Quantum interference in coherent molecular conductance

Abstract

Coherent electronic transport through individual molecules is crucially sensitive to quantum interference. Using exact diagonalization techniques, we investigate the zero-bias and zero-temperature conductance through $\pi$-conjugated annulene molecules (modeled by the Pariser-Parr-Pople and Hubbard Hamiltonians) weakly coupled to two leads. We analyze the conductance for different source-drain configurations, finding an important reduction for certain transmission channels and for particular geometries as a consequence of destructive quantum interference between states with definite momenta. When translational symmetry is broken by an external perturbation we find an abrupt increase of the conductance through those channels. Previous studies concentrated on the effect at the Fermi energy, where this effect is very small. By analysing the effect of symmetry breaking on the main transmission channels we find a much larger response thus leading to the possibility of a larger switching of the conductance through single molecules. Comment: 4 pages, 5 figures. Accepted in Phys. Rev. Lett

Full text

Page 1

arXiv:0911.4193v1 [cond-mat.str-el] 21 Nov 2009
Quantum interference in coherent molecular conductance
Juli´ an Rinc´ on, K. Hallberg
Centro At´ omico Bariloche and Instituto Balseiro,
Comisi´ on Nacional de Energ´ ıa At´ omica and CONICET, 8400 Bariloche, Argentina
S. Ramasesha
Solid State and Structural Chemistry Unit, Indian Institute of Science, Bangalore 560 012, India
(Dated: November 21, 2009)
Coherent electronic transport through individual molecules is crucially sensitive to quantum inter-
ference. Using exact diagonalization techniques, we investigate the zero-bias and zero-temperature
conductance through π-conjugated annulene molecules (modeled by the Pariser-Parr-Pople and
Hubbard Hamiltonians) weakly coupled to two leads. We analyze the conductance for different
source-drain configurations, finding an important reduction for certain transmission channels and
for particular geometries as a consequence of destructive quantum interference between states with
definite momenta. When translational symmetry is broken by an external perturbation we find an
abrupt increase of the conductance through those channels. Previous studies concentrated on the
effect at the Fermi energy, where this effect is very small. By analysing the effect of symmetry
breaking on the main transmission channels we find a much larger response thus leading to the
possibility of a larger switching of the conductance through single molecules.
The possibility of achieving controlled quantum trans-
port through single molecules has become a reality
as seen from various successful attempts in different
systems[1, 2, 3]. The electronic transport through sin-
gle π-conjugated molecules has been studied in several
theoretical [4, 5, 6, 7, 8] and experimental[9, 10, 11, 12]
works. This important step leads not only to miniatur-
ization of electronic devices but also to the possibility
of taking advantage of new physical properties in these
systems.
A fundamental physical property in quantum trans-
port is quantum interference, which could be a handle to
control conductance through such systems. Constructive
and destructive interference play a crucial role which, for
most cases, is per se independent of the structure and
composition of the molecular bridges to the leads.
Even though changing from constructive to destructive
interference might be possible for certain geometries and
molecules, disrupting the destructive interference by per-
turbing the molecule seems to be more dramatic. This
has been proposed in two recent theoretical works [6, 7]
for the so-called Quantum Interference Effect Transis-
tors (QuIET) based on single annulene molecules, includ-
ing benzene. In the first paper, the systems are mod-
eled by a Pariser-Parr-Pople (PPP) Hamiltonian[13] us-
ing the Ohno parametrization [14] and solved using the
self-consistent Hartree-Fock (HF) approximation, while
the second work resorts to ab-initio calculations within
the LDA and ab-initio approximations. The conductance
in both cases is then calculated for the strong coupling
limit using the non-equilibrium Green’s function[15] and
the Landauer-B¨ uttiker formalism[16]. Both studies con-
centrate on the equilibrium conductance at zero bias and
gate voltage (Vg = 0, i.e. the off-resonance tunnelling
regime) which coincides with the Fermi energy of the
leads. However, annulenes have a gap at this energy due
mainly to the strong Coulomb interactions present in the
molecule (N = 4n+2 annulenes like benzene, are closed-
shell molecules with no level at zero energy even in the
absence of interactions). The conductance at zero gate
voltage is finite, albeit small, only for strong coupling to
the leads[17]. For weak coupling the conductance will
be zero and will be appreciable only through the main
channels of the molecule which are a few eV away from
the Fermi energy. By analysing the QuIET at the main
transmittance channels the switching effect will be much
more pronounced and robust as we will show below.
In this work we analyze the linear conductance of a
series of annulenes at finite gate voltages and several
source-drain lead configurations.
180 degrees (the A configuration in Fig.
tem presents completely constructive interference in all
transmission channels. However, for other lead config-
urations, we find different behaviours depending on N.
For N = 4n + 2 annulenes we find a reduced conduc-
tance through particular channels due to a nearly de-
structive interferences explained below. When the molec-
ular translational invariance is broken by means of elastic
scattering or decoherence factors, we find an abrupt in-
crease in the conductance through these channels due to
disruption of destructive interference. This increase is
much larger than was found in previous works since the
focus is set on a resonance transmission channel through
the molecule.
For leads set up at
1), the sys-
However, for N = 4n annulenes the effect is even more
striking: shifting the drain lead by one lattice site from
the A to the B configuration (see Fig. 1) reduces the
transmission to zero. This is, in fact due completely to
the destructive interference for N = 4n molecules as ex-
plained below. A translational symmetry breaking per-

Page 2

2
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
R
??
??
??
??
L
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
??
A
t’t’
B
FIG. 1: Schematic representation of the molecules studied nu-
merically in the presence of a gate voltage Vg, small hybridiza-
tion to the leads t′and two source-drain lead geometries.
turbation will have, as in this case, a huge effect, since
the conductance through these channels will increase dra-
matically from zero to an appreciable value. Hence, such
molecules could have important applications. Previous
theoretical studies analyze the effect of electron correla-
tion on the structure stability of carbon rings exhibiting
competing many-body effects of aromaticity, Jahn-Teller
distortion and dimerization[18, 19]. These studies indi-
cate that, under certain conditions, some of the systems
considered below could be stabilized. In addition, ring
geometries of quantum dots are also potential devices
where such a behaviour could be observed.
In this work we consider an annular π-conjugated
molecules with N sites, weakly connected to non-
interacting leads in the A or B configurations (Fig. 1).
We consider the following Hamiltonian for the system
consisting of two leads weakly coupled to the molecule:
H = Hring+ Hleads+ Hlinks.(1)
The first term describes the isolated π-conjugated
molecule, modeled by the PPP Hamiltonian[13], with on-
site energy given by a gate voltage Vg:
Hring = −t
?
?ij?,σ
c†
iσcjσ+
?
i
Ui
?
ni↓−1
2
??
ni↑−1
2
?
−eVg
?
i,σ
c†
iσciσ+
?
i>j
Vij(ni− 1)(nj− 1)(2)
where the operators c†
tron of spin σ in the π orbital of the Carbon atom at site
i, n are the corresponding number operators and ?···?
stands for bonded pairs of Carbon atoms. The intersite
interaction potential Vij is parametrized so as to inter-
polate between U and e2/rijin the limit rij−→ ∞ [14].
In the Ohno interpolation, Vijis given by
iσ(ciσ) create (annihilate) an elec-
Vij=Ui(1 + 0.6117 r2
ij)−1/2
(3)
where the distance rij is in˚ A. The standard Hubbard
parameter for sp2Carbon is Ui= 11.26 eV and hopping
parameter t for r = 1.397˚ A is 2.4 eV [20], and all energies
are in eV. The second term in Eq.(1) corresponds to two
tight-binding semi-infinite chains for the left and right
leads. The third term in Eq. (1) describes the coupling
of the edge sites of the left and right leads with sites L
(left) and R (right) of the system respectively. When
the ground state of the isolated ring is non-degenerate,
and the coupling t′between the leads and the ring is
weak (t′≪ t), equilibrium conductance at 0 K can be
expressed to second order in t′in terms of the retarded
Green’s function for the isolated ring between sites i and
j: GR
i,j(ω). For an incident particle with energy ω =
−2tcosk and momentum ±k, the transmittance reads
[21, 22]:
T(ω,Vg) =
4t2sin2k|˜t(ω)|2
??[ω − ǫ1+ teik][ω − ǫ2+ teik] − |˜t2|??2,
where ǫ1(2)(ω) = t′2GR
play the role of a correction to the on-site energy at
the extremes of the leads and an effective hopping be-
tween them respectively.
(2e2/h)T(µ,Vg), where µ is the Fermi level, which we set
to zero (half-filled leads) and Vgenters implicitly through
the Green’s functions.
In our study, the interacting system with N sites is
solved exactly using numerical techniques (Lanczos or
Davidson) to diagonalize huge matrices, and the Green’s
functions are obtained straightforwardly. The correla-
tions are, thus, treated in an exact manner.
We now analyze the two possible scenarios for annu-
lenes with N = 4n + 2 and N = 4n with two terminals.
For the first case, the allowed total momentum quan-
tum numbers are k = 2rπ/(4n + 2) = rπ/(2n + 1), with
r an integer, while for the second case the allowed mo-
menta are k = 2rπ/4n = rπ/2n. For a two-terminal set
up, wave functions travelling through both branches of
the molecule will interfere producing different interfer-
ence patterns depending on the positions of the leads.
The phase difference will be momentum times the differ-
ence in the lenghts of the two trajectories (in units of the
C-C separation): ∆φ = k∆x.
Specializing to benzene (N = 6), for leads in the
“para” position, ∆x = 0 and the waves are in phase, in-
terfering constructively (Fig. 2). However, in the “meta”
position (∆x = 2), the interference will depend on the
k value of the particular channel. For the highest oc-
cupied molecular orbital (HOMO) and lowest unoccu-
pied molecular orbital (LUMO) the phase differences will
be ∆φ = 2π/3 and ∆φ = 4π/3 respectively and the
interference will reduce the amplitude in these chan-
nels.For benzene configuration of the leads produc-
ing a completely destructive scenario for these channels
does not exist. Previous experimental and theoretical
works [23, 24] noted that the transmission through ben-
zene in the “meta” configuration is much lower than that
through the “para” position.
There is no transmission channel at Vg = 0 since the
density of states of the bare molecule at that energy is
(4)
LL(RR)(ω), ˜t(ω) = t′2GR
LR(ω),
The conductance is G =

