STM contrast, electron-transfer chemistry, and conduction in molecules

Article abstract of DOI:10.1021/jp972510u

We have used a series of metalloporphyrin compounds to test for a relationship between the contrast of STM images and the electrochemical properties of the molecules. Molecules were tethered to a gold (111) surface by means of an isothiocyano linkage and both images and current-voltage (I-V) curves were obtained with the sample submerged in oxygen-free mesitylene. The contrast of the reducible molecules changed strongly with bias, and the corresponding I-V curves were highly asymmetric. The derivative of these curves (dI/dV) had a Gaussian-shaped peak at a voltage characteristic of the compound, although local measurements showed that there was considerable variation in this value from molecule to molecule of a given compound. These bias-dependent features were not observed in the less electroactive molecules, so the STM is capable of distinguishing electroactive molecules from non-electroactive molecules as we demonstrate with images of mixed films. We discuss one- and two-step electron-transfer mechanisms which are consistent with these observations.

Full text of DOI:10.1021/jp972510u

J. Phys. Chem. B 1997, 101, 10719-10725
10719
STM Contrast, Electron-Transfer Chemistry, and Conduction in Molecules
Wenhai Han,† Edgardo N. Durantini,‡ Thomas A. Moore,‡ Ana L. Moore,‡ Devens Gust,‡
Peter Rez,† Gerald Leatherman,† Gilbert R. Seely,‡ Nongjian Tao,§ and S.M. Lindsay*,†
Department of Physics and Astronomy and Department of Chemistry and Biochemistry,
Arizona State UniVersity, Tempe, Arizona 85287, and Department of Physics, Florida International UniVersity,
Miami, Florida 33199
ReceiVed: August 1, 1997; In Final Form: September 27, 1997X
We have used a series of metalloporphyrin compounds to test for a relationship between the contrast of STM
images and the electrochemical properties of the molecules. Molecules were tethered to a gold (111) surface
by means of an isothiocyano linkage and both images and current-voltage (I-V) curves were obtained with
the sample submerged in oxygen-free mesitylene. The contrast of the reducible molecules changed strongly
with bias, and the corresponding I-V curves were highly asymmetric. The derivative of these curves (dI/dV)
had a Gaussian-shaped peak at a voltage characteristic of the compound, although local measurements showed
that there was considerable variation in this value from molecule to molecule of a given compound. These
bias-dependent features were not observed in the less electroactive molecules, so the STM is capable of
distinguishing electroactive molecules from non-electroactive molecules as we demonstrate with images of
mixed films. We discuss one- and two-step electron-transfer mechanisms which are consistent with these
observations.
Introduction
molecules (protoporphyrins containing iron) showed a strong
potential dependence, the STM contrast increasing approxi-
The mechanism of electronic conduction in and between
molecules is a question of fundamental importance in molecular
electronics, electrochemistry, and biochemistry. It is now
possible to examine the conductivity of single molecules by
means of the scanning probe microscope, and one might expect
mechanisms quite different from metallic conduction because
electronic charge is usually localized in molecules.1 To date,
however, most scanning tunneling microscope (STM) images
of atoms2 and molecules3,4 have been successfully interpreted
in terms of modification of metallic states at the Fermi energy
caused by the interaction of the atom or molecule with the metal
surface. The electrons are assumed to be passed through the
molecule by tunnel transport,5 and charge does not accumulate
on the molecule. Current-voltage, I(V), characteristics are
linear (for small voltages), reflecting the slowly varying
electronic transmission of the tails of broadened molecular
orbitals in the HOMO-LUMO gap.5
At higher bias, however, the molecular states can be probed,
and there is evidence of their contributions in the voltage
dependence of images of benzene on graphite and MoS2.6 In a
condensed environment, it is more appropriate to consider the
relaxed states of the HOMO and LUMO. Schmickler7,8 has
suggested that the relaxed LUMO distribution (or density of
oxidized states) can be probed by an STM. In this context, the
“density of oxidized states” means the electron affinity levels
as broadened and shifted by coupling to the environment and
fluctuations. A similar use of tunneling at macroscopic
electrodes was proposed by Bennett9 and demonstrated by
Morisaki et al.10 Tao11 has studied the potential dependence
of STM images of protoporphyrin molecules adsorbed onto a
graphite electrode and kept under potential control by means
of a supporting aqueous electrolyte which covered the sample
during the STM imaging. He found that images of reducible
mately 10-fold when the substrate potential was adjusted to the
half-wave potential, E0, for the FeIII to FeII reduction. (The
relatively small difference between half-wave and formal
potentials12 will be ignored in this paper.) No such effect was
observed in the protoporphyrins that did not contain iron. Mazur
and Hipps13 have noted that the position of a peak in the
derivative of the current-voltage characteristics of a number
of metal-insulator-molecule-metal tunnel junctions coincides
approximately with the first reduction potential of the molecules.
They made this connection by assuming that the electrons were
injected at the Fermi energy of the electrode plus the applied
bias, and they used the work function of the normal hydrogen
electrode (NHE) to relate this energy to reduction potentials
measured relative to a standard reference. Further evidence for
such a correlation is found in the data of Burghard et al.14
Snyder and White15 measured enhanced tunnel current in films
of iron protoporphyrin on graphite. However, they found an
enhancement that was identical for both signs of the tip bias, a
result different from the data of Tao11 and Mazur and Hipps.13
Bumm et al.16 have used both STM and alternating-current STM
to show that a “molecular wire” (an ethyl-substituted 4,4′-di-
(phenyleneethynylene)benzothioacetate) is more conductive than
a surrounding “inert” matrix, but current-voltage characteristics
were not reported.
At first glance, a connection between a maximum in the
derivative of a tunneling current I(V) characteristic and a half-
wave potential is rather surprising. The half-wave potential does
not correspond to any molecular eigenstate energy but rather
to the equilibrium potential established when reduction and
oxidation currents are balanced at an electrode.17 In Schmick-
ler’s model, a maximum in the derivative curve occurs when
electrons are aligned with the peak in the “density of oxidized
states” (the occupied states do not contribute to the current).
Thus, in this model, the peak current is shifted away from the
half-wave potential (which lies between the peak in the reduced
and the peak in the oxidized state densities) by the amount of
the reorganization energy (λ in the terminology of Marcus18,17).
† Department of Physics and Astronomy, Arizona State University.
‡ Department of Chemistry and Biochemistry, Arizona State University.
§ Florida International University.
X Abstract published in AdVance ACS Abstracts, November 15, 1997.
S1089-5647(97)02510-8 CCC: $14.00 © 1997 American Chemical Society

10720 J. Phys. Chem. B, Vol. 101, No. 50, 1997
Han et al.
TABLE 1: Measured First Reduction (E0-1) and First
Oxidation Potentials (E0+1) for the Compounds Used in This
Work. Data for itc-TPP-Zn Are Estimated As Outlined in
the Text
compound
E0-1(SCE) (V)
E0+1(SCE) (V)
itc-TPP-H2
itc-TPP-Zn
itc-TPP-FeCl
itc-TPP-FeBr
itc-TPP-MnBr
-1.23
-1.30
-0.34
-0.24
-0.26
+0.96
+0.77
+1.13
+1.13
+1.14
was washed with 100 mL of water and separated from the water
phase, and the solvent was evaporated under vacuum. Flash
chromatography on silica gel (dichloromethane-methanol)
afforded 30 mg (89%) of itc-TPP-FeCl. MS m/z: 802 (M+)
(802 calculated for C48H33N5SFeCl), 767 (M+-Cl).
Itc-TPP-FeBr was prepared as described for itc-TPP-FeCl
using 12 mL of acetonitrile, 108 mg (0.501 mmol) of ferrous
bromide, and 22 mg (0.031 mmol) of itc-TPP in 6 mL of
chloroform. The solution was diluted with 100 mL of dichlo-
romethane and stirred with 1 M hydrobromic acid saturated with
potassium bromide (100 mL). Flash chromatography on silica
gel (dichloromethane-methanol) yielded 23 mg (88%) of itc-
TPP-FeBr. MS m/z: 846 (M+) (846 calculated for C48H33N5-
SfeBr), 767 (M+-Br).
Itc-TPP-MnBr was synthesized by refluxing 30 mg (0.042
mmol) of itc-TPP and 218 mg (1.26 mmol) of manganese(II)
acetate in 30 mL of glacial acetic acid and 20 mL of chloroform
for 4 h. After cooling, the solution was stirred with 1 M
hydrobromic acid saturated with potassium bromide (100 mL).
The organic phase was washed with 100 mL of water and
separated from the water phase, and the solvent was evaporated
under vacuum. Flash chromatography on silica gel (chloroform-
methanol) afforded the product, 33 mg (93%). MS m/z: 845
(M+) (845 calculated for C48H33N5SMnBr), 766 (M+-Br).
Itc-TPP-Zn was prepared by dissolving 17 mg (0.024 mmol)
of itc-TPP in 5 mL of dichloromethane and adding a saturated
solution of zinc(II) acetate in methanol. The mixture was stirred
for 30 min. The product, 18 mg (96%), was isolated by column
chromatography on silica gel (dichloromethane). MS m/z: 775
(M+) (775 calculated for C48H33N5SZn).
Figure 1. Structure of the molecules used in this study.
This is often a substantial fraction of an electronvolt.17 Both
Tao’s data and the experiments of Mazur and Hipps,13 on the
other hand, indicate that the maximum molecular contribution
to the current occurs quite close to the half-wave potential.
Tao’s experimental approach has the considerable advantage
that the potential of at the molecular adlayer is well established.
However, the experiments are difficult to carry out because the
adlayer must remain stable over a wide range of potential in
the supporting aqueous electrolyte. The limited available range
of tip potential further restricts conventional spectroscopic
measurements. An alternative is to attach a molecule firmly to
a substrate with a tether and carry out two-electrode spectro-
scopic measurements in an electrochemically inert environment.
We have developed a chemical tether19 which we have used to
attach a number of molecules to a gold electrode. The sample
is submerged in a dielectric fluid to permit spectroscopic, I(V),
measurements but with consequent loss of potential control.
While this introduces uncertainties, we believe that these are
on the order of a fraction of an electronvolt and are comparable
to other uncertainties, as we will discuss.
The molecules we have used are tetraphenylporphyrin deriva-
tives (TPP) into which various metals (TPP-M) have been
substituted. One of the porphyrin meso aryl rings bears an
isothiocyano group (itc-TPP-M) which tethers the molecule to
a gold substrate.19 The other three rings have methyl substit-
uents in the para positions. The structures of these molecules
are shown schematically in Figure 1. We studied the free base
(M ) 2H), iron(III), manganese(III), and zinc(II) porphyrins.
The iron and manganese atoms bear formal charges and are
bound to a halide ion ligand (itc-TPP-FeCl, itc-TPP-FeBr, and
itc-TPP-MnBr). This series of molecules is easily reduced,
while the free base and zinc porphyrins are not. We report the
results of STM imaging and spectroscopic measurements of
these molecules on Au(111) in this paper.
Freshly distilled mesitylene was used as a solvent in all STM
experiments reported (but see below).
Cyclic Voltammetry. Cyclic voltammograms of itc-TPP-
FeCl, -FeBr, and -MnBr were recorded at a glassy C electrode
in benzonitrile solutions with 0.1 M tetrabutylammonium
hexafluorophosphate electrolyte. Ag/0.01 M AgNO3 served as
a provisional reference electrode, but ferrocene (E0 ) 0.46 V
vs the saturated calomel electrode (SCE) in benzonitrile) was
used as an internal reference for correlating potentials. The
voltammograms, which covered approximately the range -2.0
V to +1.6 V vs Ag/Ag+, showed a reversible first oxidation
wave, an irreversible second oxidation wave, and two or more
reduction waves with various degrees of reversibility. Since
ligating anions dissociated to some extent upon reduction of
the trivalent metal ions, reversibility was improved by adding
some quaternary Cl- or Br- salt, and in these cases only half-
wave potentials recorded in the presence of excess anions are
reported in Table 1.
Voltammograms of itc-TPPH2 were recorded in CH2Cl2 with
the same electrolyte and glassy C electrode but were referred
directly to the SCE. Ferrocene (0.52V vs SCE in CH2Cl2) was
used to correct the values in CH2Cl2 to benzonitrile in Table 1.
Values for itc-TPPZn were not measured, but were estimated,
with good expectation of reliability, from values for itc-TPPH2,
for tetraphenylporphyrin21 and for its Zn complex22 in CH2Cl2.
Experimental Section
Samples. Itc-TPP was prepared by methods previously
reported.20,19 The composition of all samples was verified by
mass spectrometry (MS), and measured mass/charge ratios (m/
z) are listed for each compound.
Itc-TPP-FeCl. A portion of acetonitrile (20 mL) was heated
at reflux for 30 min with vigorous stirring under a stream of
nitrogen. Ferrous chloride (133 mg, 1.05 mmol) was added
and completely dissolved by stirring the mixture for several
minutes at reflux. The solution was cooled to 50 °C, and a
solution of 30 mg (0.042 mmol) of itc-TPP in nitrogen-purged
chloroform (8 mL) was added. The mixture was stirred for 15
min under nitrogen, exposed to air, diluted with 100 mL of
dichloromethane, and stirred with 1 M hydrochloric acid
saturated with sodium chloride (100 mL). The organic phase

STM Contrast in Molecules
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10721
Scanning Probe Microscopy and Spectroscopy. We used
a PicoSPM (Molecular Imaging Corp., Tempe, AZ) which
employs a single-piece Teflon fluid cell. It was necessary to
clean the cell thoroughly prior to each experiment by washing
it in a Soxhlet extractor containing a mixture of dichloromethane
(20%) and methanol (80%) for at least 4 days. We prepared
gold (111) substrates on mica as described elsewhere23 and
annealed them in a hydrogen flame immediately prior to use.
A 10-20 µM solution of the itc-TPP-M in freshly distilled
mesitylene was placed onto the gold immediately after the liquid
cell was assembled (resulting in 5-10 min of exposure to
ambient conditions in a laminar flow hood). The sample stage
was mounted on the PicoSPM and sealed into the environmental
chamber. The chamber was flushed continuously with ultrapure
nitrogen gas. The instrument was left to stabilize (1-10 h)
after which time images and spectra were collected. We used
etched Pt0.8Ir0.2 tips that had been thoroughly rinsed in ultrapure
water and dried in a stream of clean nitrogen. Images were
collected as described in the figure captions.
Current-voltage spectra were obtained starting at a set point
of 80 mV bias and 50 pA tunneling current. The sample bias
was swept over (1 V, and 256 sample points of 25 µs dwell
time were collected. Thirty consecutive sweeps were integrated
for each spectrum accumulated. On the order of 100 such
integrated spectra were obtained from randomly chosen sites
on a given sample, and the experiment was repeated several
times. The total number of integrated spectra were 672 (gold
under clean mesitylene), 848 (itc-TPP-FeCl), 598 (itc-TPP-
FeBr), 473 (itc-TPP-MnBr), 508 (itc-TPP-Zn), and 588 (itc-
TPP-H2).
Results
The images of the various molecules did not show submo-
lecular resolution, the molecules appearing as blobs with a
characteristic size of around 2nm. For that reason, it was not
possible to distinguish images of the molecules we prepared
from some forms of contamination. Thus, scrupulous attention
to cleanliness was required. In addition, failure to purge oxygen
from the samples led to complicated and generally nonreversible
current-voltage characteristics. The cleaning procedures de-
scribed were developed to yield images in which the density of
features scaled with the concentration of the molecules in the
solution. These clean conditions generally yielded reproducible
tunneling spectra. Examples of images obtained under these
conditions are shown in Figure 2.
Despite thorough cleaning of the mesitylene by repeated
distillation in clean glass, we found a low level of background
features on clean gold under mesitylene (around 10 features in
an area 100 nm by 100 nm). These features were not found
when toluene was used. However, toluene evaporated too
quickly to permit experiments of adequate duration reliably. We
repeated our measurements on itc-TPP-FeCl in toluene, finding
no difference from results obtained in mesitylene. For that
reason, we used mesitylene for all the work reported here.
Figure 2 shows that there is a strong bias dependence in the
contrast of the images of the three reducible molecules (itc-
TPP-FeCl, itc-TPP-FeBr, and itc-TPP-MnBr). There appears
to be a threshold for the onset of enhanced image contrast (of
around 0.5V), and the effect was only observed when the
substrate was biased negative with respect to the tip. Images
taken at positive sample bias are similar to the low-bias images
in Figure 2. In contrast, the less electroactive molecules, itc-
TPP-Zn and itc-TPP-H2, show no effect of bias in the range of
(1 V.
Figure 2. Typical STM images of the Au(111) surface under
mesitylene after adsorption of itc-TPP-FeBr (A), itc-TPP-FeCl (B), itc-
TPP-MnBr (C), itc-TPP-Zn (D), and itc-TPP-H2 (E). Data are shown
for low (left) and high (right) substrate bias as marked. Contrast changes
were not observed with the substrate bias positive (data not shown).
Images were acquired at a line scan rate of 3 Hz with a tunnel current
of 50 pA. A scale bar for all images is shown on the lower right of E.
that a mechanical interaction between the tip and the molecules
could play a role in the bias dependence of the images. We
show that this is unlikely to be the case with images of a mixed
film (20% itc-TPP-FeCl, 80% itc-TPP-H2) in Figure 3. The
large white spot near the middle of the image is one of the
mesitylene-induced features described earlier, but the other
features in Figure 3A are owing to the presence of porphyrin
molecules on the surface. When the negative sample bias was
increased (Figure 3B), additional features were observed on the
Lowering the bias applied to the STM tip results in the tip
being positioned closer to the metal surface, so it is conceivable

10722 J. Phys. Chem. B, Vol. 101, No. 50, 1997
Han et al.
Figure 4. Current-voltage curves averaged for many molecules
obtained over several runs as described in the text. Curves for the
reducible molecules (itc-TPP-FeCl, itc-TPP-FeBr, and itc-TPP-MnBr)
show significantly enhanced current at large negative substrate bias.
The curve for gold in mesitylene is symmetrical about the origin, while
the curves for itc-TPP-Zn and itc-TPP-H2 show some enhancement at
large positive bias.
Figure 3. Series of successive scans over a mixed adlayer of itc-TPP-
H2 and itc-TPP-FeCl with the bias of the substrate switched as shown
by the inset in the upper right of each image. A cluster of molecules
(pointed to by the arrow) lights up at high bias (B) but becomes invisible
again at low bias (C), returning again whenever the bias is high (D,
E). Tunneling conditions were as in Figure 2.
surface (one example is indicated by the arrow). These were
much brighter, having a contrast similar to the itc-TPP-FeCl
images at high negative bias (Figure 2A). When the bias was
returned to a small value, these additional features disappeared
(Figure 3C). On returning the bias to a large negative value,
the additional features were restored (Figure 3D). The cycle is
repeated in subsequent images (Figures 3E,F). The reversibility
of the effect argues against a simple mechanical interaction,
such as the tip sweeping some molecules away at low bias.
Averaged over a number of images like those shown in Figure
3, the population of molecules with bias-dependent contrast was
close to 20%, showing that these are the iron-containing
porphyrins.
Inspection of Figure 3 reveals an unexpected phenomenon
which is not evident in the images shown in Figure 2, where
the contrast has been adjusted to make all the images exhibit
approximately the same contrast at low bias. Compared to the
free base porphyrins, the iron-containing porphyrins are almost
invisible at low bias. This effect emphasizes the electronic (as
opposed to topographic) nature of the contrast.
Rather than attempt to extract quantitative data from the
images, we chose to accumulate tunneling spectra. In this mode,
the scan was disabled and the tip set to tunnel at a fixed
resistance (50 pA at 180 mV bias). The STM servo was then
transiently opened and the bias scanned from -1000 to +1000
mV over 6 ms and servo control then restored. The shape of
the spectra depends critically on the exact location of the tip in
relation to a molecule on the substrate. However, for bare gold
in mesitylene, the curves are symmetrical and the currents small,
so the integrated data taken from many scans correspond to an
average over all tip positions with respect to the molecule.
