Thanks to a breakthrough in the field of magnonics, EPFL researchers used uncharged magnetic waves to transmit and store data, rather than conventional electron streams. This discovery may solve a dilemma for energy-intensive computing technologies in the big data era.
Similar to electronics and photonics, magnonics is an engineering subfield that aims to advance information technology in terms of speed, device architecture, and energy consumption. Magnons correspond to the specific amount of energy required to change the magnetization of matter through collective excitations called spin waves (visualized above).
Because magnons interact with magnetic fields, they can be used to encode and transfer data without electron flow with energy loss due to heating of the conductors used (known as Joule heating). As data speeds and storage demands soar, energy loss is an increasingly serious barrier for electronics, as Dirk Grundler, head of the Lab for Nanoscale Magnetic Materials and Magnonics (LMGN) in the School of Engineering, explains. It is
“With the advent of AI, the use of computing technology has increased so much that energy consumption threatens its development,” says Grundler. “A major problem is the traditional computing he architecture that separates the processor and memory. The signal transformations involved in moving data between different components slows down computation and wastes energy.”
Known as the memory wall or von Neumann bottleneck, this inefficiency has led researchers to search for new computing architectures that can better support the demands of big data. And now Grundler thinks his lab may have stumbled upon such a “holy grail.”
LMGN PhD student Korbinian Baumgaertl, during other experiments with commercial wafers of the ferrimagnetic insulator yttrium iron garnet (YIG) with nanomagnetic strips on its surface, discovered that precision-engineered YIG nanomagnetism Inspired by the development of the device. With the support of the Center of MicroNanoTechnology, Baumgaertl was able to use radio frequency signals to excite spin waves in the YIG at specific gigahertz frequencies and, importantly, flip the magnetization of surface nanomagnets. .
“The two possible orientations of these nanomagnets, representing magnetic states 0 and 1, can encode and store digital information,” Grundler explains.
The road to in-memory computing
The scientists made their discovery using a conventional vector network analyzer that sent spin waves through a YIG nanomagnet device. Reversal of nanomagnets occurred only when the spin waves reached a certain amplitude, which could be used to write and read data.
“We can switch magnetic nanostructures using the same waves that we use for data processing and show that we can also have non-volatile magnetic storage within the exact same system,” Grundler explained. ” refers to stable storage. Data can be stored for long periods of time without consuming additional energy.
This ability to process and store data in the same place has the potential to change the current computing architecture paradigm by ending the energy-inefficient decoupling of processor and memory storage, achieving what is known as in-memory. is given to this technique. calculation.
Optimization on the horizon
Baumgaertl and Grundler have published breakthrough results in the journal Nature Communications, and the LMGN team is already working on optimizing their approach.
“Now that we have shown that spin waves write data by switching nanomagnets from state 0 to 1, we need to tackle the process of switching again, known as toggle switching,” says Grundler.
He also notes that, in theory, Magnonics’ approach can process data in the terahertz range of the electromagnetic spectrum (for comparison, current computers work in the slower gigahertz range). However, this has to be experimentally demonstrated.
“The potential of this technology for more sustainable computing is enormous. With this publication, we hope to increase interest in wave-based computation and introduce more young researchers to the growing field of magnonics. I hope to attract to
Original: Magnon-based computation could mark a paradigm shift in computing
Than: Lausanne Federal Institute of Technology
