THE JOURNAL OF CHEMICAL PHYSICS 124, 114708 2006

A theoretical study of molecular conduction. III. A nonequilibrium-Greensfunction-based Hartree-Fock approach

Tomomi Shimazaki
Department of Chemical System Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113-8656, Japan

Yongqiang Xue
Department of Chemistry, Northwestern University, Evanston, Ilinois 60208 and College of Nanoscale Science and Engineering, University at Albany, 255 Fuller Road, Albany, New York 12203

Mark A. Ratner
Department of Chemistry, Northwestern University, Evanston, Ilinois 60208

Koichi Yamashitaa
Department of Chemical System Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113-8656, Japan

Received 6 September 2005; accepted 24 January 2006; published online 20 March 2006 Many recent experimental and theoretical studies have paid attention to the conductivity of single molecule transport junctions, both because it is fundamentally important and because of its signicance in the development of molecular-based electronics. In this paper, we discuss a nonequilibrium Greens function NEGF -based Hartree-Fock HF approach; the NEGF method can appropriately accommodate charge distributions in molecules connected to electrodes. In addition, we show that a NEGF-based density matrix can reduce to an ordinary HF density matrix for an isolated molecule if the molecule does not interact with electrodes. This feature of the NEGF-based density matrix also means that NEGF-based Mulliken charges can be reduced to ordinary Mulliken charges in those cases. Therefore, the NEGF-based HF approach can directly compare molecules that are connected to electrodes with isolated ones, and is useful in investigating complicated features of molecular conduction. We also calculated the transmission probability and conduction for benzenedithiol under nite electrode biases. The coupling between the electrodes and molecule causes electron transfer from the molecule to the electrodes, and the applied bias modies this electron transfer. In addition, we found that the molecule responds capacitively to the applied bias, by shifting the molecular orbital energies. 2006 American Institute of Physics. DOI: 10.1063/1.2177652
I. INTRODUCTION

There have recently been many theoretical and experimental studies on the electronic properties of a molecule connected between electrodes in order to develop molecularbased electronic circuits and devices. Although the direct measurement of molecular conduction remains challenging, novel methods such as mechanical controlled break junction, the cross-wire tunnel junction technique, and measurements on single atomic monolayers SAMs by scanning tunneling electron microscopy STM can observe some features of electron transport through a single molecule.18 On the other hand, some studies have reported the rectication and switching properties in molecular conduction, and tried to develop molecular devices such as diodes, transistors, and memory cells.915 In theoretical studies, the transmission probability for electron tunneling between electrodes through a single molecule is usually investigated based on Greens function methods, and the effects of semi-innite electrodes are accommoa

Electronic mail: yamasita@chemsys.t.u-tokyo.ac.jp

dated as the self-energy.1618 The semi-innite nature of electrodes cannot be neglected in investigating molecular conduction because of the strong modulation of the molecular orbitals. There are two main modulations: the molecular orbital energies are shifted and their levels are broadened.1921 The electron can transfer from one electrode to the other through such broadened molecular orbitals, so the self-energy of semi-innite electrodes should be appropriately accommodated. We have to obtain the surface Greens function of semi-innite electrodes in order to determine the self-energy. Although recursion methods have so far been used in order to obtain the surface Greens function,2224 Sanvito et al. and Krstic et al. recently reported a novel technique that can construct the surface Greens function from Greens functions with periodic boundary conditions.25,26 Unlike the recursion methods, their techniques do not require iterative calculations for numerical convergence and can properly accommodate the evanescent state in molecular conduction. The nonequilibrium Greens function NEGF method has recently attracted much attention because it can study molecular conduction under nite biases.2737 The NEGF
2006 American Institute of Physics

0021-9606/2006/124 11 /114708/12/$23.00

124, 114708-1

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-2

Shimazaki et al.

J. Chem. Phys. 124, 114708 2006

method can give the density matrix and charge distribution on the molecule even in those situations, and the electrostatic potential can be determined from the NEGF-based charge distribution. In many studies, the potential is used in the framework of density function theory DFT in order to obtain electronic structures for a molecule connected to electrodes with nite biases. Although this NEGF-based DFT approach successfully brings deep insights into molecular conduction, we discuss in this paper a NEGF-based HartreeFock HF approach both because the theoretical clarity in the HF method enables us to investigate the features of molecular conduction and because correlation effects might subsequently be added. Galperin and Nitzan recently reported a NEGF-based HF approach.38 Although they used a constant and energy-independent self-energy term, we adopt more general formalisms for the surface Greens function of semiinnite electrodes. Other HF-based approaches to transport were reported by Mujica et al.39 and by Delaney and Greer.40 We also demonstrate a technique to obtain the NEGF-based density matrix. Our approach to the NEGF-based density matrix is based on a spectrum representation, which is different from previous other studies with an inverse matrix of effective Hamiltonian. The spectrum representation of Greens function can easily show a relationship between molecular conduction and ordinary quantum chemical properties.
II. THEORY A. Transmission probability

