Better than NAND: fast, economical memory developed

It is the ferroelectric properties of hafnium in combination with other elements that make it a promising material for memory cards and SSDs.

Instead of addressing individual segments with counted electrical pulses, as is the case with NAND flash, for example, the semiconductor presented can change its polarization when an electrical field is applied. This principle enables much faster and therefore much more efficient access to the different areas of the read-only memory.

Only the bandwidth of the applied voltage was previously too low. By adding aluminum to the hafnium oxide used, this window for possible voltages could be extended from 2 to 10 volts.

This should make it possible to construct quad-level cells (QLC) that store 4 bits in each transistor. This means that a total of 16 different states can be stored in a single unit. With this architecture, several terabytes of data can be stored on the smallest chips. On a suitable SSD, this would ultimately be significantly more.

In addition to the much faster access to this information, the voltage required is also lower than with NAND flash. Instead of the previously required 18 volts, the system works with the aforementioned 10 volts. This also reduces the power requirement for writing and reading even further.

Initial endurance tests show how far practical development has already progressed. According to the study, one million accesses have already been completed with the individual cells. And it is no coincidence that the research is being co-financed by Samsung.

Only one problem remains, and that lies in the hafnium used. As promising as the electrical properties of the element are, which is used here as hafnium aluminum oxide, the element is rare. Although it is more common than gold, it is not actually a pure mineral.

Instead, it is found in tiny traces in other minerals and is therefore extremely difficult to mine. After all, only a layer just under 25 nanometers thick is needed per cell. The extremely fast, large and super-efficient memory could also become quite expensive if no substitute material is found by then.