Quest for a solution to drift in phase change memory devices

TL;DR

The work tackles drift in phase change memory by elucidating its origin in structural relaxation of the amorphous state and by exploring three complementary avenues: (i) onset and modeling of relaxation via Gibbs and collective relaxation pictures with Meyer-Neldel corrections, (ii) nanoscale confinement using melt-quenched monatomic Sb to probe confinement effects on stability, crystallization, and drift, and (iii) a projected memory concept with an interface-controlled projection layer modeled and validated to suppress drift. The main contributions include a refined collective-relaxation framework, demonstration of Sb as a scalable, melt-quenched amorphous PCM with confinement-enhanced retention, and a compact projected-bridge model that links drift suppression to interface resistance and amorphous-length scaling, supported by FEM and experimental fits. These results reaffirm that drift is fundamentally linked to structural relaxation and show that drift mitigation is achievable through interfacial engineering, device concepts like projection, and careful material/geometry choices, with clear implications for multi-level PCM and in-memory computing. The findings advance both fundamental understanding of drift mechanisms and practical design rules for future high-density, low-power PCM architectures.

Abstract

The goal of this thesis is to gain new insights into the drift phenomenon and identify strategies to mitigate it. An extensive experimental characterization of PCM devices and in particular drift forms the foundation of each chapter. With respect to time-scales, ambient temperature, device dimensions, and combinations thereof, drift is studied under unprecedented conditions. In three studies, different aspects of drift are examined. (1) The origin of structural relaxation: Drift measurements over 9 orders of magnitude in time reveal the onset of relaxation in a melt-quenched state. The data is used to appraise two models, the Gibbs relaxation model and the collective relaxation model. Additionally, a refined version of the collective relaxation model is introduced and the consequences of a limited number of structural defects are discussed. (2) Exploiting nanoscale effects in phase change memories: Scaling devices to ever-smaller dimensions is incentivized by the requirement to achieve higher storage densities and less power consumption. Eventually, confinement and interfacial effects will govern the device characteristics. Anticipating these consequences, the feasibility to use a single element, Antimony, is assessed for the first time. The power efficiency, stability against crystallization, and drift are characterized under different degrees of confinement. (3) State-dependent drift in a projected memory cell: New device concepts are aiming to reduce drift by decoupling the cell resistance from the electronic properties of the amorphous phase. A shunt resistor scaling with the amount of amorphous material is added. Simulations and the drift characteristics of a projected device put the idealized concept to the test. The contact resistance between the phase change material and the shunt resistor is identified as a decisive parameter to achieve the desired device properties.

Figures (37)

• Figure 1: Phase change memory - Basic write operations. A PCM device can be reversibly switched between a highly resistive amorphous state and a low resistive crystalline state. The amplitude and duration of electronic pulses determine the temperature profile created in the device by joule heating. The SET operation erases the amorphous state by heating the material to a temperature range that enables recrystallization on the order of nanoseconds. To RESET the device the phase change material is molten and melt-quenched with a sharp pulse.
• Figure 2: Exemplary threshold switching characteristic. At elevated fields, the resistance of the amorphous state is exponentially field-dependent. Beyond a threshold value (Vth) the material switches to a highly conductive amorphous on-state. After threshold switching, Joule heating becomes sufficiently large to recrystallize the amorphous state. Since the PCM device is operated in series with a resistor, the voltage drop over the cell reduces in the moment of rapid switching. The inset shows the transient device voltage and current traces. In the subthreshold regime, the amorphous phase is characterized with a source meter unit (cell voltage< [1]V). Threshold switching is captured with an oscilloscope.
• Figure 3: Two basic device concepts.a, Atomic force microscopy image of a bridge cell. The phase change material is patterned to a narrow stripe connected to two metal electrodes. The material volume between the electrodes can be molten and amorphized. b, Transmission electron micrograph of a mushroom cell. The bottom electrode (heater) is patterned to a small cylinder. The highest current densities are reached in the vicinity of the heater. Above, the heater forms an amorphous dome in a half-spherical shape.
• Figure 4: Multi-level programming. a, SET window. Square-shaped programming pulses with varying amplitude and duration were applied to a doped Ge2Sb2Te5 mushroom cell in an amorphous state (RESET). The color map shows the device resistance as a function of pulse power and duration. The peak of the temperature regime of fast crystallization and thus the fastest SET is achieved with a pulse power of around [314]$\mu$W. With pulse powers below [46]$\mu$W the amorphous 'ON' state (\ref{['fig:Intro_ThresholdSwitching']}) cannot be sustained (subthreshold range). b, RESET programming. Programming pulses with varying amplitude and trailing edge were applied to a doped Ge2Sb2Te5 mushroom cell in the crystalline state (SET). The device resistance measured after programming is plotted as a function of pulse power. With increasing power higher device resistances are achieved. Longer trailing edges result in a smaller device resistance. Compared to the box pulses studied in a, an even better crystallization can be achieved with long pulse trailing edges. Error bars denote the standard deviation of five measurements.
• Figure 5: Phase change memory applications. a, Memory hierarchy. To achieve high compute performance and large memory capacity multiple memory technologies are combined. PCM devices have been proposed as a potential technology for storage class memory. This new element in the memory hierarchy is supposed to bridge the gap in access time of almost three orders of magnitude between solid-state drives and random-access memory Burr2010Fong2017. b, Crossbar implementation of deep neural network computational workloads. The information propagation from one layer of a neural network to the next through the synapses is computed by a vector multiply operation. This computation can be executed on a resistive crossbar array. The synaptic weights (W) are mapped to the device conductance (G) and the inputs from layer j (x) are mapped to a read voltage (v). The column-wise summed crossbar currents (i) correspond to the product vector (b).