Detecting elusive surface atoms with atomic force microscopy

In his famous 1959 speech titled “There is plenty of room at the bottom” (1), Nobel prize winning physicist Richard P. Feynman predicts that matter will be controlled and manipulated at the atomic scale in the foreseeable future. Furthermore, he proclaims that “it should be possible to see... individual atoms.” In many ways, Feynman's vision was realized with the invention of the scanning tunneling microscope (STM) by Gerd Binnig and Heinrich Rohrer in 1982, for which they were awarded the Nobel prize in physics (2). With the STM, quantum mechanical tunneling of electrons from a sharp metal tip to a conductive substrate is used to detect individual surface atoms with atomic resolution. The ability to “see” atoms in real space immediately solved many controversies in surface science including the complicated 7 × 7 reconstruction of the Si(111) surface (3).

One complication with the STM is that the tunneling current is a function of not only the surface topography but also the local electronic structure. On the graphite surface, there are two different types of carbon atoms in the basal plane, as distinguished by the presence (α) or absence (β) of a carbon atom in the plane immediately below the surface. Coupling of 2p_z orbitals from α carbon atoms of adjacent layers lowers their bonding energies. Thus, the 2p_z orbitals from β carbon atoms form the highest occupied (and lowest unoccupied) orbitals where the Fermi level crosses. These local electronic structure variations imply that the STM can only detect every other atom on the graphite surface. Consequently, an alternative imaging mechanism is required to detect the “hidden” α atoms on the graphite surface. The article by Hembacher et al. (4) in this issue of PNAS outlines a novel scanning probe microscope that solves the hidden atom problem. Their approach not only detects the hidden α atoms but also minimizes normal forces that would otherwise perturb the true structure of the graphite surface. This important advance may be the enabling technology that allows other soft materials to be imaged at the atomic scale.

Instead of using the STM, Hembacher et al. (4) used a close cousin of the STM called the atomic force microscope (AFM). The AFM was developed by Binnig et al. in 1986 (5) and uses atomic forces (rather than tunneling) between a sharp tip and a surface to detect atomic level variations in surface topography. Because electron current flow is not required, the AFM can be applied equally well to insulating and conductive materials. When the AFM is operated with the tip in contact with the surface at moderate loads (>10 nN), the contact area typically includes hundreds if not thousands of atoms, thus limiting the spatial resolution to ≈10 nm rather than atomic-length scales. To improve spatial resolution, the AFM tip is pulled out of contact with the surface and oscillated with an amplitude of ≈1 nm. In this case, the oscillation resonance frequency is sensitive to spatial variations in the force between the tip and sample. By using an exceptionally sharp tip and a low detection bandwidth (i.e., low data acquisition rates), atomic resolution can be achieved.

In previous noncontact mode AFM studies of graphite (6), attractive forces between the tip and sample were used for atomic resolution imaging, but the hidden α atoms were not revealed. This result is not surprising, because the physical origins of the tunneling current in STM and the attractive force in AFM are directly related (7). However, by decreasing the oscillation amplitude of the AFM tip and operating at low temperatures approaching 4 K, the more elusive repulsive force between the tip and sample was detected by Hembacher et al. (4). Because the repulsive force involves different electrons in the tungsten tip than tunneling, the hidden surface graphite atoms were revealed in this imaging mode.

Not only do Hembacher et al. (4) detect the hidden α atoms, but they also succeed in minimizing the tip-sample forces that have been known to perturb the true atomic structure of the graphite surface in previous studies (8, 9). Because the carbon–carbon spring constant perpendicular to the basal plane is only 14 N/m, one must apply forces ≤0.25 nN for the deformation to be small (≤5%) compared with the interlayer spacing in graphite. The gently oscillating cantilever in the experiments of Hembacher et al. (4) achieves a sufficiently high signal-to-noise ratio that the total interaction force between the tip and sample is controlled below the 0.25 nN threshold.

Careful inspection of their data suggests that they were able to measure forces with resolution approaching the 1-pN level. What does it mean to have this degree of control? Mate et al. (10) used an AFM to measure the friction between a tungsten tip and a graphite surface. This work is generally regarded as the first publication on nanotribology. At an applied load of 50 μN, they observed an atomic-scale friction image, with periodicity of 0.246 nm, identical to that obtained in AFM imaging of graphite (Fig. 1). Similar atomic-scale friction images have been seen on nongraphite surfaces. However, there is something unsettling about these results. Back-of-envelope calculations show that the contact must contain many thousands of atoms. More importantly, the tungsten tip is likely to have an amorphous oxide. How is it possible that Mate et al. (10) and other researchers have seen atomic-scale periodic friction behavior? One possible explanation is that tip atoms are forced into registry with the graphite surface under the applied load. Another possibility is that the large applied load dislodges a flake of graphite from the surface. If the dislodged graphite were dragged in atomic registry with the surface, the frictional force would oscillate with a periodicity of the atomic lattice. In either case, these measured forces do not represent true atomic-scale friction of the native interface. One way to solve this dilemma is to use applied loads in the sub-100 pN range. At these light loads, the vertical deformation is only a few percent of the interlayer spacing in graphite and much less in stiffer materials.

Fig. 1.

Atomic scale image of friction force caused by tungsten sliding against graphite. The applied load is 50 μN. The length of the unit vector in the hexagonal pattern is 0.246 nm. [Reproduced with permission from ref. 10 (Copyright 1987, the American Physical Society).]

Conceptually, a similar approach was used at the micrometer scale to image the stiffness variation across a given material surface. Syed Asif et al. (11) placed a sharp diamond on a sample surface and set it to vibrate at a given frequency. The amplitude and phase were measured as a function of position, from which one effectively images the variation of stiffness and energy dissipation on the surface (Fig. 2). The improvements outlined by Hembacher et al. (4) may allow comparable mechanical and tribological characterization at the atomic scale.

Fig. 2.

Two-dimensional map (10 μm × 10 μm) of the elastic modulus of a carbon fiber–epoxy composite material. Brighter regions of the image correspond to higher modulus. The map was obtained by superimposing 1 μN modulation on a steady-state 1.5–2.0 μN load. The tip radius is ≈200 nm. (Image courtesy of Kathryn J. Wahl, Naval Research Laboratory, Washington, DC.)

It should be noted that the signal-to-noise improvements outlined in ref. 4 were made possible by operating at low bandwidths, which require one to have exceptionally low thermal drifts (≈0.02 nm/h). This was obtained by working at low temperatures. The question is: can one do this at or near room temperature? For most conventional AFMs operating at room temperature, even with drift corrections, one has a difficult time to achieve drifts much better than 2–10 nm/h. Materials commonly used in AFMs have coefficients of expansion in the 10^–6 per K range. To achieve 0.02 nm/h at room temperature, we need to develop materials with a coefficient of expansion <10^–8 per K. Alternatively, one may design AFMs that are 100 times smaller. The reduced mass of a miniaturized AFM would also enable the operating frequency to be commensurately increased. With the advent of NEMS/MEMS (nano/microelectromechanical systems) technology, it is conceivable that miniaturized AFMs could have performance comparable to the 30-ton behemoth described in ref. 4, leading to an interesting parallel between the room-sized computer powered by vacuum tubes and the integrated circuit.

Although many aspects of Feynman's 1959 speech (1) have been realized with the STM and AFM, there are notable exceptions that still serve as challenges to the scientific community. In particular, Feynman explains that many fundamental biological questions will be easily understood when biological systems are imaged in real space at the atomic scale. Although the STM and AFM have been applied in biology, true atomic resolution has been difficult to achieve in most cases. However, by minimizing normal forces between the tip and sample, the article by Hembacher et al. (4) has taken an important step toward imaging soft organic and biological molecules with unprecedented spatial resolution.