Effect of Resistance Drift on the Activation Energy for Crystallization in Phase Change Memory
The crystallization properties of phase-change memory (PCM) in the presence of thermal disturbances are investigated with a novel micro-thermal stage. It is found that the recrystallization time due to thermal disturbances significantly varies depending on how the PCM cell drifts. The longer crystal...
Saved in:
| Published in | Japanese Journal of Applied Physics Vol. 51; no. 2; pp. 02BD06 - 02BD06-5 |
|---|---|
| Main Authors | , , , , , , |
| Format | Journal Article |
| Language | English |
| Published |
The Japan Society of Applied Physics
01.02.2012
|
| Online Access | Get full text |
Cover
Loading…
| Summary: | The crystallization properties of phase-change memory (PCM) in the presence of thermal disturbances are investigated with a novel micro-thermal stage. It is found that the recrystallization time due to thermal disturbances significantly varies depending on how the PCM cell drifts. The longer crystallization time is obtained following additional resistance drift, which can be described by an increase of the effective activation energy for crystallization. The possibility of achieving better retention in a PCM cell by allowing the PCM cell to drift for a longer time is demonstrated in this work. The activation energy changes at a rate of more than 1 eV/decade with varying time intervals below a second. As the ambient temperature gets higher, the effect of resistance drift on the crystallization process is diminished with respect to the dominant crystallization process which has a higher crystal growth rate at elevated ambient temperatures. |
|---|---|
| Bibliography: | The top view microscope image of the lateral PCM cell bridged with the MTS platinum (Pt) heater. The inset shows the 3D bird's eye view. The SiO x /SiN layer with thickness of about 1 μm is used for passivation between the PCM cell and the Pt heater. Electrical pulse profiles for RESET-programming, resistance reading, and MTS heating. Reading voltage ($V_{\text{READ}}$) is set at ${\sim}0.1$ V to prevent the PCM cell from being recrystallized due to Joule heating. $V_{\text{Pt}}$, the voltage amplitude of the 100 μs-long heating pulse, is changed to give different annealing temperatures ($T$). The time interval $t_{\text{p}}$ is fixed at 2 s otherwise mentioned. The effect of cell resistance drift on the crystallization time is taken into account by changing $t_{\text{p}}$. Crystallization due to thermal disturbances at $T \sim 270$ °C. Crystallization time ($t_{\text{crys}}$) is measured to be the cumulative heating time needed to form the first crystallization path within the amorphous region in the PCM cell. The inset shows the endurance characteristics with a high RESET/SET resistance ratio of ${\sim}1000$. (a) Crystallization time as a function of the time interval $t_{\text{p}}$. (b) Initial cell resistance ($R$) dependence of crystallization time for three different time intervals. The annealing temperature is maintained at $T \sim 260$ °C in both (a) and (b). The red solid line in (a) and the dashed lines in (b) are drawn as linear fits. More than 10 samples of crystallization time measurements have been averaged for each data point to give statistically reasonable values with 1-sigma uncertainties. Crystallization time as a function of temperature ($T$). The slope of the $t_{\text{crys}}$ versus $1/kT$ plot, which represents the activation energy, gets larger when we have a longer time interval $t_{\text{p}}$. The depicted lines are fits that show typical Arrhenius behavior, and the crystallization time was measured 5 times for each temperature to give statistical distribution. Change of the activation energy with different $t_{\text{p}}$ values ranging from 1 s to $10^{3}$ s, showing (a) $E_{\text{A}}$ versus $t_{\text{p}}$ and (b) $dE_{\text{A}}/dt_{\text{p}}$ versus $t_{\text{p}}$. Due to quite large statistical variations, the $E_{\text{A}}$ measurement was repeated about 20 times for each time interval. All depicted lines are guided for the eye. Activation energy $E_{\text{A}}$ increases for longer $t_{\text{p}}$, but the increase rate ($dE_{\text{A}}/dt_{\text{p}}$) of the activation energy gets smaller for larger $t_{\text{p}}$ values. Meyer--Neldel plot of measured Arrhenius parameters of activation energy ($E_{\text{A}}$) and pre-exponential factor ($\tau_{0}$) and fit with extracted parameters of $\tau_{00}$ and $T_{\text{MN}}$. $\tau_{00}$ and $T_{\text{MN}}$ are found to be $4 \times 10^{-4}$ s and 537 K. JMAK plot of $\ln\{-{\ln}[1 - x(t)]\}$ versus lnt for different activation energy values of 1.5, 2.0, 2.5, 3.0, and 3.5 eV at $T = 536$ (black), 517 (red), and 500 K (green). $x(t)$ is the volume fraction of the crystallized material at time $t$. |
| ISSN: | 0021-4922 1347-4065 |
| DOI: | 10.1143/JJAP.51.02BD06 |