As mentioned in the introduction, the STM was the first type of scanning microscope invented. The STM uses the phenomenon of quantum tunnelling.

In the STM, the point of a very sharp conducting needle is moved (using the microscope scanning system) close to the surface of a (conducting) sample and a voltage bias between the two is applied, causing a tunnelling current to flow between tip and sample. The current depends on the tunnelling probability, which varies exponentially with the tip-sample distance, and so monitoring the tunnelling current gives a very sensitive indication of the tip-sample distance. Typically a bias of 1 V or less is used, resulting in a tunnelling current of a few nA when the tip-sample separation is under 1 nm. If the tip is sharp enough, there will be one atom at the tip that is slightly higher than the surrounding atoms, and it is from this atom that most of the tunnelling current originates. The needle is then scanned across the sample surface, and the tunneling current is monitored during the scanning. The spatial resolution of STM is extremely high (sub-angstrom) in all three axes, and so the images that are obtained from STM can show individual atoms on a surface.

It has also been found that when a high voltage pulse (~10-20 V) is applied to the scanning tip, it is sometimes possible to make atoms jump from the sample to the tip or vice versa, and thus move atoms about a surface with great precision. This technique was first reported in Nature*, and naturally got a lot of people very excited. In principle, you could use this to build a electronic or mechanical device atom by atom - the ultimate in control. In reality, doing things this way is extremly slow, so don't expect anyone to start up a nanobot factory anytime soon. To see some of the atomic structures that have been built using STM systems, have a look at this gallery at IBM.

While STM is a powerful technique, there are some attendant problems. The most obvious is that the sample must be conductive, either a metal or a semiconductor. To image insulators requires coating with a layer of metal to provide the conduction path, which is of little use if the detailed surface properties are desired. While it is possible to use STM in air, most metal and semiconductor samples (and indeed the metal probe) tend to form oxide layers in air. These layers may only be a few monolayers thick, but this can be enough to necessitate dragging the tip through the oxide to bring the conductive elements close enough to establish a tunnelling current, often damaging the tip. In addition, many samples tend to acquire a few monolayers of water in ambient conditions, which can short circuit the tip-sample junction. For these reasons most STM imaging is performed in ultra-high vacuum conditions, with the sample preparation facilities and the STM system usually sharing the same vacuum system to prevent sample contamination.

It is possible to use STM for imaging magnetic samples. If the electrons in the probe are preferentially spin-polarised in one direction, the tunnelling probability will depend on the spin of the sample electrons at the surface, which depends on the magnetisation of the sample. Thus it is possible in theory to probe the magnetisation of individual surface atoms. If the probe current is polarised in one direction and then reversed, the difference between the two currents yields the component of magnetisation in that direction. The main problem with this method is controlling the spin polarisation of the probe current. One method employed by Nabhan and Suzuki (1998) is to use GaAs as the probe and illuminate it with circularly polarised laser light. The light causes the required spin majority in the conduction band of the GaAs. The polarisation direction depends on the rotation direction of the light, and thus can be reversed easily. Another method used by Bode et al (1999) is to use magnetic material on the probe itself (a thin coating of Fe in this case), and use an external field to magnetise the probe in the required direction prior to imaging (the sample imaged consisted of ~6 monolayers of Gd). Some results have been reported using both methods, but operational difficulties (particularly the requirement for UHV and clean samples) and difficult theoretical questions mean this type of magnetic microscopy is still in its infancy.