11. STM images of biological molecules and electrical measurements on single molecules
This chapter in the first edition of this book (Lindsay 1993) was devoted almost entirely to scanning tunneling microscopy. Whatever happened to the scanning tunneling microscope? Some sensational images of DNA were published (Arscott, Lee et al. 1989; Beebe, Wilson et al. 1989; Driscoll, Youngquist et al. 1990) but these were obtained on graphite substrates. It was subsequently shown that electronic artifacts on the surface of bare graphite could mimic DNA (Clemmer and Beebe 1991) even down to the level of atomic detail (Heckl and Binnig 1992). This did a great deal to discredit the technique. We had taken a somewhat different approach, using electrochemical methods to deposit DNA onto a gold electrode where it was imaged in situ. This approach gave images that clearly showed the helical repeat of the deposited DNA (Lindsay, Thundat et al. 1988; Lindsay, Thundat et al. 1989) but it was a clumsy technique, difficult to reproduce and apparently relying on co-deposition of oxidized salts. Despite success in a blind trial (Jeffrey, Jing et al. 1993) it was clear that the DNA had to be strongly attached to an electrode surface in order to be imaged (Rekesh, Lyubchenko et al. 1996). To make matters worse (see below) image interpretation was difficult. Very recently, the Kawai group at Osaka have used a pulse injection method to deposit DNA onto a clean copper surface in ultrahigh vacuum, producing beautiful images of plasmid DNA. These show the helix repeat quite clearly (Tanaka 1998). No doubt this will trigger further study, if only because the electronic sensitivity of the STM opens up the possibility of chemical identification, and this section is directed to that possibility.
Figure 25: (a) Molecular LUMOs are coupled to phonons (generalized displacement Q) giving rise to a parabolic variation in electronic energy with a corresponding Gaussian probability, P(E), of thermal activation of a given state. (b) transport through an electron accepting state proceeds by resonant tunneling into a thermally accessible part of the LUMO distribution followed by relaxation after charging by an amount 2l. The charge remains trapped in the acceptor state until another thermal fluctuation permits resonant tunneling out to the Fermi level of the second electrode.
Figure 26: (a) Schematic arrangement of a conducting AFM measurement of conduction in a single molecule. Here a carotene molecule is shown embedded in a 22 atom alkane thiol layer. (b) Current-voltage data for single carotene molecules. The behavior is ohmic and the corresponding resistance is 4.2GW, over a million times more conductive than the surrounding alkane matrix. (Data from Leatherman et al. (Leatherman, Durantini et al. 1998) with permission.)
Initial reports of STM imaging of large molecules like DNA met with skepticism from the physics community because, from a vacuum tunneling point of view, it was simply not possible; the molecule being too big (Lindsay 1993). However, Eigler et al. (Eigler, Weiss et al. 1991) showed that the contrast of an excellent atomic insulator (xenon) on a metal surface could be explained entirely by the (small) degree of hybridization between Xe atomic states and the metal surface when the xenon physisorbed. A small amount of itinerant charge from the metal is propagated out into space at the Fermi energy (which is well below the energy of the atomic state in this case). This that accounts for the contrast of the Xe atom in STM images. Such calculations are difficult to do for more complex molecules. It is clear from images of molecules that have obvious ‘feet’ attached to them that the STM ‘sees’ the contact points with the surface because this is where the hybridization occurs (Jung, Schlittler et al. 1996). Thus, in the case of DNA, one expects that the high spots in the imageare actually the low points on the molecule that lie in contact with the metal. This is something of a caricature of a complex process, but it illustrates the points that (a) identification of molecular species is difficult and (b) the image need not be related to molecular geometry in any simple manner.
Happily, there is more to the process, and in the case of molecules with electron
acceptor or donor states (redox active molecules) there is some hope for chemical
identification. Certain DNA bases are, in fact electrochemically active, albeit at
potentials which preclude simple experiments (Hinnen, Rousseau et al. 1981). However,
studies of some model compounds find a correlation between electrochemistry and STM image
contrast, a connection first pointed out by Schmickler (Schmickler 1990) and Mazur and
Hipps (Mazur and Hipps 1995). Using a series of modified porphyrins, Han et al. (Han,
Durantini et al. 1997) demonstrated that the easily reduced molecules showed strongly
enhaced contrast on a negative electrode while the non-reducible molecules did not, and
this permitted identification of the molecular species in a heterogeneous sample. Tao
carried out a particularly elegant experiment using a monolayer film of protoporphyrin on
graphite (Tao 1996). This material forms a well-organized monolayer (See Fig. 24). Further
more, molecules are available with and without a central iron atom and the two species are
entirely miscible forming a chemically heterogeneous but otherwise uniform monolayer. Tao
imaged this monolayer in an electrochemistry cell under potential control, tuning the
surface potential through the redox potential for the Fe++
Fe+++ process. He found that the image of the iron containing molecules was
brightest at the point where the redox current was maximum (shown as the peak in the
thin-film voltammograms on the right of Fig. 24).
There are currently two models of this process, illustrated schematically in Fig. 25. Schmickler has proposed that the applied electric field pulls down the LUMO energy until it coincides with the Fermi energy of the tip/substrate combination and resonant tunneling occurs. Because the state is coupled strongly to solvent and environmental fluctuations (represented schematically by a normal coordinate Q in Fig. 25a) the electronic energy has a parabolic distribution (solid line) leading to a Gaussian distribution of thermally-occupied states (P(E), the dashed line in Fig. 25a). Schmickler has proposed that STM is a way to determine this density of unoccupied states of electroactive molecules (Schmickler 1990; Schmickler and Widrig 1992). Quite another mechanism is proposed by others (Kuznetsov, Somner-Larsen et al. 1992; Sumi 1998) and it is illustrated schematically in Fig. 25b. In this view, the electron is coupled so weakly to the tip and substrate that there is time for significant charge to build up on the LUMO, and, more importantly, for environmental relaxation to reduce the energy of the charged state, resulting in trapping of the charge. This is nothing other than the process of electrochemical reduction and the trapped state is lowered in energy by a factor 2l where l is the relaxation parameter of the Marcus theory (Han, Durantini et al. 1997; Sumi 1998). The charge will reside in this state until a subsequent thermal fluctuation brings the relaxed state (EL* on Fig. 25b) up to the Fermi level of the second electrode at which point it can tunnel out. This process results in a maximum net current half way between these states, and it defines the redox potential for the process (it is a free energy so the potential is properly defined only for a standard concentration). Since l is on the order of an eV, it would seem to be trivial to distinguish the two processes, but, as of the time of writing, there is no unambiguous experimental result.
The STM suffers from the drawback that only small molecules may be used. The nature of the contact is also uncontrolled in as much as the tip will be pushed into the molecule by whatever amount is needed to achieve the set-point tunnel current. Leatherman et al. (Leatherman, Durantini et al. 1998) have used an AFM with a platinum coated probe to carry out simultaneous imaging and conductivity measurements. They studied a rather large synthetic carotene molecule embedded in a 22 carbon atom alkane-thiol film. The experimental arrangement is illustrated in Fig. 26a. In order to avoid electrochemical processes which damage the film (giving unrepeatable data), the experiment was carried out under oxygen-free and water-free toluene in an environmental chamber. Images of many molecules were analyzed to produce the current voltage data for a single molecule shown in Fig. 26b. The various curves are models based on the charge transfer theories described earlier, and it can be seen that the data is not yet good enough to resolve the theoretical debate. In fact, the curves are quite well described by a straight line with a resistance of 4.2GW. It is not clear at this stage how much of this is intrinsic and how much is related to the contact resistance between the probe and molecule. No current could be detected through the alkane film in any conditions, and its resistance was estimated to be on the order of 1016W. However, it has recently been demonstrated that much of the early data on the resistance of metallic carbon nanotubes was dominated by contact resistance by the elegant expedient of dipping one end of a nanotube into a mercury pool (Frank, Poncharal et al. 1998). Contact resistance may also be important in these studies of molecules.
