You are on page 1of 5

Chemical Physics Letters 374 (2003) 459463 www.elsevier.

com/locate/cplett

Many-body scattering formalism of quantum molecular conductance


Roi Baer
a

a,*

, Daniel Neuhauser

Institute of Chemistry and Lise Meitner Center for Quantum Chemistry, The Hebrew University of Jerusalem, Jerusalem 91904, Israel b Department of Chemistry and Biochemistry, University of California at Los Angeles, Los Angeles, CA 90095-1569, USA Received 5 April 2003

Abstract A general formalism is developed for the theory of electronic current through a molecule attached to macroscopic electrodes in a given bias. The current is calculated in a non-perturbative many-body formalism. We obtain an intuitively appealing formula. We identify two distinct limits: the Kubo linear response formula valid in the small bias regime and the Landauer formula valid for non-interacting electrons. Finally, we develop a new correlation function formula, which is more amenable to numerical computations and discuss the application of absorbing potentials in the electrodes as a possible means of calculation. 2003 Elsevier Science B.V. All rights reserved.

Passing electric currents through molecules connected to electrodes is the essence of single molecule electronics [1]. This is a relatively new basic science with the potential of considerable technological impact, due to new eects [26]. It is apparent, following the recent literature, that while experimentalists in this discipline are still struggling to dene reproducible procedures and standards of measurement, theory is also lacking rigorous results and benchmarks. Perhaps the most widely accepted theory of molecular conductance relies on the fundamental work of scientists from the eld of mesoscopic conductors [7]. However, at the molecular level the

Corresponding author. Fax: +972-2-651-3742. E-mail address: roi.baer@huji.ac.il (R. Baer).

situation seems quite dierent. Many eects that can be ignored in mesoscopic physics may become dominant on the molecular scale. It has been established that the charge distribution [8,9] and therefore molecular structure is exceedingly important. The role of electron correlation is important in molecular conductors because of the strong inhomogeneities in the external potential. Such eects can modify the conductance considerably [10]. Also, these eects may lead to new phenomena such as nite size eects and disorder [11]. The purpose of this Letter is to take a step forward towards establishing a theory of molecular scale electronics that serves as a foundation for developing new approximations and models. We present a theory of conductance, encompassing electronelectron interaction and correlation, as

0009-2614/03/$ - see front matter 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0009-2614(03)00709-7

460

R. Baer, D. Neuhauser / Chemical Physics Letters 374 (2003) 459463

well as electronnuclear interaction. Furthermore, our theory is formulated for a general setup of electrode voltages so it is not limited to small voltages. Our approach combines several concepts from dierent elds. First, we adopt Landauers point of view that conductance is quantum mechanical transmission [12]. Next, we are motivated by reactive scattering theory concepts and techniques, in particular the currentcurrent correlation approach of Miller [1315] and the negative imaginary potential approach of Neuhauser and Baer [1618]. Taking these elements via a rigorous route, we arrive at formulae which are general and taking full account of electron correlation. In the non-interacting electron limit we obtain Landauers conductance formula [12]. In the limit of zero bias we obtain Kubos currentcurrent correlation function formula for conductance [19]. The formalism presented is formally equivalent to a Greens function description of the conductivity of Meir et al. [20] in which the unknown is the full Greens function (see Fig. 1). We treat the electrodes as black-body cavities with respect to their ability to absorb and dissipate electrons. The small amount of charge that escapes in the form of electric current does not aect this equilibrium. The state of each metallic electrode is characterized by a given chemical potential ll and a temperature (for simplicity, we assume the same temperature for all electrodes, so bl b). We neglect any currentcurrent interactions and limit our discussion to cases where magnetic elds are negligible.

