Featured Story Corner

Walking Molecules

—Ludwig Bartels, University of California at Riverside

Put your weight on one foot and you can move the other one a step ahead; put the weight on that foot and you can move the first one a step ahead: Voilà, you are moving in a straight line!

These are not dance lessons but density-functional theory simulations that explain how a molecule, which we recently made, can move in a straight line on a flat surface.

There were no molecules known previously that move in a straight fashion on a highly symmetric substrate. Thus, our experimental findings were noteworthy in any case. This, however, is the story of an experimentalist whose measurements achieved widespread recognition (e.g. being part of the AIP list of the Top Physics Stories for 2005)- thanks to theoretical simulations utilizing resources of SDSC.

Our Vision

The miniaturization of conventional (macroscopic) mechanical systems (such as a truck or a syringe pump) might not find its ultimate limit before a realm is reached, in which each component (or assembly of components) is a single molecule. Nanoscale machinery will be rich both in concepts transferred down from the macroscopic world and concepts unique to the quantized nature of matter at this scale. Solid surfaces provide not only support to such machinery, but also offer a way of interfacing them with the macroscopic world (thanks to the advanced methods for surface nanopatterning developed in the context of silicon-based microelectronics).

Preliminary Research

The realization of this vision, both from the point of view of feasibility and potential application, is still shrouded in unknowns, yet the field is bristling with activity. In order to chart the possibilities this approach offers, we investigate how the interaction between organic molecules and metal substrates can determine and guide their surface behavior. We are particularly interested in thiol-based aromatic molecules, which are the main component of many of today's schemes for molecular electronics. In this context, the molecules are typically bound by one thiol group to the substrate and the aromatic group provides them with the ability to conduct charge. As a substrate, we typically use the thermodynamic equilibrium surface of copper (Cu(111)), which as-deposited copper films adopt predominantly at their surface.

Earlier we could show that attachment of a molecule by a single thiol group to the substrate allows rotation around the thiol linker even at fairly low temperatures (e.g. 60K); yet it is well known, that the thiol linkage to the substrate will not break before molecular decomposition around 400K. In collaboration with Talat S. Rahman and Sergey Stolbov of Kansas State University, who performed density functional (DFT) calculations, we could elucidate the adsorption configuration and rotational/translational behavior of such systems.

Based on this knowledge, UCR graduate student Ki-Young Kwon wondered what would happen if a rigid molecule was forced to lie flat on a surface by means of two substrate linkers attached to opposite ends. In this case, rotation around each of the sulfur linkers appears mutual exclusive, because it would require motion of the other linker. To check this hypothesis, he synthesized the molecule 9,10-diacetylthioanthracene, which can be decomposed on Cu(111) to yield 9,10 dithioanthracene (DTA). The figure shows various ways this molecule (thiol linkers drawn purple) might adsorb on a Cu(111) surface.

The Equilibrium Adsorption Site

In order to predict the dynamic behavior of the molecule, we used a variety of DFT-based methods that can both treat the substrate copper atoms and the molecule in a meaningful way. In this effort we were guided and advised by the group of Rahman, who made sure that our choices of both the computational details as well as the setup of the system were consistent. Correct description of the metallic character of the substrate and its interaction with the sulfur atoms requires the use of cyclic boundary conditions. Thus, we had to set up a repetitive supercell, which contains a certain number of metal atoms, the entire molecule, and sufficient vacuum space above it, so that when it is repeated in all three spatial directions, the interactions between the molecule in one cell and all species of neighboring cells is sufficiently low that they will not impact the results substantially.

Initially, we explored the minima of the adsorption configuration of DTA in cells that contained between 3 rows of 5 copper atoms in 2 layers and 4 rows of 6 copper atoms in 4 layers. For this purpose, we set up an initial guess how the system at minimum energy might look using the spacing of copper atoms in the bulk and the molecular geometry as expected in the gas phase. Then the used software packages varied the positions of the atoms slightly until a local minimum of the total system energy is found. In all cases we found two minimum adsorption configurations - one as shown in panel A of the figure and the other as shown in panel B. The first one is characterized by one sulfur atom of DTA being adsorbed close to a substrate hollow site, which resembles the adsorption geometry earlier found for molecules with one linker. The other thiol linker is in this case placed near an on-top site, which is energetically not favorable for it. This is balanced by optimal interaction of the anthracene body of the molecule with the substrate, which requires parallel alignment with a substrate atomic row as literature experiments on anthracene and other acenes without linkers showed. For none of the chosen supercells, this was the global minimum; however, comparison with STM measurements clearly indicated that the molecular alignment in this configuration matched the experimental data. The second adsorption minimum is characterized by the two sulfur linkers residing near the optimal hollow sites, while the acene system is misaligned with regard to the substrate, which renders its interaction less favorable. Although this adsorption minimum appeared to be favorable in the DFT simulations, in experiments it required temperatures as low as 10K to freeze a molecule out in this state (which, nevertheless, confirms that this configuration is indeed a local minimum). These results show that (the level of) DFT (we used) overestimates the more local interactions of the sulfur linkers with the substrate in comparison to the delocalized interaction of the aromatic ring system. Similar observations can be found in the literature and are commonly ascribed to the neglect of electron correlation effects in such calculations.

On the whole, the level of correlation between the experimental and the simulated data provides a gauge of the fidelity of DFT-based calculations (as performed by non-experts like us): general features of adsorbate-substrate interactions can be reproduced, yet the resultant numeric values can be somewhat off both absolutely and in relation to similar calculations.

Modeling the Diffusion Potential

Symmetry demands that if panels A and B represent minimum adsorption configurations, then also panels D and E do so. Thus, it appears that transfer of the molecule between two neighboring sites in the direction of its acene ring system visits one closely adjacent adsorption minimum after the other. In contrast, movement in any other direction does not have any favorable intermediate step. This led us to calculate the diffusion potential surface for DTA. These calculations involved successive movement of one or both of the sulfur atoms by 0.1A at a time, fixation of their x-coordinates with respect to a third layer substrate atom, and minimization of the system with regards to all atomic coordinates including the y and z coordinates of the sulfurs. Minimization after successive small steps does not only represent the real diffusive behavior in a more realistic fashion than shifts of increasing length from the initial starting position, it is also far more economical with regards to computational time: while the latter takes easily 7 hours on 8 processors (56 processor hours) using a cell of 5x3x3 substrate atoms, the former is typically achieved in 2 hours on 8 processors. For the mapping of the entire diffusion potential, we needed to explore >250 configurations and we also performed tests with larger and thicker supercell sizes to confirm our results. To simplify this task, we wrote a Mathematica program that generates new directories for each minimization that contain both the coordinate files, as well as the LoadLever instructions.

The resultant diffusion potential shows that indeed linear diffusion in the direction of the equilibrium ring axis is significantly (> x2) more facile than motion in any other direction or rotation around either of the substrate linkers. We performed STM experiments in which we measured the diffusion speed of DTA at different temperatures. Using the Arrhenius equation, such data allows estimation of the diffusion barrier, which turned out to coincide with the simulated one within 30 meV. The molecular orientation in this transition state is shown in panel C. Because we had calculated the setup of the system at various conformations, we could string together the conformations along the minimum energy path. The movie below reveals how first one sulfur atom and then the other one moves, and how each time a sulfur atom leaves a hollow site in order to perform a step, it also needs to move up from the surface. In combination, this creates the impression of bipedal locomotion/human walking as suggested by the cartoon in the bottom right, which uses the actual sulfur coordinates for the feet.

Ludwig Bartels (Ludwig.Bartels@UCR.EDU) is associated professor of chemistry at the University of California, Riverside. He is very grateful to SDSC for providing computational resources to this research project through Academic Associates Program and space for this article. These results were first published in Physical Review Letters. This project was sponsored by the US Department of Energy.