Memory made on hafnia (shorthand for hafnium oxide) could become the semiconductor industry's biggest breakthrough in materials since it adopted silicon, per a report from SciTechDaily. Though hafnia has been difficult to use as memory until recently, a study shows that it is possible to make it work and reap many benefits. The report even claims it will be fast, efficient, and cheap all at once, but there's good reason to believe it won't excite on price.
On a chemical level, hafnia is relatively simple and is merely a single hafnium atom combined with two oxygen atoms. The property that makes hafnia desirable as a material for memory chips is that it's capable of being ferroelectric, which means it can be used as a non-volatile type of memory. Non-volatile memory can retain data even without a power source, and although today's silicon-based memory can be volatile or non-volatile, virtually all RAM today is volatile.
In a study by Sobhit Singh, the scientist explains how his team of scientists from the University of Rochester and another team from the University of Tennessee, Knoxville, created a form of hafnia suitable for memory. The issue with hafnia is that it's not ferroelectric by default. Before the latest experiments, it seemed the only two ways to make hafnia ferroelectric was to either make it some nanometers thick or alloy it with yttrium, which ended up reducing many of its attractive attributes.
The breakthrough in the Rochester-Tennessee study was to use "significant pressure" to create a ferroelectric form of hafnia, which remained ferroelectric even after the pressure was removed. The same technique could also be used for antiferroelectric hafnia, which is apparently also promising for data storage, but not many details were provided.
In contrast to DRAM and MRAM, which are volatile and magnetic forms of memory., Singh says, "Ferroelectric forms of memory are robust, ultra-fast, cheaper to produce, and more energy-efficient." However, that point about price is very questionable because hafnia certainly isn't cheap right now. A report from Reuters early last year detailed that hafnium (the key component in hafnia) has seen a nearly fivefold price increase due to increased demand since 2021, raising its cost from about $1,000 per kilogram to about $5,000. Even at $1,000 a kilogram, though, hafnium is by far more expensive than silicon, which measures in the tens of dollars per kilogram.
Even if hafnia were a cheap material, though, there's little reason to believe it would remain cheap if it ever became an important material for making memory chips. More demand pushes up prices as long as supply doesn't change, and we can certainly assume hafnia isn't being produced nearly as much as silicon and silicon-based materials. Though claims about performance, efficiency, and non-volatility may hold up, price will likely prevent hafnia-based memory from being the miracle semiconductor that reports have implied it could be.
Jobs eviscerated at Chinese Arm chip design firm in wake of restrictions from TSMC — lack of access to 7nm node could cause 150 employees to be laid off
Chinese chip firms say a new round of US sanctions won’t stop China’s chip industry — Chinese government responds with its own export restrictions anyways
