Adding atoms to a semiconductor can improve its electronic properties. In an oxide, taking atoms away can have a similar electronic effect — one that could, it seems, be exploited in device applications.
By 2007, the information age will have hit a fundamental roadblock. Without major changes in technology, the spectacular improvements in computer performance that we have enjoyed for decades will cease, because transistors based on silicon and silicon dioxide will no longer be able to keep up with Gordon Moore's famous law1,2 — that the number of transistors per unit area in an integrated circuit doubles every couple of years. But these limitations might be overcome if Si and SiO2 were complemented in these devices by other materials. The candidates of choice are oxides, which are already assuming a vital role in semiconductor electronics. Now Muller et al.3 (page 657 of this issue) show that it is possible to control the electronic properties of these materials with the nanoscale precision necessary for the information industry.
Oxides offer a broad spectrum of properties — some are excellent insulators, others are superconductors. Some oxides have flippable electric or magnetic dipoles, suggesting myriad device possibilities. Indeed, oxides such as hafnium dioxide are forecast to replace SiO2 in the transistors of laptop computers within only three years1. Another oxide known as ‘Lustigem’ — alias strontiun titanate (SrTiO3) — was a popular diamond substitute in the 1960s. If some of its oxygen atoms are removed, the glittering gem turns a deep blue (Fig. 1), and changes from insulating to conducting. This change in colour and conductivity is due to electrons that are left behind: because there is a difference in charge between an oxygen ion (O2−) and an oxygen atom, for each oxygen atom removed two electrons are added to the SrTiO3 matrix. Oxygen vacancies thus function as electron-donating dopants — an effect commonly achieved in semiconductors by replacing some atoms with others that contain more or fewer electrons than the atoms for which they substitute. But can doping through vacancies be implemented and monitored in a controlled way on the atomic scale?
It seems so. Muller and colleagues3 have made an unexpected double breakthrough. With unrivalled precision, they have measured the quantity and location of oxygen vacancies in films consisting of layers of fully oxidized SrTiO3 and of SrTiO3–δ, in which some oxygen atoms are missing. Their first major advance is to have grown alternating layers of doped (δ≠0) and undoped (δ=0) SrTiO3–δ, where a layer may be as thin as three unit cells. Analogous ‘superlattices’ are used in conventional semiconductor technology to enhance the lifetime of charge carriers4; in oxide superconductors, they are used to increase the supercurrent density5. Muller et al. grew their superlattices using pulsed laser ablation — a popular research technique for depositing thin films of oxide materials. Deposition occurs when a laser beam hits a SrTiO3 target inside a vacuum chamber, vaporizing its surface into a plasma. Some of the vaporized atoms condense on a nearby substrate, again of SrTiO3, heated to 750 °C. Adjusting the oxygen pressure in the chamber controls the δ of the single crystalline SrTiO3–δ layers deposited.
To image the oxygen vacancies, the authors used a scanning transmission electron microscope (STEM). As the tightly focused electron beam of the STEM is scanned across a cross-sectional slice of the deposited superlattice, a map is made of the positions where electrons are scattered slightly by oxygen vacancies and related defects; simultaneously, the energy loss of the transmitted electrons is measured, revealing the electronic effects of the missing oxygen atoms on the surrounding atoms (that is, changes in their oxidation state). This powerful technique6 offers outstanding sensitivity in resolving and identifying columns of atoms in crystalline samples, and has been used to image individual impurity atoms in silicon7.
The team scanned their layered samples in cross-section and spotted regions in which as few as two oxygen atoms were missing. And here came the second breakthrough. The STEM images (such as the one in Fig. 2) show that the oxygen vacancy concentration can change with surprising abruptness — from a layer with no oxygen vacancies to a layer with some constant number of vacancies over a distance of only 0.4 nm (the thickness of a single unit cell of SrTiO3). At 700 °C, oxygen diffuses in minutes over many micrometres8, which would be expected to completely level out any nanometre-scale steps in the oxygen concentration profile. But it does not. That such sharp doping profiles are achievable is excellent news for the development of devices involving doped SrTiO3 layers, as it has the highest mobility of any known oxide at low temperature9. Yet the data do raise the question of why the profiles are so crisp. Are, for example, the oxygen vacancies or the sample microstructure stabilized by an as-yet-unknown mechanism, which may even be applicable to other ionic materials? No doubt Muller and colleagues will set about unravelling this puzzle too.
This work greatly broadens the options available for manipulating the electronic properties of oxides, and probably ionic materials of all sorts, at the nanometre scale. At present, the standard means of doping is to replace one cation with a different cation (that is, an impurity). But the ability to dope films without introducing impurities — thereby avoiding the risk that they might ride on the growing surface or hang around in the deposition chamber and become incorporated at undesired locations — is an enticing advantage of doping with vacancies. As films can grow by the lateral movement of steps as atoms are incorporated, even lateral doping using vacancies might be possible, analogous to the lateral superlattices and one-dimensional wires created using conventional semiconductors10. Muller and colleagues show how a view of nothing can turn gems into electronics.
About this article
Nature Materials (2005)