The evolution of magnetic data storage technology has been impressive since 1955, when IBM built the first hard disk drive featuring a storage capacity of 5MB with an areal recording density of 2 kbit/in2. The steady progress made in this area over the last fifty-five years has been driven by three figures of merit: the increase in storage density, the increase in speed of data processing, and the decrease in fabrication costs. The barrier of 100 Gbit/in2 has already been passed in 2002 [1, 2] and demonstrations with area recording densities as high as 1 Tbit/in2 are expected in the near future based on perpendicular recording technology [3, 4]. It is apparent from these data that the storage of an information bit is related to nanometer scale magnetic thin film structures. Moreover, magnetic patterned media with a single-bit-per-island recording methodology has also been considered as recording media and has successfully passed preliminary tests [5, 6]. In addition to the interest in the areal storage densities, a special emphasis is placed on the data rate of the hard disk drive. The current disk drives operate at a maximum internal data transfer rate of approximately 130 MB/s, which corresponds to a channel data rate of 1.17 Gbit/s (using an 8/9 modulation code). Therefore, the writing time for a single bit, or equivalently magnetization reversal time, is below 1 ns.
Another major research effort in the magnetic data storage has been recently devoted to magnetoresistive random access memory (MRAM), the first commercial version being produced by Freescale Semiconductors at the end of 2006. MRAM has the potential to store data at a high density, to access them with a high speed, and to be a low power consumer [7, 8]. Once the performances in these directions become comparable with the ones of the semiconductors based memories, the nonvolatility property could determine the use of MRAM as a ‘universal memory’. The most common design for MRAM uses a magnetic tunnel junction: two ferromagnetic thin films play the role of electrodes and a thin tunneling barrier separates them. The resistance of the tunneling junction is significantly modified as the magnetic moments of the ferromagnetic layers change their relative orientation. The difference in junction resistances corresponding to the stable parallel and anti-parallel orientations, respectively, makes possible the definition of the binary memory states. The ferromagnetic thin film electrodes have nanometer dimensions and the magnetization reversal time in these devices is in nanosecond regime.
In spite of the impressive progress over the years, the paradigm of magnetic data storage is now approaching its fundamental limits for areal storage density, as well as for speed in data processing [9, 10]. In a magnetic memory nano-cell, the thermal fluctuations may drive the magnetization to other energy minima and consequently, the stored information is lost. This thermal fluctuation induced phenomenon, the so-called superparamagnetic effect, is increasingly pronounced once the particle’s dimension is decreased and represents a key limitation to further improve the storage density of hard-disk-drives and magnetic random access memories [9, 11].
Consequently, there is an urgent need for reliable alternatives to current magnetic recording techniques, which should feature sub-Stoner-Wohlfarth reversal magnetic fields and sub-nanosecond magnetization reversal time. Several unconventional magnetic recording techniques, such as spin polarized current assisted recording [12-14], precessional switching [15, 16], toggle switching [17, 18], heat assisted recording [19, 20] are currently under intense research efforts. The project director has made significant contributions in characterizing the first three alternative recording techniques with respect to critical fields and switching times [21]. He has also addressed multiple challenges related to the design of magnetic field pulses suitable for recording. However, these results have been derived by completely neglecting the thermal effects. One of the main goals of the current proposal is to study the thermal stability of novel storage designs and the influence of thermal noise, as well as other types on noise, on the alternative recording techniques mentioned above. A special case is represented by heat assisted recording that tries to benefit from thermal fluctuation induced reversals by heating locally the nanocell of interest.
In addition to the interest in the disruptive influence on noise and fluctuations, a special emphasis of this project is placed on the constructive effects of noise in hysteretic systems. Since noise represents a nuisance effect in linear systems and in many nonlinear electronic devices, the potential benefits of noise seem rather counterintuitive and have been overlooked by researchers for a long period of time. However, the recent studies on stochastically driven nonlinear systems proved that such phenomena are quite common and their applications range from signal processing (dithering effect) to climate models (ice age) [22-23]. Based on several cases recently discussed in the literature [24-25] and several preliminary results that we obtained [26], it is expected for noise to play a constructive role in various hysteretic systems, activating some kind of resonance response. This phenomenon is generally known as coherence resonance, when is solely induced by noise, and stochastic resonance, when an external oscillatory signal is present. In conclusion, the stochastic analysis developed by this project is also aimed at exploring noise induced resonances in nonlinear systems and at deriving the rigorous conditions under which such resonance phenomena are possible in spintronic and semiconductor devices.
A third direction of interest for our project is the analysis of noise and fluctuation effects in nano-scale semiconductor devices. The prevalent trends in silicon electronics area have been: to increase the speed and operational frequency of analog and digital components, to decrease the power consumption of the electronic circuits, and to incorporate higher functionality on a single silicon chip. In principle, these can be further approached by scaling down the semiconductor devices and, particularly by developing ultra short channel MOSFETs [27-28]. However, it is well-known that such small devices will be very susceptible to random doping fluctuations that are inevitably present due to the random nature of ion implantation and diffusion, as well as to the fluctuation of geometrical characteristics from one device to another. These random fluctuations from one device to another or even within the device itself lead to significant fluctuations of threshold voltage, as well as of terminal and frequency response of MOSFET devices. In addition, the impact of electrical noise on the performance of semiconductor devices is increasingly pronounced with the decrease in device size [29]. For these reasons, accurate statistical analyses of random doping, geometrical fluctuation, and noise induced effects are very important for further progress in the area of semiconductor device technology [30].