Skip to main content

Full text of "Ab initio correlation effects on the electronic and transport properties of metal(II)-phthalocyanine based devices"

See other formats


Ab initio correlation effects on the electronic and 
transport properties of metal(II)-phthalocyanine 
based devices 

Arrigo Calzolari^, Andrea Ferretti^'^, Marco Buongiorno 
Nardelli^ 

^ INFM-CNR-S3 National Center on nanoStructures and bioSystems at 
Surfaces, 1-41100 Modena, Italy. 

^ Dipartimento di Fisica, Universita di Modena e Reggio Emilia, 1-41100 
Modena, Italy. 

3 CHIPS and Department of Physics, NCSU, Raleigh, NC 27695-7518 and 
CSMD, Oak Ridge National Laboratory, Oak Ridge, TN 37831-6359. 

E-mail; calzolari . arrigoOunimore . it 
Abstract. 

Using first principles calculations in the framework of Density Functional 
Theory, we investigated the electronic and transport properties of metal(II)- 
phthalocyanine {M{II)Pc) systems, both in a single molecule configuration and in 
a model-device geometry. In particular, using the Copper(II)- and Manganese(II)- 
Pc as prototypical examples, we studied how electronic correlations on the central 
metal-ion influence the analysis of the electronic structure of the system and we 
demonstrated that the choice of the exchange-correlation functional, also beyond 
the standard local or gradient corrected level, is of crucial importance for a correct 
interpretation of the data. Finally, our electronic transport simulations have 
shown that M(II)Pc-based devices can act selectively as molecular conductors, as 
in the case of Copper, or as spin valves, as in the case of Manganese, demonstrating 
once more the great potential of these systems for molecular nanoelectronics 
applications. 



PACS numbers: 

1. Introduction 

Metal(II)-phthalocyanines (MPc) have attracted a growing interest in view of 
their potential for nanoscalc appUcations, such as Hght-emitting diodes, field effect 
transistors, photovoltaic cells and single molecule devices [Ij. In particular, metal(II)- 
phthalocyanines are considered prototypical organic semiconductor materials, and 
their properties have been widely investigated in a variety of experimental conditions. 
The characteristics of the MPc/substrate interactions [21 [31 HI [5] , the effects of doping 
[HI [3 [H] , and the adsorption of additional small molecules [3] have been investigated 
with techniques such as photoemission spectroscopy [TUl[TTl[T^ll3j . scanning tunneling 
spectroscopy [HI [13 [TB] and microscopy [T71 [HI [H], electronic paramagnetic 
resonance [20], NMR [21], gas-phase electron diffraction [22], etc. However, it is 
not unfrequent that different experiments produce somewhat contradicting results 



Ab initio correlation effects on MPc-based devices 



2 




Figure 1. (color online) Atomic structure of a generic metal(II)-phthalocyanine 
(MPc) bridged between two one-dimensional gold chains in a Icad/conductor/lead 
geometry. The circle marks the phthalocyanine molecule. Different colors identify 
different chemical species: grey=C, white=H, cyan=N, gold=Au, yellow=S, 
green=M; M stands for either Copper or Manganese. In the single molecule 
configuration the Au chains are absent and the two S atoms are replaced by H 
atoms. 



[HI Uni El HH HH [2S] and a careful interpretation of these data in terms of a robust 
theoretical analysis has become mandatory. 

An accurate theoretical determination of the electronic structure of the single 
molecule is the first step towards a meaningful interpretation of these complex systems: 
understanding the sequence of the frontier molecular orbitals along with their spin 
order is of paramount importance for the analysis of photoemission, optical and 
transport experiments. 

Phthalocyanines (Pc's) are planar organic macrocycles that can host a central 
metal ion (typically a 3d transition metal) , as shown in Figure [1] Because of the 
relatively large size of the molecule, the use of ab initio approaches [26 t [27 l [28 l [29 l [30 l 
[5n [321 133] to investigate the complex metal-ligand coupling has become viable only 
recently. However, different electronic structures have been already proposed, as in 
the case of the Copper(II)-Pc (CuPc), for which a partial molecular orbital inversion 
not observed in experiments has been reported [5H1 [271 132] ■ Discrepancies like this 
can be attributed mostly to the level of description of the electronic correlations in 
the simulations. On the basis of the electron counting rules for the Pc ligands the 
central ion has a formal +2 charge. Therefore, the presence of open shell c?-metals 
requires a highly accurate description of the electron-electron interactions and of the 
spin-ordering in the MPc. 

In order to understand and resolve the ambiguities in the assignment of the 
electronic levels, we have studied how different descriptions of the correlation modify 
the electronic structure of these systems. Using the Copper(II)- and Manganese (II)- 
Pc as prototypical examples, we have demonstrated that the choice of the exchange- 
correlation (XC) functional beyond the standard local or gradient corrected level is 
of crucial importance for a correct interpretation of the data. Moreover, we have 
analyzed the effects of the electron correlation on the electronic transport through a 
model device: a M(II)Pc molecule bridged between two atomic chains of gold atoms 
in a typical lead/conductor/lead configuration, as shown in Fig. [T] This system has 
been recently proposed as a prototypical one-dimensional molecular device [141 132j . 



Ab initio correlation effects on MPc-based devices 



3 



2. Computational methodology 

Our studies are based on state-of-the-art ab initio Density Functional Theory 
calculations, as implemented in the PWSCF code [34] . At the standard level, 
the exchange and correlation functional is evaluated within the Local Spin Density 
Approximation (LSDA) |^ and the Generalized Gradient Approximation (GGA) 
as proposed in Ref. ^36J (PBE), always including spin polarization (unrestricted 
calculations). The electron-ion interaction is described by ab initio ultrasoft 
pseudopotentials J37]. For Manganese, we have included the 3s3p semicore shells in 
the valence. The electronic wavefunctions (densities) were expanded on a plane-wave 
basis set up to a kinetic energy cutoff of 25 Ry (200 Ry). The effects of the electron- 
electron correlation beyond LDA/GGA have been included using the DFT-I-U scheme, 
in the linear response approach proposed by Cococcioni and De Gironcoli in 2005 [3S\ ■ 

In all the calculations we have used supercells of dimensions 44.94 x 26.00 x 12.00 
with at least 12 A of vacuum space between adjacent replica in each direction. The 
metal/molecule/metal bridge (labeled (Au5S)2M-Pc) was modeled via one dimensional 
gold chains connected to the molecule through sulfur atoms (see Fig. 1). All the 
structures have been thoroughly relaxed. 

The transport properties of the system have been computed in a 
lead/molecule/lead geometry using the fully first principles approach implemented 
in the WanT code [321 113 • The method combines an accurate description of the 
electronic ground state, provided by ab initio DFT calculations, with the Landauer 
approach to describe transport properties of extended systems [HJ H5] . The connec- 
tion is realized by transforming the Bloch orbitals into maximally-localized Wannier 
functions [43l H^. This representation naturally introduces the ground-state elec- 
tronic structure into the real-space Green's function scheme, which is our tool for the 
evaluation of the Landauer quantum conductance. 

The WanT method has been also extended in order to include short range 
electron-electron (e-e) interactions on transport properties [13 HH H?]. The effect 
of the e-e coupling is taken into account by including a further correlation self-energy 
term in the calculation of the conductor Green's function. The e-e self-energy is 
computed here using the Three-Body Scattering formalism (3BS) [Mllinji which relies 
on an effective generalized Hubbard Hamiltonian, and it is solved in a configuration 
interaction scheme where up to three bodies are added to the non-interacting Fermi 
sea. 

3. Results and Discussions 

3.1. Copper(II)-phtalocyanine systems 

3.1.1. Single molecule. The single particle orbitals of the planar CuPc molecules can 
be classified according to the irreducible representation of the I?4/i group. In this 
representation the id orbitals split into aig {d^2), big ((ij,2„j,2), Cg {dxz,dyz) and b2g 
{dxy). One of the open problems in the interpretation of the experimental data is 
the energy position of the big orbital with respect to the ai„ (tt) and 2eg (tt*) states, 
which are the highest occupied (HOMO) and the lowest unoccupied molecular orbital 
(LUMO) of the Pc rings respectively. The big orbital has a a character and is localized 
on the copper ion, whereas the ai„ and 2eg are single and double degenerate 7r-like 
states, delocalized over the entire Pc macrocycles [see Fig. [^fa)]. 



Ab initio correlation effects on MPc-based devices 



4 




energy (eV) energy (eV) 



Figure 2. (color online) CuPc electronic structure, (a) Charge density isosurfaces 
of selected molecular orbitals and spin density {pup— pdw) for CuPc. Spin-resolved 
density of states (DOS) corresponding to (b) LDA, (c) PBE, (d) LDA+U (U= 
3eV) functionals. Shaded green areas are the projections on the atomic Cu states. 
Vertical arrows and labels identify the energy position of the peaks corresponding 
to the selected molecular orbitals, reported in panel (a). Inset of panel (d) shows 
the spin-up frontier peaks for the case U= 6.0 eV. The zero energy discriminates 
the occupied and the empty states for each system. 



In order to clarify the issue, we have calculated the electronic structure of the 
isolated CuPc molecule by using two different XC correlation functionals, namely 
LDA and PBE. The resulting density of states are reported in Fig. El^b-c). The 
analysis of the electronic structure shows the presence of mixed metal-ligand states 
in the low-energy part of the spectrum, especially between the copper ion and the 
pirole rings, the fingerprint of a direct interaction between the two subsystems. In 
Fig.O we show the big, ai„ and 2eg states that are the frontier orbitals of the overall 
CuPc molecule. In both LDA and PBE case, the aiu and the 2eg states are spin- 
degenerate {i.e. the same energy for both spin components), and the corresponding 
energy differences AE=E(2eg)-E(ai„) are very similar: 1.40 eV and 1.42 eV for LDA 
and PBE respectively. Independently from the choice of the XC functional, the CuPc 
molecule has a resulting total magnetization of 1.0 fiB, due only to the unpaired big 
state, as shown in the plot of the spin density [Fig. [D^a)]. However, the relative order 
of the states is different: in LDA [Fig. El^b)], the big orbital is the single-occupied 
HOMO (SOMO) of the overall CuPc molecule, in agreement with the results by Liao 
et al. [2S]; on the contrary, in PBE the big is the HOMO-1 state for the spin-up part, 



Ab initio correlation effects on MPc-based devices 



5 



while it is almost degenerate with 2eg orbital for the spin-down component. The 
experimental photoemission results [10] seem to indicate that the LDA configuration 
is the correct one; however, due to the small cross-section of the metallic states, no 
truly conclusive statements can be drawn [53] . It is worth noticing that this result is 
not biased by the choice of the basis set used in the calculations. We calculated the 
electronic structure of the molecule also using a localized atom-centered numeric basis 
set using the Dmol^ code [50]: with both LDA and PBE XC-functionals we obtained 
the same orbital inversion observed before. The discrepancies between LDA and PBE 
highlight the delicate balance occurring in the electronic structure of this molecule. 

In our DFT calculations the net charge transfer from Cu atom to the rest of the 
environment is symmetric and smaller than the formal one: the analysis of the Lowdin 
charges indicates that the copper atom is positively charged (5q(Cu)= -fl.02 e, and the 
Sd shell hosts ~9.4 electrons. Even though the calculated values are smaller than the 
formal ones, the correlation effects, due to the opening of the Zd shell must be taken 
carefully into account, especially if one attempts to use the ground state structure as 
the basis for the interpretation of effects (such as doping) that imply the variation of 
the oxidation/reduction state of the central ion. 

In order to gain further insights on this issue, we introduced electronic correlation 
effects beyond standard DFT by including a local Hubbard U term on the Cu ion. We 
tested several values of the U parameter (U=0.5, 2.0, 3.0, 4.0, 6.0 eV); the results for 
U=3 eV are shown in Fig. Hj^d). The Hubbard term operates mainly on the energy of 
the ion-centered states, leaving the Pc ones unchanged. We note a downward (upward) 
shift of the big state for the spin-up (spin-down) component, respectively. The net 
effect is the opening of the SOMO-LUMO gap. By increasing the value of the Hubbard 
U, the shift of the big peak increases accordingly. In particular, for U=0.5 and 2.0 
eV, both DOS resemble the LDA spectra of Fig. H^b). For U=3.0 eV, even thought 
big is still the SOMO, it is almost degenerate with the spin-up aiu state [Fig. HKd)]. 
This effective degeneracy is observed also for the other two values of the Hubbard 
parameter (U=4.0, 6.0 eV) that we considered. 

By further increasing U, we observe a slight shift of the big peak, that leads 
to an inversion of the big and ai„ states for U= 6.0 eV [see inset of Fig. [Hd)]. 
This trend confirms how the order of the single particle states is strongly related 
to the specific treatment of the electron-electron correlation. However, an on-site 
correlation parameter U=6.0 eV is a rather high value, greater than any of those 
experimentally measured for similar systems [51]. Thus, it is reasonable to conclude 
that the LDA order of the states is the correct one, while the exact position of the 
big peak wrt the Pc's states depends on the correct value of the electron correlation. 
From our calculations it results also that in the case of CuPc the PBE XC-functional 
overestimates the e-e correlation, as if a strong Hubbard potential were applied [.52j . 

3.1.2. Device configuration. The system we studied is inspired by a prototypical 
one-dimensional single- molecule device originally proposed by Nazin et al. |14j . The 
original system was realized placing a CuPc molecule between two short chains of gold 
atoms, deposited on a NiAl substrate. The electronic structure was then characterized 
by STS measurements. The metallic character of the substrate prevented the direct 
measurement of I-V characteristics through the double interface. The coherent 
transport properties of the open device have been calculated so far by Tada and 
coworkers [32], who found a quantum conductance of the order of G = 3 x 10""* 
e^/ft.. This value, smaller than the typical ones (G = 0.01 e^//i) observed for aromatic 



Ab initio correlation effects on MPc-based devices 



6 




-1 1 

energy (eV) 



-2 -1 
energy (eV) 



Figure 3. (color online) (a) Spin-polarized DOS of the periodic (Au5S)2CuPc 
system. Shaded red area (thick blue line) is the projection on the CuPc (S-Au) 
atomic states, (b) Mean field (shaded area) and correlated (straight and dashed 
lines) two-terminal transmittance of open CuPc-based device, spin-resolved. The 
zero energy references are set to Fermi level of the periodic (Au5S)2CuPc system. 



molecules [S3], is mostly due to the weak gold/CuPc interaction, i.e to the absence 
of a direct chemical bond. For these reasons we slightly modified the geometry of 
the system, connecting the molecule and the leads, through a sulfur bridge (Fig. [T]). 
The choice of the sulfur functionalization is justified by the well-known ability of thiol 
groups to anchor molecules to gold electrodes [Mj . 

Following the method f2(T described in Section [21 we have simulated an open 
device where the Au wires constitute the semi-infinite leads and the CuPc acts as the 
conductor. On the basis of the results for a single molecule, we calculated the ground 
state electronic structure using LDA XC-functional. The spin-resolved electronic 
structure of the periodic (Au5S)2CuPc building block [Fig. [SJa)] is characterized by 
a finite density of states at the Fermi level (Ei?=0 eV): while the formation of the 
nanojunction induces a realignment of the molecular and gold orbitals, the CuPc 
molecule perturbs the pure one-dimensional features of the infinite Au chain, such as 
the van Hove singularities, and the dispersive sc?-band crossing the Fermi level [55|,l56j. 
As expected, the presence of the thiol bridge at the metal/molecule interface favors the 
mixing of the molecular orbitals. Indeed, the overall SOMO for the spin-up component 
is the result of the hybridization of the dclocalized sd-like states of the pristine chain 
and the big state of the CuPc molecule. Other hybridized orbitals, which involve also 
the aiu state of the molecule, are recognizable in the energy range from E — —0.5 eV 
to Fermi level. The (Au5S)2CuPc system maintains a global magnetic polarization 
of 1.0 ^B, due to the single occupation of the big state derived from the isolated 
molecule. However, the effect of this unbalance on the overall transmission is reduced 
by the spin-unpolarized density of states of the gold wire. 

As a starting point, we calculated the electronic transport at the mean field 
level: we treated the electron correlation within DFT (LDA), by using the Landauer 
formula [32] ■ In particular, we focused on the zero bias regime; in this case the quantity 
to be calculated is the quantum transmittance T[E). The value of the transmittance 
at the Fermi level gives the quantum conductance. 

The results for the spin-up and spin-down components of the quantum 
transmittance (logarithmic scale) are shown in Fig.[3](b). The corresponding quantum 
conductance are Gup = 0.059 e^//i and Gdw = 0.064 e^//i. These values are reasonably 



Ab initio correlation effects on MPc-based devices 



7 



high, when compared with those calculated for the unbound system |32j or with the 
gold chain [57j (which gives almost G = 1.0 2e^/h). This is due to the important 
mixing of the CuPc orbitals with the sd-bands of the gold wire. In other energy 
regions, such as in the range E € [—2.0, —0.5] eV, that corresponds to the 5d band of 
the Au wire, the orbital mixing is very small, leading to a localization of the electronic 
states and a corresponding reduction of the transmittance. 

As a further test, we calculated the transport properties starting from PBE-DFT 
calculations. In that case the inversion of the CuPc states near the Fermi level is 
almost irrelevant, due to the same coupling with the unpolarized sd manyfold of the 
gold wire. 

In order to take into account the effect of the electron correlation, we evaluated the 
additional e-e self-energy operator I]e_e(a;). The latter is calculated in the framework 
of the so-called 3BS formalism [48l [49] . 3BS treats the electron correlation in a full 
many body approach beyond the single particle approximation. The resulting self- 
energy is non-hermitian and frequency-dependent and the inclusion of this quantity 
in the evaluation of quantum transmittance may introduce incoherent and dynamical 
effects on transport that are not accessible [H] within mean field techniques, like e.g 
DFT+V. 

The BBS scheme requires the inclusion of an on-site Hubbard-like term U for the 
Cu ions. We looked at the effects induced by two values of the parameter, U=3.0 and 
U=6.0 eV. The results are summarized in Fig. [DJb). Similar to the single molecule 
case, the effect of correlation does not change radically the character of the up and 
down transmittance spectra. This is in agreement with the absence of Cusd-derived 
states close to the Fermi level. The displacement of the big state is not sufficient to 
modify the conduction properties of the device. The main differences are in the region 
E G [—2.0, —0.5] eV, where an enhancement of the transmittance is observed. This is 
mainly due to the real part of the e-e self-energy, which shifts the molecular orbitals, 
leading a higher alignment with the lead states. The amount of such realignment 
depends on the value of the selected Hubbard U term, and it is not strictly related to 
a further hybridization of the conductor/lead molecular orbitals. 

3.2. Manganese(II)-phtalocyanine systems 

3.2.1. Single molecule. Let's now consider how the electronic structure of the 
molecule is modified by the presence of a highly correlated Mn central cation that 
substitute the original Copper. Due to the half occupation of the strongly localized 
Sd shell, LDA is expected to fail in the description of the electronic structure of the 
manganese atom. Therefore, we followed previous theoretical calculations [33] in using 
a gradient corrected (PBE) XC functional. 

The relaxed MnPc molecule still exhibits a planar square-like geometry, as in 
the case of CuPc and we can still label the single molecular orbitals according to 
the D/ih symmetry. However, due to different ionic radius and different metal d 
orbital occupancy, we observe a contraction (~ 1.5%) of the Mn-N bond length and a 
reduction of the metal-ligand ionicity, resulting in a stronger metal-molecule coupling. 
The spin-resolved density of states of the isolated MnPc molecule is displayed in Fig. 
ID^a). For the spin-up channel ai„ and 2eg states are the single occupied H0M0(7p 
and LUM0;7p respectively, while the hig orbital is empty. The energy difference 
AE=E(2eg)-E(ai„)=1.41 eV is very similar to the CuPc case. On the contrary, the 
spin-down axu state correspond to the HOMO-1. The spin-down HOMO and LUMO 



Ab initio correlation effects on MPc-based devices 



8 




energy (eV) energy (eV) 



Figure 4. (color online) MnPc electronic structure. Spin-resolved density of 
states (DOS) corresponding to (a) PBE calculation and (b) PBE+U (U= 3eV) 
functionals. Shaded pink areas are the projections on the atomic Mn states. 
Vertical arrows and labels identify the energy position of the peaks corresponding 
to molecular orbitals described in text. 



are given by the almost degenerate (AE=0.16 eV) duplet eig (one state occupied, 
one empty). These are 7r-like orbitals extended over both the Mn ion and the Pc 
rings. The corresponding spin-up states are instead completely occupied, where the 
double degenerate 2eg peak is shifted at higher energy and split into two single 
peaks, separated by (AE=0.07 eV). Furthermore, they both exhibit a great 3d-Mn 
component. It is worth noticing that between the eig and the 2eg duplets there are 
two other empty states, b2g and aig, fully occupied in the spin- up spectrum, that are 
(T-like orbitals localized on the central ion. Finally, the hig peak is more than 3eV 
above the spin-down HOMO. The MnPc molecule has a total magnetization of 3.0 
jiB, resulting from the unbalancing of three orbitals (eig,62g and aig) between the up 
and down components. 

The small SOMO-LUMO gap, along with the presence in the same energy region 
of several molecular orbitals mostly localized on the metallic site, deserves further 
investigation in view of the strongly correlated character of the Manganese atom. 
To this end, we improved the description of the electronic structure by including a 
Hubbard U term on the central ion (DFT-fU). We considered two cases, U=2.0 eV and 
U=3.0 eV; the results for the latter value are shown in Figure |H[b). Due to stronger 
correlations effects, even the presence of small Hubbard potential strongly modifies 
the electronic properties of the system. We observe a general downward (upward) shift 
of the occupied (empty) Mn components in the density of states. The effect of the 
correlation is weaker for spin-up states, since the frontier orbitals have a predominant 
Pc character. On the contrary, in the spin down case the on-site e-e correlation 
increases the SOMO-LUMO gap, splitting the almost degenerate eig peak with an 
energy separation that is a function of the applied U: AE=0.39 eV and AE=0.46 eV 
for U=2.0 eV and U=3.0 eV respectively. The 2eg duplet also splits in two separates 
contributes as shown in Figure [IJb) . The effect of the correlation on the 2eg states is 
more relevant for the minority spin, because it includes a greater metallic contribution. 
The &2g and aig peaks migrate towards higher energies, between the split 2eg fork, 
changing the global order of the MnPc orbitals. Higher values of the local Hubbard 
term (U> 3.0 eV) do not lead to stable (converged) electronic structures, confirming 
that the results are very sensitive to the electron correlation model. 



Ab initio correlation effects on MPc-based devices 



9 




energy (eV) energy (eV) 

Figure 5. (color online) (a) Spin-polarized DOS of periodic (Au5S)2MnPc 
system. The shaded brown area (thick blue line) is the projection on the MnPc 
(S-Au) atomic states, (b) Mean field (shaded area) and correlated (straight and 
dashed lines) two-terminal spin-resolved transmittance of an open MnPc-based 
device. The zero energy references are set to the Fermi level of the periodic 
(Au5S)2MnPc system. 



