MAKING MEMORY
A product from a UK university could merge RAM and SSDs forever
The church (or pipe) organ is one of the oldest instruments still used in European music, and has a fascinating history stretching back to Ancient Greece and a chap called Ctesibius of Alexandria in 3rd century BCE. It then took until the 19th century for it to be displaced as the most complex man-made device by the telephone exchange. It’s also a metaphor we weren’t expecting to hear from a professor of physics talking about his new invention.
But this isn’t a story about church organs. It’s about quantum resonant tunnelling, III-V materials, and how they enable a new form of data storage device that could take the place of both RAM and flash storage inside future PCs.
Companies such as Intel, with Optane phase-change memory, and Samsung with its Z-NAND technology, have been threatening to shake up the world of memory and storage for a while, but an English physics professor has secured a US patent for what he and his team call
ULTRARAM. That’s storage that’s as fast as SDRAM, but it’s non-volatile like an SSD, so it keeps its data stored when you turn off the power, all while using one per cent of the energy consumed by the DDR4 in your rig.
“The way our memory works is the same way as flash, essentially, but in the first instance we’re trying to replace
DRAM,” says that professor, Manus Hayne of Lancaster University. “We think we can match every single specification of DRAM then in addition have a memory that’s non-volatile, plus it has a lower switching energy. It should have the same speed, the same endurance, but be more compact and cheaper than DRAM. But if the working memory in your computer is non-volatile, then that would actually change how the computer works. We’re not saying we’re going to replace NAND flash straight away, because it’s just extremely cheap, getting cheaper and the density is going up all the time. At the moment we don’t see that we can compete with that. But if you have a working memory that’s non-volatile, then it may be that you don’t actually need the storage locally. Things are increasingly going on the cloud, so you might wonder how much local storage you actually need.”
This vision of PCs (or more likely handheld devices in the short term) as dumb terminals whose processing and storage is outsourced to the cloud may horrify some, while giving a case of the warm fuzzies to those who like their devices sleek, cool and with enormous battery life. Advances in 5G connectivity suggest the bandwidth is almost there to make such a device a reality, while gaming would take the Stadia approach. Meanwhile, in the data centers that currently use about two per cent of total worldwide energy, a figure that is growing inexorably, the savings could be enormous as less power is used both for running and for cooling.
Extraordinary claims, of course, require extraordinary explanations. How come, in 52 years of silicon RAM chips, we’ve not done this before? The answer, of course, is that ULTRARAM doesn’t use silicon, preferring a compound III-V semiconductor that… we’ll let Hayne explain, “III-V semiconductors are the basis for opto-electronics, lasers for example in your DVD player or light in optical fibres for the internet. Barcode readers, laser printers, all those sort of things. The semiconductors that we’re using are a little bit unusual, they’re mostly used for infrared things such as gas sensing or IR cameras, and they are fairly exotic, but the nice thing about them is you can play around with combinations and do band-gap engineering for whatever material properties you’d like to have. And in this particular case what we like to have is a certain combination of those semiconductors originally identified by researchers at the TU Berlin [Berlin Institute of Technology] as the best choice for storage time. And into that we’ve designed triple barrier resonant tunneling.”
Triple what now? We’re about as far from church organs now as it’s possible to be, but don’t worry ecclesiastical music lovers, it’s coming. “[In an ULTRARAM cell] you have two barriers, with a small region in-between in which you confine your charge. It’s a bit like an pipe organ,” says Hayne. “The sound of the note depends on the length of the pipe because there’s a certain wavelength that fits in your pipe. When you’ve got very, very thin layers the wavelength of the electron is quantized, so you’ve only got certain energies that are allowed. When you try to send an electron through a resonant tunneling barrier, an electron of the wrong energy sees the total thickness of the barrier but if you put it in at a resonant energy to the confined energy inside, then the barrier is effectively transparent.” Got that? Good.
“By putting a voltage on,” he continues, “we can change the barrier from being opaque to transparent. That gets us extremely long storage times, like flash, but with very fast, very low energy, switching.”
PATENT PENDING
Resonant tunneling is a well-known quantum mechanical effect, if you’re a physics professor, but the diodes used by Hayne’s team are slightly different, “Normally you have two barriers and one well, where the charge is confined, but in ours we have three barriers and two wells, with the wells being different thicknesses, like they play different notes. We did that because, when the thing is just sitting there, it looks like a really, really thick barrier, but put the right voltage across you can get those two different notes to move so they’re aligned. This gives us more robust storage, but still gives us the resonant tunneling effect.”
Sounds great, when can we buy one? “We’ve got a US patent, with a European one pending,” says Hayne. “We’ve already had a collaboration with IQE, the world’s largest supplier of these compound semiconductor materials, and we’ve had interest from a very large company—but I’m not allowed to tell you who they are.” Intriguing.
Ian Evenden
RESONANT TUNNELLING IS A WELL-KNOWN QUANTUM MECHANICAL EFFECT