The integrated sum of all our data is shown for all the
molecules studied in Figure 4. We also include data for the
gold substrate under clean mesitylene. The gold data are quite
symmetrical in the range of (1 V. The molecular adlayers show
some sign of additional current at the highest positive biases,
but the differences are most pronounced for high negative
substrate bias, consistent with the images (Figure 2). Note that
because the gap resistance was a set to a common value before
the acquisition of spectra, relative variations at low bias, such
as the anomalous contrast for the itc-TPP-FeCl discussed above,
are not evident in these curves.
Figure 5. (A) Derivatives (dI/dV) of the current-voltage curves shown
in Figure 4. Data for the three reducible molecules (itc-TPP-FeCl, itc-
TPP-FeBr, and itc-TPP-MnBr) are fitted by Gaussians of the form given
in eq 1 (solid lines). (B) Series of local scans (30 sweeps) over itc-
TPP-FeCl. Within the noise, each curve is well described by a Gaussian,
but the peak position is quite variable and the width is less than that of
the curves fitted to the full data set.
data according to the theory of Schmickler8 suggests that this
extra current results from tunneling via oxidized states on the
molecule. The extra current is integrated into the total as the
bias voltage is made more negative, but the derivative, dI/dV,
should show evidence of the peak in the molecular density of
states. We show the numerical derivatives of these spectra in
Figure 5(A). There are distinct peaks for the iron-containing
molecules and evidence of the onset of a maximum in the data
for itc-TPP-MnBr. We fitted the peaks with Gaussians of the
form
(V - Vp)2
dI
dV
) A exp
+ C
(1)
Most noticeable is the marked increase in tunnel current at
high negative substrate bias for the three reducible molecules,
itc-TP-FeCl, itc-TPP-FeBr, and itc-TPP-MnBr. Interpreting this
B2
[
]
where Vp is the peak voltage and B is related to the half-width

STM Contrast in Molecules
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10723
program essentially ignores the tails corresponding to high
values of Vp (see Figure 6A). When the combined data set is
averaged, these runs pull the averaged data to a slightly higher
value (Table 1). These data are useful in as much as they show
that a characteristic peak value can be determined to within
about (0.1 V from local measurements, indicating the extent
to which these measurements might be used as quantitative
nanoscale probes.
Discussion
Our results demonstrate the link between tunneling and redox
properties first proposed by Schmickler7 on theoretical grounds
and Mazur and Hipps13 on experimental grounds. The three
reducible compounds show a bias dependence in STM contrast,
while the less electroactive compounds do not. It is not just a
question of whether or not a metal is present, as demonstrated
by the behavior of the itc-TPP-Zn. Neither is the chemical
identity of the counterion of importance, a result that may reflect
a much slower dissociation in the STM experiment (where an
ionic solvent is not used) than in the electrochemical experiment
(where it is). The ring is reduced in the free base and itc-TPP-
Zn, and counterions are not involved in the same way.
The connection between bias dependence and redox properties
is consistent with Tao’s earlier experiment.11 These results
extend the work to more compounds and demonstrate the effect
in a different experiment. In our view, however, the most
important aspects of the present results lie in the degree to which
they allow us to characterize the ability of the simple two-
electrode scanning probe microscope to analyze individual
molecules. We have demonstrated that there are large local
fluctuations in the derivative peak position. In this section we
turn our attention to the theoretical basis of these phenomena.
We note that, in contrast to the earlier work of Snyder and
White,15 our results show that the current-voltage characteristics
for these porphyrin molecules are characteristically asymmetric.
However, Snyder and White worked with mobile adlayers which
adsorbed on both tip and substrate, thereby allowing electron
transfer to occur at either electrode. Here, with the molecule
bound to the substrate in a submonolayer, we expect that the
predominant interaction is with the substrate, the tip acting to
control the local field.
The problem has been treated thoroughly by Schmickler and
Widrig,8 whose notation we shall use. To attempt some
quantitative analysis, it is necessary to fix values for some of
the parameters used in that theory. Perhaps the most vexing is
the local energy of a molecular level, ꢀm + RV, where ꢀm is the
potential of the maximum in the density of unoccupied states
and RV is the energy shift caused by the tip-substrate bias, V.
Here, R is a parameter that describes the fraction of the total
potential drop that occurs across the region between substrate
and molecule. In the Marcus theory, ꢀm for the unoccupied
(oxidized) states is related to the formal potential, E0, by ꢀm )
E0 + λ, where λ is the reorganization energy (typically on the
order of 1 eV in relatively polar solvents).17 The quantities E0
and RV are calibrated under three-electrode potential control
but have to be related to the applied bias here. We use the
convention that when quantities such as V and E0 appear in
expressions for energies, units of electronvolts are implied.
When referring to measured potentials, units of volts are implied.
Electrochemical potential scales and work functions are
related by the work function of the normal hydrogen electrode
(NHE), ΦNHE ) 4.43 V,24 or for the SCE values used here ΦSCE
) 4.67 V since zero on the SCE scale lies 0.24 V positive (i.e.,
further from the vacuum) of zero on the NHE scale. With this
information, a potential on an electrochemical scale may be
Figure 6. (A) Distribution of peak potentials and (B) width parameters
obtained from fits to local scans over itc-TPP-FeCl of the sort shown
in Figure 5B.
TABLE 2: Measured Peak Positions (Vp) and Width
Parameters (B) for the Maxima in dI/dV. The Middle Row
Is an Estimate of Vp Based on Eq 2 Using the Measured
Reduction Potentials (Table 1)
itc-TPP-FeCl
Itc-TPP-MnBr
Itc-TPP-FeBr
Vp (V)
-0.89 ( 0.004 -1.05 ( 0.018 -0.89 ( 0.002
Vp predicted, λ, -0.968/R
δ ) 0
B (V)
-0.892/R
-0.873/R
0.28 ( 0.006
0.26 ( 0.012
0.26 ( 0.002
x
at half-height, ∆V, by ∆V ) B -ln 0.05 ) 0.8333B. The
constant C accounts for the ohmic part of the background tunnel
current. Fitted values for B and Vp are listed in Table 2.
The data discussed above were obtained from many runs
taken over several different experiments, and it is of interest to
know what information can be extracted from spectra obtained
over one point. Such data show considerable run-to-run
variation. However, most runs (of 30 sweeps) over the
electroactive molecules have in common the peak in dI/dV. This
is illustrated in Figure 5B, where we have selected three such
runs which illustrate peaks with low, typical, and high values
for Vp. The signals are noisy, but are well fitted by Gaussian
curves of the form of eq 1. These data suggest a significant
variable contribution to the tunneling process from the local
environment. We show histograms of the values of Vp and
width parameters B obtained from each of the individual runs
in Figure 6. The solid curve is a Gaussian fit to these
distributions, giving Vp ) -0.855 ( 0.1 V (Figure 6A) and B
) 0.18 ( 0.07 (Figure 6B). The width parameters are smaller
than those fitted to the full data set (Table 2), indicating a
significant inhomogeneous broadening of the distribution owing
to environmental fluctuations. The intrinsic width for the itc-
TPP-FeCl is about 0.15 V (0.833 × 0.18 V), while the width
of the averaged data (Table 2) is 0.23 V (0.833 × 0.28 V).
This shows that the averaged data set is broadened considerably
by spatial heterogeniety. One might expect that the average
value of the peak voltage, Vp, would be the same for the two
sets of measurements. This is not the case because our fitting

10724 J. Phys. Chem. B, Vol. 101, No. 50, 1997
Han et al.
the electron to the other electrode, returning the molecule to its
oxidized state.
Schmickler8 has evaluated some representative cases showing
that a Gaussian peak occurs in the quantity dI/dV as observed
in these experiments. The second case has not been analyzed
in the same manner, so we present a discussion below.
A simple derivation of the theory of the first process from
quantum mechanical principles has been presented by Schmick-
ler.8,17 The calculated current density, j, is
2
2
Fs(ꢀ)|Vsm| Ft(ꢀ)|Vmt|
ν
π
p
j ) -e
2Dox(ꢀ) dꢀ
(3)
∫
2
0 Fs(ꢀ)|Vsm| + Ft(ꢀ)|Vmt|
where Fs(ꢀ), Ft(ꢀ) are the densities of states of the tip and
substrate, respectively, Vsm and Vmt are the matrix elements for
tunneling from substrate to molecule and molecule to tip, and
Dox(ꢀ) is the density of oxidized states given by
Figure 7. Relationship between reduction potentials measured on the
SCE scale (right side) and the position of the Fermi energy of the metal
substrate (left side). More negative potentials on the SCE scale are
closer to the vacuum. The lack of potential control at the surface results
in uncertainty ((δ, on the order of 1 eV or less) in the potential at the
molecule. The potential difference between electrons at the Fermi energy
and a point E0 on the SCE scale, ∆V, could take on a range of values
as indicated.
(ꢀ - E0 - λ + RV)2
π
Dox(ꢀ) )
exp -
(4)
[
]
k Tλ
4λkBT
x
B
The derivative of the current with respect to voltage between
tip and substrate shows a maximum at potential corresponding
to ꢀm, the energy level of the oxidized state.8 Schmickler also
shows that the current from this process is probably larger than
that due to direct tunneling without interaction with the oxidized
states.
related to a potential (such as a Fermi energy) that is expressed
with respect to the vacuum. The situation is more complex
here. An electron is removed from the Fermi surface (at an
energy EF) of a metal to a molecule, located not outside the
metal in a vacuum, but rather in a region near the surface where
the electric field is still varying rapidly. We refer to this
interfacial region as the double layer, in accordance with surface
science convention. Thus, the electric potential at the molecule
differs from that in the vacuum by some amount, which can be
positive or negative, depending upon the orientation of local
dipoles. We call this difference between the vacuum and the
potential on the edge of the double layer (δ, and we expect it
to be small in the case of dense metals where most of the work
function arises from electron-electron interactions intrinsic to
the metal.25 (Experimental support for this approximation is
to be found in the emersion experiments of Ko¨tz et al.26) These
relationships are illustrated schematically in Figure 7. Thus,
the potential difference needed to bring an electron from the
Fermi energy to a potential E0 is
There is also the possibility that the tunneling electron stays
on the electroactive molecule long enough for the molecule to
relax into its “reduced” state. This two-step process has been
considered by Kuznetsov et al.29 Clearly, this implies that the
time for electron transfer is much longer than the vibration time
of the molecule, which is on the order 10-14 s. Neglecting
backward flowing current from tip to substrate, the observed
current would be proportional to
RsmRmt
π
(5)
(
)
kBTλ Rsm + Rmt
where
(ꢀ - E0 - λ + RV)2
4λkBT
∞
2
∆V ) RV ) ΦM - ΦSCE - E0 ( δ
(2)
Rsm
Rmt
)
)
-∞Fs(ꢀ)|Vsm| fs(ꢀ) exp -
(6)
(7)
∫
[
]
where ΦM is the work function of the metal (5.3 V for the Au-
(111) surface27) and δ is expected to be small (less than 1 eV).
