The ideal memory is low-cost and allows data to be stored and retrieved in no time. It should also be non-volatile, which means no information is lost in the case of power failure. Semiconductors companies race to find a solution that would combines it all in a single device: the universal memory. Recent research suggests a new approach that may change the game.
Memories for real-time application
Non-volatile memories have extremely diverse applications, ranging from electronics to automation, to healthcare monitoring. This represents a growing market of order 10 to 100 B$, stimulated by the rise of the Big Data and the Internet of Things.
Among the different types of available memories, one can distinguish two categories. On the one hand, extremely fast but expansive memories are used for micro-processing and short term storage. On the other hand, cheap, non-volatile memories are used for storage. Slow access to the stored data is becoming the bottleneck for reaching higher speed data manipulation. One is thus looking for a so-called universal memory that would fit every purpose at low cost.
In recent years, the MRAM (for Magnetic Random Access Memory) has emerged as a seducing candidate and has attracted the attention of the biggest players in the field including Samsung, IBM, Toshiba and ST Microelectronics. One such approach is pursued at industrial scale by Everspin Technologies, based on ferromagnetic materials. The most recent chips have access times of just 30 ns and a data retention exceeding 20 years is claimed. Importantly, such memories can be written and read electrically, which represents a key point for integration and scalability. Yet the solution is not perfect. Reading and writing a MRAM requires a current to flow through the memory cell. The process is rather energy inefficient and results in heating and thermal instability.
A new approach based on antiferromagnets
In a collaboration led by the group of Dr. Denys Makarov (Helmholtz-Zentrum Dresden-Rossendorf e.V.), researchers from Germany and Switzerland suggested a promising new approach based on an insulating magnetoelectric antiferromagnet. Their prototype device is made of a thin film of Chromium oxide (Cr2O3), where data is saved, with Platinum electrodes on top and below to write and read the information. The writing is done by applying an electric field through the cell, which results in no energy dissipation in the device. And there is more to it. Compared to ferromagnets, antiferromagnets offer greater stability, better prospects for high-density integration and faster operation speed. But this comes at a price: they are hard to read, especially when relying on electrical means.
The team of Dr. Makarov has recently developed an original measurement approach that allows reading the magnetic state of the storage layer with an extreme sensitivity. The idea is to measure a transversal electrical resistance in a nearby electrode. Using a four terminal geometry combined with an active offset compensation scheme, the team could obtain direct access to the surface magnetization of the storage layer.
Now for the first time, all the ingredients are combined in a single device. The result is a non-volatile memory prototype, which can be entirely manipulated and read electrically.
Imaging magnetic domains with color centers in diamond
The work, which is published in Nature Communications, includes images of the magnetic domains of the storage layer. Those were recorded using the diamond based imaging technology, which has been developed by Prof. Maletinsky at the University of Basel and QNAMI. Dr Makarov says “this high-resolution microscopy data was essential to back up key assumptions on the magnetic state of the Cr2O3 surface. No other experimental technique could have provided such information”.
Further work will prove that such devices can be operated at high speeds, which will then certainly trigger commercial interest for the solution.
- Purely antiferromagnetic magnetoelectric Random Access Memory,
T. Kosub et al, Nature Communications 8, 13985, (2017) (Open Access)
- Room-temperature antiferromagnetic memory resistor,
X. Marti, Nature Materials 13, 367 (2014).
Technology and markets:
- A glimpse on the non-volatile memory market forecast
- Non-volatile memory and its use in entreprise applications
- Viewpoint on the antiferromagnetic RAM