The electric current entering electrode l is dened by i ^ ^ _ ^ ^ Il qe N l qe H ; Nl : h  1

^ ^ ^ Here, qe is the electron charge, H T U VE VW is the total Hamiltonian of the system, ^ composed of kinetic energy T , electronelectron ^ repulsion U and the interaction of the electrons ^ with the nuclei of the electrodes VE and the nuclei ^ of the wire VW . We rst assume xed nuclei, although we mention later that the formulation extends to include nuclear vibrations. We will also ^ refer to the electrodes Hamiltonian HE H ^ VW . The electrodes Hamiltonian has an important property, in that it does not allow the electrons to move from one electrode to another, because of a barrier (in the absence of the wire) through which they only undergo negligible amounts of tunneling. Next, we assume that the electrons do not interact strongly in the electrodes. More precisely, we assume that deep within the electrodes, the electrons are not strongly interacting, so that one can dene asymptotic states associated with electron scattering. The number of electrons in electrode l is XZ ^l N hl rwy rws r d3 r; 2 s
s

where hl r is dened as [14] & 1 r is in electrode l; hl r 0 otherwise

Fig. 1. Schematics of the molecular wire. L and R are the electrodes, composed of atomic cores in their equilibrium position. The vertical dotted line is the surface dening the current IR . The dotted curve describes the atomic cores of the wire. The region of the wire must be large enough so that the electrodes are decoupled when there is no wire.

and we sum over both spins. It is clear that all the ^ Nl commute: Nl ; Nl0 0 for any l and l0 . However, the electrodes do not cover space and thereP fore we may not assume that l Nl is the total number of electrons in the system. In particular, when currents start to ow, there can be electrons within the wire NW . The zero-order Hamiltonian HE does not let electrons tunnel appreciably from one electrode to the next. It therefore conserves the number of electrons in each electrode HE ; Nl 0. This property would be important in the derivations below.

R. Baer, D. Neuhauser / Chemical Physics Letters 374 (2003) 459463

461

Other preliminaries are in order now. First, we assume is the electrodes are in some sense innite. Thus the Mller operator [21]
h h X lim eiHt= eiHE t= t!1

is well dened. It operates on many-electron wavepackets as follows: rst, a wavepacket at t 0 is transformed into its primitive parent by propagating it back in time to t ! 1 using the ^ uncoupled Hamiltonian HE ; next, the primitive parent is propagated back forward to t 0 by the ^ coupled Hamiltonian H . This combination transforms any initial scattering eigenstate of HE into a ^ scattering state of H with well dened asymptotic currents. The following intertwining relations hold then [22]: X ^ ^ Nl : H X X HE ; N X X 5
l

One important feature about this formula is that since the scattering states span the full Hilbert space of the problem, we are free to include a term l ~ lW NW in the ~ N sum, for any value of lW . This follows from two facts: rst, NW commutes with Nl , and second, it yields zero for any of the scattering states of HE . The freedom in the choice of the chemical potential of the wire will prove convenient in the discussion below, for example when we consider the case of near-zero bias when all the chemical potentials are equal. Using the intertwining relation equation (5) ^ l~ ^ in the expression, we obtain: IR Z 1 Treb~N Xy IR ^ ebH X . The basic expression for the current into electrode R then becomes h i ^ l ~^ 7 IR Z 1 lim Tr ebH eb~N IR t ;
t!1 h^ h ^ where IR t eiHt= IR eiHt= . This is the basic formula for quantum conductance. The interpretation is simple and intuitively appealing: the steady-state current into an electrode R is the expectation value of the corresponding current operator at t ! 1, thermally and chemically averaged over all possible scattering states. Note that we could have used the Hermitian conjugate intertwining relation HE Xy Xy H and l have the ebH trade places with e~N in Eq. (7). Thus, we see that the order of placement of operators in the formula does not aect the nal result. A similar observation was made by Miller [14] in the context of chemical reaction rates. Eq. (7) is the general equation for describing the electronic current in molecular devices. It formally takes into account all electronelectron correlations. Let us study some limiting cases. First, consider non-interacting electrons. The trace formula, Eq. (7), immediately simplies in the standard Fermi Dirac way h i X 1 ^ ^ IR 2 lim Tr1 1 ebH ll Nl IR t ; 8 l t!1 ^ ~ ^

