au, and have found that the
664: qualitative modification of the dc conductance by the ac voltage can
665: be predicted from the equilibrium transmission function. Depending on
666: the metal, the detailed structure of the contact, and the external
667: frequency, the effect can be either an increase or a decrease in the
668: conductance. At present, experiments are under way to test these
669: predictions \cite{Guhr06}.%\\
670:
671: \acknowledgments
672:
673: We acknowledge useful discussions with D. Schmidt, E. Scheer,
674: P. Leiderer, R. H. M. Smit, F. Pauly, S. Wohlthat, and M. H\"afner.
675: This work was financially supported by the Helmholtz Gemeinschaft
676: (Contract No.\ VH-NG-029) and by the DFG within the Center for Functional
677: Nanostructures.
678:
679:
680: \begin{thebibliography}{99}
681:
682: \bibitem{Tucker85}
683: J. C. Tucker and M. J. Feldman,
684: %\emph{Quantum detection at millimeter wavelengths},
685: Rev. Mod. Phys. \textbf{57}, 1055 (1985).
686:
687: \bibitem{Tien63}
688: P. K. Tien and J. P. Gordon,
689: %\emph{Multiphoton Process Observed in the Interaction of
690: % Microwave Fields with the Tunneling between
691: % Superconductor Films},
692: Phys. Rev. \textbf{129}, 647 (1963).
693:
694: \bibitem{Cuevas02}
695: J. C. Cuevas,
696: J. Heurich, A. Mart\'{\i}n-Rodero, A. Levy Yeyati, and G. Sch\"on,
697: %\emph{Subharmonic Shapiro Steps and Assisted Tunneling in Superconducting
698: %Point Contacts}
699: Phys. Rev. Lett. \textbf{88}, 157001 (2002).
700:
701: \bibitem{Chauvin06}
702: M. Chauvin, P. vom Stein, H. Pothier, P. Joyez, M. E. Huber, D. Esteve,
703: and C. Urbina,
704: %\emph{Superconducting Atomic Contacts under Microwave Irradiation}
705: Phys. Rev. Lett. \textbf{97}, 067006 (2006).
706:
707: \bibitem{Grafstrom02}
708: S. Grafstr\"om,
709: %\emph{Photoassisted scanning tunneling microscopy},
710: J. Appl. Phys. \textbf{91}, 1717 (2002).
711:
712: \bibitem{Voelcker91}
713: M. V\"olcker, W. Krieger, and H. Walther,
714: %\emph{Laser-Driven Scanning Tunneling Microscope},
715: Phys. Rev. Lett. \textbf{66}, 1717 (1991).
716:
717: \bibitem{Levy91}
718: A. Levy Yeyati and F. Flores,
719: %\emph{Photocurrent effects in the scanning tunneling microscope},
720: Phys. Rev. B \textbf{44}, 9020 (1991).
721:
722: \bibitem{Platero04}
723: G. Platero and R. Aguado,
724: %\emph{Photon-assisted transport in semiconductor nanostructures},
725: Phys. Rep. \textbf{395}, 1 (2004).
726:
727: \bibitem{Jauho94}
728: A.-P. Jauho, N. S. Wingreen, and Y. Meir,
729: %\emph{Time-dependent transport in interacting and noninteracting
730: %resonant-tunneling systems},
731: Phys. Rev. B \textbf{50}, 5528 (1994).
732:
733: \bibitem{Buttiker96}
734: M. B\"uttiker and T. Christen,
735: \emph{Mesoscopic Electron Transport},
736: eds. L. L. Sohn, L. P. Kouwenhoven and G. Sch\"on,
737: Vol 345 of NATO ASI. Series E: Applied Science
738: (Kluwer, Dordrecht, 1997), p.259.
739:
740: \bibitem{Pedersen98}
741: M. H. Pedersen and M. B\"uttiker,
742: %\emph{Scattering theory of photon-assisted electron transport},
743: Phys. Rev. B \textbf{58}, 12993 (1998).
744:
745: \bibitem{Wang99}
746: B. Wang, J. Wang, and H. Guo,
747: %\emph{Current Partition: A Nonequilibrium Green's Function Approach},
748: Phys. Rev. Lett. \textbf{82}, 398 (1999).
749:
750: \bibitem{Kohler04}
751: S. Kohler, J. Lehmann, and P. H\"anggi,
752: %\emph{Driven quantum transport on the nanoscale},
753: Phys. Rep. \textbf{406}, 379 (2005).
754:
755: \bibitem{Zheng00}
756: W. Zheng, W. Wei, J. Wang, and H. Guo,
757: %\emph{ac responce of an atomic tunnel junction},
758: Phys .Rev. B \textbf{61}, 13121 (2000).
759:
760: \bibitem{Galperin05}
761: M. Galperin and A. Nitzan,
762: %\emph{Current induced light and light induced current
763: % in molecular tunneling junctions},
764: Phys. Rev. Lett. \textbf{95}, 206802 (2005).
765:
766: \bibitem{Buker02}
767: J. Buker and G. Kirczenow,
768: %\emph{Theoretical study of photon emission from molecular wires},
769: Phys. Rev. B \textbf{66}, 245306 (2002).