FIG. 1. Molecular conduction between leads p and q.

the self-energy to the isolated molecular Hamiltonian Hmol in one-electron orbital approximation, i.e., the Fock matrix, and the retarded and advanced Greens functions can be obtained as follows: H = Hmol +
R p

R q, R p

4a
R q

ES H GR = ES Hmol GR = ES Hmol GA = GR .
R p

GR = I,

4b 4c 4d

R 1 , q

Here, the subscripts p and q represent leads p and q, respectively. The transmission probability T pq can be calculated from the following equation see Appendix A : T pq = Tr
pG R qG A

The retarded self-energy R can be determined from the R surface Greens function gsurface of electrodes and the electronic coupling matrix between the electrodes and the molecule as follows:16
R

T R gsurface

Here, p = i R A and q = i R A . The current I p p p q q through a molecule can be determined from the transmission probability and Landauer formula by using the following equations when the voltage V is applied to the molecule, as shown in Fig. 1:41,42 Ip = 2e h
+

Moreover, we can use the following relationship:

A R

T E fq E

fp E

dE,

6a 6b

R R T

,
R gsurface = R gsurface T.

2
p q

= if We adopt the following and R equations as the surface Greens function gsurface to determine 16,18 the self-energy of the electrodes:
R gsurface =

= EF eV, = EF + 1 eV.

6c

exp ikc

3a 3b

Here, EF is the Fermi energy of the electrode at zero bias, is the chemical potential, and is a parameter that varies from 0 to 1; we set = 0.5.
B. The nonequilibrium Greens function

E=

+ 2 cos kc .

Here, E is the energy of the electron wave in the electrode, is the diagonal element of the unit cell, is the nearest neighbor electronic coupling element, c is the lattice constant, and k is the wave number. Although the above equations have very simple forms, the surface Greens function derived by Sanvito et al. and Krstic et al., which can accommodate more complex systems, reduces to Eq. 3a in onedimensional problems.21 Therefore, we can investigate the fundamental features even if Eq. 3a is adopted as a primitive model for the surface Greens function. The effective Hamiltonian, which represents the system including a molecule and semi-innite electrodes, is determined by adding

The molecular orbitals that interact with the semi-innite electrodes are broadened because of interaction with the electronic states of the electrodes, and the electron transfers from one electrode to the other along such broadened molecular orbitals. On the other hand, we can control the electron occupation of the broadened molecular orbitals by applying a bias to the electrodes because the bias can change the chemical potential of the electrodes; the electron occupation number of the broadened molecular orbitals is not always an integer, unlike the situation in the isolated molecule, as seen in Fig. 2. Moreover, the occupation becomes very

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-3

Hartree-Fock study of conduction

J. Chem. Phys. 124, 114708 2006

D r1,r2 =

1 2 i

G r1,r2 ;E dE,
+

11a

G r1,r2 ;E =

exp iEt G x1,x2 dt,

t = t2 t1 . 11b

In order to numerically calculate the density matrix, we need the following steps. First, dene the spectrum function A r1 , r2 ; E as follows:
+

A r1,r2 ;E =

exp iEt A x1,x2 dt,

t = t2 t1 , 12a

A x1,x2 = i GR x1,x2 GA x1,x2

FIG. 2. Electron occupation of orbitals in a molecule connected to semiinnite electrodes.

= i G x1,x2 G x1,x2 .

12b

complicated when different biases are applied to each electrode. The NEGF methods have been widely studied to accommodate these problems, and the density matrix for molecules connected to electrodes can be calculated. In the following, the density matrix of a molecule under a nite bias is derived based on the NEGF method. In the NEGF formalism, not only the retarded and advanced Greens functions but also the following lesser and greater Greens functions must be accommodated together:31,34,4345 GR x1,x2 = i G x1,x2 = + i G x1,x2 = + i G x1,x2 = i
A

We assume that the following relation is applicable for nonequilibrium as well as equilibrium problems: G r1,r2 ;E = if E A r1,r2 ;E . 13

Here, f E is the Fermi-Dirac distribution function. We then obtain G = if E A = if E i GR GA = if E iGA GA = if E GA GR , and similarly,
1

GR

1 A

GR
R

= if E iGA Es Hmol

Es Hmol

GR 14

t1 t2 t2 t1

x1 , x1 , x1 ,

x2 x2

, ,

7a 7b 7c 7d

G = if E GR GA . Finally, the density matrix is given as D= 1 2 1 2

+

15

x2

x1

x2 .

f E GA GRdE
+

Here, x1 = r1 , t1 and x2 = r2 , t2 . These four Greens functions are not independent and have the following relationships: GR x1,x2 = GA x1,x2 = and therefore GR x1,x2 GA x1,x2 = G x1,x2 G x1,x2 . 9 t1 t2 G x1,x2 G x1,x2 , t2 t1 G x1,x2 G x1,x2 , 8a 8b

f p E GA

pG

+ f q E GA

qG

dE.