Actually, the second term in Eq. (5) is not a strict equality; there are contributions of very high energy states of HE associated with electrons localized on the wire. However, the contribution of these states is negligibly small due to the Boltzmann averaging at nite temperature. We can now write the expression for the current entering electrode R, IR , which is a thermal ^ weighted average of the expectation value of IR at t ! 1. In other words, we average over all scattering eigenstates of HE (where for each scattering state there are several electrons that come from one electrode, several from another, etc.). This yields a trace under the assumption that the scattering states are complete, i.e., that there are no bound states in the wire. The corresponding procedure is summarized in the following equation: h i ^ ^ l~ ^ IR Z 1 Tr ebHE eb~N Xy IR X ; 6 Q ^ ^ l~ where Z l Zl TrebHE eb~N is the electrode grand canonical partition function, and P ~ N l ll Nl . In brief, the expression says that l ~ we should thermally average over all states, associate an electrode-dependent chemical potential with each electron, and calculate the eventual ux from such a state.

where the trace Tr1 is over single electron states. ^ Here Nl simply acts as a projection operator on electrode l and the factor of 2 accounts for spin degeneracy. Note that Eq. (8) is equivalent to the Landauer formula [7,23]

462

R. Baer, D. Neuhauser / Chemical Physics Letters 374 (2003) 459463

IR

Z 2qe X FD bE li TlR E dE; h l

9
1

where h is Plancks constant, FD x 1 ex and TlR E is the cumulative transmission probability, given by Millers formula [24] TlR E ^ ^^ h=qe limt!1 Tr1 dE H Nl IR t. Next, consider a dierent limiting case, of interacting electrons at zero bias ll l for all l, and we also include lW . In that case the chemical potential term simplies to ~ N lN . This immel ~ diately implies that the total current should be ^ ^ zero. Indeed, since H and N commute, Eq. (7) simplies to h i ^ ^ ^ IR Z 1 lim Tr ebH lN IR t : 10
t!1

showing that the current expectation value is a correlation function between current entering the electrode R and a weighted current X ^ ^ sin hbll =2Il : 14 J 2
l

In the course of deriving Eq. (13), we used the following relation, to be proved in a more detailed account, applicable for Fermions: ^ ^ ^ ^ eaNl Il eaNl coshaIl sinhaIl ; Nl :
^ ^

15

^ But IR is a commutation relation involving H (Eq. (1)) so the trace evaluates to zero as it should. Next, consider the conductance GlR oIR =oll at zero bias. From Eq. (7) h i ^ l l ^ ^ GlRl0 bZ 1 lim Tr ebH lN Nl tIR : 11
t!1

Now that the current formula is a correlation function, we add a negative absorbing potential to each electrode. The role of this potential is to absorb any out-going ux from scattering events. The imaginary potential should be placed deep enough inside the electrodes so that it absorbs only electrons that have a negligible chance to be reected back into the molecular wire. Thus the P ^ ^ ^ non-Hermitian Hamiltonian Hc H i l Cl is formed. This Hamiltonian induces the following non-reversible Heisenberg evolution of any ob^ servable A:
^ ^ At eiHc t AeiHc t :

This can be simplied. Following Miller et al [13], we write the t ! 1 limit as an integral: R1 _ limt!1 ct 0 ct dt c0, obtaining the wellknown Kubo result, derived from linear-response theory, that the conductance is the currentcurrent correlation function Z 1 ll0 l 1 GlR bZqe Tr ebH lN Il IR t dt: 12
0

16

The electrons in each electrode are eventually R1 _ ^ ^ consumed, so 0 N t dt Nl 0, thus Z 1 Z 1 ^ ^ ^ ^ IR t dt 2 NR CR t dt NR 0: 17


0 0

Using this, it is possible to obtain the following expression: IR qe Z Z 1 h h i i ^ ^ l~ ^ l~ ^ ^ Tr ebH eb~N =2 J eb~N =2 NR CR t dt:
0 1

Once again, it is seen that our general formula Eq. (7) is compatible with known special limits. Going back to the general formalism, we now discuss how the formalism can be used in an actual calculation. While in the one-particle case the Greens function formalism is useful, in the general case, we propose complex absorbing potentials for limiting the grid size [15,18]. Before introducing these, let us convert Eq. (7) into a correlation function, as this will facilitate the calculations. We transform the limit of t ! 1 to a denite integral Z 1 h i ~ ^ l~ ^ 1 ^ l^ ^ IR qe Z Tr ebH eb~N =2 J eb~N =2 IR t dt;
0

