How sure is the future of spintronics?
In field of spintronics, we can often be very sure of ourselves about what role we play in the future of computing. MRAM, we say, will soon replace convenetional computer memory, domain wall storage will replace hard drives, and domain wall logic will replace the conventional computer processor.
But how in touch with the actual problems of the computing industry are we? Is success guaranteed? Or are these poster-children for the future of spintronics destined to go the way of magnetic bubble memory – to promise so much, but only achieve limited, niche success?
This review, published in Nature goes through the limits to improvements in computing. A look at the history of Moore’s law shows us that many impassable barriers have subsequently been overcome by technological improvement. Limitations in engineering, manufacture and device architecture have been by-passed. But could we now be enterring the last few miles of the “death march of Moore’s law”. The next 30 years of computing may be very different to the last 30.
Interconnects, rather than gates are the problem
I was surprised to discover that, for all the focus on transistors, and gates, it is interconnects – the nanoscopic cabling that connects the transistors – provide some of the greatest problems of today’s devices. For instance, it is not possible for a signal to travel across the entire processing chip within the speed of a single clock cycle (about 200ps). Indeed, you’d even struggle if you were travelling at the speed of light!
The author of this review believes that within the next 10 – 20 years Moore’s law as we know it will be over. Replacing it will be lower-powered computers which solve more specific problems, “the most successful computers are designed for the decathlon, rather than for just the sprint only”. Could this ‘more than Moore‘ approach benefit spintronics, which may struggle to compete directly with CMOS technology?
Can we deliver on the hype?
Many ‘future technologies’ are surrounded by hype which those in the industry know can’t be delivered on. Quantum computers, carbon nanotubes and photonics all come under fire for failing to deliver on scalability, fault tolerance and their interconnects. For the spintronics specialist there is some hope as “spin-states are particularly attractive because the promise high-density non-volatile storage”.
Whether the spintronics community can deliver on that promise remains to be seen, but this feature alone may be enough to maintain interest in the field for years to come.
For better or for worse, spintronics research is strongly tied to technological developments and device improvements. I think it’s important then that we do what we can to listen to what the real problems in computing are, and think about
what we and the field of spintronics can do to solve them.
Markov I.L. (2014). Limits on fundamental limits to computation, Nature, 512 (7513) 147-154. DOI: http://dx.doi.org/10.1038/nature13570