Why defective microchips do not necessarily have to be thrown away

The circuits built up from many transistors in the smallest of spaces on the ever smaller microchips are also becoming more susceptible to material defects. However, Viennese researchers are now showing in the specialist journal “Advanced Materials” why faulty chips can also function properly and should not necessarily be thrown away.

The basis of electronic circuits on chips are transistors through which current can either flow or not. Such circuits can sometimes consist of billions of transistors, according to a broadcast by the Vienna University of Technology (TU) on Wednesday. In order to pack such a number on a microchip, the conditions there are correspondingly cramped. However, the smaller the transistors become, the more likely it is that small material defects – such as individual atoms in the wrong place or unclean connections – will prevent them from fulfilling their function.

“In the case of larger components, such defects do not play such a dominant role as long as they do not occur too frequently. But with tiny transistors in the order of a few nanometers, a single defect can lead to the transistor’s characteristic curves being far outside the specified tolerance range. It is therefore considered unusable,” says Michael Waltl from the Institute for Microelectronics at the Vienna University of Technology.

But that doesn’t have to be the case, describes Waltl’s team in their work. Because it could well be that a circuit works despite several broken transistors. The decisive factor is “not whether an individual transistor meets certain abstract criteria when errors occur at the atomic level, but whether the entire circuit then still behaves correctly.” In industry, however, component tests are often based on whether and how many transistors are not working, because it is then assumed that above a certain value the entire structure will not do its job.

In experiments and computer simulations, however, the Viennese scientists have now shown that different circuits can tolerate many faulty individual components: “It may be the case, for example, that the transistor at a very specific point in the electronic circuit has to be particularly fault-free, that at another transistor in the same circuit the tolerances are larger,” says Waltl. If this is the case, you could counteract this by using two different types of transistors with different error probabilities. For this reason, the researchers believe that computer models should be used as early as the design stage, which virtually depict circuit parts or entire circuits.