3.2.2. Device configuration. We modeled the double interface (Au5S)2MnPc as 
before in the case of the CuPc. The relaxed geometry does not present any new 
characteristics, while the DFT electronic structure structure [Fig. [^a)] is more 
complex, due to the greater complexity of the MnPc molecule. While the Au 
component of the spectra is not particularly affected by the change of the metal atom, 
the molecular states display a complex reorganization. In the spin- up component, 
the molecule/metal coupling induces a broadening of the ai„ state with the sd gold 
band in the energy range Ee [—1.0, —0.25] eV, and the splitting of the pristine double 
degenerate 2eg peak into a molecular doublet, close to the empty big state. The spin- 
up HOMO and LUMO have both a sd character, derived from the gold chain, and 
are separated by an energy gap AE=0.43 eV, higher than the corresponding Cu case. 
The analysis of the single particle orbitals shows that the spin-down HOMO has also 
a pure Au character, while the spin-down LUMO is a hybrid state, stemming from the 
mixing of the sd Gold band and the ai„ state of the Phthalocyanine. The spin-down 
HOMO and LUMO are almost degenerate in energy, with a gap AE=74 meV [Fig. 
[5l[a)]. In the range EG [0.1, 0.5] eV we observe a manifold of states, which includes the 
two eig, the &2g and the aig orbitals, each partially hybridized with the Gold states. 
The split 2eg and big peaks are recognizable at higher energies. The depletion of the 
spin-down ai„ and eig states, occupied in the gas phase (Fig. [J]), leads to a further 
unbalance of ^2 electrons between the up and down components, with respect to the 
isolated molecule. This enhances the total magnetization up to ^=4.79 Bohr mag/cell, 
which may be justified in terms of charge transfer at the interface from the molecule 
towards the gold chain for the spin-down component. 

This marked spin asymmetry is reflected in the transmittance spectra, as shown in 
Fig.[5l[b). At the mean field level, the spin-up component behaves as a semiconductor 
with a vanishing conductance at zero bias. On the contrary, the non-negligible density 
of states at the Fermi energy (zero of the energy scale) in the spin-down component 
leads to a finite conductance Gdw = 0.12 e^/h. If we compare the cases of Cu- and Mn- 
Pc we observe a distinctively different behavior: in the case of Copper the reduced 
spin unbalance modifies only slightly the almost degenerate conduction properties 



Ab initio correlation effects on MPc-based devices 



10 



of the gold wires, leading to a better conductor (higher conductance) for both the 
spin channels, while for Manganese the spin asymmetry along with an enhanced 
molecule/lead coupling changes the conduction properties of the Au wire as a function 
of the spin. This is a very remarkable result that suggests that the MnPc molecular 
device could work as a spin valve. If confirmed by experiments, this would open the 
way to a new generation of devices for molecular spintronics applications. 

We have studied the correlated transmittance of the MnPc two terminal device 
for two values of the Hubbard U parameter (U=2.0, 3.0 eV) in our 3BS scheme. The 
results for the correlated transport [straight and dashed lines in Fig.[5](b)] clearly show 
that in the region near the Fermi energy the transmittance is almost insensitive to 
the inclusion of the electron-electron correlation. In particular ,the spin valve behavior 
of the MnPc-based device is stable upon the inclusion of the e-e coupling, and only 
minor modifications at EG [—2.0, —1.0] eV are induced by the realignment of the lower 
energy molecular orbitals, as in the case of CuPc system. 

This result can be understood in terms of the electronic structure [Fig. O^a)]. 
The strong molecule/gold bonding leads to a removal of the Mna^ states from the 
Fermi level to higher energies, while the only molecular components close to i? = 
eV are related to the aromatic Pc rings (ai„), only slightly affected by the electron 
correlation. In other words, the short range electron correlation truly affects transport 
when the simultaneous mixing of the delocalized states of the leads and the partially 
occupied states of the correlated conductor occurs at the Fermi energy [ISl [31] . As 
a consequence, both Cu- and Mn-Pc transmittances are almost unaffected by the e-e 
correlation. 

4. Conclusions 

In conclusion, we have demonstrated that a proper inclusion of electronic correlation 
effects is essential to develop a complete understanding of the the electronic and 
transport properties of phthalocyanine- based nanostructures. Considering different 
forms for the exchange-correlation functional (LDA, PBE, DFT-I-U) in the standard 
DFT, as well as the inclusion of on-site correlation beyond DFT (BBS), we have proved 
that the inclusion of correlation reproduces the correct orbital sequence in CuPc, as 
observed in experiments. We should stress that our results are valid for the molecule in 
its ground state, so that a comparison with experiments involving oxidized or reduced 
states of the molecule are not physically meaningful. Moreover, electronic correlation 
have clearly an even greater effect in the MnPc molecule, due to the open 3d shell of 
the Manganese ion. We do not exclude also that other kind of effects, such as those 
related to the Jahn- Teller distortions, might play an important role in the electronic 
properties of these systems, as proposed in the case of alkali doped MPc crystals |58j . 

