Toward Single Atom Chemical Analysis with STM ∗

An important function of microscopy is the chemical analysis of the sample. Chemical analysis with an atomic resolution microscope can mean two diﬀerent things. First, for a sample of known composition or adsorbed species, it means distinguishing the chemical species from the atomic image. Second, for a sample of unknown composition, it means the identiﬁcation of the chemical components from the atomic resolution image. Here I compare available methods of chemical identiﬁcation in ﬁeld ion microscopy (FIM) and scanning tunneling microscopy (STM), and report our progress in achieving true atomic resolution for a non-destructive chemical analysis of a sample surface using STM. [DOI: 10.1380/ejssnt.2003.102]


I. INTRODUCTION
Microscopy to most people means obtaining a magnified image of the sample using a microscope [1,2]. But in fact microscopy can do much more than just obtaining an image. It can be used to obtain information on the density of states, or the energy states and the band structure of the sample. One of the functions of microscopy, the identification of chemical species and their distribution in the sample, is another important aspect. In atomic resolution microscopy, at least in principle, all these functions can be achieved for individual atoms and atomic sites. In the past, however, the chemical analysis aspect is found to be much more difficult to achieve than other aspects. Here I summarize as well as compare what have been achieved for single atom chemical analysis in field ion microscopy (FIM) and in scanning tunneling microscopy (STM), and also report our progress in achieving a non-destructive chemical analysis with STM. In particular, I would like to emphasize the importance of using a chemically and thermally stable single atom sharp tip for this purpose.

II. CHEMICAL ANALYSIS FROM MICROSCOPE IMAGES
In field ion microscopy, the minor species of a twocomponent dilute alloy often give rise to vacancy-like image spots. This was thought to result from the preferential field evaporation of atoms of the minority alloy species. A possible preferential field evaporation of a chemical species from an alloy surface in the FIM was emphasized by many investigators in the mid-1960's [2]. To investigate the image contrast of alloys, Tsong and Müller reported first to ordered alloys PtCo [3][4][5]. Although this ordered alloys had L1 0 fct structure with a c/a ratio of 0.97 to 0.98, the FIM image showed a structure unexpected from the fct structure symmetry. It was interpreted that in this alloy, all the Co atoms were either not imaged, or gave rise to too dim an image to be seen * This paper was presented at The 4th International Symposium on Atomic Level Characterizations for New Materials and Devices (ALC '03), Kauai, Hawaii, USA, 5-10 October, 2003. † Corresponding author: tsongtt@phys.sinica.edu.tw in the FIM.
The non-imaged species in PtCo was first assumed to be Co atoms, not Pt atoms. It was later confirmed with a study of the asymmetric Pt 3 Co with L1 2 structure [3][4][5]. Tsong and Müller interpreted the image contrast to arise from the preferential field ionization above atoms of these two species. Similar image characteristics were later found for other ordered alloys such as MoNi 4 , WNi 4 and so on by Newman and Hren, and by many others [6]. Here I must state clearly that a partial preferential field evaporation of a species may still occur, but the important thing is that in the FIM, the chemical species of binary ordered alloys can be identified without ambiguity from the image contrast. This method was later applied to measure the long-range and short-range order parameters of samples under different heat treatments [7]. As no pc image digitizer was available in the 1960's and 1970's, it was not possible to obtain short-range order parameters with any reliability, but formation of atomic row-like clusters could be seen on partially ordered PtCo and Pt 3 Co alloy surfaces.
Similar image characteristics could be used to identify chemical species in ordered and partially ordered alloys using STM as was first reported by Schmid et al. in 1993 [8, 9]. They used PtNi and PtNi 3 in their first study but since then have extended the method to many other ordered or partially ordered alloys. They have also obtained information on order parameters as well as the structure of clusters in partially ordered alloys. These two alloys have also L1 0 and L1 2 structures as PtCo and Pt 3 Co. The advantage of STM as compared to FIM is that the sample surface can be much larger in size, thus a statistically reliable amount of data can be more easily collected. Also the formation of various types of clusters in partially ordered alloys can be much more easily seen and studied.
In STM, at least in principle, one can measure the atomic site specific density of states from which the chemical species of that atom is identified. This method is complicated by the fact that the density of states spectrum is in general very sensitive to the atomic environment of the sample atom as well as that of the tip atom. The tip atom is particularly difficult to control in STM. In the past, with great effort one may be able to start an STM experiment with a single atom sharp W tip. But after an image scanning, nobody has ever demonstrated that the tip remains single atom in sharpness. Even worse, nobody has ever demonstrated that the tip-apex atom is really a e-Journal of Surface Science and Nanotechnology W atom when atom resolved STM images are obtained. In other words, single atom chemical identification is still out of question in practical STM experiments.

III. CHEMICAL IDENTIFICATION OF UNKNOWN SAMPLES
For the purpose of true single atom chemical analysis of a sample, Müller et al. introduced a time-of-flight atom-probe field ion microscope (ToF APFIM) [10]. The atom-probe FIM is basically an FIM having a small hole covering one to a few atoms in image size at the screen. An operator can aim the hole at a particular atom or a few atoms, then use either a ns high voltage pulse or a ps laser pulse to stimulate field evaporation of these atoms. Behind the probe hole is a time of flight mass spectrometer of single ion detection sensitivity. From their flight times, these surface atoms can be chemically identified. Recently, 3D atom probe has been developed [11]. By using a position sensitive ion detector, no probe hole is needed. With the help of computer image reconstruction, in fact the spatial distribution of chemical species in the sample can now be displayed in a 3D image with near atomic spatial resolution. The composition of each surface atomic layer can be derived. Whereas this technique is now highly developed, it has some intrinsic limitations. First, the sample has to have the shape of a sharp tip. Second, although field evaporation usually occurs atom-by-atom from kink sites of metallic samples, for compounds, especially semiconductor samples, this does not occur because of the large field penetration depth. Field evaporation can then be quite random. Spatial identification becomes extremely difficult to achieve. Third, the greatest disadvantage of APFIM is that the chemical identification is done destructively, i.e. the sample is destroyed by the chemical analysis. One of the great advantages of APFIM is that atoms and atomic layers can be field evaporated, thus a 3D distribution of chemical species of the sample can be obtained.
In STM, there are at least two conceivable methods for doing single atom chemical analysis without destroying the sample, namely from the vibration spectrum of the atom or molecule of interest [12], and from the density of states distribution of a surface atom. The first method probably can be more easily achieved for adsorbed atoms and molecules where chemical bonds are covalent in nature, similar to high-resolution electron energy loss spectroscopy (HREELS). For adsorbed atoms on a metal surface, distinguishing atoms of different chemical species may involve a measurement of an atomic site specific phonon spectrum. A sufficient mass resolution would be very difficult to achieve for separating chemical species of similar masses and chemical bond strength.
The other method is the measurement of the electronic density of states distribution of a surface atom, or the surface site-specific electronic density of states. It is conceivable that the measured density of states distribution can be compared with theoretical distributions of atoms of various elements. From such a comparison, the surface atom in question can be unambiguously identified. A good possibility is that a standard table of electronic density of states distributions can be first made through laborious experiments. The table can then be used later for single atom chemical analyses by the comparison of density of states distributions, a method similar to Auger Electron Spectroscopy except now the chemical analysis is done on one particular atom.

IV. SINGLE-ATOM SHARP THERMALLY AND CHEMICALLY STABLE TIPS
In all these experiments, it is necessary that a chemically and thermally stable STM probing tip of single atom sharpness with a known apex atom is used. We have recently used a method developed by Madey and coworkers [13,14] to produce such a tip [15]. They found that when bcc W(111) or Mo(111) surface is covered with about two monolayers of Pd or Pt atoms and subsequently annealed to high temperatures, but below their thermal desorption temperature, many nanometer size pyramids can be formed on the surface. These pyramids have three (112) facets. They are either W or Mo pyramids wrapped in a physical monolayer of chemically inert noble metal atoms Pd or Pt.
In our experiments, Fu et al. [15] use (111)-oriented W tips. A tip is first carefully cleaned by flushing to ∼ 2000 • C in ultra-high vacuum. It is then further low temperature field evaporated. About two monolayers of Pd are deposited and the tip is immediately annealed to ∼1000 K for about three minutes. A subsequent imaging with the FIM reveals that a single atom sharp tip is formed which can be proven atom by atom using low temperature field evaporation. Fig. 1 shows such an experiment.
After the annealing, on turning on the imaging voltage very slowly, the image of one atom appears at the center position of the (111) facet, as shown in Fig. 1(a). When this atom is field evaporated, the image shows the next layer that is now consisted of three atoms as seen in (b). When this 3-atom layer is field evaporated, the underneath layer is found to consist of seven atoms. Whereas the numbers of atoms in the first and second layer are in good agreement with the model of single atom sharp pyramid shown in a model of Fig. 2, the number of atoms in the third layer should be 10 instead of 7 found in the experiment. Obviously three atoms (shown by doted circles in the model) are still missing. It means the tip is not yet atom perfect. On further annealing, one finds there are 1, 3, and 10 atoms in the 1st, 2nd, and 3rd layer, respectively. It means the pyramid is now atom-perfect down to at least the third layer. In fact one can continue this procedure to prove that the pyramid is now atom perfect down to several layers. This can also be confirmed by looking at the single atom sharp wedges of the pyramid (images of wedge atoms are oblong in shape because of image distortion by the diverging field ions originated at these wedges).
In fact from the much lower evaporation field of these atoms we can conclude that they are Pd atoms. The inner most atom of the 3rd layer is, however, found to be field evaporation resistant. It is not possible to field evaporate this atom without also field evaporating the 4th layer. That means this is a W atom. In a similar fashion, one can prove atom by atom and layer by layer that the structure is a W pyramid wrapped in one physical monolayer of noble metal Pd atoms. Each deposition of Pd atoms, about twenty cycles of heat treatment can result in the formation of an atom-perfect nanometer pyramid. It is important to point out here that the pyramid is thermally stable up to the temperature of it formation, or ∼1000 K. It is conceivable that this kind of tips can be used for STM probing with great stability, but at the present time, we do not have the needed accurately controlled heating capability of the STM probing tip. In addition, one would need to use either a TEM or an FIM to make certain that an atom perfect pyramid is formed before the start of an STM experimental run, and after the experiment the tip remains atom perfect also. A TEM should be most valuable for the characterization of the probe tip before and after an experimental run. We are currently building a TEM-STM system aimed for single atom chemical analysis of the STM sample.
Let me emphasize here again. In single atom chemical analysis with STM, I can envision that such a tip is used to take density of states data from different atoms at the surface. Since the tip atom is known, from the spectrum taken at a particular atom site, one should be able to compare with theoretical calculations. From such procedure, the chemical element of that surface atom is identified. Another possible procedure is that a standard table of density of states functions is first made by depositing atoms of known species on the surface. This table is then used for later use to establish the chemical species of unknown atoms on the surface, similar to a standard table in AES analysis. Of course in this type of experiments, the tip is best characterized first by FIM, and is then continuously observed in a TEM during an STM scanning.
Single atom tips, in fact, have many other applications than that for use as a probing tip of STM and scanning probe microscopes (SPM) because of their unique properties [16]. They can be used as a point electron and ion source ofÅspatial resolution. When it is used as an electron source, electrons emitted should be coherent in phase, thus can be used for holographic imaging purposes. Another possible use is field emission display provided a well-aligned self-assembled tip array can be produced. There are attempts by some investigators to achieve this goal but so far without success.