Phase Change Materials: Memory, Optics, and Glass Viscosity
Phase change materials (PCMs) have been instrumental in various technologies, from outdated optical discs to cutting-edge memory and optical computing. Their utility stems from a significant change in optical and electrical properties when transitioning between amorphous and crystalline states.
Rewritable CDs and DVDs
Early applications of PCMs include rewritable CDs and DVDs. These discs feature a laminate structure: a polycarbonate substrate, followed by a phase-change material layer sandwiched between two dielectric layers. These dielectric layers protect against oxidation and harmful interactions with the plastic substrate, especially during the high-temperature processes (around 600°C) required to melt and quench the material into an amorphous state.
The PCM used in these discs is typically Indium Antimony Telluride (AIST), a modified composition of antimony telluride with added silver and indium to modulate crystallization kinetics. A metal reflective layer above the PCM captures optical power, reducing the laser power needed for writing. An additional resin layer and polycarbonate complete the disc structure.
Data recording involves a focused laser beam. A longer, lower-power pulse induces crystallization, while a short, intense pulse melts and amorphizes the material. The crystalline state is more reflective, and the amorphous state is less reflective, allowing for the encoding of digital information (ones and zeros) based on reflectance changes. While this technology is now largely obsolete, it marked the first widespread application of PCMs in data storage.
Phase Change Memory (PCM)
Phase Change Memory (also abbreviated as PCM) represents a more modern application. The historical development of PCM began in the 1960s with Ovshinsky's discovery of threshold switching in chalcogenide compounds, a phenomenon crucial for PCM device operation, leading to these devices sometimes being called "ovonic devices."
In the late 1960s, Gordon Moore (of Moore's Law fame) developed the first PCM device, capable of storing 256 bits. While seemingly small, this demonstrated the potential for exponential growth in memory density, akin to Moore's Law. The 1980s saw a breakthrough by Yamada's group in Japan, who discovered that Germanium-Antimony-Telluride (GST) compositions exhibited fast, reversible switching, enabling the field's commercialization in the 1990s and early 2000s.
PCM Structure and Operation
A basic PCM cell consists of two metal electrodes with a layer of phase change material in between. Current passed through the electrodes generates Joule heating, which changes the material's structural state. Reading the stored information is done by measuring electrical conductivity: the crystalline state is several orders of magnitude more conductive than the amorphous state due to electron delocalization. The same electrodes can be used for both writing (applying high current) and reading (applying a low voltage to measure resistance).
Threshold Switching
A key challenge in PCM is passing current through the highly insulating amorphous state to crystallize it. This is overcome by "threshold switching." When a high voltage is applied to the amorphous state, exceeding a certain threshold, the material suddenly becomes electrically conductive. This is a purely electronic phenomenon, not involving atomic rearrangement, and is fully reversible.
The electronic mechanism involves the intermixing of states around the Fermi surface within the band gap of the amorphous semiconductor, creating empty states below and fewer states above the Fermi level, thus enabling conductivity. This allows a large current to flow, leading to Joule heating and crystallization (the "set" regime). If the material is already crystalline, a high current pulse followed by rapid cooling can amorphize it (the "reset" regime).
Comparison with Other Memory Technologies
PCMs are non-volatile, meaning data persists even after power is removed, unlike DRAM. This is because data is encoded in the material's microstructure, which doesn't require continuous energy to maintain.
A comparison chart of various non-volatile and traditional memory technologies (Memristor, STT-RAM, DRAM, Flash, Hard Drive) highlights PCM's characteristics. While such charts can be biased (e.g., by the developer of a specific technology), they reveal key trade-offs:
- Density: Continuously improving with each generation.
- Energy per bit: PCMs tend to have high energy consumption due to the heating required for phase change.
- Read time: State-of-the-art PCMs can achieve very fast read times (e.g., 2 nanoseconds) with low-voltage pulses.
- Write time: Crystallization limits write speed (around 50 nanoseconds for GST), but faster alloys are being developed.
- Retention: Can exceed 10 years.
- Endurance: Reported up to 10^11 to 10^12 cycles in literature.
3D XPoint Technology
Intel's 3D XPoint technology, developed in partnership with Micron, is a notable example of 3D stacking of PCM. Although Intel never officially confirmed it, reverse engineering suggests it uses GST-like phase change materials.
The 3D XPoint architecture, also known as a crossbar, uses perpendicular wires to individually address memory cells at each intersection. To overcome "sneak path" currents (where current flows through unintended cells), a "selector" is placed in series with each memory cell. This selector, often a chalcogenide compound like Germanium-Arsenic-Selenium, also exhibits threshold switching. By setting the operating voltage correctly, only the intended cell's selector turns on, preventing current leakage through other paths. This transistor-less, stackable design allows for high density.
Intel's initial goal for Optane (based on 3D XPoint) was to bridge the gap between fast, volatile DRAM and dense, slower NAND flash. However, despite its technical achievements, Optane faced market challenges:
- Evolving Competition: DRAM became cheaper and higher capacity, while SSDs (NAND-based) significantly improved in speed, narrowing Optane's niche.
- Cost: Manufacturing challenges kept Optane products expensive.
- Ecosystem: Lack of specific chipsets, hardware, and software optimization limited adoption.
- Business Decisions: Intel's shift in focus led to the winding down of the Optane business.
Phase Change Materials in Optics
Beyond data storage, PCMs are gaining traction in optics, leveraging their dramatic optical property contrast between amorphous and crystalline states. Unlike most materials where atomic rearrangement causes only minor changes in refractive index, PCMs exhibit a huge jump (e.g., a change of 3 for GST 225) due to changes in bonding configuration (covalent to resonant/metallic). The crystalline phase also becomes more optically absorbing due to increased electrical conductivity.