18 Practical calculation of these expressions or of Eq. (12) could proceed in several directions. Quantum Monte Carlo methods, such as Auxiliary-Field method [25] or semiclassical approaches can be used, in principle. Alternately, usual Quantum Chemical approaches, such as conguration-interaction methods, can be deployed. These methods would be quite limited in the number of electrons they can simulate if one wanted a fully correlated description; however,

13

R. Baer, D. Neuhauser / Chemical Physics Letters 374 (2003) 459463

463

with proper construction of a HartreeFock or density functional-like ground-state made of scattering states [10,26], it is possible to design a multiple-excitation approach that can in principle handle a large number of electrons, as will be shown in a future work. Finally, we remark that the same formalism is applicable even when we relax the clamped-nuclei assumption and include nuclear motion. The only dierence in formulation is that the trace has to be extended to include also all possible states of the nuclei.

Acknowledgements We thank M. Ratner, E. Rabani and R. Koslo for useful discussions. This work was supported by the Israel Science Foundation founded by the Israel Academy of Sciences and Humanities, the USA National Science Foundation and the Petroleum Research Fund.

References
[1] J. Jortner, M. Ratner, Molecular Electronics, Blackwell Science, Oxford, 1997. [2] A. Aviram, M.A. Ratner, Molecular Electronics: Science and Technology, New York Academy of Sciences, New York, 1998.

[3] J.M. Tour, M. Kozaki, J.M. Seminario, J. Am. Chem. Soc. 120 (1998) 8486. [4] R. Baer, D. Neuhauser, Chem. Phys. 281 (2002) 353. [5] R. Baer, D. Neuhauser, J. Am. Chem. Soc. 124 (2002) 4200. [6] D. Walter, D. Neuhauser, R. Baer (submitted). [7] S. Datta, Electronic Transport in Mesoscopic Systems, Cambridge University Press, Cambridge, 1995. [8] V. Mujica, A.E. Roitberg, M. Ratner, J. Chem. Phys. 112 (2000) 6834. [9] Y.Q. Xue, S. Datta, M.A. Ratner, Chem. Phys. 281 (2002) 151. [10] R. Baer, D. Neuhauser, Int. J. Quantum Chem. 91 (2003) 524. [11] Y. Imry, R. Landauer, Rev. Mod. Phys. 71 (1999) S306. [12] R. Landauer, IBM J. Res. Dev. 1 (1957) 223. [13] W.H. Miller, S.D. Schwartz, J.W. Tromp, J. Chem. Phys. 79 (1983) 4889. [14] W.H. Miller, J. Chem. Phys. 61 (1974) 1823. [15] T. Seideman, W.H. Miller, J. Chem. Phys. 96 (1992) 4412. [16] D. Neuhauser, M. Baer, J. Phys. Chem. 94 (1990) 185. [17] D. Neuhauser, M. Baer, J. Chem. Phys. 91 (1989) 4651. [18] D. Neuhasuer, M. Baer, J. Chem. Phys. 90 (1989) 4351. [19] R. Kubo, J. Phys. Soc. Jpn. 12 (1957) 570. [20] Y. Meir, N.S. Wingreen, Phys. Rev. Lett. 68 (1992) 2512. [21] C. Moller, Det. K. Danske Vidensk. Selsk. Mat.-Fys. Medd. 23 (1945). [22] J.R. Taylor, Scattering Theory: The Quantum Theory of Nonrelativistic Collisions, Krieger, Malabar, FL, 1983. [23] A. Nitzan, Annu. Rev. Phys. Chem. 52 (2001) 681. [24] W.H. Miller, J. Phys. Chem. A 102 (1998) 793. [25] R. Baer, M. Head-Gordon, D. Neuhauser, J. Chem. Phys. 109 (1998) 6219. [26] N.D. Lang, P. Avouris, Phys. Rev. Lett. 84 (2000) 358.

You might also like