Controlling the thermoelectric properties of thiophene-derived single-molecule junctions

Article abstract of DOI:10.1021/cm504254n

Thermoelectrics are famously challenging to optimize, because of inverse coupling of the Seebeck coefficient and electrical conductivity, both of which control the thermoelectric power factor. Inorganic-organic interfaces provide a promising route for realization of the strong electrical and thermal asymmetries required for thermoelectrics. In this work, transport properties of inorganic-organic interfaces are probed and understood at the molecular scale using the STM-break junction measurement technique, theory, and a class of newly synthesized molecules. We synthesized a series of disubstituted thiophene derivatives varying the length of alkylthio-linkers and the number of thiophene rings. These molecules allow the systematic tuning of electronic resonances within the junction. We observed that these molecules have a decreasing Seebeck coefficient with increasing length of the alkyl chain, while oligothiophene junctions show an increasing Seebeck coefficient with length. We find that thiophene-Au junctions have significantly higher Seebeck coefficients, compared to benzenedithiol (in the range of 7-15 μV/K). A minimal tight-binding model, including a gateway state associated with the S-Au bond, captures and explains both trends. This work identifies S-Au gateway states as being important and potentially tunable features of junction electronic structure for enhancing the power factor of organic/inorganic interfaces.

Full text of DOI:10.1021/cm504254n

Article
pubs.acs.org/cm
Controlling the Thermoelectric Properties of Thiophene-Derived
Single-Molecule Junctions
William B. Chang,†,○ Cheng-Kang Mai,‡,○ Michele Kotiuga,§,∇,○ Jeffrey B. Neaton,§,∇,⊥
Guillermo C. Bazan,‡ and Rachel A. Segalman*,†
†Department of Chemical Engineering & Materials, University of California, Santa Barbara, California 93106, United States
‡Center for Polymers and Organic Solids, Departments of Chemistry & Biochemistry and Materials, University of California,
Santa Barbara, California 93106, United States
§Department of Physics, University of California, Berkeley, California 94720, United States
∇Molecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, California 94720, United States
⊥Kavli Energy NanoSciences Institute at Berkeley, Berkeley, California United States
S
* Supporting Information
ABSTRACT: Thermoelectrics are famously challenging to optimize,
because of inverse coupling of the Seebeck coefficient and electrical
conductivity, both of which control the thermoelectric power factor.
Inorganic−organic interfaces provide a promising route for realization of
the strong electrical and thermal asymmetries required for thermo-
electrics. In this work, transport properties of inorganic−organic inter-
faces are probed and understood at the molecular scale using the STM-
break junction measurement technique, theory, and a class of newly syn-
thesized molecules. We synthesized a series of disubstituted thiophene
derivatives varying the length of alkylthio-linkers and the number of
thiophene rings. These molecules allow the systematic tuning of electronic resonances within the junction. We observed that
these molecules have a decreasing Seebeck coefficient with increasing length of the alkyl chain, while oligothiophene junctions
show an increasing Seebeck coefficient with length. We find that thiophene−Au junctions have significantly higher Seebeck
coefficients, compared to benzenedithiol (in the range of 7−15 μV/K). A minimal tight-binding model, including a gateway state
associated with the S−Au bond, captures and explains both trends. This work identifies S−Au gateway states as being important
and potentially tunable features of junction electronic structure for enhancing the power factor of organic/inorganic interfaces.
derivative of the transmission function, with respect to energy,
evaluated at EF, and increases as Γ decreases and the frontier
orbital resonance moves closer to EF.4,5 Therefore, it is possible
to couple S and G in new ways in a molecular junction with
Lorentzian resonances, to maximize the power factor with junc-
tions having a transmission function featuring a narrow frontier
orbital resonance close to EF. Experimentally, both S and G
have been shown to increase simultaneously with energy level
alignment.3 Positioning the energy of the frontier highest occu-
pied or lowest unoccupied molecular orbitals (HOMO or
LUMO) closer to the EF of gold will result in a simultaneous
increase in both S and σ.6,7,8−11 Other methods shown to control
S2G are by varying binding groups,4,8,12−17 molecular design, and
choice of metal contacts.2,18−21
INTRODUCTION
■
In recent years, single molecule junctions have provided
scientists with a unique manner to probe charge dynamics at
the molecular scale and, in turn, understand how to tune the
thermoelectric properties. A large interest lies in controlling
and maximizing the power factor of molecular junctions: S2G,
where S is the Seebeck coefficient (also known as the thermo-
power) and G is the electrical conductance. S and G are in-
versely correlated in macroscopic materials, challenging
researchers to maximize thermoelectric efficiencies by decou-
pling these two phenomena.
In junctions formed with small molecules strongly bonded to
two electrodes, transport is typically dominated by tunneling:
the so-called “Landauer transport regime”.1,2 In this coherent
tunneling regime, the relative level alignment of the frontier
orbital to the lead Fermi energy (EF) and the nature of the
bond, including the coupling of this level to the lead (Γ),
dictate the magnitude and sign of G and S. Assuming the trans-
mission function is an approximately Lorentzian function
peaked at the frontier orbital resonance energy with broadening
Γ, G increases with Γ, and as the resonance moves closer to EF.3
On the other hand, S is proportional to the negative logarithmic
In this work, the energetics of the junction were manipulated
through modification of the molecular backbone while leaving
the binding group constant. For this series of molecules, the thi-
oacetate binding group was selected, because it binds strongly
to gold. Because of the strength of the S−Au chemical bond,
Received: November 19, 2014
Published: November 22, 2014
© 2014 American Chemical Society
7229
dx.doi.org/10.1021/cm504254n | Chem. Mater. 2014, 26, 7229−7235

Chemistry of Materials
Article
there is a large charge localization at the metal/molecule
interface, leading to a gateway state, which is an intermediary
state coupling the lead to the molecular backbone. The effect of
the gateway state under variation of the molecular backbone
was examined. The electron transport through these junctions
can be separated into two sequences: the transport into the
molecule through the binding group, and transport within
the molecule itself. It has been experimentally shown that
the binding groups on each molecule dominate the coupling
of the molecule to the electrode, resulting in varying peak
breadths within the calculated transmission functions of these
junctions.22−24 The strength of the bond coupling the mole-
cule to the gold electrodes determines the energetics of the
metal/molecule interface, which has strong implications for
S and G.
The concept that differences in molecular design can cause
strong changes in the Seebeck coefficient of the molecular
junction was examined. Evidently, the molecular design dictates
the modes of electron transport within the channel. Thiophene-
based molecular junctions, shown in Figure 1, were used to
investigate this concept through two series: one consisted of
thiophene molecules with increasing alkyl linker lengths (TA
series: T1, TA2, and TA3) and the other had increased thio-
phene π-conjugated units (OT Series: T1, OT2 and OT3).
Thiophene was chosen to study the effects of a higher HOMO
energy, since thiophene has a HOMO energy closer to the EF
of gold than benzenedithiol.25,26 By increasing the alkyl linker
length in TA2 and TA3, the coupling of the molecular back-
bone and gateway state is expected to be reduced, leading to
localization of charge on the thiophene core and a weaker peak
in the transmission function gateway state. An increase in the
conjugated backbone in OT2 and OT3 was designed to investi-
gate the effects on conjugated molecular length coupling to
a gateway state for a HOMO positioned very close to EF.
Detailed synthesis of all these molecules can be found in the
Experimental Methods section. The experimental conductance
and Seebeck coefficient measurement was performed in a
custom-built STM-break junction setup, with experimental de-
tails provided in the method section below. Measurements of G
for the molecular series T1, TA2, TA3 and T1, OT2, OT3 are
presented which show a decreasing trend with length in both
cases. S was measured for the same molecular series T1, TA2,
TA3 and T1, OT2, OT3. The OT series showed a positive,
increasing S with length and the TA series showed a positive
S that decreases with length. A tight-binding model with a
gateway state was employed to capture all of the experimentally
observed trends reported herein.
Figure 1. (Top) Chemical structures of molecules measured. The
thioacetate binding group will be referred to as AcS (a) T1: (S,S′-
thiophene-2,5-diyl diethanethioate), (b) TA2: (S,S′-thiophene-2,5-
diylbis(methylene) diethanethioate), (c) TA3: (S,S′-2,2′-(thiophene-
2,5-diyl)bis(ethane-2,1-diyl) diethanethioate (d) OT2: S,S′-([2,2′-
bithiophene]-5,5′-diyl)diethanethioate (e) OT3: S,S′-([2,2:5′,2″-ter-
thiophene]-5,5″-diyl) diethanethioate. (Bottom) Electrical conduc-
tance of molecular junctions with T1, TA2, and TA3 (blue circles) and
T1, OT2 and OT3 (red squares) plotted as a function of molecular
length. The data are fit to a standard exponential β-decay model. The
βTA-decay parameter is calculated to be 3.15 nm−1, and the βOT-decay
parameter is calculated to be 2.96 nm−1; these values show that
increasing the alkyl linkers results in a β value that compares between
the β value of π-conjugated systems and nonconjugated systems. The
data points are the peak in the conductance histogram, and the error
bars represent the full width at half-maximum of the histogram peaks,
which can be found in the Suporting Information (SI), and do not
represent instrumental error, but, instead, variances in binding
geometry.
are 0.0075G0, 0.0035G0, 0.002G0, 0.0031G0, and 0.0007G0,
respectively. These length-dependent values compare well to
similar thiophene molecules in the literature.25 This shows that
T1 has a similar conductance, compared to more commonly
studied molecules, such as benzenedithiol (0.011G0) or ben-
zenediamine (0.0064G0).10 The experimental results indeed
show an exponential decay in electrical conductance with mole-
cular length, with T1, TA2, and TA3 having a βTA-decay param-
eter of 3.15 nm−1 and T1, OT2, and OT3 having a βOT-decay
parameter of 2.96 nm−1. Since it has been shown that a non-π-
conjugated system has a β-decay parameter of 9.363 nm−1,27
whiles a π-conjugated system has a β-decay parameter of
2.9 nm−1,25 it can be inferred that additional non-π-conjugated
linkers to a π-conjugated system results in behavior between
that of a fully conjugated π-conjugated system and a non-
conjugated system, because of partial delocalization of the
electron over the molecule.
EXPERIMENTAL RESULTS
■
The conductance of a single molecule junction as a function of
length is viewed as a tunneling barrier between two electrodes,
with an exponential decay parameter termed β, which differs
based on the molecular backbone. To confirm that this holds
true, the electrical conductance of molecules T1, TA2, TA3,
OT2, and OT3, as shown in Figure 1, were investigated. The
histograms of these conductance traces can be found in the SI.
Experimental results for the conductance are presented, as a
function of molecular length, and these results are compared to
β-decay parameters of known conjugated and nonconjugated
systems. The molecular length is defined to be from sulfur to
sulfur between the two thiol terminations when the molecule is
in a relaxed geometry (that is, as symmetric and straight as
possible). The conductance of T1, TA2, TA3, OT2 and OT3
Additional information about the effects of π-conjugated versus
nonconjugated molecular fragments on electron transmission can
7230
dx.doi.org/10.1021/cm504254n | Chem. Mater. 2014, 26, 7229−7235

Chemistry of Materials
Article
be obtained from measurement of the Seebeck coefficient.
Molecules in the TA series have an increasing number of alkyl
linkers between the molecular ring and the binding group. This is
intended to increase the spacing between the aromatic system
and the electrodes, in order to isolate the molecule energy levels
from the electrodes, which has been already seen to decrease
the conductance more rapidly than the addition of conjugated
units. The blue line in Figure 2 shows the Seebeck coefficient
Seebeck coefficient with increasing molecular length. The slope
of the Seebeck coefficient, with respect to molecular length, for
oligophenyldithiols is 6.9 μV/(K nm).28 The larger Seebeck
coefficient measured for T1 (6.83 μV/K), compared to ben-
zenedithiol on this system (4.8 μV/K), shows the effect that
using heterocycles with better HOMO energy alignment to the
gold EF increases S without lowering G.
THEORETICAL MODEL
■
To quantitatively model transport properties of a junction,
there are two key ingredients: an accurate representation of
how the molecule binds to the leads (junction geometry more
generally), and a quantitative description of the level alignment
of the molecular orbital resonance energies, with respect to the
lead Fermi energy (EF).4 For thiol-terminated molecules, broad
conductance histograms are measured compared to other ter-
minal binding groups,29 resulting from the sensitivity of con-
ductance, with respect to binding geometry, which is highly
variable, because of the nondirectionality of the S−Au bond,30
posing challenges to theory in developing representative junc-
tion geometries. In addition, even if suitable model geometries
existed, first-principles density functional theory (DFT) ap-
proaches are widely known to lead to incorrect level align-
ment,31,32 resulting in an artificially low tunnel barrier and an
overestimation of the conductance by an order of magnitude or
more.33 While GW-based self-energy corrections can improve
the description of level alignment in the junction, corrections
come at considerable computational expense, even for well-
defined geometries.
We use a physically motivated tight-binding model with a
minimal number of parameters, following previous work on
strongly bound molecules;25,34 our model is sufficient to explain
measured trends in junction G and S, and provides a basis for
interpreting future ab initio calculations. In our tight binding
model, the leads are coupled via a broadening Γ to a gateway
state, at energy Es, localized on the thiol; this gateway state is, in
turn, coupled to the molecular backbone at energy Eb with off-
diagonal coupling strength τ. For molecules with N thiophene
rings, we introduce N sites along the backbone with inter-ring
hopping parametrized by δ. Using these parameters, the Hamil-
tonian (/), for OT2 can be expressed as
Figure 2. Seebeck coefficient of T1, TA2, TA3 (blue circle) and T1,
OT2, OT3 (red square). T1 is in both TA and OT, and it is repre-
sented as a blue circle on a red square. Benzenedithiol was also
measured in this setup (black triangle). The slope of the alkyl-
thiophene (blue line) is −10.6 μV/(K nm) and corresponds well to
literature trends on the Seebeck coefficient of pure alkanethiol junc-
tions. The slope of the oligothiophene (red line) is 10.08 μV/(K nm),
and is nearly double that of oligophenylthiols, the nearest comparable
system.
measurements of the TA series, with respect to molecular length,
and compares these values to benzenedithiol (black). The Seebeck
coefficients for T1, TA2, and TA3 are 6.83, 3.01, and 2.2 μV/K,
respectively. By fitting these numbers to a linear regression, the
calculated slope is −10.6 μV/(K nm). In pure alkanethiol systems
with no π-conjugation in the literature, the slope in the Seebeck
coefficient is −5.6 μV/(K nm).28 This demonstrates that the
trend with the introduction of an alkyl linker to a π-conjugated
system is not the same trend observed in solely nonconjugated or
conjugated systems. As suggested with the length trend in con-
ductance, the Seebeck coefficient decrease is larger than with pure
alkane systems, and may be due to electron delocalization across
the molecule, resulting in a system between that of a purely alkyl
non-π-conjugated system and a fully π-conjugated system.28
In comparison, the Seebeck coefficients of oligothiophenes
were also measured, where the increase in molecular length was
due to additional π-conjugated systems, and not alkyl spacer
groups. The red trend in Figure 2 shows that the Seebeck co-
efficients of T1, OT2 and OT3, which have an increase in con-
jugated molecular length, compared to molecules in the blue
trend, which have increases in nonconjugated molecular length.
By increasing the molecular π-conjugation length, an increase in
the Seebeck coefficient was observed. The Seebeck coefficients
of T1, OT2, and OT3 are 6.83, 7.49, and 14.84 μV/K, respec-
tively, with a calculated slope of the Seebeck coefficient, with
respect to molecular length of 10.08 μV/(K nm). Oligophenyls
with amine and thiol binding groups, which is a similar sys-
tem studied in the literature, also demonstrate an increase in
⎛
⎜
⎜
⎜
⎜
⎜
⎜
⎞
⎟
⎟
⎟
⎟
⎟
⎟
⎛
⎜
⎞
⎟
Γ
2
E − i
τ
0
0
s
⎝
⎠
τ
0
Eb
δ
Eb
0
/ =
δ
τ
⎜
⎜
⎝
⎟
⎟
⎛
⎜
⎞
⎟
Γ
0
0
τ
Es − i
⎝
⎠
2 ⎠
As noted in previous works on related systems,25 the gateway
state arises from the strong S−Au bond formed in the absence
of the hydrogen, allowing for strong hybridization and large
charge localization. For example, gateway states can be seen in
the transmission calculations of alkanedithiols of increasing
length in the work of Li et al.35 It can be shown that the positive
and decreasing S, as a function of length, for the alkylthiophene
molecular series T1, TA2, TA3 is a direct consequence of this
gateway state. Initially the parameters in the model were fit to
the measurements of the oligothiophenedithiol−Au junction
(T1, OT2, OT3) G and S; the model was then extended to
7231
dx.doi.org/10.1021/cm504254n | Chem. Mater. 2014, 26, 7229−7235

Chemistry of Materials
Article
incorporate the addition of CH2 groups in order to the capture
the behavior observed from TA2 and TA3.
To model the OT series, the data were fit to the model
varying the parameters Γ, Es, Eb, δ, and τ, and the experimen-
tal trends are replicated. As shown in Figure 3, the Seebeck
molecule, represented by the decrease in τ, the gateway state
can interact more strongly with the gold, resulting in a shifted
gateway state energy. Using the Γ and Eb values from the OT
series, the model is fit via Es,TA2, Es,TA3, τTA2, and τTA3. In the
model, it is observed that the decrease in τ contributes pri-
marily to the decrease in the conductance as a function of
length, where the shift in energy of the gateway state is the
crucial element to capture the decrease in Seebeck coefficient,
as a function of molecular length. The conductance and Seebeck
coefficient values from the model are compared to experiment in
Figure 4b and the transmission curves in Figure 4a, which again
clearly show the resonance due to the gateway state just below
EF, whereas the HOMO resonance is a few electron volts below
EF. If a different starting point from the OT series is used, the
same trends are produced with similar error bars.
Other efforts to capture this behavior for reduced tight-
binding models without gateway states fail. With a single level
alone, the only way to achieve a decreasing G and S with length
is to move the frontier orbital away from EF by a very large
unphysical energy shift. If we introduce a second level, a set of
parameters for which S is positive and decreasing with length
can be found; however, the conductance increases with length,
which is another unphysical result. Only this tight binding
model with a gateway state is able to re-create the trends ob-
served in experiment.
CONCLUSION
■
Thiophene junctions have a higher Seebeck coefficient and
conductance, because of closer alignment of the highest
occupied molecular orbital (HOMO) energy with the gold
Fermi energy (EF). By incorporating alkane linkers of in-
creasing length between the binding group and thiophene, the
molecular orbitals are localized, resulting in both a decreasing
conductance and Seebeck coefficient. In oligothiophene junc-
tions with increased conjugated units, an increase in the Seebeck
coefficient was observed. This increase in Seebeck coefficient is
attributed to a shift in the HOMO energy closer to the Fermi
energy of gold, along with a sharpening of this peak in the
transmission function. This highlights the needs for design rules,
as an increase in molecular length will result in a decrease of
conductance, but may increase or decrease the Seebeck co-
efficient, depending on the molecular design and the presence of
a gateway state.
Figure 3. (a) Schematic of tight-binding model being used with
multiple sites along the molecular backbone. (b) The calculated trans-
mission values for the set of parameters: Γ = 5.32 eV, Es = −0.98 eV,
Eb = −5.54 eV, δ = −2.40 eV, and τ = −0.83 eV. (c) The corre-
sponding conductance values, as predicted by the model. (d) Calcu-
lated Seebeck coefficient values, compared to experiment.
coefficient increases with length and the conductance decreases
with exponentially length, with a decay constant ranging from
3−5 nm−1, comparable to reported values for other conju-
gated backbones. The T1 transmission function, shown in red
in Figure 3b, resulting from the best-fit parameters, clearly in-
dicates features associated with the gateway state just below EF.
For the molecules TA2 and TA3, the T1 Hamiltonian is
extended to incorporate more sites representing additional CH2
groups by allowing the position of the gateway state (Es), to
move up in energy toward EF and the value of τ to decrease.
The changes in these two parameters physically represent a
decrease in overlap between the charge density on the thiol and
the thiophene frontier states. Under this decoupling along the
EXPERIMENTAL METHODS
■
Synthesis. All glassware was oven-dried or flame-dried, and the
reactions were conducted under an argon atmosphere, using the
Figure 4. (a) Transmission curves for T1, TA2, TA3 with the parameters: Γ = 5.32 eV, Eb = −5.54 eV, and Es = −0.98, −0.59. −0.10 eV, τ = −0.83,
−0.68, −0.58 eV for T1, TA2, TA3 respectively. (b) Conductance predicted by the model and compared to experiment. (c) The Seebeck coefficient
from the model, compared to the experimental values.
7232
dx.doi.org/10.1021/cm504254n | Chem. Mater. 2014, 26, 7229−7235

Chemistry of Materials
Article
schlenk line technique. THF, DMF and diethyl ether were obtained via
a solvent purification system packed with alumina. Unless specifically
mentioned, all chemicals are commercially available and were used as
received. Flash chromatography was performed using 60 Å silica gel
at −78 °C for 30 min, and at 0 °C for 1 h, before cold ethylene
peroxide (1.9 mL, 38.0 mmol, 5.0 equiv) was added quickly via a
syringe. The reaction was allowed to warm to room temperature and
was stirred overnight (15 h) and quenched with saturated NH4Cl (aq,
20 mL). The aqueous layer was extracted with CHCl3 (20 mL × 3).
The combined organic layers were dried (MgSO4), filtered, and con-
centrated to provide a brown oil. Purification by flash chromatography
(1:1 hexane:EtOAc) provides 2,2′-(thiophene-2,5-diyl)diethanol as a
1
(37−75 μm). H NMR spectra were recorded at either 500 or 600
MHz, and 13C NMR spectra were recorded at 100 or 126 MHz in
CDCl3. Chemical shifts are reported in units of ppm, referenced to
residual solvent peak as follows: 7.24 ppm for 1H NMR; 77.16 ppm for
13C NMR.
1
colorless oil (616 mg, 47%). H NMR (500 MHz, CDCl3) δ 6.65 (s,
2H), 3.75 (t, J = 6.3 Hz, 4H), 2.94 (t, J = 6.3 Hz, 4H), 2.47 (s, 2H).
13C NMR (126 MHz, CDCl3): δ 139.62, 125.44, 63.38, 33.52.
GC-MS: m/z = 172.
S,S′-Thiophene-2,5-diyl Diethanethioate (T1).
2,2′-(Thiophene-2,5-diyl)diethanol (198 mg, 1.15 mmol, 1.0 equiv)
was dissolved in CHCl3 (5 mL), Et3N (480 μL, 3.45 mmol, 3.0 equiv),
p-TsCl (482 mg, 2.53 mmol, 2.1 equiv) in CHCl3 (3 mL), and
catalytic amount of DMAP (5%) were added. After stirring at room
temperature overnight (24 h), the reaction was diluted with H2O
(20 mL), and the aqueous layer was extracted with CHCl3
(20 mL × 2). The combined organic layers were dried (MgSO4),
filtered, and concentrated to provide 2,2′-(thiophene-2,5-diyl)bis-
(ethane-2,1-diyl)bis(4-methylbenzenesulfonate) as a brown oil, which
was used without further purification. The analytical sample as a
colorless oil was purified by flash chromatography (3:1 hexanes:E-
tOAc). 1H NMR (600 MHz, CDCl3): δ 7.72 (d, J = 8.1 Hz, 4H), 7.30
(d, J = 8.1 Hz, 4H), 6.56 (s, 2H), 4.14 (t, J = 6.8 Hz, 4H), 3.04 (t, J =
6.8 Hz, 4H), 2.42 (s, 6H). 13C NMR (151 MHz, CDCl3): δ 144.99,
137.25, 132.92, 129.96, 127.96, 126.05, 70.04, 29.78, 21.74. MS (ESI)
m/z calcd for C22H24O6S3Na+: 503.1; found: 503.1.
2,5-Dibromothiophene (1.98 g, 8.18 mmol, 1.0 equiv) was dissolved in
dry Et2O (40 mL), and the solution was cooled to −78 °C. n-BuLi
(2.5 M in hexanes, 7.20 mL, 18.0 mmol, 2.2 equiv) was added
dropwise via syringe, and the reaction was allowed to stir at −78 °C for
30 min. Dry sulfur (786 mg, 24.5 mmol, 3.0 equiv) powder was added
in one portion, and the reaction was stirred for another hour before
AcCl (1.45 mL, 20.4 mmol, 2.5 equiv) was added via syringe. The cold
bath was removed, and the reaction was stirred for an additional 1 h,
before pouring into water (50 mL). The aqueous layer was extracted
with Et2O (30 mL × 2), and the combined organic layers were washed
with brine (10 mL), dried over MgSO4, filtered, and concentrated to
provide a gray oil. Purification by flash chromatography with a gradient
of 1:1 hexane:CH2Cl2 to CH2Cl2 provided T1 as a colorless oil
1
(1.58 g, 83%). H NMR (500 MHz, CDCl3): δ 7.10 (s, 2H), 2.39
The crude product and KSAc (657 mg, 5.75 mmol, 5.0 equiv) were
combined in a 25-mL round-bottom flask, dry DMF (12 mL) was
added, and the suspension was stirred at 80 °C for 3 h. The reaction
was diluted with H2O (50 mL) and Et2O (50 mL), and the aqueous
layer was extracted with Et2O (50 mL × 2). The combined organic
layers were washed with H2O (30 mL × 4), brine (20 mL), dried
(MgSO4), filtered, and concentrated to provide a yellow crystalline
solid. Purification by flash chromatography (1:1 hexanes:CHCl3, then
CHCl3) provided TA3 as a colorless crystalline solid (268 mg, 81%
(s, 6H). 13C NMR (126 MHz, CDCl3): δ 193.00, 135.42, 131.17,
29.73. GC-MS: m/z = 232.
S,S′-Thiophene-2,5-diylbis(methylene) Diethanethioate (TA2).
1
over two steps). H NMR (500 MHz, CDCl3): δ 6.61 (s, 2H), 3.08
(t, J = 7.4 Hz, 4H), 2.97 (t, J = 7.1 Hz, 4H), 2.29 (s, 6H). 13C NMR
(126 MHz, CDCl3): δ 195.37, 140.93, 124.87, 30.66, 30.64, 30.21. MS
(EI): m/z = 288.03.
2,5-Dibromothiophene was converted to 2,5-diformylthiophene via
lithium-halogen exchange, followed by trapping with DMF36 and, sub-
sequently, to thiophene-2,5-diyldimethanol via reduction with NaBH4
following the known procedures.37 TA2 was obtained via Mitsunobu
reaction shown as follows.
S,S′-([2,2′-bithiophene]-5,5′-diyl) Diethanethioate (OT2).
Diisopropyl azodicarboxylate (DIAD, 1.06 mL, 3.38 mmol, 1.0 equiv)
was added dropwise via a syringe into a dry THF (10 mL) solution of
PPh3 (1.41 g, 5.38 mmol, 2.5 equiv) in an ice bath. Thiophene-2,5-
diyldimethanol (310 mg, 2.15 mmol, 1.0 equiv) was dissolved in dry
THF (12 mL), and added slowly via a syringe to the above solution,
followed by the addition of AcSH (380 μL, 5.38 mmol, 2.5 equiv). The
reaction was allowed to warm to room temperature slowly and stir
overnight (24 h), and then heated to 70 °C for 2 h. After cooling to
room temperature, the reaction mixture was concentrated, and purified
by flash chromatography (1:1 hexane:CH2Cl2) to provide TA2 as a
2,5-Dibromothiophene (405 mg, 1.25 mmol, 1.0 equiv) was dissolved
in dry THF (25 mL), and the solution was cooled to −78 °C. n-BuLi
(2.5 M in hexanes, 1.25 mL, 3.125 mmol, 2.5 equiv) was added
dropwise via a syringe, and the reaction was allowed to stir at −78 °C
for 30 min, and 0 °C for 1 h. Dry sulfur (120 mg, 3.75 mmol, 3.0
equiv) powder was added in one portion, and the reaction was stirred
at 0 °C for another hour, before AcCl (267 μL, 3.75 mmol, 3.0 equiv)
was added via a syringe. The reaction was allowed to warm to RT and
stirred overnight (15 h), before pouring into water (50 mL). The
aqueous layer was extracted with Et2O (30 mL × 2), and the com-
bined organic layers were washed with brine (10 mL), dried over
MgSO4, filtered, and concentrated to provide a brown oil. Purification
by flash chromatography with a gradient of 1:1 hexane:CHCl3 to
1
colorless oil (141 mg, 25%). H NMR (500 MHz, CDCl3): δ 6.72 (s,
2H), 4.21 (s, 4H), 2.32 (s, 6H). 13C NMR (126 MHz, CDCl3): δ
194.72, 140.09, 126.47, 30.43, 28.28. GC-MS: m/z = 260.
S,S′-Thiophene-2,5-diylbis(methylene) Diethanethioate (TA3).
1
CHCl3 provided OT2 as a yellowish solid (102 mg, 26%). H NMR
(500 MHz, CDCl3): δ 7.14 (d, J = 3.8 Hz, 2H), 7.04 (d, J = 3.8 Hz,
2H), 2.41 (s, 6H). 13C NMR (126 MHz, CDCl3): δ 194.00, 142.72,
136.66, 125.13, 125.05, 29.83.
S,S′-([2,2:5′,2″-terthiophene]-5,5″-diyl) Diethanethioate (OT3).
2,5-Dibromothiophene (1.84 g, 7.61 mmol, 1.0 equiv) was dissolved in
dry THF (38 mL), and the solution was cooled to −78 °C in a dry
ice/acetone bath. t-BuLi (9.84 mL of 1.7 M in pentane, 16.7 mmol,
2.2 equiv) was added dropwise via a syringe. The reaction was stirred
7233
dx.doi.org/10.1021/cm504254n | Chem. Mater. 2014, 26, 7229−7235

Chemistry of Materials
Article
OT3 was synthesized under similar conditions as OT2,38 and purified
by flash chromatography with 3:2 hexane:CH2Cl2 to provide a
yellowish solid (yield = 16%). H NMR (600 MHz, CDCl3) δ 7.12
(d, J = 3.7 Hz, 2H), 7.06 (s, 2H), 7.04 (d, J = 3.8 Hz, 2H), 2.41 (s,
6H). 13C NMR (151 MHz, CDCl3): δ 194.10, 143.10, 136.70, 136.26,
125.34, 124.42, 124.37, 29.75.
Measurement. Five microliters (5 μL) of a 0.05 M THF solution
with the desired molecule at concentration of 10 mg/mL was drop-cast
by micropipette onto gold films. The gold films were thermally evapo-
rated onto a freshly cleaved mica substrate to enable an atomically
smooth surface. A modified STM-break junction with a gold tip is then
crashed into the gold substrate and withdrawn, with current and
voltage measurements being acquired throughout the process. Both
the STM Au tips and the Au thin films on mica were hydrogen-flame-
annealed to ensure the removal of containments.
S = S2C − β2SL
1
and the conductance of the TA series was fit to a decaying exponential:
G = G2C exp( − β2CL)
For each series, the difference between the experimental fit and the
value from the model was simultaneously minimized for all of the data
available, with respect to the parameters in the model. The parameters
in the model for the OT series were Γ, Es, Eb, δ, and τ. The parameters
for the model of the TA series use the parameters from the OT series
but allow the coupling between the gateway state and the backbone, as
well as the position of the gateway energy, to vary for each parameter;
therefore, we minimize, with respect to Es and τ, for each TA2 and
TA3 independently and keep Γ and Eb fixed.
The measurement of the molecular conductance requires a bias of
5 mV to be applied between tip and substrate upon approach. The
STM gold tip is driven to the gold substrate with a rate of 1 nm/s until
a conductance threshold of 1G0 is reached. The tip is then withdrawn
at a rate of 0.5 nm/s, with an Au−Au junction first forming, followed
by a junction with the molecule of interest. No voltage bias is applied
upon the withdrawal process. This process is repeated until 5000 junc-
tion traces are acquired. These results are then analyzed by histogram
statistics and the resulting histogram peaks gives the molecular con-
ductance, with the full width at half maximum (fwhm) of the peak
being the error bar.
The measurement of the Seebeck coefficient (S) required a tem-
perature gradient across the substrate and tip. The substrate was
heated via a specially constructed STM cell and Peltier heater, and the
STM tip was connected to a thermal sink to ensure a constant tem-
perature of 20 °C. The power output to the Peltier heater was varied
to control the temperature of the substrate. The gold STM tip is
driven into the gold substrate until a conductance of 5G0 is measured.
The tip−substrate voltage bias is then removed via a solid-state relay,
and this cycle is repeated until 2000 junctions are acquired. Each trace
is binned into a histogram. The peak voltage is the thermoelectric
voltage at that temperature, and the fwhm is the error bar. S is calcu-
lated via a least-squares linear fit of the measured molecule, with the
temperature drop as the independent variable, the peak voltage as the
dependent variable, and the fwhm approximating the error in the
measurement. Per convention, S is the negative of the resulting slope.
A subtraction of the Seebeck coefficient of the instrument (1.6 μV/K)
is then performed.
ASSOCIATED CONTENT
* Supporting Information
Detailed experimental details, synthesis, and theory procedures
can be found in the Supporting Information. This material is
available free of charge via the Internet at http://pubs.acs.org.
■
S
AUTHOR INFORMATION
Corresponding Author
*E-mail: segalman@engineering.ucsb.edu.
■
Author Contributions
○These authors contributed equally. The manuscript was
written through contributions of all authors. All authors have
given approval to the final version of the manuscript.
Funding
This work is supported by the Air Force Office of Scientific
Research (AFOSR) (No. MURI FA9550-12-1-0002).
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Dr. Bhooshan Popere, Dr. Bryan MuCulloch, Prof.
Shannon Yee, Prof. Fabian Pauly, and Dr. Victor Ho for their
insights.
REFERENCES
Theory. To calculate the conductance and Seebeck coefficient from
■
/
, the Green’s function (G) is constructed,
(1) Di Ventra, M.; Pantelides, S.; Lang, N. Phys. Rev. Lett. 2000, 84,
the Hamiltonian
979−982.
Gr = (ε − /)−1
(2) Di Ventra, M.; Pantelides, S. T.; Lang, N. D. Appl. Phys. Lett.
2000, 76, 3448.
which, along with the coupling, is used to calculate the transmission via
the standard expression
(3) Buttiker, M.; Imry, Y.; Landauer, R.; Pinhas, S. Phys. Rev. B 1985,
̈
31, 6207−6215.
T = Tr[ΓGrΓGa]
where Gr is the retarded Green’s function and Ga is the advanced
Green’s function. The transmission is related to the conductance and
thermopower through the following relations:
(4) Dell’Angela, M.; Kladnik, G.; Cossaro, A.; Verdini, A.;
Kamenetska, M.; Tamblyn, I.; Quek, S. Y.; Neaton, J. B.; Cvetko,
D.; Morgante, A.; Venkataraman, L. Nano Lett. 2010, 10, 2470−2474.
(5) Reed, M.; Zhou, C.; Muller, C.; Burgin, T.; Tour, J. Science 1997,
278, 252−254.
(6) Reddy, P.; Jang, S. Y.; Segalman, R. A.; Majumdar, A. Science
2007, 315, 1568−1571.
G = G0T(EF)
(7) Quinn, J.; Foss, F. J. Am. Chem. Soc. 2007, 129, 6714−6715.
(8) Xiao, X. Y.; Xu, B. Q.; Tao, N. J. Nano Lett. 2004, 4, 267−271.
(9) Chen, F.; Hihath, J.; Huang, Z. F.; Li, X. L.; Tao, N. J. Annu. Rev.
Phys. Chem. 2007, 58, 535−564.
(10) Tao, N. J. J. Mater. Chem. 2005, 15, 3260−3263.
(11) Xu, B. Q.; Tao, N. J. J. Science 2003, 301, 1221−1223.
(12) Ahn, S.; Aradhya, S. V.; Klausen, R. S.; Capozzi, B.; Roy, X.;
Steigerwald, M. L.; Nuckolls, C.; Venkataraman, L. Phys. Chem. Chem.