770:
771: \bibitem{Wu05}
772: J. Wu and B. Wang and J. Wang and H. Guo,
773: %\emph{Giant enhancement of dynamic conductance in molecular devices},
774: Phys. Rev. B \textbf{72}, 195324 (2005).
775:
776: \bibitem{Agrait03}
777: For a recent review see N. Agra\"{\i}t, A. Levy Yeyati,
778: and J.M. van Ruitenbeek,
779: %\emph{Quantum properties of atomic-sized conductors}
780: Phys. Rep. \textbf{377}, 81 (2003).
781:
782: \bibitem{Guhr06}
783: D. Guhr, D. Rettinger, J. Boneberg, A. Erbe, P. Leiderer,
784: and E. Scheer,
785: %\emph{Influence of chopped laser light onto the electronic
786: %transport through atomic-sized contacts},
787: %submitted to Journal of Microscopy
788: cond-mat/0612117.
789:
790: \bibitem{Scheer98}
791: E. Scheer, N. Agra\"it, J. C. Cuevas, A. L. Yeyati, B. Ludolph,
792: A. Martin-Rodero, G. R. Bollinger, J. M. van Ruitenbeek, C. Urbina,
793: %\emph{The signature of chemical valence in the electrical
794: %conduction through a single-atom contact},
795: Nature \textbf{394}, 154 (1998).
796:
797: \bibitem{Datta92}
798: S. Datta and M. P. Anantram
799: %\emph{Steady-state transport in mesoscopic systems illuminated by
800: %alternating fields},
801: Phys. Rev. B \textbf{45}, 13761 (1992);
802: %\bibitem{Anantram95}
803: M. P. Anantram and S. Datta,
804: %\emph{Effect of phase breaking on the ac responce of mesoscopic systems},
805: Phys .Rev. B \textbf{51}, 7632 (1995).
806:
807: \bibitem{Brandes97}
808: T. Brandes,
809: %\emph{Truncation method for Green's functions in time-dependent fields},
810: Phys. Rev. B \textbf{56}, 1213 (1997).
811:
812: \bibitem{Viljas05}
813: J. K. Viljas, J. C. Cuevas, F. Pauly, and M. H\"afner,
814: %\emph{Electron-vibration interaction in atomic gold wires},
815: Phys. Rev. B \textbf{72}, 245415 (2005).
816:
817: \bibitem{Papa}
818: M. J. Mehl and D. A. Papaconstantopoulos,
819: %\emph{Applications of a tight-binding total-energy method for
820: %transition and noble metals: Elastic constants, vacancies,
821: %and surfaces of monatomic metals},
822: Phys. Rev. B \textbf{54}, 4519 (1996);
823: D. A. Papaconstantopoulos and M. J. Mehl,
824: %\emph{The Slater-Koster tight-binding method: a computationally
825: %efficient and accurate approach},
826: J. Phys. Condens. Matter \textbf{15}, R413 (2003).
827: %http://cst-www.nrl.navy.mil/bind/.
828:
829: \bibitem{Dreher05}
830: M. Dreher, F. Pauly, J. Heurich, J. C. Cuevas, E. Scheer, and P. Nielaba,
831: Phys. Rev. B \textbf{72}, 075435 (2005);
832: F. Pauly, M. Dreher, J. K. Viljas, M. H\"afner, J. C. Cuevas, and P. Nielaba,
833: Phys. Rev. B \textbf{74}, 235106 (2006).
834: %cond-mat/0607129.
835:
836: \bibitem{Note1}
837: This choice for representing the potential profile in a nonorthogonal
838: basis is discussed for example in
839: M. Brandbyge, N. Kobayashi, and M. Tsukada,
840: Phys. Rev. B \textbf{60}, 17064 (1999).
841:
842: \bibitem{Datta95}
843: S. Datta, \emph{Electronic Transport in Mesoscopic Systems}
844: (Cambridge University Press, Cambridge, 1995).
845:
846: \bibitem{Note2}
847: This is given in the usual way by
848: %$T^{eq}=\Tr[\omat{G}_{CC}^r\omat{\Gamma}_R\omat{G}_{CC}^a\omat{\Gamma}_L]$
849: $T^{eq}=\Tr[\omat{t}\omat{t}^\dagger]$,
850: where $\omat{t}=(\omat{\Gamma}_L^{eq})^{1/2}\omat{G}^{r,eq}_{CC}
851: (\omat{\Gamma}_R^{eq})^{1/2}$
852: \cite{Datta95,Viljas05,Dreher05}, and the eigenchannel
853: transmissions are defined as the eigenvalues of $\omat{t}\omat{t}^\dagger$.
854: Due to spin degeneracy, the matrices are now in a basis
855: with only a single spin species.
856:
857: \bibitem{Cuevas1998}
858: J. C. Cuevas, A. Levy Yeyati, A. Mart\'{i}n-Rodero, G. R.
859: Bollinger, C. Untiedt, and N. Agra\"{\i}t,
860: Phys. Rev. Lett. \textbf{81}, 2990 (1998).
861:
862:
863: \end{thebibliography}
864:
865: \end{document}
866: