What are Bismuth nanolines?

Bismuth nanolines are 1D features, which are very long (>500 nm), perfectly straight, and only 1.5 nm wide. Bi nanolines were discovered by accident in 1994, while we were investigating the use of Bi as a surfactant in Ge/Si growth. A sample had been left to anneal overnight, and in the morning, instead of flashing it clean as was usual, I decided to see how clean the surface was. While doing so, I noticed that the surface was unusually flat, and that long bright lines could be seen running across the surface. These were the Bi nanolines.
The Bi nanolines form by competitive adsorption and evaporation. If Bi is deposited onto the surface at around 820-870 K, small islands of Bi form. At the end of the deposition time, the islands start to evaporate over the course of 10-15 minutes. However, not all the Bi evaporates, some it remains on the surface, and forms the nanolines.
A pdf version of a presentation that I gave in December 2004, OwenNano3Biline.pdf, is available for download. Also, we have recently finished writing a review article for the Journal of Materials Science. A selection of the references from that review are available at the Bismuth nanoline references page.

Left: Bi nanoline, bright at high biases. Right: dark at low biases. Centre: a perspective view of the Haiku model of the Bi nanoline.

Properties of Bi nanolines

A large area image is shown below. This is a 600 nm square picture, showing several lines which have grown on different terraces. Since the dimers run in perpendicular directions on neighbouring terraces, the nanolines run in two different directions. By annealing for a long time, with a small Bi flux, one direction tends to win, and a single-domain surface results, with nearly all the Bi lines running in the same direction. Another interesting phenomenon of the Bi nanolines is that they repel defects. At the temperatures where these lines form, missing dimer defects move extremely rapidly, but they tend to avoid the area close to a Bi nanoline. This can be seen in the third picture below, where a highly defective surface contains regular stripes of missing dimer defects. However, broad clean patches exist around each nanoline.

Left: Bi nanolines growing in both directions. Centre: One domain dominates on vicinal surfaces. Right: Defect Exclusion Zones - patches clear of dark defects - indicate a repulsion between Bi nanoline and missing dimer defects.

What is the structure of the Bi nanoline?

The 3 proposed structural models for the Bi nanoline are shown here. Only the Bi dimers are visible in STM, and the substructure has had to be inferred from a combination of detailed STM measurements, and modelling of many possible structures. You can see more details about the Bi nanoline structure at Dave Bowler's pages. Before we came up with the Haiku model, we initially thought that the Bi nanoline had a 3-dimer structure, where Bi dimers substituted for Si dimers either side of a missing Si dimer defect. This structure became known as the Miki model. Later, a paper by Naitoh et al. in Japan demonstrated that this model must be wrong. They took a plan view image of the nanoline on the Si substrate, and marked off Si dimers to establish the registry of the substrate. By extending these markers across the nanoline, they showed that the Bi nanoline is four dimers wide, and not three, as in the Miki model. This procedure is shown in the left-hand image below. They also proposed a model, which became known as the Naitoh model, but this did not have a very good energy, and the Bi dimers were too close together. So for about 2 years, there was no stisfactory model of the Bi nanoline. We ploughed through testing literally hundreds of candidate structures, none of which were even as good as the Miki model (which is very stable). Finally, we came up with the Haiku structure. This is the only structure we have found which is more stable than the Miki model. Furthermore, it is 4 dimers wide, has the correct Bi dimer spacing, and the reconstructed core gives a simple explanation for the growth of very long, perfectly straight nanolines. Not that the Miki model is a poor structure. It is much easier to form than the Haiku model, and will form upon initial adsorption of Bismuth onto the surface. During annealing of the surface, the two structures co-exist, as is shown in the right-hand image below, marked as "background Bi" and "Bi-1DV trench". This co-existence, may provide an explanation for some of the contradictory results obtained from non-local experimental techniques, such as X-ray and Photoelectron Diffraction, purportedly for the Bi nanoline.

Left: Markers on the Si dimers show that the nanoline is 4 dimers wide. Right: STM images showing the co-existence of the Bi nanoline and the Miki structure ("Bi-1DV trench").

The Haiku model

We named the new model the "Haiku" structure, partly because it was first discovered in Japan, and partly because it comprises 5 and 7-membered rings. The clue to this structure came from a similar As/Ge step edge structure. We found that this structure was strained, but was very stable, and so we decided to try variations of the structure, to see if one would match the STM. After many attempts, we found that by putting two of the As/Ge structures together, removing some central atoms, and bonding the whole thing together, we ended up with a double-core structure. First, we rotated second and third-layer atoms to make 7-membered rings. One half of this is the As/Ge derived structure. Then we removed 4 central atoms to relieve the compressive strain of the 7-membered rings, and pull together, to make a structure under tensile strain. This structure would allow for strain relief of the Bi in the top layer of Si, since Bi is much bigger than Si, and hence generates compressive stress. Indeed, in experiment, the strain induced by this reconstruction causes a change in contrast of the Si near the nanoline (This change in contrast is reproduced by modelling of the Haiku model, but not by the other two structures). Thus the driving force for the formation of this structure is quite clear, and there is good agreement between experiment and modelling.

The Haiku structure of the Bi nanolines is unusual, extending five layers beneath the surface, and comprising 5-membered and 7-membered rings of silicon, capped with a pair of Bi dimers. This core is actually a small piece of hexagonal silicon, a structure which exists in nature as the mineral Lonsdaleite. Unlike other nanowire structures on Si(001), such as the silicide wires mentioned in our review paper, which are essentially the result of anisotropic strain in an otherwise conventional heteroepitaxial system, the Bi nanolines do not have a bulk structure. They are therefore neither a line of surface adsorbates, nor a heteroepitaxial crystal, they are more of a adsorbate-stabilized reconstruction of the Si crystal, and are therefore unique in our experience of surface structures. The image shown here shows the triangular core of the Haiku model, which believe is responsible for its structural perfection, and extreme straightness.

But is it really 3 or 4 dimers wide?

Despite the apparent success of the Haiku model, some researchers have continued to support the Miki model, see their recent paper Bismuthrefs 2005 Ref.9 and our Comment and their Reply Bismuthrefs 2006 Refs.4,5. We find this rather puzzling, as the STM data seems very clear to us. The procedure shown above, of marking lines on the Si substrate, is pretty unambiguous, and, moreover, has been reproduced multiple times by different experimental groups. Their reply focusses on the pitfalls of measuring registry from linescans, which of course is why we use the line-marking method above. As a double check, we have taken an STM image which has a clean surface, not a H-terminated one, and attempted to match up lines with lines on both Si dimer troughs and Si dimer peaks. As is evident, the nanoline is 4 dimers wide trough-to-trough, or 5 dimers from the Si dimer on one side to the Si dimer on the other side. (Note that the edge Si dimer is pulled in towards the centre of the nanoline by the tensile stress of the nanoline core.) Both of these measurements point firmly towards a 4-dimer model, i.e. the haiku model. For a 3-dimer model, the line on a Si dimer peak would have to pass through the centre of the nanoline, which it clearly doesn't. Moreover, the dark trenches on the outside of the Bi dimers matches the Haiku model far better than the Miki model. Finally, the STM evidence of co-existence of the two models, shown above, should end this controversy. However, probably this will not be resolved until some direct evidence for the subsurface reconstruction of the Haiku model is found.

Nanoline Templating

As well as being long, straight and defect-free, Bi nanolines have some useful properties which makes them useful for nanowires. They do not make nanowires themselves, as the band gap of the Bi nanoline is larger than that of the surrounding silicon. However, I am pursuing a route I have called "nanoline templating", in which the nanoline is used as a template for a masking process, which then allows other metals to be deposited on top of the nanoline to form a wire. In this way, wires of different metals, with different properties can be made. Bi nanolines are resistant to attack by atomic hydrogen, oxygen and ozone, and this chemical inertness has useful properties for an application as a nanowire. Hydrogen acts as a mask over the surrounding silicon, while oxidation of the silicon around the nanoline isolates the wire from the rest of the substrate.

During my time at ICYS in Japan since November 2004, I have tried out ammonia as an alternative masking material, deposited Ag, Au, Pd and Pt, Fe, Co, all of which form nanoclusters and In and Al which form nanowires, and thought up several hundred more experiments to do! A page about the nanoline templating is in the works, when I have time.