Next generation memory
Flash memory remains the dominant type of non-volatile (that is, storing information in the absence of electric current) memory due to its widespread use in solid-state drives (SSDs) and USB flash drives familiar to everyone. But, despite its popularity and widespread use, the technology still remains problematic, especially if the production rate drops below the 30nm process - the flash memory speed decreases. In addition to this, there is a limited number of write-erase cycles and a relatively low write speed (in milliseconds). Due to all these limitations, researchers have long been looking for an effective flash memory replacement for the production of the above products.
Currently, there are several alternative developments that could well replace the silicon flash memory, such as PRAM (phase-change RAM), FeRAM (ferroelectric RAM), MRAM(magnetoresistive RAM) and RRAM (resistance-change RAM). However, to this day, scientists from various universities and companies have not been able to successfully apply the current technological standard in the production of memory using any of these technologies — either the mode switching mechanism or the platform itself loses efficiency and speed at the “nano” level. In addition, none of these developments lacks such important characteristics in commercial production as an increase in write-erase cycles (compared to flash memory), long-term data storage in the absence of current, and a high switching speed between read / write modes. It is the qualitative and quantitative growth of these indicators that is considered the main requirement in the development of non-volatile memory of the next generation.
And also, not for fun, the researchers plan to replace this technology as a whole. A joint team of scientists from Samsung and Korean Sejong University recently published an interesting publication in the journal Nature Materials , describing the new RRAM production technology (resistance-change RAM is a technology that allows you to change the cell voltage so that its state changes from low resistance (high conductivity) to high resistance (low conductivity)) of tantalum oxide (TaO x ), which in the tests showed a huge advantage over existing technologies, breaking the results on almost all counts.
RRAM-based devices operate as follows: at sufficient voltage, a material that functions under normal conditions as an insulator (high resistance state) switches to a low resistance state. The microcircuit itself, whose structure is multilayer (sandwich), sits on the main layer of tantalum oxide (TaO 2-x ), on which a thinner layer of oxide (Ta 2 O 5-x ) is applied , acting as an insulating layer surrounded by platinum electrodes. This configuration, known as MIMB (metal-insulator-base-metal), is an insulator that can switch to a high-conductivity state by changing its configuration to MMBM (metal-metal-base-metal). Interesting, isn't it?
Actually, the nature of the process of such switching has not yet been fully studied, but the authors of the study believe that the laying of highly conductive filaments extending through a layer of Ta 2 O 5-x oxide will lead to the fact that oxygen ions begin to move along them at a sufficiently high voltage the result of the redox process.
So, in the insulating (MIMB) state, what is between the platinum electrode and tantalum oxide forms the metal-semiconductor transition, also known as the Schottky barrier, while in a highly conductive (MMBM) state the same forms an ohmic contact. The main difference between them is that the current profile (depending on the voltage) is linear and symmetrical for the ohmic contact, but for the Schottky Barrier it is non-linear and asymmetric. The presence of a barrier is also advantageous, because does not allow wandering current through an array of several devices, which is important to ensure high data storage density.
The result of the above manipulations is a memory that, with a 30nm production process and a current of 50 microamps (which is lower than the requirements of one of the alternative technologies - PRAM), beats the current flash memory performance. It was shown 10 12 (in today's flash memory, this figure varies between 10 4 -10 6) write-erase cycles with a switching time of 10 nanoseconds and an information storage period of 10 years at a temperature of 85 degrees Celsius. This is a serious enough leap forward, compared with the flash memory that is so common today. Plus, it is more stable and functions without problems in a vacuum.
It is likely that all this is too sweet and smooth to be true. Here it’s worth mentioning right away that the tests were carried out in laboratory conditions on a die capable of accommodating 64 bits of information (in this terminology, it is 64 memory modules). Before gigabyte devices made using RRAM technology can appear on the market, several more years will pass.
In the entire semiconductor industry, in order to start mass production, it is necessary to make changes to the process of nanolithography, but in this particular case, it will be necessary to fully understand the mechanism for switching resistance states. But what the researchers showed as a result is impressive. If RRAM is brought to mind, it can be used as a universal memory, suitable for both storing information and producing RAM.
For help with chemical formulas, thanks to lesch
Nature Materials via ArsTechnica