Abstract
As a sanity test for the theoretical method employed, studies on (steadystate) charge transport through molecular devices usually confine themselves to check whether the method in question satisfies the charge conservation. Another important test of the theory’s correctness is to check that the computed current does not depend on the choice of the central region (also referred to as the “extended molecule”). This work addresses this issue and demonstrates that the relevant transport and transportrelated properties are indeed invariant upon changing the size of the extended molecule, when the embedded molecule can be described within a general singleparticle picture (namely, a secondquantized Hamiltonian bilinear in the creation and annihilation operators). It is also demonstrates that the invariance of nonequilibrium properties is exhibited by the exact results but not by those computed approximately within ubiquitous wide and flatband limits (WBL and FBL, respectively). To exemplify the limitations of the latter, the phenomenon of negative differential resistance (NDR) is considered. It is shown that the exactly computed current may exhibit a substantial NDR, while the NDR effect is absent or drastically suppressed within the WBL and FBL approximations. The analysis done in conjunction with the WBLs and FBLs reveals why general studies on nonequilibrium properties require a more elaborate theoretical than studies on linear response properties (e.g., ohmic conductance and thermopower) at zero temperature. Furthermore, examples are presented that demonstrate that treating parts of electrodes adjacent to the embedded molecule and the remaining semiinfinite electrodes at different levels of theory (which is exactly what most NEGFDFT approaches do) is a procedure that yields spurious structures in nonlinear ranges of current–voltage curves.
Introduction
Even restricted to the steadystate regime, studying charge transport through molecular devices is a difficult nonequilibrium problem, and the variety of methods to approach this problem utilized in the literature [14] may be taken as a manifestation of this difficulty. As a selfconsistency test for the various approaches utilized [57], a check of whether the charge conservation condition is obeyed by the method in question is an aspect that has occasionally received consideration [2,4]. With a few exceptions [813], wherein the independence of the current of the position of the area traversed by current was also investigated, most studies of this kind only checked the fact that the current at the “left” and “right” ends of the molecule are equal [1,14,15]. Except (if at all) for simpler interface effects (e.g., those accounted for through ohmic contact resistances), conduction through macroscopic solids contacted to electrodes is determined by the properties of the solid itself, which are practically unaffected by the electrodes [16].
Things drastically change in molecular junctions. There, upon contacting to infinite electrodes, the properties of the embedded molecule can be substantially modified with respect to the isolated molecule. This is particularly true in (chemisorption) cases where the anchoring groups form covalent bonds to the electrodes. Within current approaches to molecular charge transport, mostly based on nonequilibrium Keldysh Green’s functions (NEGF) combined with density functional theory (DFT), the molecular device is partitioned into a central region (also referred to as the “extended molecule”, “transport region”, “scattering region”, or “cluster”) linked to two semiinfinite “left” (L) and “right” (R) electrodes. This partitioning is inherently arbitrary. This arbitrariness is related to the arbitrariness in choosing the size of the extended molecule, which, in addition to the physical molecule, contains adjacent atomic layers (usually up to four) of (metallic) electrodes (for more specific details, see the subsection “Impact of screening and external biases”). An important problem related to this procedure is that metal atoms belonging to the extended molecule and metal atoms belonging to the electrodes are treated at different levels of theory. An unpleasant consequence is that this procedure may be a source of unphysical scattering for the electrons traveling from one electrode to another. The behavior, discussed later, is an illustration of such spurious effects. This may affect the theoretical current, yielding spurious structures (e.g., oscillations, shoulders or inflection points like those presented below) in computed I–V characteristics. A minimal mandatory requirement (“sanity test”) for any theory is whether the transport properties are independent of how this partitioning is made as shown in Figure 1. We are not aware of previous attempts in the literature to demonstrate or even check this invariance. Addressing this issue is one of the main aims of the present work.
Results and Discussion
General considerations
The considerations that follow refer to manyelectron systems wherein correlation effects are negligible (i.e., the singleSlater determinant description applies). In such situations, the retarded Green’s function of the central region, G_{C}, linked to biased (V ≠ 0) electrodes is related to the retarded Green’s function of the isolated system, G_{0,}_{C} = (ε − H_{C})^{−1} (i.e., uncoupled to electrodes) via the Dyson equation [1,4,16]:
Here, 1_{C} is the identity matrix, H_{C} is the Hamiltonian of the central region, and are embedding retarded selfenergies, which characterize the coupling of the molecule to the left and right electrodes, respectively. Throughout this text, boldface symbols denote matrices in the electronic space of the central region (C). To simplify notation in the analysis that follows, orthonormal basis sets will be assumed throughout and matrices are referred to rather than operators. For nonorthonormal basis sets, the nondiagonal overlap matrix S_{C} would replace the identity matrix 1_{C} of Equation 1. Switching between orthonormal and nonorthonormal basis sets leaves the diagonal matrix elements (onsite energies) unchanged while renormalizing the nondiagonal matrix elements (hopping integrals). See Chapter 4.1.2 of [4], which provides further details. Whenever possible, for convenience, the energy (ε measured from the equilibrium Fermi energy of the electrodes, E_{F} = 0) is omitted as well as other arguments of various functions entering the formulas given below. In the absence of correlations, the retarded Green’s function, G_{C}, remains, like in the equilibrium case, the key quantity allowing for the expression of all the relevant nonequilibrium properties. To illustrate this, the expressions of the transmission function, , local spectral density of states, and charge densities are given below. The local densities of states (location indicated by μ) are given by the diagonal elements, , of the matrix D_{C}
The charge densities n_{μ} are expressed by the diagonal elements of the matrix [16]:
where c_{μ} and are electron annihilation and creation operators (entering the expression of the second quantized Hamiltonians, see below), and G^{<} is the socalled lesser Green’s function [16]. The transmission function is given by the trace formula [2,8,17]
where the width functions
are determined by the imaginary parts of the retarded embedding selfenergies , characterizing the molecule–electrode couplings (x = L,R). Within the Landauer approach, which provides a general framework to describe the molecular transport within the elastic, uncorrelated transport approximation, the (steadystate) current I through a molecular junction is obtained by integrating the transmission function [2,8]
The difference between the Fermi distributions f_{L,R}(ε) ≡ f(ε − µ_{L}_{,}_{R}) of the biased (V ≠ 0) electrodes characterized by unbalanced Fermi energies μ_{L,R} = ±eV/2 has an important role (albeit not the only one, see Figure 3 below) in determining the energy window of allowed (elastic) electron transfer processes contributing to the current. For simplicity, it is assumed that V > 0. Taking the limit V→0 in Equation 7, one gets the wellknown formula for conductance, , i.e., “conductance is transmission”:
where = 2e^{2}/h is the conductance quantum. The secondquantized Hamiltonian of the molecular junction considered below, which is schematically depicted in Figure 1, reads
In the equations above, a_{l} (), b_{r} (), and c_{μ} () stand for annihilation (creation) operators for electrons in the left and right electrodes and in the “small” molecule, respectively. Restricting to spinindependent interactions for convenience, the spin label can be omitted and its contribution included by a factor of 2, like in Equation 7. For simplicity, the left and right electrodes will be modeled as semiinfinite chains described as a collection of sites (l and r, respectively), each site being characterized by a single level of siteindependent energy μ_{L,R} = ±eV/2 (remember that the equilibrium Fermi energy is chosen as origin E_{F} = 0) and nearestneighbor hopping integral t_{x} (x = L,R). The coupling between molecule and the left and right electrodes will be described in terms of the transfer integrals, τ_{L} and τ_{R}, which quantify the hopping between the left and right molecular ends (“anchoring groups”, denoted by labels 1 and N, respectively) and the adjacent electrode ends (labels l = 1 and r = 1 for left and right electrodes in Equation 9). In Equation 9, the small molecule, which needs not to be one dimensional, is described by a Hermitean Hamiltonian matrix, H≡ = H^{†}, whose elements (possibly complex numbers) are arbitrary.
Real systems described within the framework provided by Equation 9 and Equation 10 include atomic chains, quantum wires, carbon nanotubes, and (possibly DNAbased) bio and larger organic molecules. For concrete cases, the model parameters (H_{μ,μ′}, τ_{L}_{,}_{R}, t_{L}_{,}_{R}) can be obtained within densityfunctionalbased, tight binding (DFTB) frameworks [1820], which represent the stateoftheart for such larger systems.
In a biased junction, they may also nontrivially depend on V and, if applicable, on gate potentials. In the description underlying Equation 9, the central region corresponds to the “small” molecule whose Hamiltonian is H. Equation 9 can be repartitioned by considering an extended molecule (Hamiltonian H_{ext}) that includes parts of adjacent electrodes (l_{0} “layers” from the left electrode and r_{0} “layers” from the right electrode)
Whether using the smallmolecule or the extendedmolecule representation (i.e., Hamiltonians H or H_{ext}(l_{0}, r_{0}), respectively), the electrodes are semiinfinite, homogeneous chains characterized by siteindependent nearestneighbor hopping integrals, t_{x}, and onsite energies, μ_{x}, (Equation 9 and Equation 11). In both aforementioned descriptions of the central region, the contact to each electrode consists of a single point, and the matrices and Γ_{L}_{,}_{R} have a single, nonvanishing element. By choosing the small molecule as the central region, the embedding selfenergy has the form
In the energy range of interest, the “surface” Green’s function, g_{x}(ε), has the form [2,8,21]:
The embedding selfenergy, , (a tilde is used for a generic, unspecified central region) can be obtained as
which via Equation 6 yields
where θ is the Heaviside step function. Notice that the relevant energy range is z_{x} = ε − μ_{x} < 2t_{x} since otherwise Γ_{x} ≡ 0 and the contributions to Equation 2, Equation 4, and Equation 5 vanish. The quantity entering Equation 15 and Equation 16 is the hopping integral at the “interface” (in fact, the contact point) between the central region and electrode x. In the small molecule representation, ≡ τ_{L,R}. In the extended molecule representation, ≡ t_{x} at the contact x where a certain number of electrode “layers” are included in the central region. The central region corresponding to the Hamiltonian operators expressed by Equation 9 and Equation 11 is different, namely, the “small” molecule (with Hamiltonian H) and the extended (“larger”) molecule (with Hamiltonian H_{ext}), respectively. However, these two total Hamiltonians describe the same total physical systems and are mathematically identical. Therefore, whether computed within the description based on the small molecule or on an extended molecule, all the physical properties (e.g., transmission, current, local density of states, charge densities) should be identical. The proof that these properties indeed coincide, which represents an important sanity test for theory, will be given below.
For demonstration, a “minimally” extended molecule is considered, obtained by adding one extra electrode “layer” to the small molecule, namely the leftmost electrode site (r = 1) of the right electrode. This extended molecule having the Hamiltonian
is schematically represented by the blue dashed rectangle in Figure 1. The demonstration goes as follows. We first show that the properties computed within the small molecule representation coincide with those based on the minimally extended molecule described above. Then, the demonstration for larger extended molecules follows immediately by induction. In the next step, the minimally extended molecule is taken as a new “small” molecule and choose the new extended molecule augmented with the next site of the right electrode. That is, H_{ext}(l_{0} = 0, r_{0} = 1)→H and H_{ext}(l_{0} = 0, r_{0} = 2)→. Notice that this is possible because, like H of Equation 10, the Hamiltonian H_{ext}(l_{0} = 0, r_{0} = 1) is bilinear in the operators and pertaining to the corresponding “small” molecule. Obviously, this procedure can be reiterated by indefinitely pushing the right end of the extended molecule toward the right. Likewise, by interchanging the “left” and “right” labels one can gradually include (in addition to sites from the left electrode also) sites belonging to the right electrode (l_{0} = 1, 2,... in Equation 12).
Demonstration that the representations based on the small molecule and minimally extended molecule yield identical physical properties
In the following, the quantities needed to compute the relevant properties are discussed for the cases where the small molecule and the minimally extended molecule are chosen as the central regions. All mathematical details and expressions needed for this explanation are given in Supporting Information File 1. Some key results are summarized in Table 1.
Table 1: Hopping integrals at the left and right contacts (), transmission function , diagonal elements of the lesser Green’s function , and the local density of states D_{μ,μ} for the cases where either the small molecule or the minimally extended molecule (as defined in the main text) are chosen to be the central region. The quantities in the same typeface (magenta, blue, and red) color are equal (cf. Equation 19, Equation 24, and Equation 20).
central region  

small molec.  τ_{L}  τ_{R}  
min. ext. molec.  τ_{L}  t_{R} 
As developed in Supporting Information File 1 from Equations S4, S8, S12 and S14 (for η = 1)
and therefore the ratio of the two transmission functions equals unity:
This fact is a direct consequence of Equation S11 in Supporting Information File 1. The equality of the local density of states computed in the small molecule and minimally extended molecule representation,
is the consequence of Equation S13a (for ξ = 1), Equation S13b (for μ = ν), Equation S4, Equation S8, and Equation S12 (1 ≤ η ≤ N) as defined in Supporting Information File 1. Equation S13a (for η = 1), S14, S4, S8, and S12 yield
For the presently considered case, Equation 4 reads
and then Equation 3 yields
To sum up, Equation 19, Equation 20, and Equation 24 demonstrate that relevant nonequilibrium properties obtained within the small molecule and the minimally extended molecule representations coincide.
Approximations of wide and flatelectrode bands and related issues
At first sight, the concept of the invariance of the transport properties upon the choice of the extended molecule size may seem of merely academic interest (possibly a part of a Ph.D. tutorial) or useful for checking the correctness of numerical code to compute transport properties (which should not change whatever the size of the central region chosen). To see that the results presented above are also relevant for more pragmatic purposes, the effect of negative differential resistance (NDR) is discussed next in conjunction with common approximations used in transport approaches. In the preceding subsection, it is shown that the invariance of the transport is the direct consequence of Equations S10 and S11 from Supporting Information File 1, which follow from the exact expressions in Equation 15 and Equation 14. Whether they are also satisfied when approximate expressions are used instead of the exact ones will be discussed below. Here, the commonly employed limits of wide and flatelectrode bands (labeled WBL and FBL, respectively) will be considered. Assuming embedding selfenergies of the form (x = L,R),
for all ε (wideband limit, WBL), all energies in the Fermi window,
contribute to the integration entering Equation 7. This is an ubiquitous approximation not only in transport studies based on model Hamiltonians [1,4]; it is also an attractive approximation for realistic calculations (particularly for more reliably but computationally much more expensive approaches beyond NEGFDFT treatments), as computation times can be radically reduced; see section “WBLbased schemes and realistic calculations” for more details. The restriction expressed by Equation 26 is imposed by the difference of the Fermi distributions, which are step functions at zero temperature, the case which is referred to below. The method based on Equation 25 and Equation 26, to which, as usual, is referred to as the wideband limit (WBL), comprises in fact two approximations. It assumes (i) featureless (flat) electrode bands, characterized by energyindependent densities of states, and therefore Σ_{x}(ε) is taken at the zerobias Fermi energy ε = μ_{x} = E_{F} = 0 and (ii) electrode bandwidths of ≈4t_{L,R} much larger than any other characteristic energies (e.g., orbital energy offsets relative to the Fermi level and widths functions Γ_{L,R}). This is the rationale to extend the integration over energy in Equation 7 up to infinity. Another possible approximation, which will be referred to as the flatband limit (FBL) below, is to consider the energyindependent embedding selfenergy, Σ_{x}, given by Equation 25 but only in the finite energy range z_{x} = ε − μ_{x} < 2t_{x}. The flatband limit (FBL) is defined by Equation 27.
By taking ε = 0 and V = 0, Equations S10 and Equation S11 from Supporting Information File 1 are satisfied when the approximate expressions of Equation 25 and Equation 27 are employed. In this case, the approximate selfenergies from Equation 25 and Equation 27 coincide with the exact one (Equation 15). Via Equation 8 and Equation 2 (see also Table 1), this implies that conductance and local density of states at equilibrium (V = 0) and zero temperature computed within the wide or flatband approximations are exact. Therefore, these quantities (as well as other properties corresponding to ε = V = 0) do not depend on the size of the central region. However, a straightforward inspection reveals that, in general (i.e., at arbitrary values of ε and V), Equations S10 and S11 from Supporting Information File 1 are no longer satisfied when the approximate expressions of Equation 25 and Equation 27 are employed. So, in general, the wide and flatband approximations do yield properties that depend on the size of the central region.
The following presents a further elaboration on this aspect, which is unphysical. In the illustrative examples presented below, the (small) molecule will be described as a single site (or level) of energy ε_{M} whose Hamiltonian is given by
where N ≡ 1, c ≡ c_{1}≡ c_{N}, in Equation 10. Notice that Equation 28 does not necessarily refer to a system consisting of a single site. It can describe a molecule wherein, as is often the case [2224], the transport is dominated by a single molecular orbital.
Impact on the I–V curves
Figure 2 depicts various I–V curves computed for this case and various sizes of the extended molecule N_{ext} ≥ N = 1. While agreeing with the exact I–V curves at lower biases (eV < 2ε_{M}), the approximate I–V curves significantly differ from the exact ones at higher biases. The wideband approximation fails to describe the NDR regime exhibited by the exact I–V curves (Figure 2a). This is due to the fact that the width of the allowed energy window within the WBL (which is the Fermi energy window of Equation 26) contributing to the current (Equation 7) continuously increases with increasing V > 0. In fact, as depicted in Figure 3, this is true only for biases eV < 2t, as shown in Figure 3a. (To simplify the analysis, t_{L} = t_{R} = t is assumed.)
For biases 2t < eV < 4t, the width of the energy window of the allowed (elastic) transitions,
is narrower than that of the Fermi energy window (eV) determined by the difference of the Fermi distributions entering Equation 7. This is the straightforward consequence of the finite electrode bandwidths, as shown in Figure 3b. The fact that this energy width Δε = 2t − V/2 − (−2t + V/2) = 4t − V decreases with increasing V is reflected in a negative differential resistance (NDR) effect, which characterizes this regime. The physics underlying Equation 29 is correctly accounted for within the flatband approximation (see the last line of Equation 27). This is why the description of the NDR effect is qualitatively correct within this approximation (cf. Figure 3b). Quantitatively, as visible in Figure 2, the FBL description of NDR is rather poor. This occurs because the energy window of the allowed transition (when correctly accounted for) is
and not the Fermi energy window of Equation 26, and thus the energy dependence of the width functions Γ_{L,R} is neglected within the FBL. As schematically shown in Figure 4, the εdependence of the width functions, which is weak at lower biases (cf. Figure 4a), becomes strong in the range defined by Equation 29, as shown in Figure 4b.
The fact that the NDR effect is overall underestimated within the FBL is due to the fact that the energydependence of Γ_{L,R} is substantial (Equation 16). This yields a significant NDR even at biases eV < 2t, as revealed by the inspection of the exact I–V curves in the ranges 2ε_{M} < eV < 2t (cf. Figure 2 and Figure 5). As illustrated by Figure 2a, there is no NDR within FBL for biases eV < 2t; in this bias range, the FBL and WBL results coincide (cf. Figure 3a, Equation 27 and Equation 25).
To complete this analysis, it is noted that the current vanishes for eV > 4t because there are no states available for elastic charge transfer processes (cf. Figure 3c and Figure 4c). While qualitatively this feature is correctly retained within the flatband approximation, it is ignored within the wideband approximation, which nonphysically predicts nonvanishing currents at these biases. Although the NDR effect per se is not the main focus of this paper, it is still noted that the NDR effect discussed above for a single site/level model is the consequence of the combined effect of the finite bandwidth and the energy dependence of the width functions (or, alternatively, the density of states at the contacts [21]). Electron correlations, which escape the conventional Landauer framework utilized here, can further enhance the NDR [2527]. Besides finite bandwidth and energydependent densities of states, for systems with more than one site, the potential drop across the molecular bridge can be an additional source of NDR [28,29]. In accord with the general arguments presented above, Figure 2b,c illustrates that currents computed within the WBL and FBL, respectively, do depend (and significantly so) on the choice of the central region. The sizes utilized in these figures (up to N_{ext} = 9) mimic current NEGFDFT transport calculations based on extended molecules including up to four adjacent layers from each electrode. Because the extended molecule considered in the present study is treated exactly, one may expect that by sufficiently increasing the size of the extended molecule, currents computed within the WBL and FBL would approach the values computed exactly. From this perspective, the results presented in Figure 5 are interesting. They show that even for extended molecules that are much larger than NEGFDFT calculations can handle (given presently available computing resources), quantitative and qualitative deviations from the exact results for current beyond the ohmic regime are substantial.
The handful of examples presented above neither aimed at an exhaustive comparison of the exact results with those deduced within the wide and flatband approximations nor at a detailed discussion of NDR effects. They mainly aimed at demonstrating that in addition to the unphysical fact of breaking the invariance of the properties upon varying the size of the extended molecule, these approximations overlook significant physical effects. Even worse, as will be discuss next and anticipated in the Introduction, the WBL and FBL can predict spurious structures in the I–V curves at higher voltages. The examples depicted in Figure 5 fully support this idea: As visible there, the features exhibited by the WBL and FBL curves at high biases (oscillations, shoulders, inflection points), which have no counterpart on the exact curves, simply represent artifacts of inadequate approaches.
Impact on the charge distribution
The unphysical dependence on the size of the extended molecule predicted by the aforementioned approximations does not only affect I–V curves and other nonequilibrium properties (which imply V ≠ 0), but also charge densities (or occupancies of molecular orbitals) at equilibrium (V = 0), to which energies other than the Fermi energy (ε = 0) contribute (cf. Equation 3) are affected. As a first example in this context one can mention the Friedel sum rule. The Friedel sum rule establishes an “exact” (see below why this word is put in quotation marks) relationship between two conceptually different quantities – ohmic conductance and level occupancy – for a nontrivial, singlelevel Hubbard–Anderson model. This model cannot be solved exactly in the general case because of the onsite Hubbard–Anderson interaction (U ≠ 0) between electrons of opposite spins occupying the same site/level, which is a source of (strong) electron correlations [3033]. The approximation made to deduce this “exact” result (which also applies in the case U = 0, when the Hubbard–Anderson model reduces to the uncorrelated model described by Equation 28) is nothing but the wideband approximation considered above. To see to what extent the “exact” Friedel sum rule is affected by the WBL one can compare the exact level occupancy with the occupancy obtained within the WBL. For the parameter values of Figure 2, the level occupancy estimated within the WBL deviates by 19% from the exact occupancy value. Although not dramatic, the error introduced by the WBL is still significant. As a second example, Figure 6 shows the occupancies of several electrode sites at distance l from the embedded (small) molecule modeled as a single level/site computed within WBL using extended molecules up to very large sizes (N_{ext} ≤ 121, i.e., up to 60 “layers” in each electrode). While the small deviations from the exact values of occupancies of the electrode sites close to the (small) molecule (l = 1,2, Figure 6a) computed within WBL are acceptable, those for more distant sites (e.g., l = 4 and l = 10, Figure 6b) are unacceptably large. They amount to effective doping levels varying within up to approx. 10%, that is, they are comparable to the largest doping levels achieved experimentally in electronic devices of nanoscopic sizes [34,35]. These deviations of from the exact values, , act as spurious charged scattering centers and are responsible for the artifacts in the I–V curves calculated within the WBL.
Additional remarks
Impact of screening and external biases
As noted in the Introduction, the threepiece partitioning in an extended molecule linked to the left and right side of semiinfinite electrodes is a mental construct that is inherently arbitrary and should by no means affect the current and other physical properties. This is a minimal mandatory requirement for any valid theoretical approach. However, although conceptually arbitrary, for practical purposes, it is convenient that the partitioning fulfills certain conditions [36]. Unnecessary, more demanding computational effort can be avoided if the junction is partitioned such that: (i) the properties of the electrodes are homogeneous and do not differ from those of the bulk materials (metals), (ii) there is no direct interaction between the left and right electrodes, and (iii) interactions between the extended molecule and the two electrodes are merely confined to the extended molecule–electrode interfaces. “Screening” is the term under which conditions (i)–(iii) are usually listed in the context of realistic (DFT) calculations. The “extended” molecule should be taken large enough so that effects of the cluster to the (Kohn–Sham) potential outside the scattering region is screened. Outside the sufficiently large, extended molecule, screening effects should be altogether negligible, the potential should smoothly evolve into that of perfectly homogeneous bulk electrodes, and the charge distribution should match across the boundary of the scattering region and leads [37]. The inclusion of a sufficient number of additional electrode layers to satisfy these screeningrelated conditions may be an issue even for metallic electrodes and is a serious challenge for nonmetallic electrodes. One should note that the results presented in Figure 6, illustrating that even at sizes computationally prohibitive for microscopic studies, the charge density at the ends of the extended molecule computed within the WBL does not properly evolve into that of the bulk electrodes. The rather general model Hamiltonian of Equation 9 does satisfy these conditions. Condition (i) is satisfied because the electrodes’ parameters (onsite energies μ_{L,R} and hopping integrals t_{L,R}) are independent of the position (l,r). Condition (ii) is satisfied because does not contain products (e.g., ) mixing operators from different electrodes. Condition (iii) is also satisfied because merely contains products involving nearest neighboring sites at the contacts, such as and by choosing the small molecule described by the Hamiltonian H as central region, or and by choosing an extended molecule described by the Hamiltonian H_{ext}(l_{0}, r_{0}) of Equation 12, which includes l_{0} “layers” of the left electrode and r_{0} “layers” of the right electrode. As discussed in subsection “Impact of screening and external biases”, in the model employed in this paper, all conditions for perfect screening are satisfied by using the “small” molecule as central region. Normally, to meet these conditions, several electrode layers are added to the “physical” molecule. For example, using the present nomenclature, the “small” molecule consists of a benzene dithiol (BDT = “physical”) molecule and several gold layers in the benchmark Au–BDT–Au junctions. For this reason, the term “small” molecule used in this paper refers to the smallest molecule (“smallest extended molecule”) satisfying the perfect screening conditions (i) to (iii) formulated in the text. The detailed demonstration exposed above clearly reveals that the invariance discussed in this paper also holds for situations wherein both the properties of the molecule and those of the interactions at molecule–electrode interfaces are affected by the applied bias V, gate potentials V_{G}, defects (implying nonvanishing imaginary parts of the matrix elements):
Although the electron spin has not been considered in order to simplify the presentation, the invariance discussed in this paper also holds in the presence of magnetic fields. One can see, for instance, that the quantities entering the relevant equations do not contain t_{L,R} and or τ_{L,R} and separately, but only t_{L,R} and τ_{L,R}. However complicated the dependencies in the right hand side of Equation 31 and Equation 32 might be, they do not break the invariance demonstrated above. So, this invariance holds regardless of the potential profile across the junction and how the (possibly very nontrivial) potential drop depends on the electrode–molecule contact interactions. This holds regardless of how involved the selfconsistent procedures needed to determine the potential landscape are. Needless to say, these are all highly nontrivial tasks for realistic/DFT approaches, for which even the correct description of the both (positive and negative) bias polarities in simpler molecular junctions is an issue (see, e.g., the discussion related to Figure 5 in [22]).
WBLbased schemes and realistic calculations
Within all treatments of uncorrelated transport based on Equation 1 and Equation 5 (which is the case of all NEGFDFT flavors), the WBL represents a computationally attractive approximation. The fact that the WBL embedding selfenergies (hence also the width functions in Equation 6) become εindependent has a twofold advantage. This scheme enables one to perform conventional DFT calculations for a finite, isolated, extended molecule (i.e., uncoupled to semiinfinite electrodes). In principle, this can be done with any common DFT package. The implementation is easy, because the postprocessing step of adding εindependent self energies (Equation 1) does not require any DFTcode modification. A further advantage is that the diagonalization can be done before performing transport calculations. This fact drastically reduces the computational effort, as emphasized recently [38]. By definition, the WBL (as well as the FBL) amounts to replacing the exact εdependent embedding selfenergy with its value at the Fermi energy of the unbiased system, . As can be seen from the inspection of Equation 1, Equation 5, and Equation 6, the transmission at the Fermi energy (zerobias conductance) computed within the WBL coincides with the exact transmission. Therefore,
provided that G_{C}_{,0} is computed exactly, otherwise
This is a general result that applies to any exact treatment of uncorrelated transport based on the trace formula of Equation 5. Two WBL schemes have been recently utilized within an NEGFDFT framework [38], termed WBLMolecule and WBLMetal; the latter corresponds to a central region including 3 to 6 electrode layers (Figure 4b of [38]). In the present paper, the counterpart of WBLMolecule is a junction wherein the central region has the Hamiltonian H (Equation 9), and the counterpart of WBLMetal is a central region having the Hamiltonian H_{ext}(l_{0}, r_{0}) (Equation 11, 3 ≤ l_{0} = r_{0} ≤ 6). In both cases, the corresponding Hamiltonians are supplemented with WBL εindependent embedding selfenergies via Equation 1.
It is instructive to inspect the transmissions computed at arbitrary energies and V = 0 within the full NEGFDFT (fullSCF in the nomenclature of [38]), the WBLMolecule, and the WBLMetal methods shown in Figure 4b of [38].
In the light of Equation 34, the fact that those methods predict transmission values at arbitrary energies that substantially differ from each other is not at all surprising. An initially surprising point in the present analysis is that (as seen in Figure 4b of [38]) even the transmissions computed at the Fermi energy via the three aforementioned methods do also significantly differ from each other, i.e.,
In fact, as expressed by Equation 33, these transmissions should have been equal, i.e.,
only if all values of G_{C}_{,0} were exact. The differences between these values,
are due to the fact that, unlike the exact model calculation presented in this paper, neither the DFTmethod employed to treat the full embedding (full NEGFDFT), nor that for the isolated molecule (WBLMolecule), or that for the molecule merely including several electrode layers (WBLMetal) are exact. Even for a small isolated molecule (here named the WBLMolecule case), the DFT results represent nothing but more or less (in)accurate approximations. The aforementioned differences also clearly reveal that, even within treatments at the same level of theory (e.g., using the same exchangecorrelation functional and basis sets), the results for different molecular sizes are affected by (absolute and relative) errors in a different way. The above analysis also emphasizes that and why, based on “realistic” DFT stateoftheart transport calculations, it is impossible to demonstrate the invariance envisaged in the present paper. These DFTbased approaches are too inaccurate for this purpose. To eliminate differences resulting from unreliable approaches one should go beyond the DFT level and resort to elaborate manybody schemes [16,39]. These manybody schemes are numerically prohibitive even at the lowest (GW [40]) level. For this reason, to be feasible, calculations cannot avoid treating electrodes within WBL(type) approximations, justifiable only at low energies/biases. From this perspective, the results reported in this paper unfortunately do not convey a very optimistic message. While substantial theoretical improvements within the linear response limit are possible, reliable results for molecular transport beyond the ohmic range cannot be expected from elaborate ab initio manybody approaches (combined with WBL methods) even at (nowadays) numerically completely prohibitive molecular sizes (Figure 5 and Figure 6).
Conclusion
The results reported in the present paper can be summarized as follows: (i) The independence on the size of the extended molecule used to calculate both the equilibrium and nonequilibrium properties of a nanojunction is a minimal mandatory requirement of any sound theory of molecular transport. It was demonstrated that this invariance property is strictly obeyed for all molecules that can be described within singleparticle pictures linked to chainlike electrodes. To the best of our knowledge, the present paper is the first rigorous study and demonstration of this invariance, which is a nontrivial result even for the simplest case of a molecule modeled as a single energy level. Real systems described within this framework include, e.g., atomic chains, quantum wires, carbon nanotubes, and (possibly DNAbased) bio and large organic molecules. To determine the model parameter values, density functional based tight binding (DFTB) frameworks [1820] represent the stateoftheart. It is worth emphasizing the generality of the demonstration given in this study. The description of the molecules considered in this paper goes beyond conventional tightbinding nearestneighbor (extended Hückel) approximations, wherein, for an Nsite molecular chain, the only nonvanishing matrix elements are the onsite energies H_{μ,μ} = α (1 ≤ μ ≤ N) and the nearestneighbor hopping integrals H_{μ,μ+1} = H_{μ+1,μ} = −β (1 ≤ μ ≤ N − 1). Moreover, the Hermitean Hamiltonian matrix H of the “physical” molecule does not need to be real (), and the molecule does not need to be one dimensional. Provided that electron correlations are neglected (they are ruled out by the bilinear form of H of Equation 10), H can include interactions with impurities, applied electric fields (e.g., source–drain bias, gate potentials), static (e.g., Peierls) distortions or proximity to other molecules from the environment. The fact that the matrix elements H_{μ,ν} may depend on the bias V (e.g., Stark shift of orbital energies) [22,4143] is particularly noteworthy. (ii) Considering a specific molecule or even a specific class of homologous molecular series, the demonstration of the envisaged invariance property would have been restricted to certain fixed values of the parameters H_{μ,ν}, possibly exhibiting specific and highly nontrivial bias dependence. From this perspective, it is important to reemphasize that the invariance demonstrated in this paper holds for arbitrary values of the matrix elements H_{μ,ν} and for arbitrary dependencies (e.g., on biases) of these matrix elements. Therefore, in particular, it is not limited to some nanojunctions based on certain molecular species. (iii) Further, it was shown that unlike the exact approach, the approximate approaches based on the limits of wide and flatelectrode bands nonphysically predict nonequilibrium properties that depend on the size of the central region utilized in calculations. In conjunction with these approximations, the effect of negative differential resistance (NDR) was discussed. It was found that, although qualitatively correct, the quantitative treatment of NDR is unsatisfactory within the FBL. Because the quantitative difference between the exact treatment and FBL is the energy dependence of the width functions Γ_{L,R} (or, equivalently, the density of states at the contacts), this finding can be reframed as an indication that achieving NDR effects stronger than obtained so far [44] may primarily be a problem of contact engineering. (iv) The analysis done in conjunction with the WBL and FBL has made it clear that studies on nonequilibrium transport properties at finite temperatures require more elaborated theoretical levels than studies on linear response (V→0) properties at zero temperature. Lowtemperature properties like the low bias conductance (and therefore also the related β factor of exponential attenuation with increasing molecular size) as well as other properties (e.g., thermopower [45]), to which only energies very close to the equilibrium Fermi level contribute, can be obtained within the WBL/FBL. However, studies on the properties away from equilibrium (besides I–V curves one can also mention current noise power [4648], for which the knowledge of transmission at the Fermi energy does not suffice [49]) should go beyond the WBL or FBL. (v) Last but not least, it was demonstrated (Figure 2b,c, and Figure 5a,b) that transport calculations with an exact treatment of electrodes adjacent to the active molecule and semiinfinite electrodes within the currently employed WBL or FBL may yield spurious “predictions” of unphysical features in theoretically calculated I–V curves at higher biases. This procedure, which is equivalent to treating electrode layers included in the extended molecule and semiinfinite electrodes at different levels of theory, resembles current procedures used in “realistic” NEGFDFT approaches. Therefore, these findings deserve consideration in molecular transport studies beyond the ohmic regime.
Supporting Information
Supporting Information File 1: Mathematical details for the demonstration that the small molecule and minimally extended molecule yield identical physical properties.  
Format: PDF  Size: 91.0 KB  Download 
Acknowledgements
This research was supported in part by the bwHPC initiative and the bwHPCC5 project [50] provided through associate computer services bwUniCluster and the JUSTUS HPC facility at the University of Ulm. Financial support provided by the Deutsche Forschungsgemeinschaft (grant BA 1799/21) is gratefully acknowledged.
References

Haug, H. J. W.; Jauho, A.P. Quantum Kinetics in Transport and Optics of Semiconductors, 2nd ed.; Springer Series in SolidState Sciences, Vol. 123; Springer: Berlin, Germany, 2008.
Return to citation in text: [1] [2] [3] [4] 
Datta, S. Electronic Transport in Mesoscopic Systems; Cambridge University Press: Cambridge, United Kingdom, 1997.
Return to citation in text: [1] [2] [3] [4] [5] 
Datta, S. Quantum Transport: Atom to Transistor; Cambridge University Press: Cambridge, United Kingdom, 2005.
Return to citation in text: [1] 
Cuevas, J. C.; Scheer, E. Molecular Electronics: An Introduction to Theory and Experiment; World Scientific Publishing: Singapore, 2010.
Return to citation in text: [1] [2] [3] [4] [5] 
Baym, G.; Kadanoff, L. P. Phys. Rev. 1961, 124, 287–299. doi:10.1103/PhysRev.124.287
Return to citation in text: [1] 
Baym, G. Phys. Rev. 1962, 127, 1391–1401. doi:10.1103/PhysRev.127.1391
Return to citation in text: [1] 
Kadanoff, L.; Baym, G. Quantum Statistical Mechanics; Benjamin: Reading, MA, U.S.A., 1962.
Return to citation in text: [1] 
Caroli, C.; Combescot, R.; Nozieres, P.; SaintJames, D. J. Phys. C: Solid State Phys. 1971, 4, 916. doi:10.1088/00223719/4/8/018
Return to citation in text: [1] [2] [3] [4] 
Caroli, C.; Combescot, R.; Lederer, D.; Nozieres, P.; SaintJames, D. J. Phys. C: Solid State Phys. 1971, 4, 2598. doi:10.1088/00223719/4/16/025
Return to citation in text: [1] 
Bâldea, I.; Köppel, H. Phys. Rev. B 2008, 78, 115315. doi:10.1103/PhysRevB.78.115315
Return to citation in text: [1] 
Bâldea, I.; Köppel, H. Phys. Rev. B 2009, 80, 165301. doi:10.1103/PhysRevB.80.165301
Return to citation in text: [1] 
Bâldea, I.; Köppel, H. Phys. Rev. B 2010, 82, 087302. doi:10.1103/PhysRevB.82.087302
Return to citation in text: [1] 
Bâldea, I.; Köppel, H. Phys. Rev. B 2011, 84, 037305. doi:10.1103/PhysRevB.84.037305
Return to citation in text: [1] 
Meir, Y.; Wingreen, N. S. Phys. Rev. Lett. 1992, 68, 2512–2515. doi:10.1103/PhysRevLett.68.2512
Return to citation in text: [1] 
Jauho, A.P.; Wingreen, N. S.; Meir, Y. Phys. Rev. B 1994, 50, 5528–5544. doi:10.1103/PhysRevB.50.5528
Return to citation in text: [1] 
Mahan, G. D. ManyParticle Physics; Plenum Press: New York, NY, U.S.A., 1990.
Return to citation in text: [1] [2] [3] [4] [5] 
Fisher, D. S.; Lee, P. A. Phys. Rev. B 1981, 23, 6851–6854. doi:10.1103/PhysRevB.23.6851
Return to citation in text: [1] 
Elstner, M.; Porezag, D.; Jungnickel, G.; Elsner, J.; Haugk, M.; Frauenheim, T.; Suhai, S.; Seifert, G. Phys. Rev. B 1998, 58, 7260–7268. doi:10.1103/PhysRevB.58.7260
Return to citation in text: [1] [2] 
Pecchia, A.; Di Carlo, A. Rep. Prog. Phys. 2004, 67, 1497. doi:10.1088/00344885/67/8/R04
Return to citation in text: [1] [2] 
DFTB codes. http://www.dftb.org/codes/ (accessed Feb 11, 2016).
Return to citation in text: [1] [2] 
Bâldea, I.; Köppel, H. Phys. Rev. B 2010, 81, 193401. doi:10.1103/PhysRevB.81.193401
Return to citation in text: [1] [2] 
Bâldea, I. Phys. Rev. B 2012, 85, 035442. doi:10.1103/PhysRevB.85.035442
Return to citation in text: [1] [2] [3] 
Bâldea, I.; Xie, Z.; Frisbie, C. D. Nanoscale 2015, 7, 10465–10471. doi:10.1039/C5NR02225H
Return to citation in text: [1] 
Xie, Z.; Bâldea, I.; Smith, C.; Wu, Y.; Frisbie, C. D. ACS Nano 2015, 9, 8022–8036. doi:10.1021/acsnano.5b01629
Return to citation in text: [1] 
Parida, P.; Pati, S. K.; Painelli, A. Phys. Rev. B 2011, 83, 165404. doi:10.1103/PhysRevB.83.165404
Return to citation in text: [1] 
Einhellinger, M.; Cojuhovschi, A.; Jeckelmann, E. Phys. Rev. B 2012, 85, 235141. doi:10.1103/PhysRevB.85.235141
Return to citation in text: [1] 
Pertsova, A.; Stamenova, M.; Sanvito, S. J. Phys.: Condens. Matter 2013, 25, 105501. doi:10.1088/09538984/25/10/105501
Return to citation in text: [1] 
Tanaka, Y.; Yonemitsu, K. J. Phys. Soc. Jpn. 2014, 83, 124704. doi:10.7566/JPSJ.83.124704
Return to citation in text: [1] 
Bâldea, I. Phys. Chem. Chem. Phys. 2015, 17, 20217–20230. doi:10.1039/C5CP02595H
Return to citation in text: [1] 
Langreth, D. C. Phys. Rev. 1966, 150, 516–518. doi:10.1103/PhysRev.150.516
Return to citation in text: [1] 
Shiba, H. Prog. Theor. Phys. 1975, 54, 967–981. doi:10.1143/PTP.54.967
Return to citation in text: [1] 
Ng, T. K.; Lee, P. A. Phys. Rev. Lett. 1988, 61, 1768–1771. doi:10.1103/PhysRevLett.61.1768
Return to citation in text: [1] 
Glazman, L. I.; Raikh, M. E. JETP Lett. 1988, 47, 452–455.
Return to citation in text: [1] 
Wang, S.; Ha, M.; Manno, M.; Frisbie, C. D.; Leighton, C. Nat. Commun. 2012, 3, No. 1210. doi:10.1038/ncomms2213
Return to citation in text: [1] 
Xie, W.; Wang, S.; Zhang, X.; Leighton, C.; Frisbie, C. D. Phys. Rev. Lett. 2014, 113, 246602. doi:10.1103/PhysRevLett.113.246602
Return to citation in text: [1] 
Reuter, M. G.; Harrison, R. J. J. Chem. Phys. 2013, 139, 114104. doi:10.1063/1.4821176
Return to citation in text: [1] 
Roland, C.; Meunier, V.; Larade, B.; Guo, H. Phys. Rev. B 2002, 66, 035332. doi:10.1103/PhysRevB.66.035332
Return to citation in text: [1] 
Verzijl, C. J. O.; Seldenthuis, J. S.; Thijssen, J. M. J. Chem. Phys. 2013, 138, 094102. doi:10.1063/1.4793259
Return to citation in text: [1] [2] [3] [4] [5] [6] 
von Niessen, W.; Schirmer, J.; Cederbaum, L. S. Comput. Phys. Rep. 1984, 1, 57–125. doi:10.1016/01677977(84)900029
Return to citation in text: [1] 
Aryasetiawan, F.; Gunnarsson, O. Rep. Prog. Phys. 1998, 61, 237–312. doi:10.1088/00344885/61/3/002
Return to citation in text: [1] 
Zahid, F.; Paulsson, M.; Datta, S. Electrical Conduction through Molecules. In Advanced Semiconductors and Organic NanoTechniques; Morkoç, H., Ed.; Academic Press: Cambridge, MA, U.S.A., 2003.
Return to citation in text: [1] 
Medvedev, I. G. Phys. Rev. B 2007, 76, 125312. doi:10.1103/PhysRevB.76.125312
Return to citation in text: [1] 
Bâldea, I. Phys. Chem. Chem. Phys. 2014, 16, 25942–25949. doi:10.1039/C4CP04316B
Return to citation in text: [1] 
Xu, B.; Dubi, Y. J. Phys.: Condens. Matter 2015, 27, 263202. doi:10.1088/09538984/27/26/263202
Return to citation in text: [1] 
Paulsson, M.; Datta, S. Phys. Rev. B 2003, 67, 241403. doi:10.1103/PhysRevB.67.241403
Return to citation in text: [1] 
Blanter, Ya. M.; Büttiker, M. Phys. Rep. 2000, 336, 1–166. doi:10.1016/S03701573(99)001234
Return to citation in text: [1] 
Djukic, D.; van Ruitenbeek, J. M. Nano Lett. 2006, 6, 789–793. doi:10.1021/nl060116e
Return to citation in text: [1] 
Kiguchi, M.; Tal, O.; Wohlthat, S.; Pauly, F.; Krieger, M.; Djukic, D.; Cuevas, J. C.; van Ruitenbeek, J. M. Phys. Rev. Lett. 2008, 101, 046801. doi:10.1103/PhysRevLett.101.046801
Return to citation in text: [1] 
Bâldea, I. J. Phys. Chem. C 2014, 118, 8676–8684. doi:10.1021/jp412675k
Return to citation in text: [1] 
bwHPCC5 Startseite. http://www.bwhpcc5.de/en/index.php (accessed Feb 11, 2016).
Return to citation in text: [1]
38.  Verzijl, C. J. O.; Seldenthuis, J. S.; Thijssen, J. M. J. Chem. Phys. 2013, 138, 094102. doi:10.1063/1.4793259 
38.  Verzijl, C. J. O.; Seldenthuis, J. S.; Thijssen, J. M. J. Chem. Phys. 2013, 138, 094102. doi:10.1063/1.4793259 
38.  Verzijl, C. J. O.; Seldenthuis, J. S.; Thijssen, J. M. J. Chem. Phys. 2013, 138, 094102. doi:10.1063/1.4793259 
1.  Haug, H. J. W.; Jauho, A.P. Quantum Kinetics in Transport and Optics of Semiconductors, 2nd ed.; Springer Series in SolidState Sciences, Vol. 123; Springer: Berlin, Germany, 2008. 
2.  Datta, S. Electronic Transport in Mesoscopic Systems; Cambridge University Press: Cambridge, United Kingdom, 1997. 
3.  Datta, S. Quantum Transport: Atom to Transistor; Cambridge University Press: Cambridge, United Kingdom, 2005. 
4.  Cuevas, J. C.; Scheer, E. Molecular Electronics: An Introduction to Theory and Experiment; World Scientific Publishing: Singapore, 2010. 
1.  Haug, H. J. W.; Jauho, A.P. Quantum Kinetics in Transport and Optics of Semiconductors, 2nd ed.; Springer Series in SolidState Sciences, Vol. 123; Springer: Berlin, Germany, 2008. 
14.  Meir, Y.; Wingreen, N. S. Phys. Rev. Lett. 1992, 68, 2512–2515. doi:10.1103/PhysRevLett.68.2512 
15.  Jauho, A.P.; Wingreen, N. S.; Meir, Y. Phys. Rev. B 1994, 50, 5528–5544. doi:10.1103/PhysRevB.50.5528 
1.  Haug, H. J. W.; Jauho, A.P. Quantum Kinetics in Transport and Optics of Semiconductors, 2nd ed.; Springer Series in SolidState Sciences, Vol. 123; Springer: Berlin, Germany, 2008. 
4.  Cuevas, J. C.; Scheer, E. Molecular Electronics: An Introduction to Theory and Experiment; World Scientific Publishing: Singapore, 2010. 
46.  Blanter, Ya. M.; Büttiker, M. Phys. Rep. 2000, 336, 1–166. doi:10.1016/S03701573(99)001234 
47.  Djukic, D.; van Ruitenbeek, J. M. Nano Lett. 2006, 6, 789–793. doi:10.1021/nl060116e 
48.  Kiguchi, M.; Tal, O.; Wohlthat, S.; Pauly, F.; Krieger, M.; Djukic, D.; Cuevas, J. C.; van Ruitenbeek, J. M. Phys. Rev. Lett. 2008, 101, 046801. doi:10.1103/PhysRevLett.101.046801 
8.  Caroli, C.; Combescot, R.; Nozieres, P.; SaintJames, D. J. Phys. C: Solid State Phys. 1971, 4, 916. doi:10.1088/00223719/4/8/018 
9.  Caroli, C.; Combescot, R.; Lederer, D.; Nozieres, P.; SaintJames, D. J. Phys. C: Solid State Phys. 1971, 4, 2598. doi:10.1088/00223719/4/16/025 
10.  Bâldea, I.; Köppel, H. Phys. Rev. B 2008, 78, 115315. doi:10.1103/PhysRevB.78.115315 
11.  Bâldea, I.; Köppel, H. Phys. Rev. B 2009, 80, 165301. doi:10.1103/PhysRevB.80.165301 
12.  Bâldea, I.; Köppel, H. Phys. Rev. B 2010, 82, 087302. doi:10.1103/PhysRevB.82.087302 
13.  Bâldea, I.; Köppel, H. Phys. Rev. B 2011, 84, 037305. doi:10.1103/PhysRevB.84.037305 
22.  Bâldea, I. Phys. Rev. B 2012, 85, 035442. doi:10.1103/PhysRevB.85.035442 
23.  Bâldea, I.; Xie, Z.; Frisbie, C. D. Nanoscale 2015, 7, 10465–10471. doi:10.1039/C5NR02225H 
24.  Xie, Z.; Bâldea, I.; Smith, C.; Wu, Y.; Frisbie, C. D. ACS Nano 2015, 9, 8022–8036. doi:10.1021/acsnano.5b01629 
2.  Datta, S. Electronic Transport in Mesoscopic Systems; Cambridge University Press: Cambridge, United Kingdom, 1997. 
4.  Cuevas, J. C.; Scheer, E. Molecular Electronics: An Introduction to Theory and Experiment; World Scientific Publishing: Singapore, 2010. 
18.  Elstner, M.; Porezag, D.; Jungnickel, G.; Elsner, J.; Haugk, M.; Frauenheim, T.; Suhai, S.; Seifert, G. Phys. Rev. B 1998, 58, 7260–7268. doi:10.1103/PhysRevB.58.7260 
19.  Pecchia, A.; Di Carlo, A. Rep. Prog. Phys. 2004, 67, 1497. doi:10.1088/00344885/67/8/R04 
20.  DFTB codes. http://www.dftb.org/codes/ (accessed Feb 11, 2016). 
44.  Xu, B.; Dubi, Y. J. Phys.: Condens. Matter 2015, 27, 263202. doi:10.1088/09538984/27/26/263202 
5.  Baym, G.; Kadanoff, L. P. Phys. Rev. 1961, 124, 287–299. doi:10.1103/PhysRev.124.287 
6.  Baym, G. Phys. Rev. 1962, 127, 1391–1401. doi:10.1103/PhysRev.127.1391 
7.  Kadanoff, L.; Baym, G. Quantum Statistical Mechanics; Benjamin: Reading, MA, U.S.A., 1962. 
2.  Datta, S. Electronic Transport in Mesoscopic Systems; Cambridge University Press: Cambridge, United Kingdom, 1997. 
8.  Caroli, C.; Combescot, R.; Nozieres, P.; SaintJames, D. J. Phys. C: Solid State Phys. 1971, 4, 916. doi:10.1088/00223719/4/8/018 
21.  Bâldea, I.; Köppel, H. Phys. Rev. B 2010, 81, 193401. doi:10.1103/PhysRevB.81.193401 
45.  Paulsson, M.; Datta, S. Phys. Rev. B 2003, 67, 241403. doi:10.1103/PhysRevB.67.241403 
16.  Mahan, G. D. ManyParticle Physics; Plenum Press: New York, NY, U.S.A., 1990. 
2.  Datta, S. Electronic Transport in Mesoscopic Systems; Cambridge University Press: Cambridge, United Kingdom, 1997. 
8.  Caroli, C.; Combescot, R.; Nozieres, P.; SaintJames, D. J. Phys. C: Solid State Phys. 1971, 4, 916. doi:10.1088/00223719/4/8/018 
17.  Fisher, D. S.; Lee, P. A. Phys. Rev. B 1981, 23, 6851–6854. doi:10.1103/PhysRevB.23.6851 
18.  Elstner, M.; Porezag, D.; Jungnickel, G.; Elsner, J.; Haugk, M.; Frauenheim, T.; Suhai, S.; Seifert, G. Phys. Rev. B 1998, 58, 7260–7268. doi:10.1103/PhysRevB.58.7260 
19.  Pecchia, A.; Di Carlo, A. Rep. Prog. Phys. 2004, 67, 1497. doi:10.1088/00344885/67/8/R04 
20.  DFTB codes. http://www.dftb.org/codes/ (accessed Feb 11, 2016). 
4.  Cuevas, J. C.; Scheer, E. Molecular Electronics: An Introduction to Theory and Experiment; World Scientific Publishing: Singapore, 2010. 
2.  Datta, S. Electronic Transport in Mesoscopic Systems; Cambridge University Press: Cambridge, United Kingdom, 1997. 
8.  Caroli, C.; Combescot, R.; Nozieres, P.; SaintJames, D. J. Phys. C: Solid State Phys. 1971, 4, 916. doi:10.1088/00223719/4/8/018 
22.  Bâldea, I. Phys. Rev. B 2012, 85, 035442. doi:10.1103/PhysRevB.85.035442 
41.  Zahid, F.; Paulsson, M.; Datta, S. Electrical Conduction through Molecules. In Advanced Semiconductors and Organic NanoTechniques; Morkoç, H., Ed.; Academic Press: Cambridge, MA, U.S.A., 2003. 
42.  Medvedev, I. G. Phys. Rev. B 2007, 76, 125312. doi:10.1103/PhysRevB.76.125312 
43.  Bâldea, I. Phys. Chem. Chem. Phys. 2014, 16, 25942–25949. doi:10.1039/C4CP04316B 
1.  Haug, H. J. W.; Jauho, A.P. Quantum Kinetics in Transport and Optics of Semiconductors, 2nd ed.; Springer Series in SolidState Sciences, Vol. 123; Springer: Berlin, Germany, 2008. 
4.  Cuevas, J. C.; Scheer, E. Molecular Electronics: An Introduction to Theory and Experiment; World Scientific Publishing: Singapore, 2010. 
16.  Mahan, G. D. ManyParticle Physics; Plenum Press: New York, NY, U.S.A., 1990. 
16.  Mahan, G. D. ManyParticle Physics; Plenum Press: New York, NY, U.S.A., 1990. 
39.  von Niessen, W.; Schirmer, J.; Cederbaum, L. S. Comput. Phys. Rep. 1984, 1, 57–125. doi:10.1016/01677977(84)900029 
16.  Mahan, G. D. ManyParticle Physics; Plenum Press: New York, NY, U.S.A., 1990. 
16.  Mahan, G. D. ManyParticle Physics; Plenum Press: New York, NY, U.S.A., 1990. 
40.  Aryasetiawan, F.; Gunnarsson, O. Rep. Prog. Phys. 1998, 61, 237–312. doi:10.1088/00344885/61/3/002 
28.  Tanaka, Y.; Yonemitsu, K. J. Phys. Soc. Jpn. 2014, 83, 124704. doi:10.7566/JPSJ.83.124704 
29.  Bâldea, I. Phys. Chem. Chem. Phys. 2015, 17, 20217–20230. doi:10.1039/C5CP02595H 
21.  Bâldea, I.; Köppel, H. Phys. Rev. B 2010, 81, 193401. doi:10.1103/PhysRevB.81.193401 
50.  bwHPCC5 Startseite. http://www.bwhpcc5.de/en/index.php (accessed Feb 11, 2016). 
25.  Parida, P.; Pati, S. K.; Painelli, A. Phys. Rev. B 2011, 83, 165404. doi:10.1103/PhysRevB.83.165404 
26.  Einhellinger, M.; Cojuhovschi, A.; Jeckelmann, E. Phys. Rev. B 2012, 85, 235141. doi:10.1103/PhysRevB.85.235141 
27.  Pertsova, A.; Stamenova, M.; Sanvito, S. J. Phys.: Condens. Matter 2013, 25, 105501. doi:10.1088/09538984/25/10/105501 
38.  Verzijl, C. J. O.; Seldenthuis, J. S.; Thijssen, J. M. J. Chem. Phys. 2013, 138, 094102. doi:10.1063/1.4793259 
38.  Verzijl, C. J. O.; Seldenthuis, J. S.; Thijssen, J. M. J. Chem. Phys. 2013, 138, 094102. doi:10.1063/1.4793259 
38.  Verzijl, C. J. O.; Seldenthuis, J. S.; Thijssen, J. M. J. Chem. Phys. 2013, 138, 094102. doi:10.1063/1.4793259 
36.  Reuter, M. G.; Harrison, R. J. J. Chem. Phys. 2013, 139, 114104. doi:10.1063/1.4821176 
37.  Roland, C.; Meunier, V.; Larade, B.; Guo, H. Phys. Rev. B 2002, 66, 035332. doi:10.1103/PhysRevB.66.035332 
30.  Langreth, D. C. Phys. Rev. 1966, 150, 516–518. doi:10.1103/PhysRev.150.516 
31.  Shiba, H. Prog. Theor. Phys. 1975, 54, 967–981. doi:10.1143/PTP.54.967 
32.  Ng, T. K.; Lee, P. A. Phys. Rev. Lett. 1988, 61, 1768–1771. doi:10.1103/PhysRevLett.61.1768 
33.  Glazman, L. I.; Raikh, M. E. JETP Lett. 1988, 47, 452–455. 
34.  Wang, S.; Ha, M.; Manno, M.; Frisbie, C. D.; Leighton, C. Nat. Commun. 2012, 3, No. 1210. doi:10.1038/ncomms2213 
35.  Xie, W.; Wang, S.; Zhang, X.; Leighton, C.; Frisbie, C. D. Phys. Rev. Lett. 2014, 113, 246602. doi:10.1103/PhysRevLett.113.246602 
© 2016 Bâldea; licensee BeilsteinInstitut.
This is an Open Access article under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The license is subject to the Beilstein Journal of Nanotechnology terms and conditions: (http://www.beilsteinjournals.org/bjnano)