PhD Alumni


Mantegazza Davide

Present position: Device Reliability Engineer - Intel Corporation
 

Thesis title:  Statistical analysis and modeling of programming and retention in Phase Change Memory arrays
Advisor:  Andrea Leonardo Lacaita
Research area:  Non Volatile Memories
Thesis abstract:  
In this Thesis, Phase Change Memory programming and retention are
investigated at array-level, showing the impact of (1) memory-array de-
sign and of (2) the technological cell integration scheme on programming
operation, the effect of (3) cells crystallization statistics on the program-
ming window and data retention capability.
Electrical cell parameters and characteristics are associated to each array
cell in order to support PCM statistical analyses, permitting to study
anomalous cells and intrinsic spread in the reset/set programming and in
data retention. From an array design point of view, the bit-line parasitic
resistance (voltage programming) and the bit-line RC delay significantly
reduce the reading window, thus suggesting optimized array design and
programming scheme. Cells with low reset capabilities due to process
impurities are electrically characterized and explained, permitting their
removal at process level. As regard cells programming, the reset opera-
tion is characterized by the critical quenching time (i.e. the maximum
reset pulse falling edge duration in order to avoid crystallization), the
set by the critical set time (i.e. the minimum set pulse duration re-
quired to crystallize the amorphous volume). Data retention (at a fixed
temperature) is associated with the critical annealing time (i.e. the min-
imum annealing time required to crystallize the amorphous volume).
These critical times are investigated at array level. The analysis of the
crystallization statistics indicates that the distributions of quenching,
set and annealing critical times are log-normal distributed, revealing no
anomalous crystallization statistics for the statistical range considered.
Notwithstanding the absence of anomalous crystallization behavior, cells
with lower critical time reduce the programming window and data re-
tention capabilities, due to the non linear crystallization characteristic
of the PCM cell. We further show that the quenching, set and annealing
critical times are correlated and their spread is not due to area variation.
We finally provide 1) an analytical, physics-based model for the quench
and set critical times correlation; 2) an analytical, empirical model for
set, reset and retention statistics and 3) a Monte Carlo, physics-based
retention model based on crystallization in presence of composition fluc-
tuations at the nm scale of the active material. The models allow pre-
dicting programming/retention distributions at multi Gb array level and
for scaled technologies.

Curriculum: