Flexible PCM Addresses Switching Barriers

Article By : Gary Hilson

Stanford researchers have been using superlattice materials and flexible substrate to lower PCM (phase-change memory) switch power.

Combining superlattice materials and a flexible substrate may solve one of the key drawbacks of phase-change memory (PCM).

Already considered a promising candidate for data storage in flexible electronics, PCM’s high switching current and power remain a barrier to commercialized, high volume production of PCM. But researchers at Stanford University who came together working in separate areas have been able to demonstrate a lower switching current density of ~0.1 mega-ampere per square centimeter in flexible superlattice PCM, which is one to two orders of magnitude lower than in conventional PCM on flexible or silicon substrates.

Flexible substrate with memory devices (Courtesy: Stanford University)

As laid out in their recently published paper, the Stanford researchers enabled these characteristics by heat confinement in a superlattice material within a pore-type device on an ultralow–thermal conductivity substrate. The resulting PCM devices display multilevel capability with low resistance drift, indicating promise for emerging in-memory computing applications, as well as data storage flexible Internet of Things (IoT) electronics. Even without the flexibility benefits, the research has also yielded thermal engineering insight for conventional PCM on commercial, rigid silicon substrates.

In a briefing with EE Times, Eric Pop, professor of electrical engineering, along with PhD student Asir Khan and post-doc Alwin Daus outlined how the reduced switching current density is enabled by heat confinement in the superlattice material, assisted by current confinement in a pore-type device and the thermally insulating flexible substrate. “There are other groups that had worked on superlattice before, but it’s been very difficult to work with,” said Pope.

Khan was able to essentially use the superlattice of PCM to demonstrate the switching, while  post-doc Daus was able to bring his expertise in flexible electronics; the group set out to see if they could combine superlattice PCM and flexible electronics to make memory work on a flexible substrate. “Not only did it work, but it worked better than we expected,” said Pope. “Usually when you make electronics on flexible substrates, they usually tend to be worse because the fabrication process is poor because the lithography is coarse.” Efforts at deposition with superlattices go back 15 years, but conditions are tricky, he said. Once the Stanford team worked out the deposition conditions it was reproducible.

Daus said that although deposition has been previously demonstrated by other groups, strict temperature requirements make it difficult to replicate results. “The calibration of the recipe is not straight forward. You need your own tuning and your own toys.” That’s where Khan’s expertise came into play, he said.

Khan said the team’s success can be attributed to taking a different approach than previous work with superlattices by focusing on creating an architecture that would trap heat better, which was further enhanced by the flexible substrate. Even more interesting was that the PCM technology the team placed on a flexible substrate performed better than when it’s on a silicon substrate. “A flexible substrate itself provides thermal insulation for the phase change to occur at the lower power,” said Pope.

The potential for a bendable memory to be used in flexible electronics is attention grabbing, but the energy efficiency is just as significant, if not more so. “It has been 100 times more energy efficient than a conventional phase-change memory,” said Khan. “That’s a happy breakthrough from two perspectives.” Flexible electronics value energy efficiency, but there’s been little work done to build in a memory that’s also flexible, he said, let alone PCM that has the retention, stability, and low power. “The breakthrough we’ve made can equally be applied to non-flexible electronics as well.”

Flexible electronic devices use as paper, plastic, or metal foil as substrates in combination with active thin-film semiconductor materials such as organics or metal oxides or amorphous silicon. Advantages over crystalline silicon include thinness, conformability, and low manufacturing costs. From a real-world product perspective, the first thing that likely come so mind are foldable screens on smartphones that are starting to hit the market, but in additional to potential for flexible memory, flexible processors may also be feasible.

Composite image showing (left) flexible memory substrate held by tweezer and (diagonal) a sequence of the memory substrates while being bent (Courtesy: Stanford University)

A group within the semiconductor company Arm has now managed to implement one of the company’s smaller embedded designs using flexible silicon. Dubbed “PlasticArm,” Arm worked with PragmatIC Semiconductor to implement a version of its Cortex M0+ processor, a 32-bit processor that can execute a simplified subset of the Arm “thumb” instructions. The processors use small bits of memory known as “registers” to store data that’s being worked on. These registers are in a reserve section of the RAM. Like Stanford’s PCM, PlasticArm optimized for small sizes, low power use and embedded use cases.

One potential use case for Stanford’s flexible PCM memory would be a small RFID tag for storing spoilage data in food supplies or other manufacturing data for a variety of goods, said Pope, as they all require some sort of power source, logic, and processing power, a well as a small antenna transmitting the data that still needs to be data stored locally. “We haven’t found a lot of options for data storage on cheap plastic substrates, and certainly not data storage that gets better when you put it on the cheap plastic substrate,” he said. “This technology gets better because phase change memory benefits from thermal insulation plastic is an excellent thermal insulate. It’s essentially a funny win-win.”

This article was originally published on EE Times.

Gary Hilson is a general contributing editor with a focus on memory and flash technologies for EE Times.

Subscribe to Newsletter

Leave a comment