Surface Chemistry of Materials, Spring 2013

LECTURE 10---SURFACE TOPOGRAPHY and CHEMICAL IDENTIFICATION USING SCANNING TUNNELING MICROSCOPY/SCANNING TUNNELING SPECTROSCOPY

From our previous lectures, we have seen that by bringing a metal tip into close proximity to a conductive surface, and by applying a voltage bias between tip and sample, we can obtain a tunneling current. Further, this tunneling current (I) is exponentially dependent on the tip-sample separation Z:

(1)I = I0 e-2jZ ,where j = (2m[V0-E]/ 2)1/2

(Remember, we now identify E in (1) as the bias voltage, while V0 is the local barrier height or work function—a quantity which depends on the atomic structure and composition of the surface.)

G. Binnig and H. Rohrer were able to apply these quantum mechanical effects—understood since the 1930’s—to develop an instrument capable of atomic resolution of the topography of surfaces in the early to mid-1980’s (e.g., G. Binnig, et al, Phys. Rev. Lett. 50, 120 (1983); G. Binnig and H. Rohrer, Rev. Mod. Phys.59,615 (1987)). The 5-decade time lag between theory and application was due entirely to the fact that the necessary electronics, computer, and materials technology had not developed. This is an important lesson that science (especially analytical chemistry) depends as much on technology as the other way ‘round.

We will consider here the use of STM at the solid/vacuum interface. A lecture on chemistry at the liquid/solid interface is posted on the website.

Let’s consider what happens when we move a tunneling tip parallel to a perfectly flat surface (we are not worried yet about atomic resolution, so let’s just think of our surface as flat, without atomic corrugations.) We apply a bias potential between tip and surface. (The polarity of the bias does not matter, but let’s suppose the tip is positive relative to the sample, so that electrons tunnel from the surface to the tip.) This situation is shown schematically in figure 1. The tip is moving at constant velocity from left to right, and encounters a step, height H, H< Z. To the left of the step, the current (IL) will be (eqn.1)

(2)IL = I0 e-2jZ

As the tip approaches the step, the current will increase, and to the right of the step, the current (IR) will be:

(3)IR = I0 e-2j(Z-H)


Figure 1: tip moving parallel to a surface approaching a step.

For typical values of bias voltages and work functions, j ~ 1 Å-1. Let’s assume Z is about 5 angstroms, and that our step is about 1 angstrom in height. From (2) and (3), we then have:

IR/IL = [I0 e-2jZ]/[ I0 e-2j(Z-H)] = e2jH7.4

This means that we should observe a 740% increase in tunneling current going over a step that is of atomic dimensions! Therefore, if we can get a tip close to a surface (and hold it steady), we could scan over a surface and get an atomic-resolution topographical map of the surface. Again it is the exponential change in tunneling current with tip-sample separation that permits us to use tunneling as a tool for surface topography.

There is a small problem with the above picture: it is in practice extremely difficult to move a tip at constant height over a real surface. Real surfaces—even those of single crystals—often have steps and defects tens of angstroms in height. In addition, it is often difficult to move a tip parallel even with a very flat surface, since even a small misalignment with the plane of the sample can completely distort the measurement. The above events will sooner or later result in a crash of the tip into the surface.

One way around such difficulties is to use a constant current feedback mechanism that is based on keeping the tunneling current constant. Such a mechanism is shown in figure 2.


Figure 2 Schematic of tip motion under constant current feedback

As shown in figure 2, the tunneling current (ITunneling) is fed from the tip into the computer. The computer then adjusts the tip/sample separation such that ITunneling is maintained at a constant value set by the experimenter. This is done by sending the required voltage signal to the piezoelectric drive (more on this later) which can physically move the tip with near-atomic precision in x, y and z directions. The computer can then record the tip motion during scanning. As the tip approaches a step, the constant current feedback will raise the tip in order to maintain a constant tip/current separation and ITunneling. Repeated line scans of the surface can then be added together (this is all done in computer software) to yield a topographic image of the surface as shown (rather crudely) in figure 3.

Fig 3. Individual line scans added together to proved the image of a step.

Since the original development of STM, imaging software has become increasingly sophisticated to take advantage of improving computer hardware. Modern imaging packages (sold with STMs by such companies as Omicron, RHK and Digital Imaging/VEECO) typically use finely adjusted vertical brightness scales to provide “pictures” of surfaces such as that shown in figure 4.


Fig. 4. Atomic resolution image of a Fe(111) surface with pseudo-rectangular surface reconstruction. See J. S. Lin, B. Ekstrom, S. G. Addepalli, H. Cabibil and J. A. Kelber, Langmuir 14, 4843 (1998)

In this way, STM allows one to “see” atoms at a surface by building up maps of the tip motion across the surface under constant current feedback. It is tempting to associate the individual bumps in the image with atomic positions. Most of the time, such an association is perfectly correct. It is important to remember, however, that what the tip is actually measuring is a tunneling current, which is actually dependent on the electron charge density. On polar surfaces, atoms may be divided into donor sites, and acceptor sites. The acceptor atoms will have filled valence orbitals, while the donor sites will have empty valence orbitals. This is what generally happens at the surfaces of compound semiconductors. If the charge donation is large enough, the tip will image only donor or only acceptor sites. This type of situation is shown in figure 4.

Figure 4: STM image of a polar surface may include only donor or only acceptor sites, depending on tip/sample polarity

  1. This type of effect can even be observed for relatively non-polar metal surfaces, where the size spatial extent of the charge cloud on one site can obscure that of a nearest neighbor.. One example is Ni3Al(111). This surface consists of Al atoms each isolated from each other by a coordination shell of Ni nearest neighbors. Because of charge distribution effects, however, the apparent atomic resolution image of the surface yields a nearest-neighbor distance of just twice the actual distance. Only the Al sites are imaged. (See fig. 5 and S. G. Adepalli, B. Ekstrom, N. P. Magtoto, J.-S. Lin, and J. A. KelberSurface Science 442, 385 (1999)


Fig. 5, Atomic resolution image of Ni3Al(111)

In Fig. 6, we see an empty states and filled states image for a TiO2 surface. The same area of the surface is imaged with tip biased negative wrt the sample (empty states) and tip biased positive (filled states).

The band structure for TiO2 (Fig. 7) indicates that the filled states correspond to O sites, and the empty states to cation sites.


Electronic Effects and Chemical Information

In an earlier lecture, we saw how for VBias < V0m ,

ITunneling VBias

for a tip in proximity to a metallic surface. What happens if the tip is in proximity to a semiconducting or insulating surface. Obviously, if the insulating film is very thick, no tunneling will be observed. (Why?) For a semiconductor or very thin insulating oxide, however, STM spectra can be obtained. In order to understand this, consider the situation of a tip in proximity to a semiconductor, with band gap Eg, as in figure 8

Figure 8. Metallic tip in contact with a semiconducting or insulating surface.

Note that at zero bias (fig. 8), an electron cannot tunnel from the Fermi level of the tip to sample, because there are no acceptor states on the surface. (We are assuming no impurity states in the band gap.) Similarly, no electrons can tunnel from the surface to an empty state on the tip because there are no electronic states in the gap. In fact, we will observe zero tunneling current until  VBias > Eg/2. For VBias > Eg/2, electrons can tunnel from the top of the semiconductor valence band into empty states on the tip. For VBias < Eg/2, electrons can tunnel from the Fermi level of the tip to the empty valence band states on the semiconductor. The I/V curve for a semiconductor looks like (fig. 9):

Fig. 9 I/V curve for a semiconductor

Similar curves will be obtained for very thin oxide films on metal substrates, even though such oxides may have very large bulk bandgaps. The reason this is possible is that if an oxide is sufficiently thin, wave functions of atoms at the oxide/metal interface will mix, leading to a band gap that is much less than the nominal value. The result is that an I/V curve for a thin oxide film or island on a metal substrate will also look like figure 9.

One can then use STM I/V curves to give some limited chemical information on the surface immediately beneath the tip. Is it an insulator (semiconductor), or a metal? This is sometimes called scanning tunneling spectroscopy (STS). Using STS allows one to probe the local electronic structure of the surface beneath the tip.

A real example is displayed in figure 10 which shows an STM image and I/V curve for Ni3Al(111) compared to that for the same sample exposed to O2 so as to form an extremely thin ( ~ 5 Å) insulating Al2O3 overlayer.


Questions:

  1. Figure 9 indicates that for a tip in proximity to an insulator/semiconductor surface, tunneling will be observed for a bias voltage with a magnitude greater than Eg/2. What happens, however, if Eg/2 > V0, the local barrier height?
  1. You are using an STM to image the InAs(100) surface, which is quite polar. Should the tip be biased positive or negative with respect to the surface in order to image the indium atoms?

3.A tip at a distance of 1 Å from the surface and a bias of 0.1 V results in a tunneling current of 1 nA. The local barrier height is 4.0 eV. The tip, as it scans, moves over a region where the local barrier height is decreased to 2.0 eV. Assuming the constant current feedback loop is on, what will happen to the tip/sample distance?

1