Page 3

3
03
6
9
Ed
0.01
0.02
0.03
0.04
0.05
HOMO integral
0.2
0.4
0.6
0.8
1
para (A)
meta (B)
-10
-5
0
5
10
Vg[eV]
0
0.05
0.1
0.15
0.2
T
Ed = 0
Ed = 2
Ed = 4
Ed = 8
HOMO
LUMO
N = 6
FIG. 2: Top: Transmittance through a benzene molecule
for the “para” (A substitution in Fig. 1) (broken line) and
“meta” (B substitution in Fig. 1) (full line) configurations
as a function of the gate voltage measured from the Fermi
energy. Here t′= 0.4. Bottom: Same for the “meta” config-
uration in the presence of different on-site potentials which
break the translational invariance. Here a finite Lorentzian
width η = 0.03 and t′= 0.1 have been taken for visualization
purposes. Inset: evolution of the area under the transmission
peak through the HOMO as a function of local energy.
zero. A detailed analysis of quantum interference effects
in benzene was carried out for the large coupling regime
[25] where interesting effects appear at zero energy and
finite bias [26, 27] at weak coupling.
What would happen if the translational symmetry is
broken by an external perturbation? This question was
first addressed in [6] by introducing a local energy (Σ),
at one site in benzene, the real part of Σ would produce
elastic scattering and its imaginary part, decoherence.
In that study, the focus was at the Fermi energy of the
leads (set to zero) where the observed effect is small.
However, by studying the effect of external perturbations
on the main transmittance channels such as the HOMO
and LUMO, we find a much larger response. We show
these results in the bottom part of figure 2, where an
additional diagonal energy is added to the site to the right
of the B (“meta”) position (the effect is not qualitatively
dependent on this position). It is clearly seen that, in
this case, the small peak corresponding to the HOMO
level develops and grows as the local energy is increased,
disrupting the translational symmetry responsible for the
destructive interference (see inset).
For larger 4n + 2 rings, the momentum of the HOMO
level approaches π/2 and the interference becomes more
destructive. This is shown for a 10 site annulene in figure
3, where the peaks for the B position of the leads are
much smaller than those corresponding to benzene (Fig.
2, top). Growth of the HOMO peak with a local potential
energy is larger in this case than in benzene.
However, for annulenes with 4n atoms, depending on
the total momentum of the particular channel, the inter-
0.2
0.4
0.6
0.8
1
A
B
024
6
8
Ed
0
0.01
0.02
0.03
0.04
0.05
HOMO integral
-10
-5
0
5
10
Vg[eV]
0
0.05
0.1
0.15
0.2
T
Ed = 0
Ed = 3
Ed = 6
Ed = 9
HOMO LUMO
N = 10
FIG. 3: Same as figure 2, for N = 10 annulene.
ference can be totally destructive. For these annulenes
the interference will be completely constructive in the A
configuration and totally destructive if one lead is shifted
by one site (B configuration, see Fig. 1). In these cases
we can observe the emergence of a transmission channel,
if the translational invariance is disrupted.
The effect of symmetry breaking by a local external
perturbation is much more striking in annulenes with
4n sites. As discussed above, we expect a new peak to
emerge when the translational symmetry is broken. In
figure 4 we show results for conductance through an 8-
site annulene represented by the PPP model. In the top
figure, the HOMO and LUMO channels, fully formed for
the A configuration, completely vanish in the B case due
to quantum interference. When a local perturbation is
0.2
0.4
0.6
0.8
1
A
B
00.20.4
0.6
0.81
ϕ
0
0.01
0.02
HOMO integral
-10
-5
0
5
10
Vg[eV]
0
0.05
0.1
0.15
0.2
T
ϕ = 0
ϕ = 0.25
ϕ = 0.5
ϕ = 0.75
ϕ = 1
N = 8
HOMO LUMO
FIG. 4: Top: Transmittance through an [8]-annulene molecule
for the A (broken line) and B (full line) configurations as a
function of the gate voltage measured from the Fermi energy.
Here t′= 0.4. Bottom: Transmittance for the B configuration
in the presence of a local perturbation of the hopping of the
form t(1 − ϕ) which breaks the translational invariance (η =
0.03 and t′= 0.1). Inset: evolution of the HOMO integral as
a function of ϕ.

Page 4

4
0 0.20.4
0.6
0.81
ϕ
0
0.01
0.02
HOMO integral
0 0.01 0.02 0.03
δ
0
0.01
0.02
A
B
-6
-4-2024
6
Vg[eV]
0
0.04
0.08
0.12
T
δ = 0
δ = 0.01
δ = 0.02
δ = 0.04
δ = 0.07
N = 12
FIG. 5: Top: HOMO integral of a [12]-annulene molecule for
the A (broken line) and B (full line) configurations as a func-
tion of the locally perturbed hopping t(1−ϕ) (left panel) and
the dimerization parameter δ (right panel). Bottom: Trans-
mittance for the B configuration in the presence of dimeriza-
tion.
applied to the molecule, in this case a different hopping
on the bond to the right of the B lead, t(1 − ϕ), the ab-
sent peak emerges as shown in the inset of the bottom
figure. A similar effect occurs also for a local potential
Finally, in figure 5 we present results for a [12]-
annulene. In this case also the HOMO and LUMO peaks
for the B configuration are non-existent due to destruc-
tive interference. Here we show how the HOMO peak
arises as a function of a different hopping parameter in
one lead of the form t(1−ϕ) (top left) and as a function
of dimerization, δ, for transfer integral for t(1 − (−1)iδ)
of the bond between sites i and i + 1 (top right). The
transmittance in the A configuration doesn’t change in
the presence of dimerization, while in case B, the HOMO
and LUMO channels develop suddenly already for a small
δ (bottom panel). Dimerization of the annulenes lifts the
two-fold degeneracy of the HOMOs as well as the LUMOs
in large annulenes. Thus, the condition for destructive
interference between the two paths is lifted and one ob-
serves transmission through these channels.
In conclusion, by analysing the resonant conductance
through the HOMO and LUMO channels in the weak
lead-molecule coupling regime we find a strong depen-
dence on the source-drain configuration and on the
molecular geometry due to quantum interference. This
effect is more robust and striking than in the strong-
coupling case since in the latter the destructive interfer-
ence occurs only at zero gate voltage and can be masked
in an experiment, while in the former a whole channel can
appear or disappear depending on the geometry. We’d
also like to stress that these results are not affected by
molecular vibrations at room temperature since modes
that can cause decoherence are excited at temperatures
higher than 500K[6].
In [4n + 2]-annulenes like benzene in the B (“meta”)
configuration (Fig. 1) the interference will be nearly de-
structive. The larger the molecule the closer the phase
difference between different paths is to π and hence
the interference becomes more destructive.
annulenes, on the other hand, the interference for these
configurations can be completely destructive. We have
shown the effect of symmetry-breaking perturbations on
different transmission channels. For [4n]-annulenes the
effect is striking since the transmission changes from zero
to a large finite value for small perturbations. These ef-
fects should be seen in substituted annulenes and should
also be appreciable in transport through rings of quan-
tum dots.
We thank A. A. Aligia and B. Alascio for useful discus-
sions. This work was done in the framework of projects
PIP 5254 of CONICET, PICT 2006/483 of the ANPCyT
and the ARG/RPO-041/2006 Indo-Argentine collabora-
tion. SR thanks DST for support through JC Bose fel-
lowship.
For [4n]-
[1] A. Aviram, M. A. Ratner, Chem. Phys. Lett. 29, 277
(1974).
[2] M. A. Reed, et al., Science 278, 252 (1997).
[3] G. Cuniberti, G. Fagas, K. Richter (eds), Introducing
Molecular Electronics, Springer, Berlin 2005.
[4] A. Nitzan, M. A. Ratner, Science 300, 1384 (2003).
[5] N. Tao, Nature Nanotech. 1, 173-181 (2006).
[6] D. Cardamone, et al., Nano Lett. 6, 2422 (2006).
[7] S-H. Ke, et al., Nano Lett. 8, 3257 (2008).
[8] S. Yeganeh, et al., Nano Lett. 9, 1770 (2009).
[9] J. Park, et al., Nature 417, 722-725 (2002).
[10] B. Venkataraman, et al., Nature 442, 904 (2006).
[11] A. V. Danilov, et al., Nano Lett. 8, 1 (2008).
[12] T. Dadosh, et al., Nature 436, 677 (2005).
[13] R. Pariser, R. Parr, J. Chem. Phys. 21, 466 (1953); J. A.
Pople, Trans. of the Faraday Soc., 49, 1375 (1953).
[14] K. Ohno, Theor. Chim. Acta 2, 219 (1964).
[15] K. Hirose, ”First-principles calculations in real space for-
malism: Electronic configurations and transport proper-
ties of nanostructures”, Imperial College, London 2005.
[16] M. B¨ uttiker, et al., Phys. Rev. B 31, 6207 (1985).
[17] For certain geometries (e.g. the meta configuration) the
conductance at Vg = 0 for the strong coupling regime can
be zero due to quantum interference effects[25], but not
between paths with k = π/2 as stated in [6, 7], since this
is not an allowed quantum number for benzene molecules.
[18] T. Torelli, L. Mitas, Phys. Rev. Lett. 85, 1702 (2000).
[19] F. Sondheimer, Acc. Chem. Res. 5, 81 (1972).
[20] Z. G. Soos, and S. Ramasesha, Phys. Rev. B 29, 5410
(1984).
[21] E. Jagla, C. Balseiro, Phys. Rev. Lett. 70, 639 (1993).
[22] A. A. Aligia, et al., Phys. Rev. Lett. 93, 076801 (2004).
[23] S. N. Yaliraki and M. A. Ratner, Ann. N. Y. Acad. Sci.
960, 153 (2002).
[24] D. Walter, et al., Chem. Phys. 299, 139 (2004).
[25] G. Solomon, et al., Chem. Phys. 129, 054701 (2008).
[26] G. Begemann, et al., Phys. Rev. B 77, 201406(R) (2008);
78, 089901(E) (2008).

Page 5

5
[27] M. H. Hettler, et al., Phys. Rev. Lett. 90, 076805 (2003).