Phys. 2012, 14, 13841−13845.
2
π2kB T
d
dE
S = −
ln T(E = EF)
3e2
where G0 is the quantum of conductance.
In order to fit the parameters in the model, the experimental data
were fit either to a linear or exponential trend. When fitting the experi-
mental data, the thermopower of the OT series was fit to a line with a
positive slope and intercept:
(13) Chen, W.; Widawsky, J. J. Am. Chem. Soc. 2011, 17160−17163.
(14) Cheng, Z.-L.; Skouta, R.; Vazquez, H.; Widawsky, J. R.;
Schneebeli, S.; Chen, W.; Hybertsen, M. S.; Breslow, R.;
Venkataraman, L. Nat. Nanotechnol. 2011, 6, 353−357.
S = S1C + β1SL
The thermopower of the TA series was fit to a line with a negative
slope and positive intercept:
7234
dx.doi.org/10.1021/cm504254n | Chem. Mater. 2014, 26, 7229−7235

Chemistry of Materials
Article
(15) Kamenetska, M.; Quek, S. Y.; Whalley, A. C.; Steigerwald, M. L.;
Choi, H. J.; Louie, S. G.; Nuckolls, C.; Hybertsen, M. S.; Neaton, J. B.;
Venkataraman, L. J. Am. Chem. Soc. 2010, 132, 6817−6821.
(16) Kiguchi, M.; Miura, S.; Hara, K.; Sawamura, M.; Murakoshi, K.
Appl. Phys. Lett. 2006, 89.
(17) Lindsay, S. J. Chem. Educ. 2005, 82, 727−733.
(18) Beebe, J.; Kim, B.; Gadzuk, J.; Frisbie, C. D.; Kushmerick, J.
Phys. Rev. Lett. 2006, 97, 026801.
(19) Kiguchi, M.; Murakoshi, K. J. Phys. Chem. C 2008, 112, 8140−
8143.
(20) Schneebeli, S. T.; Kamenetska, M.; Cheng, Z. L.; Skouta, R.;
Friesner, R. A.; Venkataraman, L.; Breslow, R. J. Am. Chem. Soc. 2011,
133, 2136−2139.
(21) Yee, S. K.; Sun, J. B.; Darancet, P.; Tilley, T. D.; Majumdar, A.;
Neaton, J. B.; Segalman, R. A. ACS Nano 2011, 5, 9256−9263.
(22) Malen, J. A.; Yee, S. K.; Majumdar, A.; Segalman, R. A. Chem.
Phys. Lett. 2010, 491, 109−122.
(23) Venkataraman, L.; Park, Y. S.; Whalley, A. C.; Nuckolls, C.;
Hybertsen, M. S.; Steigerwald, M. L. Nano Lett. 2007, 7, 502−506.
(24) Park, Y.; Whalley, A. J. Am. Chem. Soc. 2007, 129, 15768−
15769.
(25) Capozzi, B.; Dell, E. J.; Berkelbach, T. C.; Reichman, D. R.;
Venkataraman, L.; Campos, L. M. J. Am. Chem. Soc. 2014, 136,
10486−10492.
(26) Tsutsui, M.; Taniguchi, M. Sensors (Basel) 2012, 12, 7259−
7298.
(27) Beebe, J. M.; Engelkes, V. B.; Miller, L. L.; Frisbie, C. D. J. Am.
Chem. Soc. 2002, 124, 11268−11269.
(28) Malen, J. A.; Doak, P.; Baheti, K.; Tilley, T. D.; Segalman, R. A.;
Majumdar, A. Nano Lett. 2009, 9, 1164−1169.
(29) Venkataraman, L.; Klare, J. E.; Tam, I. W.; Nuckolls, C.;
Hybertsen, M. S.; Steigerwald, M. L. Nano Lett. 2006, 6, 458−462.
(30) French, W. R.; Iacovella, C. R.; Rungger, I.; Souza, A. M.;
Sanvito, S.; Cummings, P. T. J. Phys. Chem. Lett. 2013, 4, 887−891.
(31) Toher, C.; Sanvito, S. Phys. Rev. B 2008, 77, 155402.
(32) Pontes, R. B.; Rocha, A. R.; Sanvito, S.; Fazzio, A.; Roque da
Silva, A. J. ACS Nano 2011, 5, 795−804.
(33) Quek, S. Y.; Choi, H. J.; Louie, S. G.; Neaton, J. B. Nano Lett.
2009, 9, 3949−3953.
́
(34) Widawsky, J.; Chen, W.; Vazquez, H.; Kim, T.; Breslow, R.;
Hybertsen, M. S.; Venkataraman, L. Nano Lett. 2013, 13, 2889−2894.
(35) Li, C.; Pobelov, I.; Wandlowski, T. J. Am. Chem. Soc. 2008, 130,
318−326.
(36) Mitsumori, T.; Inoue, K. J. Am. Chem. Soc. 1995, 2467−2478.
(37) Ikeda, T.; Higuchi, M.; Sato, A.; Kurth, D. Org. Lett. 2008,
2004−2007.
(38) Boer, B. De; Meng, H.; Perepichka, D.; Zheng, J.; Frank, M. M.;
Chabal, Y. J.; Bao, Z. Langmuir 2003, 4272−4284.
7235
dx.doi.org/10.1021/cm504254n | Chem. Mater. 2014, 26, 7229−7235