Although we can’t pin down the exact date for the funeral, silicon chips are set to die out just as bipolar transistors did some 25 years before them. And with their death, the era of spintronics will truly be born.Let me explain.

If you compare transistors produced using today’s 65nm fabrication facilities with 22nm ones that will be the norm in about 10 years, you can easily see the exact dimensions of the transistor become increasingly variable as it gets smaller. The trouble with variable transistors is that they behave unpredictably, perhaps leaking more current, or taking longer to work than others on the same chip, and with a higher percentage failing completely.

Another problem relates to the size of atoms. A 22nm silicon process makes gate lines that are 13nm wide and with a physical gate length of 9nm. However, the insulating layers used are only 1nm or 2nm thick, or just a few atomic layers. So it’s clear that silicon lithography cannot be extended much further than 22nm.

Which brings us to the future. Whereas traditional electronic circuits transport charge carriers – electrons – through a conductor, spintronic circuits harness a different property of electrons to do more work with less effort.

It turns out that electrons spin in one of two directions – normally referred to as spin-up and spin-down, which correspond to the electron spinning clockwise or counterclockwise. This has been known for about 80 years but it required an additional discovery to make a useful device.

IBM research into nanoscale devices suggests that in 10 to 20 years spintronics could be used to make non-volatile RAM chips. It is speculated that far in the future an individual electron spin could be used to store one bit of data. Likewise, it seems spin could also be used to speed interfaces, boost processing power and possibly even underpin quantum computing.

Other scientists are working on an alternative to spintronics called molecular electronics (ME). In a nutshell, researchers found that individual atoms of some chemicals can be used to build electronic components. Dr Heike Riel of IBM’s Zurich lab said the idea is to integrate a rectifier, for example, into a molecule.

Again this work is at an early stage, but ME devices could harness individual molecules to make memory cells or transistors.

Recently IBM researchers made a single-molecule memory element that could retain its data for several minutes. Besides the difficulty of making the right chemicals, it turns out the hardest part of this was to connect wires to each end of the molecule. Now where did I leave my bulldog clips?