Tunnel Junction Memory Research Articles

Optimizing Write Current and Write Power Consumption in Magnetoresistive Random Access Memories by Using Generalized Asteroid Curves

Theoretical expressions for minimizing write current and write power consumption in magnetoresistive random access memories (MRAMs) are presented. The expressions are derived based on a generalized asteroid curve which approximately represents the minimum switching magnetic fields of magnetic tunnel junction (MTJ) memory cells recently developed for smaller write currents and better cell selectivity. The generalized asteroid curve has a shape parameter to fit the curve to the switching characteristics of MTJ cells. The derived expressions contain the shape parameter as well as other cell and array structural parameters, and thus give theoretical limits of the write current and power from physical points of view including MTJ shape.

Noise-Shaping Sense Amplifier for MRAM Cross-Point Arrays

A sensing technique using a voltage-mode architecture, noise-shaping modulator, and digital filter (a counter) is presented for use with cross-point MRAM arrays and magnetic tunnel junction memory cells. The presented technique eliminates the need for precision components, the use of calibrations, and reduces the effects of power supply noise. To obviate the effects of cell-to-cell variations in the array, a digital self-referencing scheme using the counter is presented. Measured experimental results in a 180-nm CMOS process indicate an RMS sensing noise of 20 /spl mu/V for a 5-/spl mu/s sense time. Further increases in sense time are shown to increase the signal-to-noise ratio. The current used by the sense amplifier and counter was measured as 10 /spl mu/A when running at 100 MHz or 10 mA when 1000 sense amplifiers are used with a memory subarray having 1000 bitlines.

Advanced HSPICE Macromodel for Magnetic Tunnel Junction

Magnetoresistive random access memory (MRAM) using a magnetic tunnel junction (MTJ) memory cell has the potential to revolutionize both high-density and high-speed memory applications with devices exhibiting non-volatility and good write endurance. This work presents a macromodel for an MTJ that is applicable to an HSPICE circuit simulator. The macromodel is realized as a five-terminal subcircuit that reproduces the characteristics of an MTJ including hysteresis and asteroid curves with thermal variation, and the R–V characteristics.

Magnetically pinned ring dots for spin valve or magnetic tunnel junction memory cells

Ni–Fe/Mn–Ir asymmetric ring dots with partially planed outer sides are investigated in order to confirm a method for obtaining pinned layers in magnetic memories with asymmetric ring shapes. Magnetic force microscopy revealed that the direction of vortical magnetization is pinned in Ni–Fe/Mn–Ir asymmetric ring dots despite the direction of the magnetic fields. This investigation shows that the Ni–Fe/Mn–Ir asymmetric ring dots can be applied to pinned layers in magnetic memories with asymmetric ring shapes.

Demonstrated Reliability of 4-Mb MRAM

The successful commercialization of MRAM will rely on providing customers with a robust and reliable memory product. The intrinsic reliability of magnetoresistive tunnel junction (MTJ) memory bits and the metal interconnect system of MRAM are two areas of great interest due to the new materials involved in this emerging technology. Time dependent dielectric breakdown (TDDB) and resistance drift were the two main failure mechanisms identified for intrinsic memory bit reliability. Results indicated that a lifetime over 10 years is achievable under the operating condition. For metal interconnect system, the initial results of Cu with magnetic cladding have met the reliability performance of typical nonclad Cu backend process in electromigration (EM) and iso-thermal annealing (ITA). Finally data retention is demonstrated over times orders of magnitude longer than 10 years.

Reduced master equation analysis of multiple–tunnel junction single-electron memory device

We employ a master equation (ME) approach in the charge transport analysis across a uniform multiple–tunnel junction (MTJ) memory trap, using a much-reduced state list derived from circuit symmetry, and previous assumptions by Jensen and Martinis. This enables all significant single tunneling and higher-order cotunneling sequences to be accounted for, while avoiding the computational cost of the full ME method. The reduced ME method is conceptually simpler and yields greater accuracy, compared with previous approximations based on tunneling probabilities. For an MTJ trap with zero stray capacitance C0, the results obtained are found to agree very closely with the full ME results up to a temperature of T≈3T0/10, where T0=e2/kBC, whereas previous methods break down at T≈T0/10. Furthermore, unlike the earlier methods, the reduced ME approach can be applied to the realistic but less symmetric case of a trap with finite C0, and remains valid up to the trap’s maximum operating temperature of T≈T0/100. Finally, our reduced ME results are in close agreement with available experimental data at T<T0/200.

The magnetic stability of spin-dependent tunneling devices

The tunneling resistance between two ferromagnetic metal layers that are separated by a thin insulator depends on the relative orientation of the magnetization M of each layer. In a memory device, independent switching of the magnetically soft layer is achieved by making the other layer either exchange-biased or magnetically hard. The repeated reversal of M of the soft layer by field cycling can demagnetize the other magnetically hard layer and thus erase the tunnel junction's memory. The M of exchange-biased structures was stable at least to 10(7) cycles, whereas in hard structures, M generally decayed logarithmically with the number of field cycles. The decay was very sensitive to the thickness of the hard layer and the composition of the soft layer. However, no decay was observed when the moment reversal was accomplished by coherent rotation, establishing that demagnetization results from the formation and motion of domain walls in the soft layer during field cycling.