The Write Stuff
Researchers have developed an energy-efficient MRAM using a multiferroic heterostructure that could enable breakthroughs in AI and beyond.
No matter how fast computing technologies get, we always manage to find a way to push them beyond their limits. Today, artificial intelligence algorithms are pushing us to the ends of our capabilities more than anything else. But this is not just a problem, it is also a strong motivator for further innovation. Because we are running up against a wall, researchers in industry and academia are feverishly working to develop new technologies that will help us to break through this present plateau and go on to even bigger and better things.
One of the major components holding back forward progress is random access memory (RAM). We need higher-capacity technologies that are both faster and more energy-efficient to support cutting-edge applications. Magnetoresistive random access memory (MRAM) is a promising type of non-volatile memory offering advantages like fast operation, high storage capacity, durability, and compatibility with existing CMOS technologies. It stores data in the magnetization vector configurations of magnetic tunnel junctions (MTJs), rather than the transistors and capacitors of traditional RAM technologies.
Sounds great, right? So why is it not being deployed everywhere, you wonder? Unfortunately, current MRAM systems are not perfect. A large current is required to switch the MTJs during a write operation, which means MRAM draws too much power to be practical for most high-performance applications. That may not be the case in the future, thanks to some innovative thinking by a team of researchers at Osaka University in Japan. They have developed a far more energy-efficient MRAM solution that requires little energy for write operations.
The researchers achieved this feat by introducing a novel component that enables electric field-based data writing in MRAM. The key lies in a multiferroic heterostructure, which allows magnetization vectors to be switched directly by an electric field, eliminating the need for large currents. The effectiveness of the heterostructure is measured by its converse magnetoelectric (CME) coupling coefficient, with higher values indicating stronger magnetization responses to electric fields. While previous designs achieved a high CME coefficient, structural inconsistencies in the ferromagnetic layer hindered reliable magnetic anisotropy and efficient operation.
To address this, the researchers developed a new material configuration by inserting an ultra-thin vanadium layer between the ferromagnetic and piezoelectric layers. This insertion created a clear interface, enabling precise control of magnetic anisotropy and enhancing the CME effect beyond the performance of previous devices. The improved structure also achieved a key breakthrough: the realization of two distinct and stable magnetic states at zero electric field, enabling non-volatile binary data storage.
With potential applications in artificial intelligence, and any other area requiring energy-efficient, persistent, and reliable memory solutions, the team’s MRAM has the potential to make a big impact in the world of computing. But at this point, the technology has not yet emerged from research labs. Only time will tell if it proves to be practical for real-world applications.