Let’s be honest, Moore’s Law may not be completely dead, but it will be on life support for a while. Gone are the days when performance could be improved “easily” by improving manufacturing processes or simply shrinking transistor size. Recently, R&D teams trying to improve the performance of semiconductors year by year must extend their reach not only to the transistor architecture itself, but also to materials engineering. Exploring the elements present in the periodic table could give semiconductors designs that are smarter, smaller and more energy efficient than silicon. Currently, the research team with the Stanford University School of Engineering is turned into one of the most precious metals on earth – Palladium – as a potential location for faster, cooler and more energy efficient memory architectures.
The research team’s approach is based on the premise that we have moved on from the Internet age and are already entering the AI age. According to Shang Wang, an engineering professor at Stanford University, Leland T. Edwards, the idea behind the study is “the edge of training locally on your home computer, cell phone, or smartwatch in case of a heart attack or similar. Enable the above AI.” detection or speech recognition. ”
Still, to make it happen, our current crop of tech can’t do it, he added. We need a type of memory that can be turned on and off in the press without a single product coming to market, a memristor. Memristor is a type of non-volatile memory that does not require constant current to retain data in the memory banks (check here for more information on memristor technology). Mahendra DC, lead author of this paper, explains this best. “We’re hitting a wall with current technology,” he says DC. “So we have to think about what other options we have.”
To break through this barrier, researchers singled out a compound of palladium, manganese palladium 3, as possessing the properties necessary to finally allow memristor designs to see the light of day. Part of the checklist concerns current semiconductor manufacturing technology. Materials being considered for new semiconductor designs must address the fact that tools and supply lines are rooted in the realm of silicon. Ideally, therefore, alternative compounds can be incorporated into current manufacturing processes without the need for extensive (and expensive) process changes.
Manganese Palladium 3 fits that requirement, but the most important property of this compound is the way its particles are manipulated into a memory device. The information here is obtained by manipulating the voltage, not from the voltage state as is the case with standard memory devices such as NAND flash. spin The orientation of electrons in a compound. Essentially, researchers can manipulate an electron’s magnetic field to make its center shuttle back and forth between the north and south poles. One of these spin directions (North-South or South-North) is then made to represent either 1 or 0, unlocking the binary number system behind computing as we know it. Called “spin-orbit torque magnetoresistive random access memory,” or SOT-MRAM for the necessary shorthand, this method of spin data storage has the potential to store data faster and more efficiently than current technology allows. and potentially store data at even higher densities.
Of course, all this research comes with one big caveat. Palladium is now almost as expensive as gold, but it is also a much rarer commodity. To make matters worse, his 40% of the world’s palladium is now mined in South Africa, and another 44% in Russia. Palladium availability itself can therefore be problematic from a logistical and geopolitical standpoint. Moreover, as demand for palladium for SOT-MRAM increases, its price will certainly rise further. All of these must be considered when pursuing a new transistor design.
Overall, the research is promising, but like all memoir star news of the past decade (and more), we still seem to be at the beginning of the proverbial yellow brick road. The future will tell us if we finally see the castle or not.