16

Now we can clearly distinguish the contributions from leads p and q in the density matrix under the condition of a nite bias on a molecule.

C. Analysis of the NEGF-based density matrix by an approximate Greens function

The charge density of the electron and the hole can be determined by the lesser and greater Greens functions, respectively, ne x1 = nh x1 =

Equations 14 and 15 can also be expressed as follows: G = if E A = if E i GR GA = 2if E Im GR . 17

x1 x1

x1 = iG x1,x1 , x1 = iG x1,x1 .

10a 10b

Therefore, another expression for the density matrix can be obtained as follows: D= 1
+

The density matrix can be calculated from the lesser Greens function as follows:

f E Im GRdE.

18

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-4

Shimazaki et al.

J. Chem. Phys. 124, 114708 2006

In order to analyze Eq. 18 we derive an approximate Greens function that includes the rst-order energy correction based on perturbation theory,20,21 Hmol + = Hmol
1 a 0 a 0 a R 0 a 1 a 0 a R

q,a

0 a

0 a

26b

1 a

+
0 a

Here, a = p,a + q,a. Then the approximate density matrix becomes see also Appendix B D= where GR =
a

+ =

+
0 a 0 a

1 a

+ ,

19a 19b

1 2

fp E p,a +

p,a q,a

fq E p,q +

q,a q,a

Im gRdE, a

27

, .

0 a

19c

gR , a C 0,aC 0 ,a E
0 a

28a

We remark that is the one-electron orbital. Then the following approximate Greens function from the above equations is obtained: GR r1,r2 =
a 0 a

gR a and note Im
1 a

r1

0 a

Re

0* r2 a 1 i Im a

Re

1 a

i Im

1 a

28b

1 a

20

The matrix representation of the approximate Greens function and approximate density matrix can be expressed by the linear combination of atomic orbital LCAO approximation of molecular orbitals a0 = C 0,a , where are the atomic orbitals, GR =
a +

1 2

1 2

p,a

1 2

q,a .

29

C 0,aC 0 ,a E f E
a 0 a

Re

1 a

i Im

1 a

21

C 0,aC 0 ,a
a 0 a

1 2 E

Re

1 2 a

a/2

dE.

22

An electron from lead p or from lead q occupies molecular orbital a according to the ratios p,a / p,a + q,a and 7 q,a / p,a + q,a , respectively. These relations show that the electron in molecular orbital a has no scattering interactions with phonons or other electrons; in other words, Eq. 27 represents a charge distribution for ballistic electron transport. Therefore, the spectrum representation of effective Greens function can easily show a relationship between p q ,a and the NEGF-based density matrix unlike the inverse matrix of effective Hamiltonian. Moreover, we remark that the approximated NEGF-based density matrix Eqs. 22 and 27 can well reproduce the exact one.

In order to obtain Eq. 22 , we used Im

1 a

0 a

Im

0 a

1 2

0 a

0 a

1 2

a.

D. The NEGF-based density matrix

23 If we assume that there is no interaction between the molecule and the electrode, the rst-order energy correction in Eq. 22 becomes zero, that is, Re a1 0 and Im a1 0 when R 0, lim
r0

1 2 E

/2

24

Next, we accommodate the exact Greens function by spectrum representation in order to analyze the density matrix given by Eq. 16 . The effective Hamiltonian, in which the self-energy is added to the analysis of the isolated molecular Hamiltonian Fock matrix , is not Hermitian, so we cannot use ordinary orthogonal relations. However, we can use the following orthogonal relations by considering a dual space:16,46,47 Hmol + Hmol + r
R a

= =

Then, the NEGF-based density matrix reduces to the ordinary density matrix that is used in the usual Hartree-Fock method, D = lim
R0

a a, * a a,

30a 30b 30c

f E Im GR dE

* a

r dr =

ab ,

occ

=
a

C 0,aC 0 E ,a

0 a

dE.

25

and therefore the exact spectrum representation of the Greens function, GR r1,r2 =
a a

In order to investigate the properties of the NEGF-based density matrix, we substitute the approximate Greens functions in Eq. 16 , and we use the following equations:
p,a

r1 E

* a a

r2

31

0 a

0 a

26a

where a and a are expanded by the LCAO approximation, CR ,a , and a = CA ,a . a=

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-5

Hartree-Fock study of conduction

J. Chem. Phys. 124, 114708 2006

Now Eqs. 30a and 30b can be written as follows: Hmol +

R

CR a

R aSCa ,

32a 32b 32c 32d

GA =
a

CACR a a E
* a

CACR E
*

CA CR E i /2

, 33b

R R R CR = C1,aC2,a Cn,a T , a

Hmol +

CA = a

* A aSCa ,

= + 0,

33c

A A A CA = C1,aC2,a Cn,a T , a

and the matrix representations of the Greens functions become GR =

a

CRCA a a = E a

CRCA + E

CR CA E + i /2

, 33a

denotes molecular orbital energies where the subscript with an imaginary part and the subscript , and CR and CA denote those without an imaginary part. We remark that is innitely small, and the imaginary part of the last terms in Eqs. 33a and 33b can be replaced with a delta function see Eq. 24 . By substituting these relations in Eq. 14 , the lesser Greens function becomes

G = if E

C AC R E C AC R
,

CA CR E

/2 C AC R

C RC A + E CR CA E . + i

CR CA E + if E
,

+ i

/2 CA CR E i

= if E

C RC A + if E E CA CR

E CR CA

/2

/2

C RC A E

+ if E
,

/2

+ i

/2

34

The molecular orbital energy without imaginary part means that the molecular orbital has no interactions with the semi innite electrodes, then CR = CA see Appendix C . By noting that CR = CA =C and CR SCA = ab, the second term in Eq. a b 34 becomes zero as follows: second term = if E
,

C AC R E

* i Hmol +

Hmol

CR CA E

+ i

/2
A

= if E
,

iCACR Hmol +

CR CA CACR Hmol +
*

CR CA

+ i

/2

= if E
,

i CACR SCA CR i CACR SCA CR E

*

+ i

/2

= 0.

35

We can argue similarly for the third term in Eq. 34 and obtain the following equations as the lesser Greens function and the density matrix:

1 D= 2 +

f E
+ ,

C AC R E C C

C RC A dE E dE

f E
+

1 2 E

/2

G = if E
,

C AC R E

C RC A E C C

1 = 2 C C /2 E

f E
+ occ ,

C AC R E

C RC A dE E dE. 37

+ if E

+ i

/2

, 36

C C

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-6

Shimazaki et al.

J. Chem. Phys. 124, 114708 2006

FIG. 5. Change of total charges on C6H4S2.

g
FIG. 3. SCF loop in the NEGF-based HF approach.

r1

r2

1 r1 r2

r1

r2 dr1dr2 , 38c

Equation 24 is used to change the rst line to the second line in Eq. 37 . In our calculations Eq. 37 is used as the NEGF-based density matrix.

III. THE NEGF-BASED HARTREE-FOCK APPROACH

In this paper, we calculate the Fock matrix by using the Hartree-Fock method and the density matrix based on Eq. 37 . In these calculations, we determine the Fock matrix in the external electric eld for electrodes with different applied biases. The NEGF-based Fock matrix Hmol can be determined from the following equations by using the NEGFbased density matrix of Eq. 37 : Hmol =h + D g g , 38a 1 r 2 r
A

where h includes the kinetic energy, the nucleus-electron interaction, and the interaction with the external electric eld. g is an electron-electron Coulomb repulsion operator, F is the external electric eld, and ZA is the atomic number of nucleus A. In our calculations, we perform iteratively the following steps until the NEGF-based density matrix converges, as seen in Fig. 3. The Fock matrix, which includes the interaction with the external electric eld, is calculated based on the NEGF-based density matrix and Eq. 38a , and then the extrapolated Fock matrix is obtained by the direct inversion in the iterative subspace DIIS method.48 Next, the eigenvectors and eigenvalues are calculated from the extrapolated Fock matrix. Then, the NEGF-based density matrix is recalculated from the extrapolated Fock matrix and the selfenergy; the eigenvectors and eigenvalues have to be deterTABLE I. Mulliken charges on C6H4S2. Bias V 0.0 1.0 Without NEGF 0.0287 0.0645 0.0781 0.0698 0.0726 0.0764 0.0829 0.0857 0.0000 NEGF 0.0666 0.0616 0.0730 0.0652 0.0689 0.0824 0.0888 0.1227 0.1250 2.0 3.0

= +

r dr rr r dr, 38b

ZA RA r

r dr F

S1 C1 C2 C6 C3 C5 C4 H1 H4 H2 H3 S2 Total

0.0572 0.0685 0.0740 0.0740 0.0685 0.0797 0.0797 0.0572 0.0000

0.0003 0.0606 0.0823 0.0657 0.0768 0.0732 0.0861 0.1143 0.0000

0.0279 0.0568 0.0863 0.0616 0.0810 0.0699 0.0894 0.1429 0.0000

S1 C1 C2 C6 C3 C5 C4 H1 H4 H2 H3 S2 Total FIG. 4. Coordinates in our calculations.

0.0873 0.0655 0.0700 0.0700 0.0655 0.0846 0.0846 0.0873 0.1016

0.0606 0.0570 0.0739 0.0587 0.0708 0.0826 0.0952 0.1727 0.1960

0.0652 0.0531 0.0732 0.0519 0.0716 0.0847 0.1034 0.2316 0.2981

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-7

Hartree-Fock study of conduction

J. Chem. Phys. 124, 114708 2006

TABLE II. Relationship between HOMO energy and chemical potentials of electrodes. Bias V
0 HOMO 1 a HOMO 1 a HOMO 0 HOMO + Re a p a q Emoleculeb

0.0 5.79 0.03 0.201 5.82 4.77 4.77 1021.413

1.0 5.86 0.03 0.201 5.89 5.27 4.27 1021.409

2.0 6.07 0.04 0.200 6.11 5.77 3.77 1021.394

3.0 6.36 0.06 0.198 6.42 6.27 3.27 1021.373

Re Im

1 a HOMO

In eV. In a.u.; dened in Eq. 40 .

FIG. 7. Change of energy of molecule according to the applied bias.

mined at each energy of the electron wave because the selfenergy term depends on its energy. Here, we note that the second term in Eq. 37 is, approximated by using the eigenvectors of the extrapolated Fock matrix. If the calculated NEGF-based density matrix does not converge, we return to the rst procedure again and continue the above steps until the density matrix converges. The NEGF-based Mulliken population and the charge QA on nucleus A can be calculated by the following equations:
AO AO A= A

D S ,
A,

39a 39b is calculated based on Eq. 37 .

QA = ZA

where the density matrix D

IV. CALCULATIONS FOR BENZENEDITHIOL

In this section, we discuss the calculated result of electron transfer through benzenedithiol by using the NEGFbased HF approach. In our calculations, diffuse functions are added to the STO-6G basis in order to let the molecule respond to the external electric eld. The parameters for the surface Greens function in Eq. 3b are set to be = 4.77 eV, = 3.0 eV, and c = 3.5334 . We also assume that the electrodes can interact only with the 3py orbital pi type of the sulfur atom of the C6H4S2 species. Here, the z

axis goes through the S1 and S2 sulfur atoms of C6H4S2, and the x axis is perpendicular to the z axis and in the plane of C6H4S2 see Fig. 4 . Although highest occupied molecular orbital HOMO , HOMO-4, and lowest unoccupied molecular orbital LUMO can interact with the electrodes under this assumption, the HOMO is expected to dominate the total current through benzenedithiol because it is closest to the Fermi level. We calculated both the ordinary free molecule and NEGF-based Mulliken charges see Fig. 5 . Table I shows that the total sum of the NEGF-based Mulliken charges is not zero, unlike ordinary Mulliken charges, because of electron transfer from C6H4S2 to the electrodes. In addition, the NEGF-based Mulliken charges change according to the applied voltage on the electrodes; Fig. 5 shows the change of total charge on C6H4S2. The chemical potentials p and q, respectively, for leads p and q, are summarized in Table II 0 with the HOMO energy HOMO and the rst energy correc1 1 tions Re HOMO and Im HOMO calculated by using simultaneous equations.20,21 Figure 6 shows the relationship between these energies, which are tted to a Lorentz function in order to explain the transmission probability. When the bias increases, lead p extracts the electron from C6H4S2 as is clearly seen in the relationship between these energies. On the other hand, the contribution of electron occupation from lead q induces only a small change in the total electron occupation because the electron from lead q almost lls the HOMO even in the case of a low bias, and more positive charges resulting on C6H4S2 with a high bias. Table II also

FIG. 6. Relationship between HOMO energy and chemical potentials of the electrodes.

FIG. 8. Change of HOMO energy according to the applied bias.

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-8

Shimazaki et al.

J. Chem. Phys. 124, 114708 2006

FIG. 9. Potential energies of C6H4S2 at a bias voltage of 3.0 V: a by EGF for y = 0.0 surface, b by EGF for y = 1.0 surface, c by NEGF for y = 0.0 surface, and d by NEGF for y = 1.0 surface.

shows that the HOMO energy shifts in the negative direction with an increasing bias. In order to investigate why the molecular orbital shifts in the negative direction, we analyzed the NEGF-based energy for C6H4S2 by the following equation:
n

Emolecule =
,

D h g

1 + 2

D D
, , ,

g 40

+ Vnn ,

where D and D are the NEGF-based density matrices in Eq. 37 , and Vnn is the repulsion energy between nuclei. The results are summarized in Table II and the change in Emolecule is shown in Fig. 7 with the energy of the molecule with equilibrium Greens function EGF for comparison. The increase in the NEGF-based energy indicates that the electron transfer to the electrodes destabilizes C6H4S2, so the molecule responds to prevent transport of an electron from itself to the electrode by lowering the HOMO energy. The chemical potential of lead p and the HOMO energy with respect to the applied bias are shown in Fig. 8. The HOMO energy

lowers with the decrease of the chemical potential of lead p, although the HOMO energy with EGF is almost constant. Therefore, there are two reasons why the molecular orbital shifts: one is the effect of self-energy representing semiinnite electrodes, and the other is the charge transfer from a molecule to the electrode. The computed electrostatic potentials, expressed as energies, calculated from NEGF-based Mulliken charges and the three-dimensional Poisson equation are shown in Figs. 9 c and 9 d ; Figs. 9 a and 9 b are for ordinary Mulliken charges with EGF. The NEGF-based potential energies on the sulfur atoms in C6H4S2 are particularly different from those with EGF due to electron transfer from molecule to electrodes. Lastly, theoretical currents through C6H4S2 based on Landauers formula and its conductance are respectively, given in Figs. 10 a and 10 b ; the currents and conductance in Fig. 10 a are determined with EGF, and those in Fig. 10 b are determined by the NEGF-based HF approach. From these gures, we can easily overview the inuence of electron transfer from the molecule to the electrodes.

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-9

Hartree-Fock study of conduction

J. Chem. Phys. 124, 114708 2006

simplied tight-binding model for the electrodes. Calculations by the NEGF approach gave electron transfer from the molecule to the electrodes and shifted the molecular orbital energy. The electron transfer to the electrodes destabilizes C6H4S2, and the HOMO shifts in a direction so as not to increase the electron transfer. The calculations for other properties by NEGF-based HF approach are in progress. Those results will be reported elsewhere.

ACKNOWLEDGMENTS

This research was supported by a Grant-in-Aid for The 21st Century COE Program for Frontiers in Fundamental Chemistry, and for Scientic Research KAKENHI in Priority Area Molecular Nano Dynamics, from the Ministry of Education Science Sports and Culture of Japan. The authors thank the Computer Center of the Institute for Molecular Science for the use of computers. Two of the authors M. R. and Y. X. thank the MURI/DURINT program of the NSF for support.

FIG. 10. Theoretical currents through benzenedithiol and conductance a by EGF- and b NEGF-based HF approaches.

APPENDIX A: DERIVATION OF LANDAUER FORMULA

V. SUMMARY

In this appendix, we derive the form of the current in Eq. 6a from the Greens functions.16,34,44 The current operator can be expressed as follows: t = e d ne j dt =e = i H,ne = e HG G H h , A1

In this paper, we have discussed a NEGF-based HF approach. Although the NEGF method has so far been mainly adopted with DFT, the clarity of the HF approach can easily give a deep understanding of molecular conduction. In both DFT and HF, the NEGF-based density matrix can reduce to an ordinary density matrix if there are no interactions between a molecule and an electrode, and NEGF-based Mulliken charges have similar properties to those of free molecules. Moreover, the only difference in the self-consistent eld SCF cycles between the ordinary and NEGF-based HF approaches is a routine to calculate the density matrices. That is, the NEGF-based HF approach is the same as the ordinary approach if the molecule does not interact with the electrodes. These characteristics of the NEGF-based HF approach are useful in the study of complicated phenomena in molecular conduction because we can directly compare molecules connected to electrodes with isolated ones. On the other hand, although the electron-electron correlations are considered in the DFT method, the choice of functional remains arbitrary. If NEGF-based post-HF methods, such as conguration interaction CI methods40 and Mller-Plesset methods, are adopted, we will be able to more easily investigate the effects of electron-electron correlations and will gain deeper insights into electron-electron correlation in molecular conduction. The NEGF-based HF approach is a basis of these post-HF methods for molecular conduction. We also calculated the transmission probability for transport junction using C6H4S2 based on the NEGF-based HF approach. In our calculations, we paid attention especially to electron transfer through the HOMO because it dominates for C6H4S2, at biases below 2 eV. We used a highly over-

e HmolG + G G Hmol G h

where H = Hmol + and the charge density of electron ne is expressed by the lesser Greens function see Eq. 10a . We convert the current operator from the time domain to the frequency domain energy domain by Fourier transformation. The lesser Greens function can be expressed by using the self-energy as follows: G = GR GA . A2

Substituting Eq. A2 in Eq. A1 gives E = e HmolGR j h GA + G GR GAHmol G . A3 We obtain the following equations from Eq. 4b : HmolGR = ESGR
R

GR I,
A

A4a A4b

GAHmol = GASE GA

I,

and substitution of the above equations in Eq. A3 gives

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-10

Shimazaki et al.

J. Chem. Phys. 124, 114708 2006

E = e ESGR j h GR =

GA

GR GA

GA
A

GA + G G

e jq E = i Tr h

qG

A .

A7c

GASE + GR
R

+ GR

Noting from Eqs. 13 and A2 = i f, we obtain A5 A8

e ESG h
A

G G

GA + G G SE .

+G

+G

We can calculate the current from a trace operation, jE e = Tr E = Tr j h e = i Tr G h A ,

R

e j p E = i Tr h if p
pG

if pGR +

pG

+ if qGR

qG A

e GA = Tr h

pG

qG

fq A9

G +

GR GA A6

fp .
APPENDIX B: DERIVATION OF APPROXIMATE DENSITY MATRIX

where we use Eqs. 12a and 12b and the relation Tr AB = p + q and = p + q, Eq. A6 becomes = Tr BA . If j E = j p E + jq E , where e j p E = i Tr h
pG

A7a

We derive the equation for the density matrix when we use approximate Greens functions. This method is useful if it is necessary to accommodate large systems, such as DNA or proteins. We obtain the following equation from Eq. 23 : Im
1 a

