SWS2140 MTJ sample
Fitting parameter Good fitting finally, but how do we get it?
Sample Dimensions Row: A – U (1 - 21) Active sample: B – U (2 - 21) Measured: B – T (2 – 20) Change with shape and Area Column: 2 – 30 Active sample: 2 – 30 Measured: 2 – 30 Change with MgO thickness
Column vs. R A good start: R decreases with column number, but still noise. Each dashed line is for a different row. Now we need to filter out the resistance out side the trend. Therefore, we plot R vs. Column for every single row in next slide.
R vs. Column for each row Before filtering After filtering How to filter: (Use Row J as an example) Get the max and min Resistance of Row J, scale and plot R ~ Column.(left plot) Based on the plot, get rid of the max two resistances, then replot. (right plot) Now resistances of Row J scale nicely.
Good Resistance Data And if we fit R_middlevs 1 / Area, we will get the nice fitting now.
Column vsMgO thickness Here is the designed MgO thickness variation. The thickness varies from 0.73 to 1.29 nm of MgO across 150 mm, the diameter of the substrate. Column 30: position around 0 mm, Column 1: position around -12.7 mm (-0.5 in) t: thickness in nm
RA ~ MgO thickness From our work on test pieces, 0.8 nm MgO gives an RA product of 3.2 Ohm µm^2 and 1.10 nm MgO gives 30 Ohm µm^2. Log RA varies linearly with Mg0 thickness (see the attached figure).
Matching with experiments • Typical RA value (choose column 16): • Calculation: • Experiment: • RA changing with column: • Calculation: • Experiment:
Pulse results • Can find stable switching region. • Apply pulse with up to 2V. • Two ways of failure: • Break down – reason B • Change hysteresis loop – reason C
Die g3 Reason C: change hysteresis Reason B: break down
Die d3 Reason C: change hysteresis Reason B: break down
P26 – 60 x 180 H 1970 points have Rdc in between 300 tries each point 71 points Run 200 ps – 500 ps to check. Only consider > 800 as AP Only show # of pulse > 50 Clearer curve
P26 – 60 x 180 H Amplitude = 0.7 V