1: \documentclass[twocolumn,prl,showpacs]{revtex4}
2: %\documentclass[preprint,prl,showpacs]{revtex4}
3: \usepackage{psfrag}
4: \usepackage{graphicx}
5: \usepackage{amsmath}
6: \usepackage{amssymb}
7:
8: \begin{document}
9: \title{Conduction Mechanism in a Molecular Hydrogen Contact}
10: \author{K. S. Thygesen}
11: \affiliation{Center for Atomic-scale Materials Physics, \\
12: Department of Physics, Technical University of Denmark, DK - 2800
13: Kgs. Lyngby, Denmark}
14: \author{K. W. Jacobsen}
15: \affiliation{Center for Atomic-scale Materials Physics, \\
16: Department of Physics, Technical University of Denmark, DK - 2800
17: Kgs. Lyngby, Denmark}
18: \date{\today}
19:
20: \begin{abstract}
21: We present first principles calculations for the conductance of a
22: hydrogen molecule bridging a pair of Pt electrodes. The transmission
23: function has a wide plateau with $T\approx 1$ which extends across
24: the Fermi level and indicates the existence of a single, robust
25: conductance channel with nearly perfect transmission. Through a
26: detailed Wannier function analysis we show that the $\text{H}_2$
27: bonding state is not involved in the transport and that the plateau
28: forms due to strong hybridization between the $\text{H}_2$
29: anti-bonding state and states on the adjacent Pt atoms. The Wannier
30: functions furthermore allow us to derive a resonant-level model for
31: the system with all parameters determined from the fully
32: self-consistent Kohn-Sham Hamiltonian.
33: \end{abstract}
34:
35: \pacs{73.63.Rt,73.20.Hb,73.40.Gk} \maketitle The study of electron
36: transport through single molecules has evolved during the last decade
37: as new experimental techniques have made it possible to produce
38: atomic-scale contacts with a few or even a single molecule suspended
39: between macroscopic
40: electrodes~\cite{joachim95,reed97,reichert_weber02,tao03}. At the
41: same time theoretical efforts have been made to describe and
42: understand the experiments from first
43: principles~\cite{taylor02,damle01,diventra02}. The connection between
44: experiment and theory, however, has been complicated by the crucial
45: but in practice uncontrollable atomistic details of the contact region
46: between the molecule and the leads. While the majority of previously
47: investigated molecules have shown a conductance much lower than the
48: quantum unit, $G_0=2e^2/h$, Smit.~\emph{et al.} recently measured a
49: conductance close to $1G_0$ for a hydrogen molecule bridging a pair of
50: Pt electrodes~\cite{smit02}. The result immediately raises the
51: question: how can a hydrogen molecule which has a closed shell
52: configuration and a large energy gap be conducting? Despite the
53: simplicity of the system, there are still considerable disagreements
54: among the reported calculations for the conductance of the hydrogen
55: bridge. Quantitatively, values of
56: $0.9G_0$~\cite{smit02,cuevas_heurich03} and
57: $(0.2-0.5)G_0$~\cite{garcia_palacios04} have been published by
58: different group using similar methods. Perhaps even more importantly,
59: the physical explanations for the obtained results are very different.
60: Indeed, both the bonding~\cite{cuevas_heurich03,garcia_palacios04} as
61: well as the anti-bonding~\cite{smit02} state of the $\text{H}_2$
62: molecule have been proposed as the current-carrying state.
63:
64: In this letter we present conductance calculations based on Density
65: Functional Theory (DFT) showing that a hydrogen molecule bridging a
66: pair of Pt contacts can have a conductance close to $1G_0$ and we explain the physical mechanism
67: behind this result. The transmission function is found to have a characteristic
68: plateau with $T\approx 1$ in an energy window of 4 eV around the Fermi
69: level, indicating the existence of a single, very robust conductance
70: channel with nearly perfect transmission.
71: By performing a Wannier function (WF) analysis we can
72: directly study the transmission through the $\text{H}_2$ bonding and
73: anti-bonding states separately.
74: The results clearly demonstrate that the bonding
75: state takes almost no part in the transport and that the plateau is a
76: result of a strong hybridization between the $\text{H}_2$ anti-bonding
77: state and a combination of $d$- and $s$-like orbitals located on
78: the neighboring Pt atoms. The analysis furthermore
79: allows us to determine characteristic model parameters from first
80: principles which in turn provides a very simple description of the system.
81:
82: To describe the molecular contact we use the supercell shown in the
83: inset of Fig.~\ref{fig1}. It contains the $\text{H}_2$ molecule
84: anchored between two 4-atom Pt pyramids which again are attached to
85: Pt(111) surfaces~\cite{calc}. We calculate the conductance of the
86: relaxed structures assuming that the electrons move phase coherently
87: through the contact and are influenced only by the self-consistent
88: Kohn-Sham potential. In this case the conductance is given by
89: $G=G_0T(\varepsilon_F)$, where $T(\varepsilon_F)$ is the transmission
90: function at the Fermi level~\cite{datta_book}. The transmission
91: function is found using the Green's function method described in
92: Refs.~\cite{thygesen_bollinger03,datta02,brandbyge02}. In this
93: approach the system is divided into three regions: A left lead, $L$, a
94: right lead, $R$, and a central region, $C$. The leads are assumed to
95: be periodic such that all scattering takes place in $C$. In our case
96: $C$ coincides with the supercell of the DFT calculation and the leads
97: are bulk Pt(fcc) described in a supercell containing $3\times 3$ atoms
98: in the transverse plane to match the central region at the interfaces.
99: The transmission function is then given by the
100: formula~\cite{meirwingreen}
101: \begin{equation}\label{eq.condformula}
102: T(\varepsilon)=\text{Tr}\big [G^r_C(\varepsilon) \Gamma_L(\varepsilon) G^a_C(\varepsilon)\Gamma_R(\varepsilon)\big],
103: \end{equation}
104: where $G_C^r$ is the retarded Green's function of the scattering region
105: \begin{equation}
106: G^r_C(\varepsilon)=[(\varepsilon+i\eta^{+})S-\Sigma_L(\varepsilon)-\Sigma_R(\varepsilon)-H_C]^{-1}.
107: \end{equation}
108: Here $H_C$ and $S$ are the Hamiltonian and overlap
109: matrices of the central region, $\eta^+$ is a positive infinitesimal and
110: $\Sigma_{\alpha}$ is the self-energy from lead $\alpha$. The coupling strength of lead
111: $\alpha$ is given by
112: $\Gamma_{\alpha}=i(\Sigma_{\alpha}-\Sigma_{\alpha}^{\dagger})$.
113:
114: We use partly occupied Wannier functions, $\{\phi_{n\alpha}\}$, (see
115: below) as basis functions in each of the three regions
116: ($\alpha=L,R,C$). Due to the limited size of the
117: supercell in the plane perpendicular to the transport direction the
118: conductance should be calculated as an integral over the Brillouin
119: zone in the corresponding plane. We thus form the Bloch states
120: $\psi_{\bold k_{\perp}n\alpha}(\bold r)=\sum_{\bold
121: R_{\perp}}e^{i\bold k_{\perp}\cdot \bold
122: R_{\perp}}\phi_{n\alpha}(\bold r-\bold R_{\perp})$, where $\bold
123: R_{\perp}$ runs over supercells in the transverse plane. For each
124: $\bold k_{\perp}$ we obtain a Hamiltonian matrix $H(\bold
125: k_{\perp})_{n\alpha,m\beta}=\langle \bold k_{\perp} n\alpha|H|\bold
126: k_{\perp}m\beta\rangle$ which in turn leads to a conductance $G(\bold
127: k_{\perp})$ through Eq.~(\ref{eq.condformula}). The integrated
128: conductance can then be approximated by the finite sum $\sum_{\bold
129: k_{\perp}}w(\bold k_{\perp})G(\bold k_{\perp})$, where $w(\bold
130: k_{\perp})$ are appropriate weight factors.
131:
132: \begin{figure}[!b]
133: \includegraphics[width=1.0\linewidth]{fig1.eps}
134: \caption[cap.wavefct]{\label{fig1} Calculated transmission for the
135: molecular hydrogen contact shown in the inset. For comparison both the $\bold
136: k$-point sampled transmission and the $\Gamma$-point transmission are
137: shown. The $\Gamma$-point transmission has more structure, however, the qualitative
138: features of the curves are essentially the same. The wide
139: plateau with $T\approx 1$ extending across the Fermi level indicates
140: a single, very robust conductance channel with nearly perfect transmission.} \end{figure}
141:
142: We focus on a single, fully relaxed contact characterized by the bond
143: lengths $d_{\text{H-H}}=1.0$~\AA~and $d_{\text{Pt-H}}=1.76$~\AA.
144: The vibrational modes of the hydrogen molecule in this configuration
145: are in fair agreement with new experimental results~\cite{djukic04}.
146: In Fig.~\ref{fig1} we show the transmission function calculated using
147: 8 irreducible $\bold k$-points to sample the transverse BZ. The same
148: curve calculated within the widely used $\Gamma$-point approximation
149: is shown for comparison. The two curves have essentially the same
150: features, however, the $\Gamma$-point curve has more structure. This
151: is because the $\bold k$-point sampling provides the
152: correct smearing of the electronic structure in the leads which effectively
153: washes out features related to single points in the
154: transverse plane of the lead Brillouin zone. An interesting feature of
155: the transmission function is the wide plateau with $T\approx 1$
156: extending across the Fermi level. We shall refer to this plateau as the $1G_0$-plateau.
157:
158: \begin{figure}[!b]
159: \includegraphics[width=0.77\linewidth,angle=270]{fig2.ps}
160: \caption[cap.wavefct]{\label{fig2} Calculated transmission for the
161: structure shown in the inset of Fig.~\ref{fig1} when all coupling to
162: the bonding, respectively, the anti-bonding $\text{H}_2$ state has
163: been cut. The full transmission has been repeated for
164: comparison. The narrow peak around $-7$~eV is clearly due to the
165: bonding state, while the peak at $-2$~eV and the wide plateau around
166: the Fermi level almost
167: exclusively involve the anti-bonding state.}
168: \end{figure}
169:
170: To gain insight into the formation of the $1G_0$-plateau we perform a
171: Wannier function analysis. The WFs are defined as linear combinations
172: of the Kohn-Sham eigenstates with the expansion coefficients chosen to
173: make the WFs orthogonal and maximally localized. By including selected
174: unoccupied eigenstates in this construction we can obtain good
175: localization properties of the WFs also for metallic
176: systems~\cite{souza01, partlyoccwfs}. We stress that the minimal WF
177: basis set retains the accuracy of the plane wave DFT calculation since
178: the WFs by construction span the eigenstates below a certain energy
179: which has been set to 4~eV above the Fermi level in the present
180: calculation. The transformation results in the following set of WFs:
181: For each Pt we obtain 5 $d$-orbitals centered at the atom and a single
182: $\sigma$-orbital located at an interstitial site. For each hydrogen we
183: find an $s$-orbital, $|i\rangle$ ($i=1,2$), which is slightly
184: elongated towards the contacting Pt atom. We proceed by transforming
185: the hydrogen $s$-orbitals into bonding and anti-bonding combinations
186: $|b\rangle=(|1\rangle+|2\rangle)/\sqrt{2}$ and
187: $|a\rangle=(|1\rangle-|2\rangle)/\sqrt{2}$. $|a\rangle$ and
188: $|b\rangle$ are the only states with significant weight on the
189: molecule and provide two conductance channels well separated in
190: energy. The on-site energies are $\langle b|H|b\rangle=-6.4$~eV and
191: $\langle a|H|a\rangle=0.1$~eV relative to the Fermi level of the
192: metal. By cutting all coupling matrix elements involving the bonding,
193: respectively, the anti-bonding state we can directly test their
194: individual contributions to the total conductance when interference is
195: neglected. The result is shown in Fig.~\ref{fig2}. The narrow peak
196: just below $-7$~eV is completely gone when the bonding state is
197: removed but is not affected by the absence of the anti-bonding state.
198: The peak is thus clearly due to transmission through the bonding
199: channel which is in good agreement with the calculated on-site energy
200: of $|b\rangle$. In the energy regime -6 -- -4 eV both the bonding and
201: anti-bonding states contribute to the transmission. For energies above
202: -3 eV the removal of the bonding state has little effect on the
203: transmission which must therefore be ascribed to the anti-bonding
204: state. A small exception to this is the narrow peak at -1 eV which is
205: caused by hybridization of the bonding state with Pt
206: $d_{z^2}$-orbitals on the contacting atoms. Overall we can conclude
207: that the peak at -2 eV and the $1G_0$-plateau which determines the
208: conductance are due to transmission through the anti-bonding state.
209:
210: \begin{figure}[!b]
211: \includegraphics[width=0.62\linewidth]{fig3.ps}
212: \caption[cap.wavefct]{\label{fig3} Contour plots of the orbitals
213: determining the transport properties of the hydrogen contact: the
214: $\text{H}_2$ anti-bonding state, $|a\rangle$, and the corresponding
215: left and right group orbitals, $|g_L\rangle,|g_R\rangle$. The left
216: (right) group orbital has been constructed by applying the
217: DFT Hamiltonian to $|a\rangle$ and then projecting onto the
218: Wannier functions of the left (right) part of the contact.}
219: \end{figure}
220:
221: The fact that the bonding state takes almost no part in the
222: transmission around the Fermi level allows us to describe the contact
223: by a resonant level model~\cite{newns69} with all parameters
224: determined from first principles. In the resonant level model we
225: consider a single level, $|a\rangle$, of energy $\varepsilon_a=\langle
226: a|H|a\rangle$ coupled to infinite leads via the matrix elements
227: $t_{k\alpha}=\langle k\alpha|H|a\rangle$, where $\{|k\alpha\rangle\}$
228: is a basis of lead $\alpha$. The model has served as starting point
229: for many more advanced studies such as shot noise, electron-electron
230: and electron-phonon interactions in resonant tunneling
231: systems~\cite{thielmann03,ng88,hyldgaard94}. A particularly useful
232: formulation of the model can be obtained if we introduce the group
233: orbital of lead $\alpha$ by
234: $|g_{\alpha}\rangle=c_{\alpha}P_{\alpha}H|a\rangle$, where
235: $P_{\alpha}$ is the orthogonal projection onto lead $\alpha$ and
236: $c_{\alpha}$ is a normalization constant. By supplementing the group
237: orbital by orthonormal states $\{|\tilde k\alpha\rangle \}$ we obtain
238: a new basis with the key property $\langle \tilde
239: k\alpha|H|a\rangle=0$ for all $\tilde k$. The level is thus coupled to
240: the lead via the group orbital only. Since the contact is symmetric
241: $\langle g_L|H|a\rangle=\langle g_R|H|a\rangle\equiv V$. The
242: imaginary part of the level self-energy, $\Delta_a$, is directly
243: related to the DOS of the group orbitals calculated with $V=0$:
244: $\Delta_a=\pi|V|^2(\rho^0_L+\rho^0_R)=(\Gamma_L+\Gamma_R)/4$. The
245: real part of the self-energy is the Hilbert transform of $\Delta_a$.
246: For a symmetric contact we have $\rho^0_L=\rho^0_R\equiv\rho^0_g$ and
247: the transmission in Eq.~(\ref{eq.condformula}) takes the simple form
248: \begin{equation}\label{simpletrans}
249: T(\varepsilon)=2\pi^2|V|^2\rho^0_g(\varepsilon)\rho_a(\varepsilon).
250: \end{equation}
251: Since $\rho_a$ can be obtained from $\Sigma_a$ and $\varepsilon_a$
252: this expression shows that the transmission is determined by the three
253: quantities $\rho^0_g$, $V$ and $\varepsilon_a$.
254:
255: \begin{figure}[!b]
256: \includegraphics[width=0.8\linewidth,angle=270]{fig4.ps}
257: \caption[cap.wavefct]{\label{fig4} Upper panel shows the transmission
258: together with the projected density of states (DOS) for the
259: $\text{H}_2$ anti-bonding state and the corresponding left group
260: orbital.
261: The lower panel shows the same quantities obtained from the
262: single-level model when
263: $\rho_g^0$ is approximated by a semi-elliptical band and we use the
264: coupling, $V$, and on-site energy, $\varepsilon_a$, from the first
265: principles calculation.}
266: \end{figure}
267:
268:
269: By applying the DFT Hamiltonian to $|a\rangle$ in the WF basis we
270: construct the group orbitals of the $\text{H}_2$ anti-bonding state.
271: Contour plots of the orbitals are shown in Fig.~\ref{fig3}. The group
272: orbital is mainly composed of the $d_{z^2}$-orbital of the apex Pt
273: atom and three $\sigma$-orbitals centered within the Pt pyramid. We
274: calculate $\rho^0_g$ for the uncoupled system by cutting all coupling
275: matrix elements to $|a\rangle$. The result is shown in the upper
276: panel of Fig.~\ref{fig4} together with $\rho_a$ and the transmission
277: function. The pronounced peak at -1 eV is due to the
278: $d_{z^2}$-orbitals on the apex Pt atoms.
279:
280: If we neglect the narrow peak at $-1$~eV, $\rho^0_g$ can be described
281: by a semi-elliptical band on top of a flat background, see lower panel
282: of Fig.~\ref{fig4}. The coupling and level energy can be directly
283: read off the Hamiltonian matrix and we find $V=1.9$~eV and
284: $\varepsilon_a=0.1$~eV relative to the Fermi level. It should be
285: noticed that the coupling which is relevant for the adsorption of the
286: hydrogen molecule to the contact is $\sqrt{2}V=2.7$~eV since the level
287: is coupled by $V$ to \emph{both} leads. From these parameters we can
288: determine the Green's function for the level which in turn yields
289: $\rho_a$ and $T$. The result is summarized in the lower panel of
290: Fig.~\ref{fig4}. Based on the good agreement with the first principles
291: results we conclude that the simple model indeed gives a realistic
292: description of the system. It is then clear that the peak at $-2$~eV
293: represents the bonding combination between $|a\rangle$ and the Pt band
294: and that the $1G_0$-plateau forms because: (i) $\varepsilon_a$ lies
295: close to the Fermi level and well inside the relevant Pt band as
296: defined by the group orbital. (ii) the width of the renormalized level
297: ($\Delta_a$) is comparable to the band width, i.e. the limit of
298: strong chemisorption~\cite{newns69}.
299:
300: The crucial point in the proposed mechanism is the strong
301: hybridization of the $\text{H}_2$ anti-bonding state with the Pt bands
302: around the Fermi level. This picture agrees well with the conventional
303: understanding of hydrogen dissociation on simple and transition metal
304: surfaces which has been established on the basis of DFT
305: calculations~\cite{hammer95,norskov_houmoller81}. The
306: bonding and antibonding states of a hydrogen molecule at a simple
307: metal surface are broadened and furthermore shifted down due to the
308: hybridization with the metal $s$- and $p$-states. During the
309: dissociation process the antibonding resonance crosses the Fermi level
310: and becomes gradually filled with the result that the
311: hydrogen-hydrogen bond is weakened. For the transition metal the same
312: general picture applies, but the hybridization with the $d$-states
313: further affects the antibonding resonance. The fact that the
314: antibonding state in the calculations for the bridging hydrogen
315: molecule between Pt contacts is close to the Fermi level is thus an
316: indication that the hydrogen-hydrogen bond is weakened by the coupling
317: to the metal in agreement with the resulting increase of the
318: hydrogen-hydrogen bond length. The values we find for the positions of
319: the bonding and antibonding molecular levels, $\varepsilon_b= -6.4$~eV, $\varepsilon_a=
320: 0.1$~eV are in fact quite close to the ones used by Hammer and
321: N{\o}rskov~\cite{hammer95} ($\varepsilon_b= -7$~eV, $\varepsilon_a=1$~eV) to describe hydrogen
322: in the dissociative transition state on metals. This is in clear
323: contrast to the studies by Cuevas \emph{et
324: al.}~\cite{cuevas_heurich03} and Garc{\'i}a~\emph{et
325: al.}~\cite{garcia_palacios04} who consider the hydrogen molecule in
326: the bridging position to have almost the same bond legnth as the free
327: molecule and who report very large bonding-antibonding splittings of
328: 23-24 eV which even exceed the DFT-PW91 value of 10.4 eV for a free
329: molecule.
330:
331: In summary, we have presented first principles conductance
332: calculations showing that a hydrogen molecule suspended between Pt
333: contacts can have a conductance close to $1G_0$. Through a detailed
334: Wannier function analysis we have identified the conduction mechanism
335: as being due to a strong hybridization between the $\text{H}_2$
336: anti-bonding state and certain Pt bands. A resonant level model with all
337: parameters determined from the self-consistent DFT Hamiltonian was
338: shown to account for the important features of the first principles
339: transmission function.
340:
341: We would like to thank J. van Ruitenbeek and D. Djukic for many inspiring
342: discussions. We acknowledge support
343: from the Danish Center for Scientific Computing through Grant No.
344: HDW-1101-05.
345:
346: %%%%%%% References
347:
348: \bibliographystyle{apsrev}
349: \begin{thebibliography}{24}
350:
351:
352: \bibitem{joachim95} C.~Joachim, J.~K. Gimzewski, R.~R. Schlittler and
353: C.~Chavy Phys. Rev. Lett. {\bf 74}, 2102 (1995).
354:
355: \bibitem{reed97} M.~A. Reed, C.~Zhou, C.~J. Muller, T.~P. Burgin and
356: J.~M. Tour Science {\bf 278}, 252 (1997).
357:
358: \bibitem{reichert_weber02} J.~Reichert, R. Ochs, D. Beckman, H.~B.
359: Weber, M. Mayor and H.~v. L{\"ohneysen} Phys. Rev. Lett. {\bf 88},
360: 176804 (2002).
361:
362:
363: \bibitem{tao03} B.~Xu and N.~J. Tao Science {\bf 301}, 1221 (2003).
364:
365:
366: \bibitem{taylor02} J.~Taylor, M.~Brandbyge and K.~Stokbro Phys. Rev.
367: Lett. {\bf 89}, 138301 (2002).
368:
369: \bibitem{damle01} P.~S. Damle, A.~W. Ghosh and S. Datta Phys. Rev. B
370: {\bf 64}, 201403 (2001).
371:
372: \bibitem{diventra02} M.~Di Ventra, S.~T. Pantelides and N.~D. Lang
373: Phys. Rev. Lett. {\bf 88}, 046801 (2002).
374:
375:
376: \bibitem{smit02} R.~H.~M.~Smit, Y.~Noat, C.~Untiedt, N.~D.~Lang,
377: M.~C.~van Hemert and J.~M.~van Ruitenbeek, Nature {\bf 419}, 906
378: (2002).
379:
380:
381: \bibitem{cuevas_heurich03} J.~C. Cuevas, J. Heurich, F. Pauly, W.
382: Wenzel, and G. Sch{\"o}n, Nanotechnology, {\bf 14}, R29 (2003).
383:
384: \bibitem{garcia_palacios04} Y. Garc{\'i}a, J.~J. Palacios, E.
385: SanFabi{\'a}n, J.~A. Verg{\'e}s, A.~J. P{\'e}rez-Jim{\'e}nez and E.
386: Louis, Phys. Rev. B {\bf 69}, 041402(R) (2004).
387:
388: \bibitem{calc} The surface is modeled as a slab with a thickness of 4
389: atomic layers and the supercell contains $3\times 3$ atoms in the
390: surface plane. Both the Pt pyramids and the H atoms have been
391: relaxed using a plane wave DFT code~\protect\cite{dacapo} to obtain
392: the most stable contact geometry. We use an energy cut-off of 25~Ry
393: for the plane wave expansion and describe the ion cores by ultrasoft
394: pseudopotentials~\protect\cite{vanderbilt90}. To treat exchange and
395: correlation we use the PW91 functional~\protect\cite{pw91}. The
396: Brillouin zone is sampled by a single $\bold k$-point along the
397: contact axis ($z$-axis) and $4\times 4$ $\bold k$-points in the
398: transverse plane.
399:
400: \bibitem{dacapo} B. Hammer, L.B. Hansen, and J.K. N{\o}rskov, Phys.\
401: Rev.\ B {\bf 59}, 7413 (1999); S.R. Bahn and K.W. Jacobsen, Comp.\
402: Sci.\ Eng.\ {\bf 4}, 56 (2002); The Dacapo code can be downloaded at
403: http://www.fysik.dtu.dk/campos.
404:
405: \bibitem{vanderbilt90} D.~Vanderbilt, Phys. Rev. B {\bf 41}, 7892
406: (1990).
407:
408: \bibitem{pw91} J.~P.~Perdew, J.~A.~Chevary, S.~H.~Vosko,
409: K.~A.~Jackson, M.~R.~Pederson, D.~J.~Singh and C.~Fiolhais, Phys.
410: Rev. B {\bf 46}, 6671 (1992).
411:
412: \bibitem{datta_book} S.~Datta,\emph{Electronic Transport in Mesoscopic
413: Systems} Cambridge (1995).
414:
415:
416: \bibitem{thygesen_bollinger03} K.~S.~Thygesen, M.~V.~Bollinger and
417: K.~W.~Jacobsen, Phys. Rev. B {\bf 67}, 115404 (2003).
418:
419: \bibitem{datta02} Y. Xue, S. Datta and M.~A. Ratner Chem. Phys. {\bf
420: 281}, 151 (2002)
421:
422: \bibitem{brandbyge02} M. Brandbyge, J.~L. Mozos, P. Ordej{\'o}n, J.
423: Taylor and K. Stokbro Phys. Rev. B {\bf 65}, 165401 (2002).
424:
425: \bibitem{meirwingreen} Y.~Meir and N.~S.~Wingreen, Phys. Rev. Lett.
426: {\bf 68}, 2512 (1992).
427:
428: \bibitem{djukic04} D. Djukic and J. M. van Ruitenbeek, private
429: communication
430:
431: \bibitem{souza01} I.~Souza, N.~Marzari and D.~Vanderbilt, Phys. Rev. B
432: {\bf 65}, 035109 (2001).
433:
434: \bibitem{partlyoccwfs} K. S. Thygesen and K. W. Jacobsen,
435: arXiv:cond-mat/0411086.
436:
437: \bibitem{newns69} D.~M.~Newns, Phys. Rev. {\bf 178}, 1123 (1969).
438:
439: \bibitem{thielmann03} A.~Thielmann, M.~H.~Hettler, J.~K{\"o}nig and
440: G.~Sch{\"o}n Phys. Rev. B {\bf 68}, 115105 (2003).
441:
442: \bibitem{ng88} T.~K.~Ng and P.~A.~Lee Phys. Rev. Lett. {\bf 61}, 1768
443: (1988).
444:
445: \bibitem{hyldgaard94} P.~Hyldgaard, S.~Hershfield, J.~H.~Davies and
446: J.~W.~Wilkins Ann. Phys. {\bf 236}, 1 (1994).
447:
448: \bibitem{hammer95} B.~Hammer and J.~K.~N{\o}rskov, Surface Science
449: {\bf 343}, 211 (1995).
450:
451: \bibitem{norskov_houmoller81} J.~K. N{\o}rskov, A. Houm{\o}ller, P.~K.
452: Johansson and B.~I. Lundqvist, Phys. Rev. Lett. {\bf 46}, 257
453: (1981).
454:
455: \end{thebibliography}
456:
457: \end{document}
458:
459: