Shrinking ferroelectric RAM

2 min read

A clearer understanding of how shrinking the design of ferroelectric random access memory (RAM) affects its performance could lead to a range of better-performing products.

So hope engineers at Cranfield University and the National Physical Laboratory (NPL) who are beginning a three-year research project to develop new techniques for investigating ferroelectric RAM materials at the subnanometre level.

Ferroelectric RAM is different to other kinds of RAM because it uses a ferroelectric layer instead of a dielectric layer for non-volatile storage, which is computer memory that can retain stored information even when not powered.

Most dielectric RAM is constructed with a silicon layer or silicon dioxide layer, whereas ferroelectric RAM is usually made of functional material such as barium titanate or lead zirconate titanate (PLZT).

Ferroelectric RAM was originally developed for space applications at NASA because of its robustness and ability to withstand radiation, but it has been integrated into consumer products such as Sony’s PlayStation 2 in recent years.

Steve Dunn, a nanotechnology specialist at Cranfield University, said engineers would like to find more applications for ferroelectric RAM but this will require making the systems more dense.

‘It’s difficult to get many megabit RAM systems out of ferroelectric RAM at the moment,’ he added. ‘There is lots of research currently focused on trying to shrink down the footprint of each device to try and get more applications elsewhere.

'However, when you start to size constrain something, anomalous behaviour starts to appear.’

The scanning probe microscopes currently used to study functional material are not sensitive enough to visualise this behaviour on the subnanometre level. Dunn and colleagues at the NPL are working to modify the nanostructured tips of the probes to enhance the resolution at the surface of the material.

It’s a challenging task, Dunn said, because the nature of ferroelectric material makes it difficult to scan.

A ferroelectric RAM is based on a binary system made up of positively or negatively charged surfaces. Dunn explained these positive or negative surfaces, represented as 1 or 0, depend on the state of the material’s internal dipole and electric field.

‘When you go to smaller tips, you change your electric field and you change the interaction between the surface,’ he added. ‘The problem with a ferroelectric is that, when you get to a certain field level, the ferroelectric will spontaneously change its dipole orientation. It might do that locally and we might not be able to control that.’

Dunn said the solution to the problem will be to change the way the probes interact with the surface of the ferroelectric material.

‘We might try and use very small tips such as those of carbon nanotubes or other conductive graphite whiskers,’ he added.

The team's goal will be to reduce the tip ‘interaction volume’ from a 30nm radius to at least 2nm.

Dunn said this will give engineers an increased understanding of fatigue. ‘It will allow us to understand how you can make it do 1014 or 1018 cycles rather than only 1010 or 108 cycles,’ he added.

‘By having a look at how charge is injected down at very fine resolutions, we can start to see how the ferroelectric behaviour is changing in response to the electric field that we’re putting in.’

According to Dunn, the Cranfield and NPL team is hopeful it will be able visualise this in three years.

‘Once we understand what the problem is, we might be able to feed into some of the really exciting work that’s coming out of other areas in the world and the UK in actual device design and materials fabrication,’ he said.

Dunn added that the result should be electronic devices with longer lifespans, higher data densities and less toxic materials. ‘If you could have a product that lasts a bit longer and is easier to get rid of, then there is obviously a benefit,’ he said.

Siobhan Wagner