The transport properties of these M(II)Pc-based devices are less sensitive to 
the inclusion of electron correlation effects than the corresponding single molecule 
electronic structures. The e-e interaction seems in fact to be partially quenched by 
the strong coupling with the uncorrelated gold leads. This observation confirms the 
fact that, due to the reduced number of electronic states, the transport properties of 
a molecular device can not be directly inferred from the electronic structure of the 
isolated molecule. On the basis of our transport results, we can not exclude a priori, 
that effects of the correlation might be important in other device configurations, where 
the molecule/lead coupling is different. 

Finally, our electronic transport simulations have shown (irrespective on the 



Ab initio correlation effects on MPc-based devices 



11 



details of the calculations) that M(II)Pc-based devices can act selectively as 
molecular conductors, as in the case of Copper, or as spin valves, as in the case 
of Manganese, demonstrating once more the great potential of these systems for 
molecular nanoelectronics applications. 



Acknowledgments 

Carlo Cavazzoni, Franca Manghi and Silvia Picozzi are gratefully acknowledged for 
fruitful collaborations and discussions. Funding was provided in part by the Regional 
Laboratory of EmiliaRomagna "Nanofaber" ; by the Italian MIUR through PRIN 2006, 
and by the Department of Energy of the US government. Computing time at the 
CINECA supercomputing facilities was provided by INFM-CNR. 

References 

[1] Kadish K M, Smith K M and Guilard R, cds 2003 Application of phtlialocyaniries vol 19 of The 

Porphyrin Handbook (Academy Press, San Diego) 
[2] Aristov V Y, Molodtsova O V, Zhilin V M, Vyalikh D V and Knupfer M 2005 Phys. Rev. B 72 

165318 

[3] Lei S B, Deng K, Yang D L, Zeng Q D and Wang C 2006 J. Phys. Chem. B 110 1256-1260 
[4] Cossaro A, Cvetko D, Bavdek G, Floreano L, Cotter R, Morgante A, Evangelista F and Ruocco 

A 2004 J. Phys. Chem. B 108 14671 
[5] Suzuki T, Kurahashi M, Ju X and Yamauchi Y 2002 J. Phys. Chem. B 106 11553-11556 
[6] Craciun M F, Rogge S and Morpurgo A F 2005 ,7. Am. Chem. Sac. 127 12210-12211 
[7] Ding H and Gao Y 2006 Appl. Surf. Sci. 252 3913 3947 
[8] Gao Y and Yan L 2003 Chem. Phys. Lett. 380 451-455 
[9] Szczpaniak B and Bragiel P 1995 Vacuum 46(5) 465-467 
[10] Downes J E, McGuinness C, Glans P A, Learmonth T, Pu D, Sheridan P and Smith K E 2004 

Chem. Phys. Lett. 390 203-207 
[11] Kurmaev E Z, Shamin S N, Galakhov V R, Moewes A, Otsuka T, Koizume S, Endo K, Katz 

H E, Bach M, Neumann M, Ederer D L and Iwami M 2001 Phys. Rev. B 64 045211 
[12] Lozzi L, Granato V, Picozzi S, Simeoni M, La Rosa S, Delly B and Santucci S 2006 

J. Vac. Sci. Technol. A 24(4) 1668-1675 
[13] Schwieger T, Peisert H, Golden M S, Knupfer M and Fink J 2002 Phys. Rev. B 66 155207 
[14] GVNazin, Qiu X H and Ho W 2003 Saerice 302 77-81 
[15] Nazin G V, Wu S W and Ho W 2005 PNAS 102(25) 8832-8837 
[16] Mikaelian G, Ogawa N, Wu X and Ho W 2006 .J. Chem. Phys. 124 131101 
[17] Lippel P H, Wilson R J, Miller M D, WoU C and Chiang S 1989 Phys. Rev. Lett. 62(2) 171-174 
[18] Lu X, Hipps K W, Wang X D and Mazur U 1996 J. Am. Chem. Soc. 118 7197-7202 
[19] Hipps K W, Lu X, Wang X D and Mazur U 1996 ,/. F'hys. Chem. 100 11207-11210 
[20] Finazzo C, Calle C, StoU S, Van Doorslaer S and Schwcigcr A 2006 Phys. Chem. Chem. Phys. 8 
1942-1953 

[21] Filibian M, Carretta P, Miyake T, Taguchi Y and Iwasa Y 2007 Phys. Rev. B 75(8) 085107 
[22] Mastryukov V, , Ruan C Y, Fink M, wang Z and Pachter R 2000 J. Mol. Struct. 556 225-237 
[23] GiovaneUi L, Vilmercati P, Castellarin-Cudia C, Themhn J M, Porte L and Goldoni A 2007 

J. Chem. Phys. 126(4) 044709 
[24] Molodtsova O V, Zhihn V M, Vyalikh D V, Aristov V Y and Knupfer M 2005 J. Appl. Phys. 98 

093702 

[25] Schwieger T, Knupfer M, Gao W and Kahn A 2003 Appl. Phys. Lett. 83(3) 500-502 

[26] Day P N, Wang Z and Pachter R 1998 J. Mol. Struct.-Theochem 455 33-50 

[27] Rosa A and Baerends E J 1994 Inorg. Chem. 33 584-595 

[28] Liao M S and Scheiner S 2001 J. Chem. Phys. 114(22) 9780-9791 

[29] Lozzi L, Santucci S, La Rosa S, Delley B and Picozzi S 2004 J. Chem. Phys. 121(4) 1883-1889 
[30] Nguyen K A and Pachter R 2001 J. Chem. Phys. 114(24) 10757-10767 
[31] Bialck B, Kim I G and Lee J I 2003 Thin Solid Films 436 107-114 

[32] Tada T, Hamayama S, Kondo M and Yosliizawa K 2005 J. Phys. Chem. B 109 12443-12448 

[33] Liao M S, Watts J D and Huang M J 2005 Inoiy. Chem. 44 1941-1949 

[34] Baroni S, Dal Corso A, de Gironcoli S and Giannozzi P 2001 http://www.pwscf.org 



Ab initio correlation effects on MPc-based devices 



12 



[35] Perdew J P and Zunger A 1981 Phys. Rev. B 23(10) 5048-5079 

[36] Perdew J P, Burke K and Ernzerhof M 1996 Phys. Rev. Lett. 77(18) 3865-3868 

[37] Vanderbilt D 1990 Phys. Rev. B 41(11) R7892-R7895 

[38] Cococcioni M and de Gironcoli S 2005 Phys. Rev. B 71(3) 035105 

[39] Calzolari A, Ferretti A, Cavazzoni C, Marzari N and Buongiorno Nardelli M 2005 WanT code, 

http: //www. wannier-transport . org. 
[40] Calzolari A, Marzari N, Souza I and Buongiorno Nardelli M 2004 Phys. Rev. B 69(3) 035108 
[41] Landauer R 1970 Philos. Mag. 21(172) 863-867 

[42] Datta S 1995 Electronic transport in mesoscopic systems (Cambridge University Press, 
Cambridge) 

[43] Marzari N and Vanderbilt D 1997 Phys. Rev. B 56(20) 12847-12865 

[44] Ferretti A, Calzolari A, Bonforroni B and Di Felice R 2007 J. Phys. Condens. Matter. 19 036215 
[45] Ferretti A, Calzolari A, Di Felice R, Manghi F, Caldas M J, Buongiorno Nardelli M and Molinari 

E 2005 Phys. Rev. Lett. 94(11) 116802 
[46] Ferretti A, Calzolari A, Di Felice R and Manghi F 2005 Phys. Rev. B 72(12) 125114 
[47] Darancet P, Ferretti A, Mayou D and Olevano V 2007 Phys. Rev. B 75(7) 075102 
[48] Calandra C and Manghi F 1994 Phys. Rev. B 50(4) 2061-2073 
[49] Ferretti A, Boni V and Manghi F 2006 http://www.s3.infni.it/3bs . 
[50] Delley B J 2000 Comp. Mater. Sci. 17 122-126 

[51] The screened value of the local Hubbard potential may be evaluated in the self consistent 
calculation as the derivative of the total energy with respect to the number of electrons, 
as described in Ref. 1381 . However this goes beyond the aim of the present work. Here, we are 
not interested in the exact evaluation of the e-e interaction, but in the qualitative description 
of the effects of the electron correlation. 

[52] We observed this effect in this specific CuPc molecular structure, thus this conclusions can not 
be straightforwardly extended to other copper-based system, without deep further analysis. 

[53] Xu B and Tao N J 2003 Science 301 1221-1223 

[54] Schreiber F 2000 Prog. Surf. Sci. 65 151-256 

[55] Calzolari A, Cavazzoni C and Buongiorno Nardelli M 2004 Phys. Rev. Lett. 93(9) 096404 

[56] Calzolari A and Buongiorno Nardelli M 2005 Phys. Rev. B 72 045416 

[57] Agrait N, Yeyati A L and van Ruitenbeek J M 2003 Phys. Rep. 377(2-3) 81-279 

[58] Tosatti E, Fabrizio M, Tobik J and Santoro G E 2004 Phys. Rev. Lett. 93(11) 117002