Macromolecular Memory

Transcription

1 Jason Sherrer Nanoelectronics Macromolecular Memory There are many different technologies being developed to replace present day computer memory, each with their different strengths and weaknesses. Currently it seems very unlikely that any of them are going to be able to out perform present day memory for quite some time as reflected in the current ITRS report. One of these new technologies with a slight chance is Macromolecular Memory also referred to as Organic Memory. It exhibits non-volatile memory characteristics meaning it doesn t need any electrical current passing through it to retain its data. Macromolecular Memory is a very broad term covering many different types of devices. Over the course of roughly a year and a half the Department of Materials Science and Engineering at the University of California published three different papers showing the results of their experiments with this new technology. None of these papers refer the same device; each is entirely different as far as its material selections and conduction mechanisms. Organic memory is still very much in the exploratory stage of its development to the point where some of the data recorded from these devices can t fully be explained. Macromolecular memory has some potential benefits which are worth continuing to look into. For example it has the potential for low cost due to large area fabrication by printing techniques, its mechanically flexible, and has the possibility to produce very small memory cells. To get an understanding of the kind of devices possible using organic materials, the following pages will explain how a select few work and provide a comparison between what is in development and what is already employed for non-volatile memory.

2 Before delving deeper into what Organic Memory is and how it works, lets look at what is currently being used. For many years now we ve used Electrically Erasable Programmable Read-Only Memory (EEPROM). A special type of EEPROM, known as Flash Memory is widely used because of its quick read and write times, as well as its cheap manufacturing. From a paper in 2005 by the University of Cambridge the listed read time is about 20 ns with a write time of about 10 µs. Flash memory is also very durable and can withstand between 10 4 and 10 6 rewrite cycles before structural failure. Furthermore it is capable of having a large memory density. For example a microsd card has an area of roughly 1.5 cm 2, is about 1 mm thick and can store approximately 16 GB of information. Flash memory is used in just about every portable electronic device purchasable nowadays. Just to name a few, a GPS, hand held gaming systems like the Nintendo DS, and Digital Cameras. Not only do these devices utilize it internally but they typically have a SD card slot to increase their memory even further. It is now possible to purchase a hard drive for a computer which uses flash memory technology. These are great for laptops; you don t need to worry about damaging the hard drive by moving it while its spinning and they are a bit more resilient to sudden impacts. By far the most common use of flash memory is the USB flash drive everyone owns. As mentioned earlier, macromolecular memory is still in the exploratory stage and so there are many different types of devices using a wide array of conduction schemes to fulfill the purpose of memory. University of California s initial device was a two terminal electronic bistable device back in The structure of this device, shown below in Figure 1, is simply a thin organic film pressed between two 200 µm thick

3 aluminum electrodes. The organic film is a combination of 8-hydroxyquinoline (C 9 H 7 NO), polystyrene, and gold nanoparticles protected by 1-dodecanethiol (CH 3 (CH 2 ) 11 SH). The memory cell had a total area of 200 x 200 µm 2. The cutoff Figure 1. Structure of the Al-polymer thin film-a1 memory device voltages for the electrical transitions and the current in the high-conductivity state were constant in both a vacuum and atmospheric environment; however the current in the low conductivity state showed a one order of magnitude difference being higher in the atmosphere and was attributed to water and other substances in the air. A write-readerase-read cycle, switching between the high and low conductance states while checking the states between changes, was conducted to determine the required voltages needed to perform these tasks as well as switching time in a normal atmosphere environment with results shown below in Figure 2. A voltage of 5 V switched the cell to its high conductivity state while the cell required -2.3 V to switch it to the low conductivity state. The reading phase performed to check the device s state required a voltage of 1.1 V across the cell. The resultant currents were in the range of 10-6 to 10-9 A depending on what part of the switching test was being performed. The device required a pulse length of 25 ns to switch from low to high. It is mentioned that the device may be faster than that and this measurement was restricted due to measuring equipment limitations. Due to

4 the speed at which the memory cell can switch between it s high and low conductivity states, it is believed it is caused by a field-induced electronic process. Figure 2. Results from Write-read-erase-read cycle. Top curve is the applied voltage and the bottom curve is the response current. W, R, and E stand for write, read, and erase on the bias curve while 1 and 0 on the current curve stand for the device in its high and low conductance states. During the analysis of this device, the specific method of conduction became a topic of debate and is believed to undergo two different methods. In the low state, the conduction is most likely from impurities within the device or hot electron injection. The high state conduction method is a bit of an enigma. The cell was set in its high conductivity state and plotted the resultant I-V curve for temperatures ranging between room temperature and liquid nitrogen temperatures, around 293 K and 75 K respectively. Results showed the I-V curve to be unaffected by temperature. Because of this temperature independence, the main conduction method in the high state is thought to be

5 a combination of Direct Tunneling and Fowler-Nordheim tunneling, supported by a paper in 2003 from Yale University. The paper explored the conduction mechanisms in alkanethiol monolayer devices, specifically the same dodecanethiol compound used to encompass the gold nanoparticles in this device. It looked at 4 specific ways tunneling can occur - Direct Tunneling, Fowler-Nordheim Tunneling, Thermionic emission, and Hopping conduction, of which only the first two show the same temperature independent I-V curves. The major difference between Direct Tunneling and Fowler-Nordheim Tunneling is Direct Tunneling is when there is a square barrier being pierced by the electron while Fowler-Nordheim is a triangular barrier. The only way to match the experimentally determined I-V curve was to combine the above mentioned types of tunneling. At lower voltages direct tunneling is the primary method of conduction and as the voltage increases, Fowler-Nordheim switches to become the dominant method. Combining the two in this way matched the cell s curve nicely shown in Figure 3. Figure 3. I-V curve for device in high state. The square points are the experimental results, the solid black line is the fit line combining Direct Tunneling and Fowler-Nordheim tunneling, and the dashed line is just Fowler-Nordheim tunneling.

6 The unique aspect of this device is the switching process. As mentioned above it is believed to be due to a field-induced electronic process because of the rapid rate, 25ns, at which it flips from its low to high state. It also suggests a cause for the different conduction states in the device. This all stems from the reaction between the gold nanoparticles encrusted in 1-dodecanethiol and the 8-hydroxyquinoline molecules in the organic film. In the low state there is little to no interaction between these molecules which explains there being few charge carriers and hence a very low current. As the device is switching into the high conductive state, an electron on the highest occupied orbital of an 8-hydroxyquinoline molecule may gain enough energy to break off and enter the gold nanoparticle. This requires the electron to tunnel through the 1-dodecanethiol. This results in the gold becoming slightly negatively charged because of the extra electron and the 8-hydroxyquinoline to be slightly positive. This creates more carriers and the current increases. As the device enters the high state, the main conduction method is charge tunneling between the molecules. It is speculated the electrons pass through the gaps in between the molecules instead of the molecules themselves because the gaps in the molecules crystalline structure are so much smaller than the gaps between two individual molecules. As said above as the bias increases Fowler-Nordheim tunneling becomes the dominant conduction method in the cell. It s possible the barrier between the 8-hydroxyquinoline molecules changes from a square to a triangle resulting in the change of tunneling styles however it is unclear.

7 Figure 4. Design of the Cu-organic layer-buffer layer-cu device About six months after the previous paper, another one from the same college concerning another organic memory device was published. However the method at which this device was created is different and it s suggested switching process is completely different. Firstly, this device has a different stack arrangement. Two copper plates, the anode and cathode, sandwich an organic layer and buffer layer and the entire thing sits on a glass substrate layer as is shown above in Figure 4. The 4nm thick buffer layer is a dielectric material like Lithium fluoride or Aluminum Oxide, but never specifies which for this particular device. The 100 nm organic layer is a combination of the following materials: 2-amino-4,5-imidazoledicarbonitrile (AIDCN), Tris-8- (hydroxyquinoline) aluminum (Alq3), and zinc 2,9,16,23-tetra-tert-butyl-29H,31Hphthalocyanine(ZnPc). Through observation it was determined the device enters the high conductivity state when 0.7 V are applied and switches back to the low conductivity state when 2 V are applied. The recorded currents for the high and low conductivity states were around 10-2 amps and 10-9 amps respectively. A voltage lower than 0.7 V with a resulting current of 10-3 amps can be used to read the device to determine its current conductivity state. It was also determined the device had a retention time of over 6 months in both states under no voltages and could continue to function completely at 110 degrees C for several hours. This device showed a peculiarity during the switching process; when the 0.7 volts was applied to switch the cell to the high conductivity state

8 there was a 28 ms lag time from when the voltage was applied till the device reacted shown in Figure 5. Switching from the high state to the low state happened within nanoseconds however. Figure 5. Device response to applied voltage. It took the device 28 ms to respond to the applied voltage and enter the high conductive state. This lag time encountered going from the low state to the high state can be explained by the switching mechanism. Basically, the switching occurs when copper ions metallisize the organic layer. In the off state the copper is bordering the organic film. During the switching process from low to high, the copper anode becomes partially ionized forming Cu+ ions. These ions migrate toward the cathode, penetrating the film in the process and cause metallization of the film. Once these ions reach the cathode a consistent distribution of Cu+ ions are within the film and the device switches. Therefore, once the voltage is applied it takes 28 ms for the ions to drift through the film and reach the cathode. An article from the IEEE Transactions on Electron Devices journal studied this copper drift phenomenon. It s noted that this doesn t just happen in organic films, but also in silicon. The rate at which the Cu+ can drift through the

9 material is largely based on the density of the cross-links in the chemical makeup. These cross-links are the bonds that connect the large chains of the polymer together. The more cross-links, the less spaces the Cu+ ions have to push through the material and so have a slower rate of diffusion while a material with less cross-links allow the Cu+ to diffuse faster. When switching the device off, the Cu+ ion flow is interrupted and the ions in the organic layer continue to head towards the cathode. Once this gap forms in the organic layer the device switches off. This happens so quickly because it doesn t take much time for this separation in the ions to occur. It is unclear why applying a higher voltage across the cell causes this interruption in Cu+ ion generation. All that is concretely known is in the off state the organic layer shows no traces of Cu+ ions. Below is a table comparing some of the crucial attributes for non-volatile memory. It can be seen that organic memory can t quite compete with our current forms of nonvolatile memory, especially in the data retention areas and endurance areas. If organic memory is going to become viable for large scale use it has a long way to go. Up to this point there has been no mention of an actual memory device using this technology aside from the very basic one bit cell in a lab. They also have a handful of additional problems. Aside from not currently being very competitive with flash memory they are difficult to assemble, have a low yield percentage of usable devices per batch, and the devices that are functional typically don t have consistent characteristics. They do however have the potential to become competitive because of their relatively fast speeds, potentially cheap to mass produce and ease of manufacturing. For example the thin film device can possibly be built in 3-dimensional arrays to vastly improve memory density. Another device created by the University of California is a three terminal organic non-

10 volatile memory cell which demonstrates a different way to determine if the cell is on or off. It checks the bias across a third terminal, which is unaffected by size of the cell. They theorize this would allow organic memory to be made much smaller then current memory and greatly increase its memory density. If macromolecular memory is going to work, it will need to take the strengths of these many different individual organic memory devices and compile them into one device. Organic memory still has a far way to go if it is going to be used in commercial devices, especially because it s biggest problems involve creating a cell that works and has the same functionality as another cell of the identical design. EEPROM Flash EEPROM thin film device cu ion device read time 50 ns 20 ns 25 ns ns range write time 1 ms 10 micros 25 ns 28 ms endurance 10^2 to 10^6 10^4 to 10^6 Not tested Not tested data retention > 40 years > 100 years Not tested > 6 months max voltage required 5 V 5 V 5 V 2 V max response current 10^-6 10^-2 Table 1. Compilation of some important numbers for memory devices

11 References J Ouyang, C.W Chu, C.R. Szmanda, L. Ma, Y. Yang. Programmable polymer thin film and non-volatile memory device. Los Angeles, CA University of California W. Wang, T. Lee, M. A. Reed. Mechanism of electron conduction in self-assembled alkanethiol monolayer devices. New Haven, CT Yale University L. Ma, Q. Xu, Y. Yang. Organic nonvolatile memory by controlling the dynamic copperion concentration within organic layer. Los Angeles, CA University of California J. He, L. Ma, J. Wu, Y. Yang. Three-terminal organic memory devices. Los Angeles, CA University of California S. Skorobogatov. Data Remanence in Flash Memory Devices. Cambridge, United Kingdom University of Cambridge A.L.S. Loke, J.T. Wetzel, P.H. Townsend, T. Tanabe, R.N. Vritis, M.P. Zussman, D. Kumar, C. Ryu, S.S. Wong. Kinetics of Copper Drift in Low-k Polymer Interlevel Dielectrics. IEEE Transactions on Electron Devices Vol. 46, Nov C. Woodford. How flash memory works. 4/23/2007