What is nanoline templating?

Patterning on the sub-10 nm scale is very difficult to achieve using conventional optical lithography, and other techniques such as electron-beam lithography, SPM lithography, atom-by-atom placement et al. do not scale to a large number of pattern elements, because the writing time is proportional to the number of structures written.

By using self-assembled nanolines as templates, these problems can be overcome. Self-assembly will occur all over the sample, and hence the writing time is independent of the number of structures formed. The template will be much more uniform, and if nanowires can be grown epitaxially onto the template, then the wires produced will be much more uniform, with fewer grain boundaries and other defects. In this way, it should be possible to break through the 10-nm barrier, and produce a high-density array of templates across a whole sample without having to do any patterning.

The structure and properties of the Bi nanolines are discussed more fully on the Bi nanolines page. Briefly, they are 1.54 nm wide, and can be grown to lengths exceeding 1 um. They are perfectly straight, without kinks or defects. They are not metallic themselves, having a local band gap rather larger than the surrounding silicon, and are chemically passive, so that a masking layer can be put down on the silicon, leaving the Bismuth nanoline exposed. As such, they make good candidates for templates.

The various processes which are involved in nanoline templating are shown schematically below. Surface metal atoms can either diffuse to the nanoline, or else join another metal atom on the passivated background, to form background clusters. This is equivalent to the situation in epitaxial growth where atoms can diffuse to substrate steps rather than forming islands. These background clusters may either form homogeneously (A in Figure 1) or else heterogeneously on a defect in the passivating layer (B). For those metal atoms which diffuse to the nanoline (C to D), this process may or may not involve a diffusion barrier, and the nanoline itself may be an adsorption potential well for the metal atoms, compared to the background. Once on the nanoline, metals atoms will diffuse to form nanowires (E) or clusters (F) (or diffuse off the nanoline). The interaction and relative strengths of these different processes for the different metals, will determine the efficacy of the nanoline as a template, and the morphology of the resulting nanostructure.

Realising the Concept

• The first step is to grow the Bi nanolines. The basic technique is to deposit Bismuth onto a Si(001) surface, which is set at around 850 K. After an anneal of a few minutes, the Bi nanolines form spontaneously. However, the growth process is a little more involved in order to produce a clean, well-ordered template.
• The next step is to passivate the silicon background, so that the deposited species do not stick to the background, but are forced to stick onto the template. This is done by covering the silicon in a masking layer, so far I have used ammonia and hydrogen. These species do not attack the Bi nanoline due to its different chemical nature.
• Having completed the clean template, the final step is to deposit material onto the surface, at a suitable flux and temperature, so that it diffuses to the Bi nanoline, forming nanostructures.

How to make the template

Bi will evaporate from the Si(001) surface at temperatures above 750 K. At 750 K, this process takes over an hour, whereas at 850 K, it will take only a few minutes. The Bi nanolines can be formed anywhere within this temperature window, with the Bi flux and substrate temperature controlling the detailed morphology, such as the average length, and the size of the domains of parallel nanolines.

Bi is deposited onto the surface at a flux of around 0.1 Monolayers (ML) / min. for 10-20 mins. The surface temperature is then maintained at high temperature, and the surface is annealed under STM observation to follow the development of the nanolines. Initially, the Bi will enter a metastable state, where Bi dimers are associated with missing dimer defects which relieve their strain. These complexes will often order into rough trenches, leading to the formation of short strips of Bi dimers on the surface. The "background" Bi which remains in the metastable state acts as a surface reservoir of Bi. The nanolines are assumed to nucleate from these background Bi features, although the details of the nucleation mechanism remain unclear.

There is clearly a large activation barrier to the nucleation of the Bi nanolines, as there is a considerable incubation time at 800 K, before they appear. At 850 K, nanolines have formed, and are growing already, by the end of the Bi deposition. With a higher substrate temperature, the nanolines are longer, and more well-ordered, and the surface is generally cleaner. With control over the substrate temperature, and the total amount of Bi deposited, the density of nanolines can be controlled. By long, slow Bi deposition over 30 mins. or more, something close to complete coverage of the surface in Bi, can be achieved.

The background Bi has a local structure very similar to that of the Bi nanoline. In fact, this structure was the original suggestion for the structure of the nanoline. It will act as an alternative adsorption site for deposited metal, and therefore for templating purposes, it is necessary to minimise the density of this Bi. The anneal should therefore be continued until all of this background Bi has evaporated back into vacuum, or has been converted into nanolines. However, once the background Bi reservoir has become depleted, Bi will start to evaporate from the nanolines, and within a minute, less than the time taken to scan one STM image, all the nanolines will evaporate. It is necessary, therefore, to anneal until just before all the background Bi has been removed to protect the integrity of the nanolines.

Masking the background Silicon

The difference between the Bi and the Si dimers will result in different chemical activity. While Bi is a Gr. V element, with a large lone pair, which might be expected to react with electron accepting groups, such as BF₃, in fact, it is almost entirely passive. This is because the hybridization of the Bi in the Bi dimers is p³, with 90 degree bond angles, and thus the lone pair has s character. This orbital sits some distance below the Fermi level, and thus is not available for covalent interactions. The Bi nanoline is therefore inert against attack by atomic H, ammonia, oxygen at up to 750 K, and ozone at room temperature. However, it appears to be susceptible to attack by water. As a result of this chemical passivity, exposure to atomic H leads to termination of the background in H, leaving the Bi nanolines exposed. Metal or other species can then be deposited onto the substrate and will diffuse to the nanolines, forming nanostructures along a straight line.

Deposition of Gr.III metals

As would be expected, Gr. III metals such as In and Al have a strong interaction with Bi, which is Gr. V. The result of the interaction is that a fascinating III-V 1-D compound is formed, with a zigzag structure. In inserts into Bi-Si backbonds, and the Bi-Bi dimer bonds, so that a chain of alternating In and Bi atoms is formed, each with two bonds to an unlike neighbour, and one to the Si substrate. The Bi is able to maintain its preferred p³ configuration, while In adopts its preferred sp² planar configuration. An image of the single and double zigzag structures is shown aboce.

We have determined the physical and electronic structure of this 1-D chain. It is a semiconductor, with a smaller band gap than the Bi nanoline. If In is inserted into both sides of the nanoline, a double zigzag is produced, which has 6 possible isomers, the most stable of which is shown. The isomer which has the lowest energy appears to be the one that we see in experiment, although the filled-states STM cannot distinguish between the lowest two isomers. The only difference between these two is the position of some In atoms, so empty-states images are likely to be required in order to pin down the structure completely.

The second layer of In breaks up the zigzag structure, forming a complete layer of In. The Bi floats to the top of the wire, and oval and hexagonal features are seen. This is the limiting thickness for metal deposition onto the templates. More metal will start to form spherical clusters, instead of forming a 3rd layer. Thus this is 1-D Stranski-Krastanov growth. Al behaves in a similar fashion to In, forming a two-layer wire. However, the well-ordered zigzag structures are not seen. This may be due to the greater size difference between the two metals, than in the case of In and Bi.

Deposition of noble metals

Noble metals would be preferable materials for the growth of nanostructures on these templates, as they would be stable in air. Several noble metals have been deposited, with similar results. In all cases, the interaction of the metal with itself is stronger than the interaction of the metal with the template, and hence metal droplets are formed. This can be seen as non-wetting behaviour, in contrast to the wetting behaviour of the Gr. III metals.

Ag has been studied most extensively amongst the noble metals. At low fluxes, and at room temperature, a small density of nanoclusters around 0.5 nm high are formed. As deposition continues, the number density of these clusters increases, while their size does not. By warming the substrate slightly, to an estimated 400 K, larger clusters start to form, 1 nm or more. The largest that have been grown are 5 nm in diameter, but this is limited only by the experimental time, as the flux must be kept low for the templating process to work.

Growing Atomic Chains

The next step is to try to form single atomic chains on the Si(001) surface, using the Bi nanoline again as a template. A number of different routes to achieve this are possible, as is summarised in the figure below. Some metals, such as Mn, will form atomic chains across the surface, as a diluted-dimer row, similar to the epitaxial growth of Si or Ge on Si(001), except with a single Mn atom replacing the Si ad-dimers. With 7 valence electrons, Mn is likely to have spare nonbonded electrons, which would give rise to interesting electronic properties for the atomic chains. Secondly, many transition metals find an energy minimum when embedded into a 5-membered ring at a 1DV site. There is a similar 5-membered ring site along each side of the Haiku structure, which suggests that this might be a preferential site for the adsorption of metals. Since these sites are in a straight line, such adsorption would generate an atomic chain of metal atoms. Modelling papers suggest that for Fe, this site is indeed stable, although in practise the Haiku structure may be destroyed in the process. Lastly, in a series of papers leading on from my Ag STM paper, it has been suggested that by adsorption at low temperature, the pathway to form clusters for noble metals may be blocked, and instead flat islands of metal within the trough of the nanoline, or atomic chains along each side, may be possible. Here the nanoline is acting purely as a physical template. This is an exciting possibility, as these chains could potentially grow tens or hundreds of atoms in length, unlike those of, for example, Mn on CuN, where the chain length is limited by the size of the CuN domains, to perhaps 10-15 atoms.

All of these possibilities will be investigated while I am in Geneva, using a new Omicron LT-STM, which will arrive imminently. Also some of the experiments will be performed in collaboration with Dr. Miki at NIMS.