Real-Time Switching Dynamics in STT-MRAM

Abstract

In this paper a new experimental technique for measuring the switching dynamics and extracting the energy consumption of Spin Transfer Torque MRAM (STT-MRAM) device is presented. This technique is performed by a real-time current reading while a pulsed bias is applied. The switching from a high resistive state, anti-parallel (AP) alignment, to a low resistive state, parallel (P) alignment, is investigated as well as the impact of the cell diameter on the switching parameters. We demonstrate that preswitching and switching times and energies have a log-linear relationship with the applied voltage. Increasing the applied voltage leads to a higher spin torque on the free layer in a shorter time. This decreases the time needed to change the magnetization orientation of this layer, thus the time required before the switching occurs. We have also shown that for a given applied voltage, the smaller the cell the longer the time before switching. For low applied voltages, the preswitching time increases exponentially dominating the whole reversal time. The longer switching times can be explained by a lower Joule heating not sufficient to induce the thermally activated reversal process. This phenomenon is accentuated for smaller cells, where the heating is more significant and the time before switching is shorter than for larger cells.

Experimental analysis on stochastic behavior of preswitching time in STT-MRAM

“…This is coherent with the fact that the AP state is known to be less stable than P state [16,17]. We also notice that the activation energy slightly decreases with the applied voltage since less energy is required to reach the switch for high applied voltages [13]. For the P to AP transition at low applied voltage (VA=-0.45V), the activation energy is higher (above 4eV) showing a different mechanism, probably due to self-heating of the device as evidenced in [18].…”

“…This parameter depends on the applied voltage and temperature. A previous study [13] on the preswitching time for the AP to P transition at room temperature, showed a large variability on t , which seems random for a given V * . This paper will focus on demonstrating the stochasticity of both transitions.…”

“…Our characterization is performed on purpose to extract the switching parameters of the cell. The two switching operations, P to AP and AP to P, are obtained by respectively applying negative and positive bias voltage between -0.8V and 0.6V [13]. In this paper we show results carried out by quasistatic and dynamic characterizations.…”

“…In order to limit and verify the variability each measurement has been repeated thirty times for each voltage at each temperature. The experimental setup is thus composed of both Source Measure Units (SMU) and Waveform Generator Fast Measurement Unit (WGFMU) implemented in the Keysight B1500 semiconductor device analyzer [13,15].…”

“…This confirms that the P state is more stable, even at high temperature. From our previous work [13] and based on the experimental results exhibited in Fig. 5, for a given temperature the relation between the preswitching time and the applied voltage is given by the Equation (1) :…”