Changing memory

Materials World magazine
,
1 Dec 2017

Ellis Davies investigates the potential future of digital memory.

Digital memory has become unrecognisable since its invention in the late 1940s. Compared to the earliest forms of memory technology – punch cards and, later, mercury delays and cathode ray tubes – today’s tiny SSD cards and terabyte hard drives are exponentially more powerful, affordable and practical.

Moore’s Law – the observation that the number of transistors in a dense integrated circuit doubles approximately every two years to improve both cost and capacity – has held true since Gordon Moore’s 1965 paper and the invention of the integrated circuit in 1958. However, Moore’s Law could slow in the future. With the capabilities of modern memories and materials being tested, new methods and approaches may need to be adopted. Examples of this can be seen in recent developments, including a new material with ferroelectric properties and polymer sequences 100 times smaller than current hard drives.

Materials in memory

Looking back, materials have played a key role in the development of digital memory. Early examples include mercury delay lines – tubes of mercury around 1.5m in length – that could store around 1,000 bytes in a loop. Cathode ray tubes, which used phosphor to hold data, were employed around the same time (1940s). The main issue with these materials was cost, around £100,000 of mercury for 1,000 bytes, which would plague further developments.

In the late 50s, core memory emerged, as Martin Campbell-Kelly, computer historian and trustee of The National Museum of Computing, UK, told Materials World. ‘Core memory worked using small, magnetic ceramic toroids, which looked rather like polo mints strung on a tennis racket. This was a permanent memory, so it did not need constant recycling. It was very expensive – memory for a small computer typically cost around £20,000. Compared with the price of a house, it was phenomenally expensive.’ The early examples of cores were quite large physically, around a quarter of an inch wide, but later on the core become very small, almost impossible to see with the naked eye. They had to be assembled by hand, which added to the cost.

The invention of semi conductive memory (SCM) helped mediate the large costs associated with digital memory in the early 70s, and has led to modern day memory. SCM is significantly smaller than previous examples, and became easy to mass-produce once the original dyes were set up. Made from silicon transistors, SCM can be seen as the point that memory became commercially viable. ‘The reduction in the price of memory is really the miracle of the semiconductor industry,’ said Campbell-Kelly.

The key trend in the development of digital memory has been greater capacity, less money. SCM has been at the forefront of this move for more than 40 years, but researchers have begun to look to new materials to push the sector forward.

Molecular memory

Researchers at Linköping University, Sweden, have developed a new material that can be switched on and off using ferroelectric polarisation, which could make it possible to construct extremely small digital memories with very high information density. The material contains organic molecules that conduct electricity and contain dipoles – a dipole has one end with a positive charge and one with a negative charge, and changes its orientation (switches) depending on the voltage applied to it.

The dipoles in a thin film of the material can be switched at exactly the same time, which means that the film changes its polarisation. This is known as ferroelectricity – a characteristic of certain materials that have a spontaneous electric polarisation that can be reversed by the application of an external electric field.

Professor Martijn Kemerink of Linköping University explained to Materials World why the material has potential for digital memory. ‘The collective hysteretic behaviour of the dipoles creates a stable binary state – the bit can be 0 or 1 – and the electrical conductivity provides an easy way to read out that state. In a normal ferroelectric memory you need to switch the polarisation to know whether the bit was 0 or 1, in other words you had to destroy the information to read it, which is annoying as you have to re-write it every time you read it,’ he said. 

The problem with conventional ferroelectric memories is that there is no reliable way to statically read out or see the polarisation state. ‘Traditionally, the polarisation would be discovered by applying a large voltage, for instance positive,' explained Kemerink. ‘If the device was poled negatively, one detects a short current pulse related to the polarisation switching – if the device was already poled positively, nothing happens. After that voltage, one knows whether the polarisation is positive or negative, but the information has also been erased if it happened to be negative, so it has to be re-written by applying a negative bias.’ In the new material, information can be read-out using a small voltage. ‘The trick is that the small voltage is enough to measure the conductivity but not enough to change the polarisation. Therefore, there is no need for re-writing after read-out,’ highlighted Kemerink.

Trifluoromethyl and vinylidene fluoride were synthesised using size-exclusion chromatography in a solution of 5% aluminium chloride to create the final organic molecules – PBI-oVDF and Pc-oVDF. 

These spontaneously place themselves on top of each other to form a stack or a supramolecular wire, with a diameter of a few nanometres. The resulting wires can then be placed into a matrix, in which each junction constitutes one bit of information.

The synthesis of the molecules is currently too complicated for practical use, but the team aims to tackle this issue. Kemerink said, ‘Questions like on/off ratio, retention, magnitude of the conductivity and the underlying physics are [our key concerns]. The potential prospect of a single molecular stack acting as a memory bit is very inspiring.’ 

The size of the matrices makes them applicable for small digital memories with very high information density, largely for use in flexible electronics and other smaller devices.

One for the archives

Secondary storage devices such as floppy disks and external hard drives have also followed the trend of greater capacity, less money. From drum stores – a rotating drum coating with a magnetic iron oxide, where data was read and written like a tape recorder – to modern examples boasting 500–1,000GB in a small casing. With a large amount of data stored around the world, external archives are important for record keeping.

Taking a step away from the standard use of SCM, researchers at Charles Sadron Institute and the Institute of Radical Chemistry, France, are working on long sequence polymers as a possible alternative for archival data storage. The study shows the potential of polymers to provide data storage on a scale 100 times smaller than that of current hard drives, with a life expectancy of thousands of years. 

Previous studies have used biopolymers that can only be read in short chains, meaning fewer bytes of data can be stored. To read the information from biopolymers using a mass spectrometer (an analytical tool that ionises chemical species and sorts the ions based on their mass-to-charge ratio), the sequences need to be broken. Synthetic polymers therefore are better suited because their structure can be designed to allow in-situ breaks, whereas a biopolymer’s structure cannot really be controlled as it is fixed by evolution. 

Jean-Francois Lutz, Research Director at the Charles Sadron Institute, told Materials World, ‘We designed the polymer to break in a mass spectrometer. This makes the technique easy because we can use intact macromolecules with more that 80 monomers – containing around eight bytes – and have the full sequence read out. You cannot do this with biopolymers.’

The polymer sequences are made up of two kinds of monomers – either a propyl phosphate or dimethylpropyl phosphate and synthon. After every eight monomers, a molecular separator was added. Lutz explained the synthesising process. ‘We use automatic synthesisers to produce the bits of information, the monomers, one by one. We assemble chains by putting the monomers one after another, which is time consuming, usually in the range of a few hours. We imagine that this will decrease in the future,’ he said. For each monomer a chemical reaction is needed, followed by multi-reactions to build the sequence. 

To test the polymer, the team recorded the word sequence in ASCII code, which assigns a unique byte to each letter and punctuation mark. The word was decoded using mass spectrometry, which set a new record for molecule length read using this technique. 

These polymer sequences are best suited for archival use, storing films, books and other generic data that does not need to be accessed regularly. ‘The polymer is a powder and is therefore stable at room temperature. It does not degrade quickly, unlike hard drives that could crash or need replacing after a few years,’ said Lutz.

Digital memories have become part of our everyday lives, be it the computer you work at all day, or the terabytes of storage devices keeping more and more information in archives. SCM has championed the world of memory for close to half a century, but the future could hold a new era of storage.