Evaluation of the local potential shift is complicated, involving
polarization of the surrounding medium and transitions within
the molecule which may not be strictly proportional to the
applied voltage. Nonetheless, we expect 0 < R < 1. We note
in passing that Snyder and White have proposed that the
electrode may float to an equilibrium potential controlled by
the molecular adsorbate.15 This cannot be the case here where
the electrodes are connected by low-impedance circuitry.
The electron-transfer process may be described in one of two
ways which involve the states of the molecule explicitly. In
one model due to Schmickler8,28 electrons tunnel from tip to
substrate (or vice versa) by a resonant transition through the
unoccupied (oxidized) states. The molecule itself does not
acquire a charge over a time scale comparable to or longer than
the phonon relaxation time in the process. In another process,
discussed by Kuznetsov,29 the electron transfers to the molecule
where it is trapped by relaxation (i.e., it reduces the molecule).
A second thermal fluctuation then allows the molecule to transfer
∞
2
-∞Fs(ꢀ + eV)|Vmt| (1 - ft(ꢀ + eV) ×
∫
(ꢀ - E0 - λ + RV)2
4λkBT
exp -
[
]
and fs(ꢀ), ft(ꢀ) are the Fermi functions for the substrate and tip,
respectively. The functions are, of course, identical, but are
labeled here in order to clarify the direction of tunneling.
We have evaluated dI/dV for the two models (model I, eqs 3
and 4; model II, eqs 5, 6, and 7) and the following parameter
values: E0 ) 0.6 V, λ ) 0.5 V, and R ) 0.5. Here, E0 is taken
to be a negative voltage referenced to the metal Fermi energy
(taken as zero). We also used the broad-band approximation,
taking the matrix elements and the densities of states of the
metals to be constant over the energy range of interest. The
results of the calculations are shown in Figure 8 (crosses for
model I, dots for model II). As expected, model I shows a
maximum in dI/dV at (E0 + λ)/R, and the points are well fitted
by a Gaussian of the form of eq 1 (solid line).

STM Contrast in Molecules
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10725
nate among electroactive and non-electroactive molecules could
have applications in the design of molecular-scale electronic
devices. The nature of the electron transfer through the
molecules is not clear. Both one-step (resonant tunneling) and
two-step (reduction/oxidation) processes are consistent with our
observations, and further experiments, such as measurement of
the temperature dependence of the currents, are required to
elucidate this point. One might expect the resonant tunneling
contribution to be larger (because it is a first-order process),
but electron interactions and thermal fluctuations may favor the
two-step process. We note that the relative contrast of elec-
troactive molecules far from the reduction potential appears to
be less than that of similar but non-electroactive molecules.
Furthermore, the molecules appear to behave as though all of
the applied potential is dropped across them. Further work is
required to understand these observations.
Figure 8. Calculated dI/dV for the one-step tunneling process (eqs 3
and 4) shown by the crosses and the two-step process (eqs 5, 6, and 7)
shown by the dots. Parameters were E0 ) -0.6 V, λ ) 0.5 V, and R
) 0.5. The solid lines are fits to Gaussians of the form of eq 1.
Model II is more complicated, containing the product of two
molecular densities of states separated by 2λ. The corresponding
maximum is not centered between these peaks but is pulled to
somewhat higher voltages by the rapid increase in current at
high bias. For these parameters, dI/dV is a maximum at about
0.2 V lower bias than for model I. The form of dI/dV is clearly
not Gaussian, but the data are still fitted quite well by a simple
Gaussian (solid line). Thus, within experimental error, we
cannot distinguish these two processes on the basis of our data.
We can use eq 2 and the measured reduction potentials (Table
1) to estimate to position of the peaks in dI/dV as referenced to
the gold Fermi energy. For simplicity, we have used model I
and taken λ to be zero. The results are expressed in terms of
R in Table 2.
Acknowledgment. This work was supported in part by
grants from the National Science Foundation (BIR 9513233),
Molecular Imaging Corporation (TCL96-157C), and AFSOR
(F49620-96-1-0346) (N.T.). We are grateful to Steve Woodward
for assistance in the lab.
References and Notes
(1) Duke, C. B. Can. J. Chem. 1985, 63, 236-241.
(2) Eigler, D. M.; Weiss, P. S.; Schweizer, E. K.; Lang, N. D. Phys.
ReV. Letts. 1991, 66, 1189-1192.
(3) Sautet, P.; Joachim, C. Ultramicroscopy 1992, 42-44, 115-121.
(4) Joachim, C.; Sautet, P. Electron tunneling through a molecule. In
Scanning Tunneling Microscopy and Related Methods; Behm, R. J., Ed.;
Kluwer: Netherlands, 1989; pp 377-389.
(5) Joachim, C.; Gimzewski, J. K. Europhys. Lett. 1995, 30, 409-
414.
The agreement with the measured values is good if R is taken
to be unity and λ very small. This result is unexpected, but
significant, since departure from the assumptions used here
would push the peak to higher voltages, not lower. We note
that Mazur and Hipps13 analyzed their data as though the full
potential difference were applied across the molecule (R ) 1).
This phenomenon requires further investigation. It should be
pointed out, however, that a value of R close to unity is required
for the observation of asymmetric tunnel-current characteristics
(0.5 would give symmetric curves and zero would give no
effect).
A small value for λ is also consistent with the results of Tao.11
In general, electron transfer according to model II (the two-
step process) would be consistent with a small effective value
for λ, but our data do not discriminate between the models.
Finally, we note that we have not considered the effects of
the matrix elements between the electrodes and the molecules.
These can make large differences, particularly when internal
molecular structure is resolved. This has been demonstrated
beautifully in a series of experiments by Lu et al.,30,31 who show
that, for metallophthalocyanines with states close to the Fermi
energy, the metal is visible only if dz states interact with the
gold substrate.
(6) Fisher, A. J.; Blo¨chl, P. E. Phys. ReV. Lett. 1993, 70, 3263-3260.
(7) Schmickler, W. J. Electroanal. Chem. 1990, 296, 283-289.
(8) Schmickler, W.; Widrig, C. J. Electroanal. Chem. 1992, 336, 213-
221.
(9) Bennett, A. J. J. Electroanal. Chem. 1975, 60, 125-131.
(10) Morisaki, H.; Ono, H.; Yazawa, K. J. Electrochem. Soc. 1988, 135,
381-383.
(11) Tao, N. Phys. ReV. Lett. 1996, 76, 4066-4069.
(12) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Funda-
mentals and Applications; John Wiley: New York, 1980.
(13) Mazur, U.; Hipps, K. W. J. Phys. Chem. 1995, 99, 6684-6688.
(14) Burghard, M.; Fischer, C. M.; Roth, S.; Schlick, U.; Hanack, M.
Synth. Met. 1996, 76, 241-244.
(15) Snyder, S. R.; White, H. S. J. Electronanal. Chem. 1995, 394, 177-
185.
(16) Bumm, L. A.; Arnold, J. J.; Cygan, M. T.; Dunbar, T. D.; Burgin,
T. P.; Jones, L.; Allara, D. L.; Tour, J. M.; Weiss, P. S. Science 1996, 271,
1705-1707.
(17) Schmickler, W. Interfacial Electrochemistry; Oxford University
Press: Oxford, 1996.
(18) Marcus, R. A. J. Phys. Chem. 1965, 43, 679-701.
(19) Han, W.; Li, S.; Lindsay, S. M.; Gust, D.; Moore, T. A.; Moore,
A. L. Langmuir 1996, 12, 5742 - 5744.
(20) Gust, D.; Moore, T. A.; Moore, A. L.; Liddell, P. A. Meth. Enzymol.
1992, 213, 87-100.
(21) Tokel, N. E.; Keszthelyi, C. P.; Bard, A. J. J. Am. Chem. Soc. 1972,
94, 4872-4877.
(22) Seely, G. R.; Gust, D.; Moore, T. A.; Moore, A. L. J. Phys. Chem.
1994, 98, 10659-10664.
(23) DeRose, J. A.; Thundat, T.; Nagahara, L. A.; Lindsay, S. M. Surf.
Sci. 1991, 256 , 102-108.
Conclusions
(24) Reiss, H.; Heller, A. J. Phys. Chem. 1985, 4207-4213.
(25) Lang, N. D.; Kohn, W. Phys. ReV. 1971, B3, 1215-1223.
(26) Ko¨tz, E. R.; Neff, H.; Mu¨ller, K. J. Electroanal. Chem. 1986, 215,
331-344.
We have shown that the bias dependence of the current
through individual molecules can be used to discriminate
between electroactive and non-electroactive molecules imaged
in a scanning probe microscope. Local fluctuations and lack
of a strict correlation with macroscopic electrochemistry
preclude the use of the microscope as a precise analytical probe.
Nonetheless, certain rather characteristic signals are observed
which could aid identification of molecules among a limited
range of candidate species. In addition, the ability to discrimi-
(27) Weast, R. C.; Astle, M. J. CRC Handbook of Physics and Chemistry;
CRC Press: Boca Raton, 1979.
(28) Schmickler, W. Surf. Sci. 1993, 295, 43-56.
(29) Kuznetsov, A.; Somner-Larsen, P.; Ulstrup, J. Surf. Sci. 1992, 275,
52-64.
(30) Lu, X.; Hipps, K. W.; Wang, D.; Mazur, U. J. Am. Chem. Soc.
1996, 118, 7197-7202.
(31) Lu, X.; Hipps, K. W. J. Phys. Chem. in press.