In this podcast episode, MRS Bulletin’s Laura Leay interviews Eric Pop, Xiangjin Wu, and Asir Intisar Khan from Stanford University about their work building a phase-change memory superlattice at the nanoscale. They created the superlattice by alternating layers of antimony-tellurium nanoclusters with a nanocomposite made from germanium, antimony, and tellurium (GST467). Each layer is ~2 nm thick and the superlattice consists of 15 periods of these alternating layers. The microstructural properties of GST467 and its high crystallization temperature facilitate both faster switching speed and improved stability. The device operates at low voltage and shows promise for high-density multi-level data storage. This work was published in a recent issue of Nature Communications.
In this podcast episode, MRS Bulletin’s Laura Leay interviews Eric Pop, Xiangjin Wu, and Asir Intisar Khan from Stanford University about their work building a phase-change memory superlattice at the nanoscale. They created the superlattice by alternating layers of antimony-tellurium nanoclusters with a nanocomposite made from germanium, antimony, and tellurium (GST467). Each layer is ~2 nm thick and the superlattice consists of 15 periods of these alternating layers. The microstructural properties of GST467 and its high crystallization temperature facilitate both faster switching speed and improved stability. The device operates at low voltage and shows promise for high-density multi-level data storage. This work was published in a recent issue of Nature Communications.
LAURA LEAY: Welcome to MRS Bulletin’s Materials News Podcast, providing breakthrough news & interviews with researchers on the hot topics in materials research. My name is Laura Leay. Modern life and scientific advances are both becoming ever more reliant on data. Aspects such as big data, high-performance computing and artificial intelligence are driving a demand for robust, energy-efficient memory. New research has led to the development of a type of computer memory that has low power requirement, can be stable for years, is nonvolatile which means that it doesn’t need to be continuously powered in order to store data, is high density, and fast. The development involves phase-change memory, where a material sandwiched between two electrodes changes between amorphous and crystalline states. Prof. Eric Pop from Stanford University explains.
ERIC POP: Phase-change memory is a type of memory or data storage in which we encode the data in the change in resistance, or rather actually, in the resistance of a material. Most memory structures that we’re familiar with such as in our phones, laptops, and so on typically store charge on very, very tiny capacitors. So in phase-change memory we’re actually changing the structure of a small resistor. This is literally a nanoscale resistor.
LAURA LEAY: To build the phase-change memory, a superlattice is created by alternating layers of antimony-tellurium nanoclusters with a nanocomposite known as GST467 which is made from germanium, antimony, and tellurium. Each layer is around 2 nm – or just a few atoms – thick and the superlattice consists of 15 periods of these alternating layers. The unique microstructural properties of GST467 and its high crystallization temperature facilitate both faster switching speed and improved stability. The nanoclusters serve as a precursor for crystallization and contribute to the fast switching speed of GST467.
ERIC POP: Some of the ideas came from our collaborators at University of Maryland. This is Heshan Yu and Ichiro Takeuchi. They’re the ones who essentially discovered this germanium-antimony-telluride 457 compound. Our contribution was essentially taking that unique composition that they had discovered and putting it into our superlattices. We said, not only can we put into a superlattice but we can shrink the lateral dimensions of that superlattice down to 40 nm. Nobody had demonstrated any kind of superlattice technology with 40 nm lateral dimension.
LAURA LEAY: Collaborators at the Taiwan Semiconductor Manufacturing Company supplied the 40 nm bottom contacts to create these nanoscale devices and the National Institute of Standards and Technology, NIST, contributed to some of the analysis of the structure. Remarkably, these tiny devices simultaneously exhibit low resistance drift, good endurance, and fast switching. First author Xiangjin Wu from Stanford University explains the significance.
XIANGJIN WU: It’s been long-known in this memory technology that there is a trade-off between the thermal stability and the speed that you’re able to operate the memory device. It’s sort of what nature gives us; there’s no free lunch for most materials combinations. But in our technology we’re able to, sort of, “break” this trade-off, or even, like, push this trade-off to the better corner with the combination of the superlattice and nanocomposite.
LAURA LEAY: The superlattice enables efficient switching. To effect a phase change in the material, and hence change the resistance, a pulse of current is passed through the material. This causes localized heating and the heat is confined by the layers of the superlattice. The superlattice is also vital for minimizing low-resistance drift which is where there is a gradual change of the resistance states over time.
ASIR INTISAR KHAN: We’ve been working on the superlattice for a while, like the last 5-6 years or maybe even beyond that. But this work has reached the peak of any of our previous work in the sense that having a low power technology is great, having a faster memory technology is great, having a memory technology that can switch millions of times is great, but if you really want to have them together in one memory device and want to see the best of all these worlds – not like, one world, two worlds, many different worlds – and want to combine into one world that has the best scenario; it is really challenging. We cannot say that this is the ideal case but at least this work shows the peak of all those metrics dumped into one cell. After all these years of development, based on our knowledgeable work, we were able to get to a point where we really see all these things coming along together still in one device.
LAURA LEAY: That was Dr. Asir Khan, visiting post-doctoral scholar at Stanford University. Xiang explains some of the challenges the group overcame to arrive at this remarkable breakthrough.
XIANGJIN WU: Individual layers are only five atoms thick, for example antimony-telluride. How to, you know, controllably deposit all these layers without them, you know, becoming intermixed, or like, growing into each other, that’s one of the challenges we overcame. So we did, you know, a lot of material optimizations like the growth temperature, how much sputtering power goes into optimize the material layering, and also what is the growth pressure. So these are some of the challenges that we overcame.
ERIC POP: We’ve demonstrated operation below 1 volt. Getting any kind of memory to work at 1 volt or below is good news. The reason is that modern computer processors all operate at about 1 volt or less and many types of memory don’t work well at 1 volt. You can see how resistance changed with voltage and you can see some major changes happening below 1 volt. And these are repeatable and reversible. So that’s one thing. The other thing that I wanted to mention, we essentially show 8 resistance states which are completely distinguishable. They’re essentially storing 3 bits of data in the same physical memory cell that’s a much higher density than can be done with other technologies. The important part here is that the resistance states are distinct and they are stable with respect to time.
LAURA LEAY: So, the research has led to a device that is smaller than anything created so far, it involves a new way of combining materials, shows improvements in tradeoffs between speed and stability, operates at low voltage, and shows promise for high-density multi-level data storage. The research shows how fundamental materials science can interface with circuit and systems engineering to eventually create a new product.
ERIC POP: It’s a very exciting time because we’re actually seeing this potentially scale up into systems. And I think, once there’s academic demonstration this can be scaled into systems and it’s really worth doing it for things like artificial intelligence applications. Our main job is to essentially show the technology value to the world and then maybe convince companies that it’s sufficiently interesting to pursue. Where we sit, we see both sort of like the really fun materials questions but I think we also sit in a place where, because we interact with computer scientists and with industry we sort of see their needs and what they need to make this work.
ASIR INTISAR KHAN: Some of these industries are looking for phase change material that can withstand a state for years even at an elevated temperature like 150°C. In this work we showed that, while we have the GST467, there are, of course, other materials combinations within the stack and by tweaking the other material layers in that stack, for example titanium-telluride, we can engineer or tweak the retention temperature so we can use some of those thermally stable materials within the stack that might help the technology to be operated even at – I think we showed – up to 130°C the ultimate goal is to achieve 150°C, there must be some magic materials combination.
LAURA LEAY: This work was published in a recent issue of Nature Communications. My name is Laura Leay from the Materials Research Society. For more news, log onto the MRS Bulletin website at mrsbulletin.org and follow us on twitter, @MRSBulletin. Don’t miss the next episode of MRS Bulletin Materials News – subscribe now. Thank you for listening.