1 2

C 0,a
,

C 0,a ,

B1

A ,

A7b

and another expression for Eq. 21 ,

Im G

C 0,aC 0 ,a =
a

1 2

C 0,a
1 2

C 0,a
,

0 a

Re

1 2 a

C 0,a
,

C 0,a

=
a

1 2

C 0,aC 0,a C 0,a E

C 0,aC 0 ,a
0 a

E
1 2

0 a

Re

1 a

i 2
1 a

C 0,a

Re

1 a

i 2
1 a

C 0,a
1 2

C 0,a gA gR , a a B2

C 0,aC 0,a
a ,

C 0,aC 0 ,a
1 a

0 a

Re

+ i Im

0 a

Re

1 a

i Im

where C 0,aC 0 ,a E
0 a

APPENDIX C: RELATION IN MOLECULAR ORBITALS

gR a

Re

1 a

i Im

1 a

B3a

We derive the relation CR = CA in this appendix. We consider the case of Im = 0, Hmol +

R

CR =

SCR .

C1

gA a

C 0,aC 0,a E
0 a

Re

1 a

+ i Im

1 a

B3b

Then, the approximate density matrix is given as 1 1 Im G = 2

R

The self-energy R can be obtained from Eq. 1 , and we consider one-dimensional electrodes in order to simplify the derivations. The self-energy can be divided into two parts:
R R 1 R 2 T a a a= ,

D=

gA a
a

gR . a

B4

1
a

exp ikaca

a,

C2a

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-11

Hartree-Fock study of conduction

J. Chem. Phys. 124, 114708 2006

1

R 1

cos k c

sin k c

= 0,

C2b

= C0 =

R 0 2C T T

C0

u uT C 0 exp ik c
T T

R 2

cos k c + i 0,

sin k c

, C2c

uT C 0

u C 0 exp ik c .

C11

sin k c

where a = t1,at2,a tn,a . Then, the rst-order correction of orbital energies can be obtained as Hmol + =
1 a 0 a R 1

Equation C11 has the similar formula with Eq. C4 , therefore, we can derive the same relations as Eqs. C8 and C9 .
M. Dorogi, J. Gomez, R. Osifchin, R. P. Andress, and R. Reifenberger, Phys. Rev. B 52, 9071 1995 . 2 L. A. Bumm, J. J. Arnold, M. T. Cygan, T. D. Dunbar, T. P. Burgin II, D. L. Allara, J. M. Tour, and P. S. Weiss, Science 271, 1705 1996 . 3 M. A. Reed, C. Zhou, C. J. Muller, T. P. Burgin, and J. M. Tour, Science 278, 252 1997 . 4 J. Chen, M. A. Reed, A. M. Rawlett, and J. M. Tour, Science 286, 1550 1999 . 5 J. K. Gimzewski and C. Joachim, Science 283, 1683 1999 . 6 X. D. Cui, A. Primak, X. Zarate, J. Tomfohr, O. F. Sankey, A. L. Moore, T. A. Moore, D. Gust, G. Harris, and S. M. Lindsay, Science 294, 571 2001 . 7 J. G. Kushmerick, D. B. Holt, J. C. Yang, J. Naciri, M. H. Moore, and R. Shashidhar, Phys. Rev. Lett. 89, 086802 2002 . 8 J. Reichert, R. Ochs, D. Beckmann, H. B. Weber, M. Mayor, and H. V. Leohneysen, Phys. Rev. Lett. 88, 176804 2002 . 9 C. P. Collier, E. W. Wong, M. Belohradsky, F. M. Raymo, J. F. Stoddart, P. J. Kuekes, R. S. Williams, and J. R. Heath, Science 285, 391 1999 . 10 C. P. Collier, G. Mattersteig, E. W. Wong, Y. Luo, K. Beverly, J. Sampaio, F. M. Raymo, J. F. Stoddart, and J. R. Heath, Science 289, 1172 2002 . 11 T. Rueckes, K. Kim, E. Joselevich, G. Y. Tseng, C.-L. Cheung, and C. M. Lieber, Science 289, 94 2000 . 12 C. Zhou, J. Kong, E. Yenilmez, and H. Dai, Science 290, 1552 2000 . 13 Z. J. Donhauser, B. A. Mantooth, K. F. Kelly, L. A. Bumm, J. D. Monnell, J. J. Stapleton, Jr., A. M. Rawlett, D. L. Allara, J. M. Tour, and P. S. Weiss, Science 292, 2303 2001 . 14 M. A. Reed, J. Chen, A. M. Rawlett, D. W. Price, and J. M. Tour, Appl. Phys. Lett. 78, 3735 2001 . 15 C. Li, D. Zhang, X. Liu et al., Appl. Phys. Lett. 82, 645 2003 . 16 S. Datta, Electron Transport in Mesoscopic System Cambridge University Press, Cambridge, 1995 . 17 S. Datta, W. Tian, S. Hong, R. Reifenberger, J. I. Henderson, and C. P. Kubiak, Phys. Rev. Lett. 79, 2530 1997 . 18 D. K. Ferry and S. M. Goodnick, Transport in Nanostructures Cambridge University Press, Cambridge, 1997 . 19 A. Nitzan and M. A. Ratner, Science 300, 1384 2003 . 20 T. Shimazaki and K. Yamashita, Int. J. Quantum Chem. 106, 803 2006 . 21 T. Shimazaki, H. Maruyama, Y. Asai, and K. Yamashita, J. Chem. Phys. 123, 164111 2005 . 22 F. Guinea, C. Tejedor, and F. Flores, Phys. Rev. B 28, 4397 1983 . 23 M. P. Lopez-Sancho, J. M. Lopez-Sancho, and J. Ribio, J. Phys. F: Met. Phys. 14, 1205 1984 . 24 M. P. Lopez-Sancho, J. M. Lopez-Sancho, and J. Ribio, J. Phys. F: Met. Phys. 15, 851 1985 . 25 S. Sanvito, C. J. Lambert, J. H. Jefferson, and A. M. Bratkovsky, Phys. Rev. B 59, 11936 1999 . 26 P. S. Krstic, X.-G. Zhang, and W. H. Butler, Phys. Rev. B 66, 205319 2002 . 27 W. Tian, S. Datta, J. I. Henderson, and C. P. Kubiak, J. Chem. Phys. 109, 2874 1998 . 28 P. S. Damle, A. W. Ghosh, and S. Datta, Phys. Rev. B 64, 201403 2001 . 29 B. Larade, J. Taylor, H. Mehrez, and H. Guo, Phys. Rev. B 64, 075420 2001 . 30 J. Taylor, H. Guo, and J. Wang, Phys. Rev. B 63, 121104 R 2001 . 31 J. Taylor, H. Guo, and J. Wang, Phys. Rev. B 63, 245407 2001 . 32 M. Brandbyge, J.-L. Mozos, P. Ordejon, J. Taylor, and K. Stokbro, Phys. Rev. B 65, 165401 2002 . 33 P. Damle, A. W. Ghosh, and S. Datta, Chem. Phys. 281, 171 2002 . 34 Y. Xue, S. Datta, and M. A. Ratner, Chem. Phys. 281, 151 2002 . 35 Y. Xue and M. A. Ratner, Phys. Rev. B 68, 115406 2003 .
1

C0 C3a

SC 0 ,
T T R 0 2C

= C0 = C0

Re

R 0 2C

+i C0

Im

R 0 2C 1

C3b becomes as

With the case of Eqs. C2b and C2c , follows:

1

= C0 = 1
n

R 0 2C n 2 0 Cr, r 2

cos k c

tr,

+i

sin k c

0 Cr, r

tr,

C4

where C0
T 0 0 0 = C1, C2, Cn,

C5

Therefore, the following relationship can be derived in the case of Im R = 0:

n 0 Cr, r 2

tr,

= 0,

C6

and
n R 0 2C

=
a r,s n

tr,ats,a 1
a

1
a

0 exp ikaca Cs, n

=
a r

tr,a

exp ikaca
s

0 ts,aCs,

= 0. C7

Finally, Hmol + Hmol +

R

C 0 = Hmol + C 0 = Hmol +

R 1 A 1

C0 = C0 =

SC 0 , SC 0 .

C8 C9

Therefore, CR = CA = C 0 by noting R = A. 1 1 In multidimensional cases, the surface Greens function can be obtained as follows:
R gsurface = a

uauT exp ikaca , a

1

C10 is

where ua is the real column vector, and

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

114708-12
36

Shimazaki et al.
43

J. Chem. Phys. 124, 114708 2006 L. P. Kadanoff and G. Baym, Quantum Statical Mechanics W. A. Benjamin, Inc., New York, 1962 . 44 H. Haug and A.-P. Jauho, Quantum Kinetics in Transport and Optics of Semiconductors Springer, New York, 1996 . 45 G. D. Mahan, Many-Particle Physics, 3rd ed. Kluwer Academic, New York/Plenum, New York, 2000 . 46 P. M. Morse and H. Feshbach, Methods of Theoretical Physics McGrawHill, New York, 1953 . 47 W. V. Haeringen, B. Farid, and D. Lenstra, Phys. Scr., T T19, 282 1987 . 48 P. Pulay, J. Comput. Chem. 3, 556 1982 .

P. Havu, V. Havu, M. J. Puska, and R. M. Nieminen, Phys. Rev. B 69, 115325 2004 . 37 W. C. J. Bauschlicher, A. Ricca, Y. Xue, and M. A. Ratner, Chem. Phys. Lett. 390, 246 2004 . 38 M. Galperin and A. Nitzan, Ann. N.Y. Acad. Sci. 1006, 48 2003 . 39 V. Mujica, M. Kemp, A. Roitberg, and M. Ratner, J. Chem. Phys. 104, 7296 1996 . 40 P. Delaney and J. C. Greer, Phys. Rev. Lett. 93, 036805 2004 . 41 R. Landauer, IBM J. Res. Dev. 1, 223 1957 . 42 R. Landauer, Phys. Lett. 85A, 91 1981 .

Downloaded 28 May 2009 to 129.105.215.213. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp