An Exploration Into Mram Technology

Computers have become an imperative part of not only our everyday lives but also our methods of understanding nature through mathematics and physics. One of the most fundamental ideas behind them is the concept of storing binary data which the computer translates into useful information. This has been made possible by giving the computer 'memory'. With many years of research and development the best physical model we have is a device bearing the abbreviation 'RAM' - Random Access Memory - a combination of circuitry that allows data to be stored and assimilated in an electronic device. With the constant advancements in physics, many improvements have been made, increasing capacity, speed and durability, with one of the most groundbreaking ones being MRAM. Based on the magnetoresistive effects found in ferromagnets, this design can be the next major step in vastly improving computer performance.

Technology is constantly advancing; when it comes to computers the rate at which it is doing so can sometimes be overwhelming, but it never fails to spark interest and excitement. More importantly each new step into the world of computing advances many other branches of science, such as physics and mathematics. When it comes to computing, one crucial aspect that is constantly under focus is the ability of a computer to store information. The more information a computer can intake at a time the better; it allows us to compute harder calculations, examine mass amounts of data faster and overall, allow the computer to do more at once. The hardware unit responsible for this is called RAM - random access memory - the name comes from the fact that the information stored can be accessed in any order. RAM modules appeared on the market for the first time in 1951. Many improvements were made which resulted to other types of RAM being created, a crucial one being DRAM (Dynamic RAM). In the late 1980s a new era of data storage began to form after the discovery of the "giant magnetoresistive effect" made by Albert Fert and Peter Grünberg, who won the Nobel Prize for it in 2007 [1] After more than twenty years of research and development, the Magnetoresistive Random Access Memory technology is finally beginning to steal the spotlight from DRAM.

In general RAM units (an integrated circuit) consist of millions of miniature capacitors and transistors. Each capacitor holds one bit of data (either a 0 or a 1) in the form of electrons. To store a 1, the capacitor is charged. To store a 0, the capacitor is discharged. However, a problem occurs with this process. Once a capacitor is charged, the charge begins to leak away which leads to loss of data. The way around this is to constantly keep charging the capacitors storing a 1. This is done by either by the CPU unit or a 'memory controller'. The data is read by one of these devices and immediately re-written. The capacitors discharge themselves in a matter of milliseconds, which means the whole process of reading and re-writing has to take place thousands of times per second, slowing down data accessibility. This continuous process is where Dynamic RAM gets its name from. The role of the transistor is to tell the circuit whether to change the state of the capacitor or to read the capacitor. Many improvements have been made to DRAM. Most notably a model called DDR SDRAM (Double Data Rate Synchronous Dynamic RAM) which utilises the idea that the

CPU of a computer needs data which is in a sequence. Instead of constantly switching rows and columns it stays on a row until it has finished checking all the memory cells in each column that corresponds to that row [2]. The process behind MRAM is entirely different. Where conventional DRAM relies on electron charge, MRAM exploits electron spin using a setup called 'spin valves' [3]. The device consists of two ferromagnetic plates, sandwiching an insulating layer, often called a 'tunnelling barrier', which is only a few atoms thick. One of the plates is a permanent magnet with a specific polarity. The other can be altered so that it matches an external field, which allows it to store memory. These plates, with the insulator, are sometimes referred to as Spin Dependent Tunnel Junctions. According to the predominant electron spin in permanent plate, the junction produces a large resistance. A current flows through the whole junction, changing the resistance. The non-permanent plate can have its polarity changed by changing the magnetic field around it; this, in turn, changes the spin on the electrons allowing for 1s and 0s to be stored. Data is read in the form of resistance. If both plates have identical electron spins then the resistance is low, which produces a 0. If the spins are opposite, resistance is high and the reader reads a 1 [4, 5, 6]. The system is displayed in Figure 1. 3 Comparison with mainstream RAM systems The current mainstream unit used in most computers is DRAM. It offers fast read and write speeds, high density (storage capacity per area) and unlimited endurance. However, its main drawback is that it is a volatile system; when the device is not being powered it does not retain any information. This means that start-up times for computers are not remarkably fast. Another disadvantage is that the 'cell leakage' is very high. As discussed previously, the capacitors in each cell discharge very quickly, meaning the system has to spend time constantly refreshing the capacitors which is detrimental to the speed. Other types of RAM exist which attempt to tackle such problems, the most notable being SRAM (Static RAM).

SRAM has many advantages over DRAM. Its design features multiple transistors for each capacitor which allows the information to stay for much longer in the capacitor. However, due to the more complicated design, the cell density is typically low. This is why it is usually used for specific tasks rather than a general source of RAM [8]. SRAM can also refer to the term 'Shadow RAM'. Shadow RAM's main purpose is to improve the performance of hardware by copying the specific code, which is in ROM (read-only-memory) and utilizing it with RAM instead. This process is called shadowing and though it limits the RAM available to other applications, it can improve the speed at which certain hardware works because the code related to each "shadowed" device will load faster through RAM [9]. Flash RAM or Flash memory is an electronic device. It was developed from EEPROM (electrically erasable programmable read-only memory) and it has some interesting advantages. Data can be reprogrammed in blocks instead of individual bytes. This allows for very flexible updating of information. It is used in modems and BIOS which often need their protocol to be changed. The problem with flash memory is that endurance is rather limited; each unit has a limited amount of read/write cycles that can occur, making the option incongruous for general purpose computers [10]. Although SRAM performs at speeds faster than MRAM, its cell density is quite low which makes it unsuitable to be the next main RAM component of modern computers. MRAM excels in every category; its non-volatile spin valve system allows for very flexible reading and writing of data, with recent improvements allowing write speeds of up to 1 nanosecond. What isn't shown in the table is power consumption. MRAM is significantly less power hungry than DRAM, an important factor to consider when needing mass amounts of RAM (for large systems).

The properties of MRAM clearly outshine the current competitors. However, there are some constrictions in its path to the mainstream which most would consider asinine. Current RAM manufactures usually construct their products with the latest motherboard and chip infrastructures in mind. In simpler terms, when a dominant motherboard or chip manufacturer produces a product, it creates guidelines for the manufacturers of additional components. For a RAM module to work, it must be made compatible with the more important component (e.g.: motherboard); the important factor here being the infrastructure of the coding. Therefore, RAM manufacturers are reluctant to begin production of MRAM since they either have to spend money to make it compatible with current generation technology or wait until the next generation of circuitry comes into fashion. Without being part of such company it is easy to think that the reasons presented are inadequate when compared to the potential of this technology, but from a capitalist point of view the only sound option is to wait until conventional motherboards and chips are ready to begin a change of phase into next-gen circuitry, simultaneously making other components compatible too.

MRAM addresses many issues with current data systems but this does not mean that other methods should not be investigated. At the moment, the two main alternatives, which are not just purely experimental, are FRAM (Ferroelectric RAM) and PRAM (Phase-change RAM). FRAM's properties can be seen in Figure 2. It performs relatively well in most categories - most importantly it is non-volatile - but it still has room for much improvement. It works using a ferroelectric film of lead zirconate titanate, often abbreviated as PZT. Under the influence of an electric field the Zi/Ti atoms can change polarity. Due to the crystal lattice nature of the PZT film, the polarity remains the same after the electric field is switched off, making the system non-volatile and power efficient [11]. PRAM uses a few different concepts together and utilises them in a very efficient way. Its read/write method is similar to MRAM: it reads resistance levels to denote 1s and 0s. However changing the resistance works completely differently. It uses a 'chalcogenide' chemical, meaning the compound contains at least one element from the chalcogen/oxygen group and one other electropositive element. When the material is in an amorphous state the resistance is high, denoting a 0. When the material changes state to poly-crystalline the resistance is low, which signifies a 1. This is another non-volatile system, capable of speeds faster than some types of Flash RAM. Until recently, its limitations were endurance - only about 100 000 cycles - now, a company called Micron claim they have achieved cycles of up to a million. PRAM is mostly used in mobile phones and portable devices because it allows for very fast start-up times, and it can be made to be compact whilst still retaining a good memory capacity. The limits on its endurance however make it somewhat unsuitable for more complex machines; a problem which rival companies like Samsung and IBM-Hynix are working on [12].

The scientific ideas and discoveries behind MRAM have managed to deliver technology that is potentially revolutionary. The evidence already at hand show that MRAM excels in every category when it comes to performance. With such high potential the question of why this development is not receiving more funding arises. The current constrictions that disallow MRAM to enter mass production are only almost only a matter of time; when dealing with billions of dollars, companies find it hard to make the first move - and why should they - their current strategy (DRAM) is working well. Once MRAM is developed to existing cell density standards it will most likely be sought after by most manufacturers. The importance of its implementations is much deeper than a simple case of improving modern computers (or profits, depending where you stand). Memory is a fundamental part of computers; improvements will lead to progression in many branches of physics other sciences. The development and research of this technology itself promotes funding into physical and mathematical theories that can revolutionise all sorts of other technological concepts.

References 1. Professor Jon Goff, "Special Lecture - Spin Valves" 2. Jeff Tyson, Dave Coustan, "How RAM Works", pages 1,2 and 4 http://computer.howstuffworks.com/ram.htm Review Article - P P Gyurov 5 3. IBM-Stanford Spintronic Science and Applications Center, "Magnetoelectronics and Spintronics & SpinAps" 4. R.C. Sousa, I.L. Prejbeanu, "Non-volatile magnetic random access memory", http://www.crocus-technology.com/pdf/MRAM_CR_v5a.pdf 5. Anton Pankratov, "Magnetoresistive Random-Access Memory (MRAM), http://www.chipdip.ru/en/video/id000307650/ 6. NVE Spintronics, "How MRAM Works" http://www.nve-spintronics.com/mram-operation.php 7. Honeywell, "Honeywell Introduces MRAM" http://www.youtube.com/watch?v=_MVxpj_iFhM 8. Jeff Tyson and Dave Coustan, "How RAM Works" pages 3 and 4 http://computer.howstuffworks.com/ram.htm 9. Microsoft, "Shadow RAM Basics" http://support.microsoft.com/kb/78528#rtDisclaimer 10. Webopedia, "Flash Memory" http://www.webopedia.com/TERM/F/flash_memory.html 11. Ramtron, "What is F-RAM" http://www.ramtron.com/about-us/what-is-f-ram.aspx?AspxAutoDetectCookieSupport=1 12. Chris Mellor "Look out, Flash! Phase-change RAM IS HERE ... in Nokia mobiles" http://www.theregister.co.uk/2012/12/17/micron_pcm_asha/