Reproducible Measurement of Single-Molecule Conductivity

See allHide authors and affiliations

Science  19 Oct 2001:
Vol. 294, Issue 5542, pp. 571-574
DOI: 10.1126/science.1064354


A reliable method has been developed for making through-bond electrical contacts to molecules. Current-voltage curves are quantized as integer multiples of one fundamental curve, an observation used to identify single-molecule contacts. The resistance of a single octanedithiol molecule was 900 ± 50 megohms, based on measurements on more than 1000 single molecules. In contrast, nonbonded contacts to octanethiol monolayers were at least four orders of magnitude more resistive, less reproducible, and had a different voltage dependence, demonstrating that the measurement of intrinsic molecular properties requires chemically bonded contacts.

Wiring a single molecule into an electrical circuit by chemically bonding each end to a metal conductor is a key requirement for molecule-based electronics. Although conceptually simple, this goal has proven elusive. The variety of methods for contacting molecules includes bonding dithiolated molecules into break junctions (1), dipping nanotubes into a mercury pool (2), touching molecules in an insulating matrix with a conducting atomic force microscope (AFM) (3), using a scanning tunneling microscope (STM) to connect to gold particles attached to dithiolated monolayers (4, 5), and contacting two monolayers together with a mercury-drop electrode (6). These pioneering experiments have demonstrated that unambiguous contact to a single molecule is difficult to achieve, as shown by large disparities in conductivities reported for identical (7, 8) or similar (1, 9) molecules. Measured currents can be very sensitive to applied stress (3, 10, 11), and calculated conductivity can disagree with experimental results by several orders of magnitude (12). In many cases, electrical connections to the molecules have been made via nonbonded mechanical contacts rather than chemical bonds, and it is likely that this may account for some of the discrepancies.

Here we report a reliable method for chemically bonding metal contacts to either end of an isolated molecule and measuring the current-voltageI(V) characteristics of the resulting circuit. Molecules of 1,8-octanedithiol were inserted into an octanethiol monolayer [on Au(111)] using a replacement reaction (13) whereby one of the two thiol groups becomes chemically bound to the gold substrate (14–17). The octanethiol monolayer acts as a molecular insulator, isolating the dithiol molecules from one another. The thiol groups at the top of the film were derivatized by incubating the monolayer with a suspension of gold nanoparticles (17). A gold-coated conducting AFM probe was used to locate and contact individual particles bonded to the monolayer (Fig. 1A).

Figure 1

(A) Schematic representation of the experiment. The sulfur atoms (red dots) of octanethiols bind to a sheet of gold atoms (yellow dots), and the octyl chains (black dots) form a monolayer. The second sulfur atom of a 1,8-octanedithiol molecule inserted into the monolayer binds to a gold nanoparticle, which in turn is contacted by the gold tip of the conducting AFM. (B) I(V) curves measured with the apparatus diagrammed in (A). The five curves shown are representative of distinct families,NI(V), that are integer multiples of a fundamental curve, I(V) (N = 1, 2, 3, 4, and 5). (C) Curves from (B) divided by 1, 2, 3, 4, and 5. (D) Histogram of values of a divisor, X(a continuous parameter), chosen to minimize the variance between any one curve and the fundamental curve,I(V). It is sharply peaked at integer values 1.00 ± 0.07 (1256 curves), 2.00 ± 0.14 (932 curves), 3.00 ± 0.10 (1002 curves), 4.00 ± 0.10 (396 curves) and 5.00 ± 0.13 (993 curves). (Spreads are ±1 SD.) Of 4579 randomly chosen curves, over 25% correspond to the X = 1 (single-molecule) peak. No obvious correlation was noted between particle size and number of molecules contacted. Conducting atomic force microscopy data were acquired with a PicoSPM microscope (Molecular Imaging) using silicon cantilevers (spring constant, 0.35 N/m) sputter-coated with 5 nm of chromium followed by 50 nm of gold. Imaging was done under toluene in a nitrogen atmosphere.

I(V) measurements made on over 4000 nanoparticles produced only five distinct families of curves. Representative curves from each family are shown in Fig. 1B. The curves correspond to multiples of a fundamental curve, lying on this fundamental curve when divided by the appropriate integer (Fig. 1C). To test for this property in all of the measured curves, we found values of a continuous divisor, X, that minimized the variance between the fundamental data set and all others. A histogram of 4600 values of X (Fig. 1D) shows that it is sharply peaked at the integer values 1, 2, 3, 4, and 5 (with a negligible number of higher values not shown). The fundamental set, containing more than 1000 curves, is ascribed to assemblies in which a single dithiol molecule links the gold nanoparticle to the underlying gold substrate.

The I(V) characteristics of the dithiol molecules bound to the gold nanoparticles differ dramatically from those of alkanethiols that are contacted through nonbonded interactions with the AFM tip. Figure 2A shows a measured I(V) curve for a bonded contact to a single dithiol molecule (current is on a log scale). AnI(V) curve measured with a nonbonded contact to the octanethiol monolayer is also shown. The current through the bonded contact is not only much larger but also has a very different voltage dependence. For nonbonded contacts, the experimentally observed I(V) characteristics are dominated by the contact rather than by intrinsic molecular properties. In the ohmic region (between ±0.1 V) the single molecule has a resistance of 900 (±50) megohm. In the same voltage range, measurement of the resistance of the nonbonded contact is limited by noise, but the resistance is at least 104 gigohm (at a contact stress of 1 GPa).

Figure 2

(A) Current (on a log scale) as a function of voltage as calculated from first principles with no adjustable parameters (dashed line), as measured for a bonded single molecule (solid line), and for a nonbonded contact (contact force = 6 nN) to the surrounding octanethiol matrix (broken line) (noise-dominated data at low bias are suppressed). (B) Current and force measured as a conducting AFM cantilever (biased at +1 V) is moved toward the sample surface. Nominal contact is achieved at 0 distance, and negative values imply continued motion toward the Au(111) interface. When the probe approaches a gold nanoparticle, the current (dashed line) through the nanoparticle jumps to its final value on contact. When the surrounding octanethiol matrix is contacted, current (dots) is much smaller and depends on force (heavy solid line).

A theoretical single-molecule I(V) curve was calculated with scattering theory (18–20) carried out self-consistently so as to avoid the need for fitting parameters such as the energy difference between the metallic Fermi energy and a molecular orbital. The results of this calculation are shown on a log scale in Fig. 2A. The shapes of the calculated and measured curves are in agreement, and the absolute values of current agree to within a factor of 6. This represents a remarkable improvement relative to previous comparisons of calculated and measured molecular conductivity [for example, a factor of 500 (12)]. Simulations were also carried out for a nonbonded monothiol with the terminal methyl group positioned near a gold surface. The results were similar to the dithiol simulation, rather than the measured curve (Fig. 2A), but depend strongly on the exact placement of the methyl group with respect to the gold. This placement is difficult to determine theoretically in the absence of bond formation.

A key factor permitting identification of the fundamental curves inFig. 1 is the lack of dependence on contact force for these chemically bonded contacts, as shown in Fig. 2B (dashed line). In contrast, nonbonded contacts made by moving the tip onto the alkanethiol monolayer (dotted line) show the strong force dependence previously reported (3, 10, 11). The stress on the monolayer is probably somewhat higher when contact is made through a gold nanoparticle rather than by direct contact with the AFM, because the particles are smaller than the end radius of the AFM probe (measured to be about 10 nm by scanning electron microscopy). Thus, the monolayer must undergo substantial deformation in both cases. The lack of force dependence of the chemically bonded contact implies that (i) interatomic distances within a molecule do not change much as the film is stressed, which is consistent with simulations of deformation in alkanethiol monolayers (21); and (ii) the bonds between the molecule and the metal do not change substantially either.

The physical structures of the monolayers investigated here were characterized by means of scanning tunneling microscopy. An image of the alkanethiol monolayer with inserted dithiol molecules (bright spots) is shown in Fig. 3A. It has been suggested that desorption of alkanethiols from a gold surface might occur via a disulfide (22). Although this process could lead to the insertion of a pair of octanedithiols, some of these molecules could in turn desorb in pairwise fashion with an adjacent octanethiol, still leading to a (perhaps small) single-molecule “fundamental” peak shown in Fig. 1C. This possibility, coupled with the narrow peak widths and the population distribution (see legend toFig. 3), render remote the possibility that the fundamental curve inFig. 1 represents two molecules. After incubation with the suspension of gold nanoparticles, isolated conducting particles of gold attached to the monolayer were observed (Fig. 3B). Control experiments showed that a pure 1,8-octanedithiol monolayer (Fig. 3C) became covered by large gold aggregates when treated with the nanoparticle suspension (Fig. 3D), whereas the pure octanethiol monolayer was unaffected (Fig. 3E). Thus, the gold particles only attached to the inserted octanedithiol molecules in the mixed monolayer.

Figure 3

STM images of (A) 1,8-octanedithiol molecules inserted into an octanethiol monolayer. Arrows point to the protruding thiols (shown at molecular resolution in the inset). The bright spots increase in density, but not in apparent size, with increasing dithiol exposure. (B) A mixed monolayer similar to that shown in (A) after incubation with gold nanoparticles that attached to the protruding thiols (the apparent diameter of the particles is increased owing to the finite size of the STM tip). The density of attached gold particles is of the same order of magnitude as the density of protruding thiols measured in (A). Controls are shown in (C) through (E). (C) Pure 1,8-octanedithiol monolayer. (D) The same monolayer after incubation with the gold nanoparticle suspension and rinsing, showing that the whole surface is covered with large gold aggregates. (E) A pure octanethiol monolayer subjected to the same treatment, showing that no gold particles are bound. All STM images were obtained with electrochemically etched PtIr tips using a PicoSPM with the sample under toluene and operating in a dry nitrogen environment.

Our approach to measuring molecular conductivity avoids the effects of variations in contact force and other problems encountered with nonbonded contacts. Nonbonded molecular contacts are found to be highly resistive and to have I(V) characteristics that are quite different from those of molecules with bonded contacts, showing that the nonbonded contact dominates electrical properties. In contrast, the bonded contacts are highly reproducible and lead to measurements that are in much better agreement with first-principles simulations without adjustable parameters. This approach is straightforward and easily applied to other types of molecules, opening a new avenue for exploring molecular electronics.


Stay Connected to Science

Navigate This Article