Stopping Resistance Drift in Phase Change Memory Cells and
Analysis of Charge Transport in Stable Amorphous Ge2Sb2Te5
Md Tashfiq Bin Kashem*, Raihan Sayeed Khan*, ABM Hasan Talukder, Faruk Dirisaglik and Ali Gokirmak
Department of Electrical and Computer Engineering, University of Connecticut, Storrs, CT 06269, USA. *Equal contribution.
Email: ali.gokirmak@uconn.edu
Abstract—We stabilize resistance of melt-quenched amorphous
Ge2Sb2Te5 (a-GST) phase change memory (PCM) line cells by
substantially accelerating resistance drift and bringing it to a
stop within a few minutes with application of high electric field
stresses. The acceleration of drift is clearly observable at electric
fields > 26 MV/m at all temperatures (85 K – 300 K) and is
independent of the current forced through the device, which is a
strong function of temperature. The low-field (< 21 MV/m) I-V
characteristics of the stabilized cells measured in 85 K – 300 K
range fit well to a 2D thermally-activated hopping transport
model, yielding hopping distances in the direction of the field
and activation energies ranging from 2 nm and 0.2 eV at 85 K
to 6 nm and 0.4 eV at 300 K. Hopping transport appears to be
better aligned with the field direction at higher temperatures.
The high-field current response to voltage is significantly
stronger and displays a distinctly different characteristic: the
differential resistances at different temperatures extrapolate to a
single point (8.9×10-8 ohm.cm), comparable to the resistivity of
copper at 60 K, at 65.6 ± 0.4 MV/m. The physical mechanisms
that give rise to the substantial increase in current in the high-
field regime also accelerate resistance drift. We constructed
field and temperature dependent conduction models based on
the experimental results and integrated it with our electro-
thermal finite element device simulation framework to analyze
reset, set and read operations of PCM devices.
I. INTRODUCTION
Spontaneous increase of resistance with time in amorphous
phase of phase change materials, known as resistance drift, can
cause erroneous inter-mixing of intermediate states in multi-
level cells and thus act as a bottleneck for denser storage in PCM
[1], [2]. A significant effort has been devoted so far to identify
the cause of drift and minimize it [3]–[8], however a
comprehensive solution has not been produced yet. There is also
a number of different transport mechanisms proposed for
amorphous phase change materials [9]–[13]. In this work, (i) we
experimentally demonstrate substantial acceleration of
resistance drift and stabilization of device resistance in melt-
quenched a-GST line-cells with application of high-field stresses
(> 26 MV/m) in 85 K to 200 K range, (ii) characterize field and
temperature dependent current conduction in stabilized devices,
(iii) construct a 2D temperature dependent hopping transport
model for low-field regime and extract the hopping distances,
hopping angles and activation energies associated with
percolation transport and (iv) construct an empirical model for
electronic conductivity in the high-field regime. We integrate the
resulting electric field and temperature dependent electrical
conductivity for stable a-GST with our finite element simulation
framework to model reset, set and read operations of PCM
devices.
II. DEVICE FABRICATION AND CHARACTERIZATION
GST line cells used for this study are fabricated by co-
sputtering from elemental targets on thermally grown SiO2 atop
Si substrates with tungsten back contacts, patterned using
photo-lithography and reactive ion etching, and capped by
Si3N4, as described in [14] (Fig. 1a). The dimensions of the cells
are: length (l) × width (w) × thickness (th) = ~600-700 nm ×
~70-150 nm × ~50±5 nm. The cells are first crystallized to the
hexagonal close packed (hcp) phase by annealing at 675 K and
then amorphized using a single 100 ns pulse with 50 ns rise and
fall times (Fig. 1b,c) in 85 K to 300 K temperature range in a
Janis ST-500-UHT cryogenic probe station under vacuum
(~0.01 mTorr). Pulse width and rise/fall times are chosen to
minimize reflections and parasitic contributions in the
measurement setup while ensuring amorphization of the cells
without substantial distortion of the waveforms. After the
amorphizing reset pulse, five DC I-V sweeps are performed at
each temperature using an Agilent 4156C parameter analyzer
(0 V to 25 V and back to 0 V in 0.1 V steps) with current
compliance set to 50 nA. The width of the devices in the
analysis and the construction of the models is ~152 nm.