You are on page 1of 8

Chemical Physics xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Chemical Physics
journal homepage: www.elsevier.com/locate/chemphys

Vibrational properties of fractionally charged molecules and their


relevance for molecular electronics and electrochemistry q
Ioan Bâldea ⇑
Theoretische Chemie, Universität Heidelberg, Im Neuenheimer Feld 229, D-69120 Heidelberg, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Important insight into the charge transfer across interfaces can be gained in situations wherein, for given
Available online xxxx adsorbate and substrate species, the (fractional) charge state of the adsorbed molecules can be varied in a
controlled way. Applied biases can continuously tune the charge of molecules embedded in nanojunc-
Keywords: tions and/or in electrochemical setups but information on the fractional charges of the corresponding
Fractionally charged molecules partial oxidized/reduced states cannot be directly accessed in experiments. Here, we present theoretical
Vibrational spectroscopy results revealing that information on the fractional molecular charge can be obtained by monitoring
Surface enhanced Raman scattering (SERS)
molecular vibrational properties, which can be measured by means of surface enhanced Raman
Surface enhanced infrared absorption
spectroscopy (SEIRAS)
spectroscopy (SERS). To this aim, we performed DFT calculations for the benchmark 1,4-benzenedithiol
Redox processes molecule. The changes in the vibrational frequencies are considerably larger than those recently
Molecular electronics measured in combined transport-SERS studies on molecular junctions based on fullerene. We believe that
Electrochemical environment this theoretical result is an encouraging message to experimentalists.
Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction estimated recently [7] from changes in the surface enhanced


Raman spectra (SERS) [11]. A maximum bias-driven change in
Charge transfer between adsorbed molecules and substrate is a molecular charge comparable to that aforementioned has been
process of broad interest for interface science. In most cases, the reported in an SERS study [12] on macroscopic systems much older
amount of charge transferred between adsorbate and substrate is than Ref. [7] (Dq ¼ 0:171  0:066 ’ 11%, cf., Fig. 5 of Ref. [12]). Ref.
merely determined by their chemical nature. Important insight [12] demonstrated the possibility of shifting vibrational
into this fundamental process can be gained in situations wherein, frequencies of pyrazine molecules adsorbed on a silver electrode
for given adsorbate and substrate species, the charge transfer can by changing the electrode potential. In later work by the same
be smoothly controlled by continuously varying external group [13,14], variations in the electrode potential have been mod-
parameters, e.g., applied biases. Molecular electronics, a field of eled by introducing the concept of effective charge.
intensive and extensive investigations in the last decade [1–6], To the best of our knowledge, there is not experimental indica-
offers the rare opportunity of preparing and studying molecules tion on a bias-induced change of molecular charge exceeding the
(adsorbed to electrodes) accommodating a non-integer average aforementioned value of the redox efficiency (Dq  10%). Based
number of electrons. As pointed out recently [7,8], the average on parameters extracted from available molecular transport data
number of electrons in a molecule linked to electrodes to form a [15], we demonstrated recently the possibility of achieving an
single-molecule junction or related (nano-) experimental almost complete redox process (j Dq j K 1) [8,16].
platforms [9,10] can be varied by applied biases. In two-terminal The electromigrated junctions used in Ref. [7] are characterized
setups, an applied source-drain (labels S and D, respectively) bias by a nearly symmetric couplings CS;D of the molecule to the two
can to some extent change this number. Under such a two- (source and drain) electrodes. For such cases (CS  CD ), we demon-
terminal experimental platform, an excess charge transferred to strated that, whatever high is the source-drain bias, the (average)
the embedded molecule of up to one tenth of electron has been electron fraction q transferred to the embedded molecule cannot
exceed 50% [16].
q In addition, we showed that increasing this fraction (q = 50%) is
It is a great pleasure for me to dedicated this article to Professor Lorenz S.
Cederbaum on the occasion of his 70th birthday. possible [8,16] by using two-terminal platforms characterized by
⇑ Address: Institute of Space Sciences, National Institute for Lasers, Plasmas and highly asymmetric couplings (say, CD  CS Þ), as it turns out to be
Radiation Physics, RO 077125 Bucharest-Măgurele, Romania. the case in STM (scanning tunneling microscope) break junctions
E-mail address: ioan.baldea@pci.uni-heidelberg.de

http://dx.doi.org/10.1016/j.chemphys.2016.08.024
0301-0104/Ó 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024
2 I. Bâldea / Chemical Physics xxx (2016) xxx–xxx

[17], or in three-terminal setups based on electrolyte gating with SIESTA [40,41], a software allowing such studies. The DFT
(EC-STM junctions under electrochemical control (EC)) implementation utilized by SIESTA for non-integer charge states
[18–20,15,21,9,22,10,23–25]. The latter experimental platform is is that proposed in Ref. [42].
particularly appealing; by varying the overpotential (which is the We conducted calculations running the SIESTA [41] trunk-462
‘‘gate potential” in electrochemical environment), q can be contin- package [40] in parallel [43] using the generalized gradient
uously tuned in the whole range of a complete redox process exchange correlation functional of Perdew, Burke, and Ernzerhof
(0 <j q j6 1). Importantly for the applicability of the results (GGA-PBE) [44] and split-valence TZP (triple zeta polarized) basis
reported below, obtained within calculations to systems in equilib- sets for all atoms. The numerical results presented below have
rium, experimental realizations of such situations are compatible been obtained using three dimensional real-space grids with a
with low source-drain biases, justifying thereby an equilibrium- mesh cutoff of 800 Ry and cubic 30  30  30 Å3 cells. Employing
based description. (See the end of Section 4 for more details.). cubic cells in the calculations for the nearly planar BDT molecule
The fraction q of electrons removed from (0 6 q 6 1, the case might be unusual but is essential. As the total energy of a charged
considered in this paper) or added to (1 6 q 6 0, a case discussed system converges very slowly with respect to the cell size, for cubic
elsewhere [8,16]) the embedded molecule cannot be directly mea- cells, a Madelung correction can/is applied to the energy, which
sured in experiments. Ref. [7] represented the first study on molec- substantially improves convergence.
ular junctions substantiating the idea that experimental SIESTA computes vibrational properties by obtaining the
information on the quantity q can be obtained by its impact on dynamical matrix with a finite difference calculations of the forces
the molecular vibrational properties. The fact that full oxidation [45] via the VIBRA utility, which is distributed as part of the SIESTA
(i.e., q = +1) and full reduction processes (i.e., q = 1) have a signif- package [40]. This represents an alternative to the more popular
icant impact on the vibrational properties of molecular species of method of obtaining phonons, which uses linear response DFT
interest for molecular electronics has been pointed out recently schemes [46,47].
[26,8]. Although this is not the main issue here, we note that, even
The idea of quantitatively relating the changes in molecular attempting to make the SIESTA and GAUSSIAN settings as close
vibrational properties to the molecular average redox state (thence q) as possible, certain differences between the results obtained by
has been theoretically elaborated in Ref. [8]. There, a change in SIESTA (with keywords PAO.BasisSize TZP, xc.functional GGA, xc.
the vibrational properties of a fractionally charge state (N þ q) rep- authors PBE) and GAUSSIAN 09 (with keywords PBEPBE/TZVP)
resenting a linear interpolation between those corresponding to exist. Examples are presented in Table 1 in the Appendix, where
the molecular states corresponding to N and N þ 1 electrons has frequency values computed for the neutral BDT molecule (q = 0)
been conjectured. Validating this conjecture by means of micro- by using GAUSSIAN 09 are included along the SIESTA/GGA-PBE/
scopic calculations based on the density functional theory (DFT) TZP-based ones used in the analysis that follows. However, in spite
represents one significant aim of the present paper. of such differences, which are similar to those mentioned earlier
The 1,4-benzenedithiol molecule (BDT) considered in this paper [7], we think that SIESTA provides a reasonable framework to treat
(Fig. 1a, generated using XCrySDen [27]) is a benchmark molecule molecular systems with non-integer charges.
for studies of molecular electronics [28–33] and studies on vibra- To further elaborate, let us first refer to the eigenvalue EKSHOMO
tional properties in related surface enhanced Raman spectroscopy corresponding to the Kohn–Sham (KS) highest occupied molecular
(SERS) setups [34]. Since charge transport through BDT molecular orbital (HOMO). SIESTA GGA/TZP-DFT calculations for q = 0, i.e., the
junctions is known to be of p-type, i.e., mediated by the highest ðnÞ
neutral (label n) BDT molecule yield EKSHOMO ¼ 4:841 eV (value
occupied molecular orbital (HOMO) [32,37,33,38], we will consider extracted from SIESTA’s *.EIG file). For DFT approaches using expli-
charge species continuously spanning the range from perfectly cit density functionals (as the case for GGA), the HOMO energy can
neutral (q ¼ 0) to completely oxidized (q ¼ þ1) states. also be expressed in terms of the derivative of the total energy EðqÞ
in the limit of the vanishing electron charge (label c) deficit q ! 0þ :
2. Method ðcÞ
EKSHOMO ¼ @EðqÞ=@qjq!0þ [48–50]. The value deduced in this way
(expressed by the slope of the tangent at origin shown in Fig. 1c) is
Studies on vibrational properties of molecules containing a non- ðcÞ ðcÞ
EKSHOMO ¼ 4:844 eV. Since, for concrete numerical reasons, EHOMO
integer number of electrons, like the present one, are faced with an
involves calculations away from charge neutrality (i.e., small but
important problem; widely employed quantum chemical software,
finite q).
like the GAUSSIAN 09 package [39] (also utilized in this study), do
not allow calculations of this kind. This imposes a drastic limitation
ðcÞ 0<q1 EðqÞ  Eðq ¼ 0Þ
on the method to be adopted. Therefore, all calculations for the EKSHOMO !  ð1Þ
fractionally charged molecular states considered were performed q

(c)
deduced from the slope: EHOMO=-4.84 eV
BDT total energy (eV)

0
0 0.2 0.4 0.6 0.8 1
Neutral Fractional molecular charge q Cation

Fig. 1. 1,4-Benzenedithiol (panel a) is a benchmark molecule used to fabricate molecular junctions. Its HOMO spatial distribution (panel b), as obtained from natural orbital
CFOUR [35] calculations and plotted by using Gabedit [36], is delocalized over the whole molecule. The energy of the Kohn–Sham HOMO can be expressed in terms of the
slope of the total energy of the molecule at infinitesimally small oxidation q ! þ0, Eq. (1) (panel c). See the main text for details.

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024
I. Bâldea / Chemical Physics xxx (2016) xxx–xxx 3

ðcÞ ðnÞ inaccuracies/noise, which appear to be hardly avoidable in SIESTA


the excellent agreement between EKSHOMO and EKSHOMO represents a
reliability test for the calculations with fractional charges. This is a studies on phonons.
relevant aspect also in the context of the problems of DFT The results reported below have been obtained by using relaxed
approaches to systems with a fractional number of particles, as geometries at all fractional molecular charges by imposing that
amply discussed in the literature [42,51–57]. forces on all atoms are less than 0.1 meV/Å and a mesh cutoff of
ðnÞ ðcÞ
The values EKSHOMO ’ EKSHOMO reasonably agree with the value 800 Ry. In addition to the very tight condition for relaxation men-
tioned above, we have verified that there is no substantial
EKSHOMO ¼ 5:041 eV obtained from DFT/PBEPBE/TZVP calcula-
improvement (not even in this respect) by varying the displace-
tions with GAUSSIAN 09. (Notice that the slightly different SIESTA
ment used for the numerical computation of the force constant
and GAUSSIAN 09 optimized geometries also contribute to this
matrix (SIESTA keyword MD.FCDispl) from 0.02 bohr to 0.04 bohr,
small difference.) As illustrated by Figs. 3, 4, 5, 7, and 9, differences
playing around with pseudopotentials (that is, using others than
between the GAUSSIAN frequencies obtained for neutral and catio-
those actually utilized, obtained from translating the ABINIT pseu-
nic species and the SIESTA frequencies corresponding to q = 0 and
do’s database available to download [40]), or by increasing the
q = 1, respectively, are larger that the aforementioned difference
mesh cutoff from 300 Ry to 1000 Ry. In principle, by increasing
in HOMO. SIESTA’s different manner of constructing (strictly local-
the mesh cutoff parameter, spurious egg-box effects [40,41] could
ized) orbitals, which strictly vanish beyond a certain radius [41], as
be reduced. The rather unconvincing impact of this parameter in
well as the usage of numerical (as opposed to analytical) gradients
improving the numerical accuracy is illustrated in Fig. 2. There,
to compute force constants are possible sources of these differ-
the lowest six frequencies calculated numerically for several mesh
ences. Still, the various SIESTA-based and GAUSSIAN-based values
cutoff values are depicted. As they correspond to the three rigid
of vibrational frequencies can be reasonably correlated among
translations and rotations of the molecule, these frequencies
themselves. In the context of assessing the differences between
should be strictly zero. However, this is not the case; a tendency
SIESTA- and GAUSSIAN-based values as being a less dramatic
of diminishing the values of six ‘‘should-be-zero” frequencies by
aspect one should remind that, for quantitative comparison with
increasing the mesh cutoff can hardly be concluded by comparing
experiment, theoretically computed vibrational frequencies need
among themselves the various panels of Fig. 2.
to be scaled using empirical scaling factors [58].
In Fig. 3, the vibrational mode whose frequency (cf. Fig. 3a and
Table 1 in the Appendix) is most affected (j dx j J 100 cm1) by
3. Results and discussion charge removal is presented. This vibration, depicted in Fig. 3c
for the neutral molecule (q = 0) and in Fig. 3d for the completely
Often molecular geometry relaxation studies can reliably be oxidized species (q = +1), corresponds to the nondegenerate mode
done using default force tolerance (0.04 eV/Å in SIESTA and 3 in both Wilson [59,60] and Herzberg [61] numbering of the par-
0.023 eV/Å in GAUSSIAN 09) and a mesh cutoff of 300 Ry, but ent benzene molecule, having A2 g symmetry in the point group
phonon calculations are usually much more demanding D6 h. It can be assigned to an in-plane C–H bending. Unlike its
computationally than relaxation. In spite of having taken a series counterpart in benzene, this mode has in BDT a significant Raman
of precautions (see below), we could not eliminate some numerical activity. The Raman activity is enhanced by oxidation, as visible in

80 80
(a) mesh cutoff = 300 Ry mesh cutoff = 500 Ry (b)
Vibrational frequency (cm )

Vibrational frequency (cm )


-1

-1

60 60

40 40

20 20

0 0

-20 -20

80 80
mesh cutoff = 800 Ry (c) mesh cutoff = 1000 Ry (d)
Vibrational frequency (cm )

Vibrational frequency (cm )


-1

-1

60 60

40 40

20 20

0 0

-20 -20
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
q q

Fig. 2. The six lowest vibrational frequency of the BDT molecule having an electron deficit 0 6 q 6 1. The fact that these frequencies, corresponding to the three rigid
translations and rotations of the molecule as a whole, do not strictly vanish or are even imaginary (case depicted by negative ordinates) illustrates SIESTA’s numerical
inaccuracies that cannot be avoided, e.g., by increasing the mesh cutoff (indicated in the legend). Lines are guide to the eye.

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024
4 I. Bâldea / Chemical Physics xxx (2016) xxx–xxx

(a) BDT (b)


+
1550 1550 BDT

Frequency (cm )

Frequency (cm )
-1

-1
1500 1500
SIESTA
GAUSSIAN 09
linear fit

1450 1450
0 0.2 0.4 0.6 0.8 1 0 4 8
q Raman activity (arb. u.)

Fig. 3. The Raman active mode whose frequency is most affected by oxidation (panel a). Raman spectral lines of the neutral and cationic species (panel b). This vibrational
mode is shown in panel c for the neutral species (q = 0) and in panel d for the completely oxidized species (q = +1). It corresponds to an in-plane C–H bending (panels c and d)
and is related to benzene’s A2 g-mode 3 in both Wilson [59,60] and Herzberg [61] numbering.

630 630
SIESTA (a)
GAUSSIAN 09 BDT
620 linear fit 620 (b) +
BDT
Frequency (cm )

Frequency (cm )
-1

-1

610 610

600 600

590 590

580 580
0 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2
q Raman activity (arb. u.)

Fig. 4. A Raman active mode (counterpart of Wilson’s mode 6b or Herzberg’s mode 18 of benzene) whose frequency (panel a) decreases upon oxidation (charge removal).
Raman spectral lines of the neutral and cationic species (panel b).

Fig. 3b by comparing the GAUSSIAN results (only available) for Raman activity of the mode 6b decreases; compare Figs. 3b and
q = 0 and q = 1. Notice that in spite of the significantly different 4b, respectively.
vibrational properties of the neutral (Fig. 3c) and oxidized Vibrational modes that have no counterpart in the parent ben-
(Fig. 3d) species (visible by inspecting the magnitudes and orienta- zene molecule are also affected by oxidation. Examples are pre-
tions of the forces depicted by light green arrows), the geometrical sented in Fig. 5 and 6. The infrared (IR) active mode depicted in
variations upon increasing charge (q ¼ 0 ! q ¼ 1) are practically Fig. 5b can be assigned to an in-plane out of phase C–S stretching
invisible by comparing Fig. 3c and Fig. 3d among themselves. (one C–S bond is elongated while the other C–S bond is shortened).
What we show next (Fig. 4) is another Raman active mode, In this case, the IR absorption intensity is much less sensitive to
again an in-plane mode, which also has a counterpart in the ben- (almost unaffected by) the oxidation state than the Raman activity
zene molecule. This is the mode 6b in Wilson notation, which discussed above; compare Fig. 5b with Figs. 3b and 4b. Unlike the
traces back to the twofold degenerate mode 18 of E2g symmetry Raman active modes shown in Figs. 3 and 4, which exhibit red
of benzene’s D6 h point group in Herzberg numbering [61]. This shifts, this mode is blue shifted upon oxidation.
mode can be assigned to an in-plane ring deformation The aforementioned should not be understood that, upon
(cf. Fig. 4c). Its vibrational frequency is not so strongly affected charge removal, vibrational frequencies of all Raman active modes
by electron removal as the mode shown in Fig. 3 (j dx j 25 cm1 are red shifted while all IR active modes are blue shifted. To illus-
vs. j dx j 100 cm1). Albeit the frequencies of both modes 3 and trate this, we present in Fig. 6 two (almost degenerate) vibrational
6b are shifted in the same direction (red shift, dx < 0), oxidation modes, which are both red shifted by oxidation (cf. Fig. 7); one is IR
has an opposite impact on the spectral Raman intensities. While active and can be assigned to an in-plane out-of-phase S–H
the Raman activity of the mode 3 increases upon oxidation, the stretching (i.e., one S-H bond is elongated while the other S–H bond

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024
I. Bâldea / Chemical Physics xxx (2016) xxx–xxx 5

(a) (b)
550 550
Frequency (cm )

Frequency (cm )
-1

-1
BDT
540 540 +
BDT

530 530
SIESTA
GAUSSIAN 09
520 linear fit 520

0 0.2 0.4 0.6 0.8 1 0 100 200 300


q IR intensity (arb. u.)

Fig. 5. An IR active mode (in-plane out of phase C–S stretching, cf. panel c), which is blue shifted upon oxidation (panel a). Raman spectral lines of the neutral and cationic
species (panel b).

Fig. 6. Two modes without counterpart in the parent benzene molecule, corre-
sponding to in-plane in-phase (Raman active) and out-of-phase (IR active) S–H
stretching vibrations, which are affected by oxidation (cf. Fig. 7).
Fig. 8. Out-of-plane vibrational modes related to C–H bending vibrations affected
is shortened), another is Raman active and corresponds to an by charge removal as shown in Fig. 9: one mode corresponding to benzene’s mode
in-plane in-phase S–H stretching (i.e., both S–H bonds are simulta- 17b in Wilson notation or mode 19 in Herzberg labeling (panel a), the other mode
corresponding to benzene’s nondegenerate mode 5 in Wilson notation or mode 7 in
neously elongated or shortened). Qualitatively, the spectral inten- Herzberg labeling (panel b).
sities of these red-shifted modes behave similarly; both the Raman
intensity of the first mode and the IR intensity of the second mode
are enhanced; quantitatively, the IR enhancement is more pro- 4. Additional remarks
nounced than the Raman enhancement.
On the contrary, the two vibrational modes without counterpart In spite of specific SIESTA’s numerical difficulties in dealing
in benzene depicted in Fig. 8 are both blue shifted by charge with phonons discussed in Section 2 and the numerical noise more
removal (cf. Fig. 9); one mode corresponds to an in-plane out- or less present in all the figures presented in this paper, a clear
of-phase S–H bending vibration and is IR active, the other mode trend of (reasonably) linear dependencies of the phonon frequen-
can be assigned to an in-plane in-phase S–H bending vibration cies on the electron charge deficit q can be observed in all figures
and is Raman active. presented above. This linear trend is visible even for modes for
Out-of-plane vibrational modes are also affected by oxidation. which frequency changes upon charge removal is very small

Examples are depicted in Fig. 8. They represent the counterpart ( xq¼1  xq¼0  10 cm1), irrespective whether these frequencies
of Wilson’s mode 17b or mode 19 in Herzberg notation and are blue shifted or red shifted by oxidation. To exemplify, in
Wilson’s 5 mode or mode 7 in Herzberg numbering (panels a and Fig. 10 we present two Raman active modes. The vibrational mode
b, respectively). As visible in Fig. 8, they are related to C–H bending depicted in the upper half is dominated by Wilson’s mode 4 (mode
vibrations. The impact of oxidation on their vibrational properties 8 in Herzberg numbering of B2 g symmetry; out-of-plane ring
is shown in Fig. 9. The spectral intensities of these blue-shifted deformation). The mode presented in the lower half is related to
modes (cf. Fig. 9a) are affected in a similar way. Oxidation reduces Wilson’s mode 8b (mode 17 in Herzberg notation of E2 g symme-
both the Raman intensity of the first mode and the IR intensity of try; in-plane C–C stretch).
the second mode; see panels b and c of Fig. 9.

(a) (b) (c)


2620 2620 2620
Frequency (cm )

Frequency (cm )

Frequency (cm )
-1

-1

-1

2600 2600 2600

2580 2580 2580


SIESTA BDT BDT
SIESTA + +
linear fit (Raman)
BDT BDT
2560 linear fit (IR) 2560 2560
GAUSSIAN 09 (Raman)
GAUSSIAN 09 (IR)
2540 2540 2540
0 0.2 0.4 0.6 0.8 1 0 20 40 60 80 0 500 1000 1500
q Raman activity (arb. u.) IR activity (arb. u.)

Fig. 7. The frequencies (panel a) of two almost degenerate modes, corresponding to in-plane in-phase (Raman active) and out-of-phase (IR active) S–H stretching vibrations
depicted in Fig. 6, affected by oxidation. The corresponding Raman (panel b) and IR (panel c) spectral lines of the neutral and cationic species.

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024
6 I. Bâldea / Chemical Physics xxx (2016) xxx–xxx

930
(a) (b) (c)
920 920 920
910
Frequency (cm )

Frequency (cm )

Frequency (cm )
-1

-1

-1
900 900 BDT 900
+ BDT
BDT +
890 BDT
SIESTA
880 SIESTA 880 880
GAUSSIAN 09
870 GAUSSIAN 09
linear fit (Raman)
860 linear fit (IR) 860 860
850
840 840 840
0 0.2 0.4 0.6 0.8 1 0 1 2 3 1 10
q Raman activity (arb. u.) IR activity (arb. u.)

Fig. 9. Impact of oxidation on the frequencies (panel a) of the out-of-plane modes depicted in Fig. 8 and the corresponding Raman (panel b) and IR (panel c) spectral lines.

678 (a) SIESTA


linear fit
676
Frequency (cm )
-1

674

672

670

668

0 0.2 0.4 0.6 0.8 1


q

1612 SIESTA (c)


linear fit
Frequency (cm )
-1

1608

1604

0 0.2 0.4 0.6 0.8 1


q

Fig. 10. The two modes depicted in panels b and d reveal that even in cases where oxidation yields small changes of vibrational frequencies (cf. panels a and c), a (roughly)
linear trend is visible. Upper half: Wilson’s mode 4 (out-of-plane mode 8 in Herzberg numbering). Lower half: Wilson’s mode 8b (mode 17 in Herzberg notation).

The aforementioned (roughly) linear dependence on q indicates for natural populations created by CFOUR [35] as input file for
that, for the BDT molecule considered here, it is reasonable to Gabedit [36].
assume that charge removal from the highest occupied molecular We have resorted to this method because, for inspecting molec-
orbital (HOMO) acts as a first-order perturbation on the molecular ular orbital spatial densities, the Hartree–Fock (HF) orbitals may
vibrational properties. One may relate this behavior with the represent a too crude approximation. Furthermore, considering
spatial distribution of the HOMO, which is the molecular orbital Kohn–Sham (KS) orbitals may be problematic even for the highest
relevant for the charge removal discussed in this paper. occupied molecular orbital (HOMO); KS-HOMO provides a quanti-
To obtain this spatial distribution, we have calculated the tative description of the lowest ionization only if the exact
natural orbital expansion of the corresponding reduced density exchange correlation functional is known [63]. The fact that, for
matrices obtained from elaborate IP-EOM-CCSD quantum chemical typical molecules used to fabricate molecular junctions, the eigen-
calculations of the lowest ionization potential (IP) at the value (‘‘energy”) corresponding to the KS-HOMO substantially dif-
EOM-CCSD (equation-of-motion coupled-cluster singles and dou- fers [62] from the lowest ionization energy taken with reversed
bles) level [35]; see, e.g., Refs. [62,17] for details. Doing so, we sign [62] demonstrates how challenging is to find the exact
found that the hole created by ionizing the neutral molecule is exchange correlation functional. This is also the case of the BDT
almost entirely (> 95%) concentrated in a single natural orbital. molecule considered in this paper. The lowest ionization potential
(This would be the ‘‘HOMO” if the single-particle picture were with reversed sign (IP ¼ 7:717 eV) obtained by us at IP-EOM-
valid.) To generate Fig. 1 we employed the MOLDEN-formatted file CCSD level and tzp (triple zeta with polarization within CFOUR

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024
I. Bâldea / Chemical Physics xxx (2016) xxx–xxx 7

notation) basis sets [35] differs from the KS-HOMO eigenvalues described by correlation functions computed at equilibrium (fluc-
mentioned in Section 2 by almost 3 eV. tuation–dissipation theorem).
Before ending this section, let us address the question on the
relevance of the results reported in the present paper. 5. Summary and conclusion
Theoretically modeling the dependence of vibrational molecu-
lar properties on the charge (q) of a molecule linked to electrodes Using the benchmark 1,4-benzenedithiol (BDT) molecule, in
to form a molecular junction under applied (source-drain) bias V this paper we have demonstrated that vibrational frequencies
represents a difficult theoretical problem; it refers to a system xm ¼ xm ðqÞ can be tuned by continuously changing the average
out of equilibrium. This particularly applies to two terminal plat- redox state from the neutral (q = 0) and fully oxidized (q = 1) limits
forms (like that of Ref. [7]) where substantial charge transfer within ranges substantially broader than achieved in recent exper-
implies high source-drain voltages cf. Ref. [16] (typically higher iments on fullerene junctions [7]. So, we believe that this is an
than real molecular junctions can withstand), implying, in turn, encouraging message to experimentalists; observing experimen-
theoretical treatments of systems far away from equilibrium. tally the effects studied theoretically in this paper should be
Applying results obtained for systems at equilibrium (to which feasible.
the calculations done within the present study refer) to such Explicit calculations for the case examined have confirmed the
nonequilibrium situations may be problematic. linear dependence xm ðqÞ vs. q conjectured recently by us [8]. This
Fortunately, the present results, obtained via methods assuming provides a reasonable framework for interpreting experimental
systems in equilibrium, are relevant for the more appealing three- data, especially if one takes into account that, in order to make
terminal platforms. As demonstrated in our recent works [8,16], in the DFT-computed values comparable with the measured vibra-
three-terminal setups, it is the gate potential that controls the tional frequencies, empirical scaling factors have to be applied
molecular charge, which can be tuned between q = 0 and [58], and these scaling factors may be different for different charge
j q j K 1, but the gate potential is not a source of nonequilibrium. species of the same molecular unit [64]. For the vibrational modes
In three-terminal setups, a nearly complete redox process can analyzed above, typical corrections (amounting to a few percents)
occur even if the source-drain voltage, which is the source of introduced by using such scaling factors are comparable with (or
nonequilibrium, is very low (e.g., V = 0.05 V, cf., Fig. 3 of Ref. [8]). even larger than) the differences between the frequencies com-
At such low V’s, the system is only slightly perturbed from equilib- puted within the DFT flavors listed in Table 1.
rium, and therefore the equilibrium description represents a legit- Let us finally mention that, as visible in Figs. 3, 4, 5, 7, and 9,
imate approximation. One can remind in this context that the charge removal has a substantial impact not only of the vibrational
linear response characterizing a slightly perturbed system can be frequencies but also on the vibrational Raman and IR spectral

Table 1
The values in cm1 of the (harmonic) frequencies of the 36 vibrational modes of the BDT molecule computed with SIESTA/GGA-PBE/TZP (labeled S/GGA) as described in the main
text for several fractional charges q. For comparison, the values computed with GAUSSIAN 09 [39] using the GGA-PBE and LDA functionals (GAUSSIAN keywords PBEPBE/TZVP and
SVWN/TZVP, labeled G09/GGA and G09/LDA, respectively) for the neutral BDT are also presented.

Method G09/LDA G09/GGA S/GGA S/GGA S/GGA S/GGA S/GGA S/GGA


q 0 0 0 0.2 0.4 0.6 0.8 1
Mode 1 88.50 64.95 90.85 90.97 88.49 88.36 85.95 82.18
Mode 2 100.30 72.51 110.93 171.19 194.95 196.30 197.99 199.35
Mode 3 119.79 102.87 128.93 181.89 216.91 247.96 250.41 244.45
Mode 4 195.11 195.70 192.46 193.27 225.38 263.53 288.86 297.42
Mode 5 273.79 272.67 276.77 273.03 269.65 277.57 297.28 304.38
Mode 6 311.71 309.79 302.27 302.89 302.02 298.43 305.41 329.05
Mode 7 333.73 329.18 328.58 328.21 329.00 328.16 327.86 333.87
Mode 8 385.72 382.60 380.93 376.88 373.25 372.82 374.40 383.13
Mode 9 444.79 446.93 460.89 462.05 462.36 464.05 464.58 465.04
Mode 10 495.85 520.06 528.94 534.76 540.35 543.90 544.37 547.98
Mode 11 540.57 553.58 603.62 599.75 594.28 590.21 583.60 579.31
Mode 12 630.88 626.64 667.93 669.75 671.40 673.84 675.93 678.09
Mode 13 733.10 732.35 730.94 734.01 733.50 735.70 734.75 732.65
Mode 14 750.68 739.72 762.61 766.77 768.93 768.13 766.68 766.03
Mode 15 776.87 780.31 770.35 769.98 770.39 775.67 779.63 783.98
Mode 16 831.86 845.18 860.69 855.87 858.81 860.68 865.13 866.34
Mode 17 860.63 863.81 869.00 863.16 866.07 869.49 875.94 878.38
Mode 18 880.79 887.35 882.45 890.70 898.56 906.49 913.84 921.70
Mode 19 893.20 900.19 898.42 904.23 909.67 915.15 919.98 925.27
Mode 20 1002.77 997.34 972.28 971.76 969.48 965.93 965.07 961.30
Mode 21 1107.32 1079.98 1085.39 1091.12 1088.14 1091.98 1091.52 1085.71
Mode 22 1115.83 1096.62 1102.77 1103.38 1098.01 1097.13 1092.54 1095.36
Mode 23 1117.27 1115.78 1109.12 1110.86 1106.82 1106.99 1106.73 1105.31
Mode 24 1173.79 1183.73 1148.46 1153.69 1156.00 1157.94 1160.81 1166.11
Mode 25 1273.32 1289.10 1248.41 1250.18 1247.33 1246.07 1246.26 1248.62
Mode 26 1359.54 1315.00 1336.58 1336.42 1334.92 1330.19 1330.17 1325.45
Mode 27 1424.51 1397.09 1412.24 1423.90 1433.55 1431.42 1421.20 1414.03
Mode 28 1471.94 1464.47 1452.64 1446.76 1440.61 1444.48 1453.06 1462.95
Mode 29 1587.16 1550.95 1568.98 1551.90 1533.19 1512.57 1490.18 1467.40
Mode 30 1620.67 1581.54 1611.96 1608.96 1606.33 1604.54 1602.51 1602.29
Mode 31 2620.52 2619.49 2591.66 2585.38 2577.13 2572.74 2565.23 2558.63
Mode 32 2620.58 2619.49 2592.21 2588.19 2581.33 2579.22 2570.93 2562.76
Mode 33 3098.95 3103.35 3074.68 3077.52 3083.27 3084.18 3089.72 3093.49
Mode 34 3099.56 3103.96 3076.81 3081.58 3086.14 3087.61 3093.97 3096.05
Mode 35 3112.53 3117.68 3083.54 3090.11 3094.28 3095.57 3097.95 3101.55
Mode 36 3115.45 3120.45 3088.04 3094.55 3097.20 3101.81 3107.29 3105.97

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024
8 I. Bâldea / Chemical Physics xxx (2016) xxx–xxx

intensities. How these spectral intensities change by continuously [27] A. Kokalj, Comp. Mater. Sci. 28 (2003) 155–168. Proceedings of the Symposium
on Software Development for Process and Materials Design.
varying the oxidation degree q in the range 0 < q < 1 could not be
[28] M.A. Reed, C. Zhou, C.J. Muller, T.P. Burgin, J.M. Tour, Science 278 (1997) 252–
discussed in this paper. SIESTA can in principle be used to compute 254.
IR intensities; unfortunately, the impact of noise is much stronger [29] Xu Xiao, N.J. Tao, Nano Lett. 4 (2004) 267–271.
than on vibrational frequencies and prevents a decent analysis in [30] A.W. Ghosh, T. Rakshit, S. Datta, Nano Lett. 4 (2004) 565–568.
[31] P. Reddy, S.-Y. Jang, R.A. Segalman, A. Majumdar, Science 315 (2007) 1568–
quantitative terms. The current SIESTA implementation does not 1571.
allow computation of Raman activities. [32] H. Song, Y. Kim, Y.H. Jang, H. Jeong, M.A. Reed, T. Lee, Nature 462 (2009) 1039–
1043.
[33] Z. Xie, I. Bâldea, C. Smith, Y. Wu, C.D. Frisbie, ACS Nano 9 (2015) 8022–8036.
Acknowledgments [34] X.-M. Zhao, X.-R. Tian, S.-S. Liu, Y.-Z. Li, M.-D. Chen, Chin. J. Chem. Phys. 24
(2011) 665.
The author is pleased to thank Lenz Cederbaum for numerous [35] CFOUR, Coupled-Cluster techniques for Computational Chemistry, a quantum-
chemical program package by J.F. Stanton, J. Gauss, M.E. Harding, P.G. Szalay
invaluable discussions and a fruitful collaboration extended over with contributions from A.A. Auer, R.J. Bartlett, U. Benedikt, C. Berger, D.E.
more than two decades. Financial support for this research pro- Bernholdt, Y.J. Bomble, L. Cheng, O. Christiansen, M. Heckert, O. Heun, C.
vided by the Deutsche Forschungsgemeinschaft (grant BA Huber, T.-C. Jagau, D. Jonsson, J. Jusélius, K. Klein, W.J. Lauderdale, D.A.
Matthews, T. Metzroth, L.A. Mück, D.P. O’Neill, D.R. Price, E. Prochnow, C.
1799/3-1) is gratefully acknowledged. This work was partially per- Puzzarini, K. Ruud, F. Schiffmann, W. Schwalbach, C. Simmons, S. Stopkowicz,
formed on the computational resources bwUniCluster and A. Tajti, J. Vázquez, F. Wang, J.D. Watts and the integral packages MOLECULE (J.
bwForCluster/JUSTUS HPC facility funded by the Ministry of Almlöf and P.R. Taylor), PROPS (P.R. Taylor), ABACUS (T. Helgaker, H.J. Aa.
Jensen, P. Jrgensen, and J. Olsen), and ECP routines by A.V. Mitin and C. van
Science, Research and the Arts Baden-Württemberg and the
Wüllen. For the current version, see http://www.cfour.de.
Universities of the State of Baden-Württemberg, Germany, within [36] A.-R. Allouche, J. Comput. Chem. 32 (2011) 174–182.
the framework program bwHPC and the bwHPC-C5 project [43]. [37] H. Song, M.A. Reed, T. Lee, Adv. Mater. 23 (2011) 1583–1608.
[38] M. Tsutsui, T. Morikawa, Y. He, A. Arima, M. Taniguchi, Sci. Rep. 5 (2015)
11519.
Appendix A [39] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman,
G. Scalmani, V. Barone, B. Mennucci, G.A. Petersson, H. Nakatsuji, M. Caricato,
X. Li, H.P. Hratchian, A.F. Izmaylov, J. Bloino, G. Zheng, J.L. Sonnenberg, M.
Table 1 Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y.
Honda, O. Kitao, H. Nakai, T. Vreven, J.A. Montgomery, Jr., J.E. Peralta, F. Ogliaro,
References M. Bearpark, J.J. Heyd, E. Brothers, K.N. Kudin, V.N. Staroverov, T. Keith, R.
Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J.C. Burant, S.S. Iyengar, J.
Tomasi, M. Cossi, N. Rega, J.M. Millam, M. Klene, J.E. Knox, J.B. Cross, V. Bakken,
[1] S. Datta, Quantum Transport: Atom to Transistor, Cambridge Univ. Press,
C. Adamo, J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. Austin, R.
Cambridge, 2005.
Cammi, C. Pomelli, J.W. Ochterski, R.L. Martin, K. Morokuma, V.G. Zakrzewski,
[2] J.C. Cuevas, E. Scheer, Molecular Electronics: An Introduction to Theory and
G.A. Voth, P. Salvador, J.J. Dannenberg, S. Dapprich, A.D. Daniels, O. Farkas, J.B.
Experiment, World Scientific Publishers, 2010.
Foresman, J.V. Ortiz, J. Cioslowski, and D.J. Fox, Gaussian Inc, Wallingford CT,
[3] T. Seideman (Ed.), Current-Driven Phenomena in Nanoelectronics, Pan
2010 Gaussian 09, Revision B.01.
Stanford, 2011.
[40] Available at http://www.icmab.es/dmmis/leem/siesta/.
[4] K. Hirose, N. Kobayashi, Quantum Transport Calculations for Nanosystems, Pan
[41] J.M. Soler, E. Artacho, J.D. Gale, A. Garcia, J. Junquera, P. Ordejon, D. Sanchez-
Stanford, 2014.
Portal, J. Phys.: Cond. Matter 14 (2002) 2745–2779.
[5] I. Bâldea (Ed.), Molecular Electronics: An Experimental and Theoretical
[42] J.P. Perdew, R.G. Parr, M. Levy, J.L. Balduz, Phys. Rev. Lett. 49 (1982) 1691–
Approach, Pan Stanford, 2015.
1694.
[6] K. Moth-Poulsen (Ed.), Single-Molecule Electronics, Pan Stanford, 2016.
[43] bwHPC and bwHPC-C5 (http://www.bwhpc-c5.de/.) funded by the Ministry of
[7] Y. Li, P. Doak, L. Kronik, J.B. Neaton, D. Natelson, Proc. Natl. Acad. Sci. 111
Science, Research and Arts Baden-Württemberg and the Universities of the
(2014) 1282–1287.
State of Baden-Württemberg, Germany, and the German Research Foundation
[8] I. Bâldea, Phys. Chem. Chem. Phys. 16 (2014) 25942–25949.
(DFG).
[9] S. Wang, M. Ha, M. Manno, C.D. Frisbie, C. Leighton, Nat. Commun. 3 (2012)
[44] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865–3868.
1210.
[45] G.J. Ackland, M.C. Warren, S.J. Clark, J. Phys. Condens. Matter 9 (1997) 7861–
[10] W. Xie, S. Wang, X. Zhang, C. Leighton, C.D. Frisbie, Phys. Rev. Lett. 113 (2014)
7872.
246602.
[46] S. Baroni, P. Giannozzi, A. Testa, Phys. Rev. Lett. 58 (1987) 1861–1864.
[11] M. Takase, S. Yasuda, K. Murakoshi, Front. Phys. 11 (2016) 1–12.
[47] M.C. Palenik, B.I. Dunlap, J. Chem. Phys. 143 (2015) 044115.
[12] J. Soto, D.J. Fernández, S.P. Centeno, I.L. Tocón, J.C. Otero, Langmuir 18 (2002)
[48] J.F. Janak, Phys. Rev. B 18 (1978) 7165–7168.
3100–3104.
[49] G.K.-L. Chan, J. Chem. Phys. 110 (1999) 4710–4723.
[13] F. Avila, D.J. Fernandez, J.F. Arenas, J.C. Otero, J. Soto, Chem. Commun. 47
[50] A.J. Cohen, P. Mori-Sánchez, W. Yang, Phys. Rev. B 77 (2008) 115123.
(2011) 4210–4212.
[51] Y. Zhang, W. Yang, J. Chem. Phys. 109 (1998) 2604–2608.
[14] F. Avila, C. Ruano, I.L. Tocón, J.F. Arenas, J. Soto, J.C. Otero, Chem. Commun. 47
[52] P. Mori-Sánchez, A.J. Cohen, W. Yang, J. Chem. Phys. 125 (2006) 201102.
(2011) 4213–4215.
[53] A. Ruzsinszky, J.P. Perdew, G.I. Csonka, O.A. Vydrov, G.E. Scuseria, J. Chem.
[15] I.V. Pobelov, Z. Li, T. Wandlowski, J. Am. Chem. Soc. 130 (2008) 16045–16054.
Phys. 126 (2007) 104102.
[16] I. Bâldea, Phys. Chem. Chem. Phys. 17 (2015) 15756–15763.
[54] A.J. Cohen, P. Mori-Sánchez, W. Yang, J. Chem. Phys. 126 (2007) 191109.
[17] I. Bâldea, Nanotechnology 25 (2014) 455202.
[55] J.-W. Song, M.A. Watson, K. Hirao, J. Chem. Phys. 131 (2009) 144108.
[18] N.J. Tao, Phys. Rev. Lett. 76 (1996) 4066–4069.
[56] J.-D. Chai, J. Chem. Phys. 136 (2012) 154104.
[19] A. Alessandrini, M. Salerno, S. Frabboni, P. Facci, Appl. Phys. Lett. 86 (2005)
[57] T. Stein, J. Autschbach, N. Govind, L. Kronik, R. Baer, J. Phys. Chem. Lett. 3
133902.
(2012) 3740–3744.
[20] A. Alessandrini, S. Corni, P. Facci, Phys. Chem. Chem. Phys. 8 (2006) 4383–
[58] A.P. Scott, L. Radom, J. Phys. Chem. 100 (1996) 16502–16513.
4397.
[59] E.B. Wilson, Phys. Rev. 45 (1934) 706–714.
[21] J.M. Artés, M. López-Martínez, A. Giraudet, I. Díez-Pérez, F. Sanz, P. Gorostiza, J.
[60] G. Varsanyi, in: I. Land (Ed.), Assignments for Vibrational Spectra of Seven
Am. Chem. Soc. 134 (2012) 20218–20221.
Hundred Benzene Derivatives, vol. 1, Adam Hilger, London, 1974, p. 1.
[22] I. Bâldea, J. Phys. Chem. C 117 (2013) 25798–25804.
[61] G. Herzberg, Molecular spectra and molecular structure. Vol. 2: Infrared and
[23] J.M. Artés-Vivancos, J. Hihath, I. Díez-Pérez, I. Bâldea (Eds.), Molecular
Raman spectra of polyatomic molecules, Van Nostrand, Reinhold, New York,
Electronics: An Experimental and Theoretical Approach, Pan Stanford, 2015,
1945. For mode numbering in the benzene molecule, see page 18.
pp. 281–324.
[62] I. Bâldea, Faraday Discuss. 174 (2014) 37–56.
[24] A. Alessandrini, P. Facci, in: I. Bâldea (Ed.), Molecular Electronics: An
[63] C.-O. Almbladh, U. von Barth, Phys. Rev. B 31 (1985) 3231–3244.
Experimental and Theoretical Approach, Pan Stanford, 2015, pp. 325–352.
[64] I. Bâldea, H. Köppel, W. Wenzel, Phys. Chem. Chem. Phys. 15 (2013) 1918–
[25] A. Alessandrini, P. Facci, Accepted manuscript, Eur. Polym. J. (2016), http://dx.
1928.
doi.org/10.1016/j.eurpolymj.2016.03.028.
[26] F. Mirjani, J.M. Thijssen, M.A. Ratner, J. Phys. Chem. C 116 (2012) 23120–
23129.

Please cite this article in press as: I. Bâldea, Chem. Phys. (2016), http://dx.doi.org/10.1016/j.chemphys.2016.08.024

You might also like