Optical Memories and Analog Computing
This property contrast enables optical memories integrated into photonic circuits. A small patch of GST on an optical waveguide can switch between absorbing (crystalline) and transparent (amorphous) states, encoding digital information in transmitted light intensity. Researchers have demonstrated storing multiple bits (e.g., 3-bit, 5-bit, 6-bit) in a single cell by controlling the fraction of crystalline vs. amorphous material.
One promising application is optical analog computing, particularly for AI workloads. Traditional Von Neumann architecture suffers from a "memory wall" bottleneck due to extensive data transfer between CPU and memory. In-memory computing, where memory itself acts as a computing unit, can mitigate this. Optical memories can perform operations like matrix multiplication and accumulation (MAC) directly. By encoding input vectors as light intensities at different wavelengths and passing them through PCM cells (whose transmittance represents weights), the sum of products can be obtained automatically by a single photodetector, offering extremely fast, scalable computation.
Reconfigurable Optical Media
PCMs also enable reconfigurable optical devices. Unlike conventional optics with fixed properties (e.g., a lens with a fixed focal length), PCMs allow post-fabrication tuning. Examples include optical switches for light steering.
Another application is dynamic color displays. A thin film of PCM on a dielectric stack can change color (e.g., pink to yellowish) when switched between amorphous and crystalline states due to light interference. By patterning small, individually switchable pixels and interlacing different color-switching pixels, full-color reflective displays can be achieved. Similar to e-readers like Kindle, these displays are non-volatile and extremely power-efficient, offering color electronic paper.
Transparent Phase Change Materials
While many PCMs exhibit increased absorption in the crystalline state, some, like certain Ge-Sb-Se-Te alloys, can switch between two transparent states while still showing a large refractive index change. This opens up new applications where transparency is crucial, such as reconfigurable lenses that can shift focus by changing the phase of light at different locations.
Viscosity of Glass: Liquid or Solid?
The question of whether glass is a liquid or a solid is complex, depending heavily on temperature and the timescale of observation. While intuitively glass seems solid, its rheological behavior can blur the lines.
Viscosity vs. Elasticity
- Elasticity: Describes instantaneous and transient deformation. Strain is proportional to stress (Hooke's Law). Materials immediately deform under force and recover when the force is removed. It arises from changes in interatomic spacing and the balance of attractive/repulsive forces. Elastic constants show weak temperature dependence.
- Viscosity: Describes time-dependent flow. Shear stress is proportional to shear strain rate (Newtonian fluid). It represents internal friction, where deformation requires molecules or atoms to move past each other. Viscous flow only occurs under shear motion. Viscosity is highly temperature-dependent, as molecular motion is thermally activated.
The "Pitch Drop Experiment" and Medieval Windows
The famous "pitch drop experiment" demonstrates that even seemingly solid materials like pitch (asphalt) can flow over extremely long timescales due to very high viscosity. Similarly, the urban legend of medieval church windows being thicker at the bottom due to glass flow is often cited. However, this is largely a myth; historical glass manufacturing processes often produced uneven panes.
Modern claims, such as Hitachi's silica glass memory storing data for 300 million years, raise questions about glass flow over geological timescales. Understanding viscosity is crucial to address these claims.
Temperature Dependence of Viscosity
Viscosity is exponentially dependent on temperature, often described by the Arrhenius equation. This is because molecular motion, which underlies viscous flow, is a thermally activated process. Materials with stronger intermolecular interactions tend to have higher viscosity and higher activation energies.
The Angell Plot: Strong vs. Fragile Liquids
The Angell plot illustrates the temperature dependence of viscosity for different glass-forming liquids. The horizontal axis is inverse temperature normalized by the glass transition temperature (Tg), defined as the temperature at which viscosity reaches 10^12 Pascal-seconds (or 10^13 Poise). All glass-forming liquids converge at this point by definition. The vertical axis is logarithmic viscosity.
- Strong Liquids: Exhibit Arrhenius behavior, meaning viscosity follows an exponential dependence on temperature, appearing as a relatively straight line on the Angell plot (e.g., silica, germania). Their activation energy doesn't change dramatically with temperature, suggesting the microscopic deformation mechanism remains consistent.
- Fragile Liquids: Show a much more dramatic change in viscosity with temperature, deviating significantly from Arrhenius behavior (e.g., ionic or molecular liquids). Their activation energy is temperature-dependent, implying that the size and mobility of molecular clusters involved in flow change with temperature.
Fragility is quantitatively measured by a "fragility parameter," which is the slope of the curve on the Angell plot at Tg.
Practical Implications of Fragility
For glass processing, strong liquids are generally preferred. Fragile liquids' viscosity changes much more dramatically with temperature, making process control (e.g., for flow and deformation) highly sensitive to temperature fluctuations. Strong liquids offer more stable processing windows.
Microscopically, strong liquids maintain their atomic network even at high temperatures, leading to consistent deformation mechanisms. Fragile liquids, however, form smaller, more mobile clusters at higher temperatures, reducing the activation energy barrier for flow.
Quantitative Description: VFT and MYEGA Equations
- Vogel-Fulcher-Tammann (VFT) Equation (or Williams-Landel-Ferry (WLF) Equation for polymers): A semi-empirical modification of the Arrhenius equation used for fragile liquids. It includes an additional parameter (TVFT) to capture temperature-dependent activation energy. However, it diverges at low temperatures, indicating its empirical nature.
- MYEGA Equation: A more recent (circa 2008) equation derived from first principles using statistical mechanics. It accurately describes liquid viscosity across different fragilities with only three fitting parameters, offering a more fundamental understanding.
Anomalous Liquids and Liquid-to-Liquid Phase Transformations
Some liquids, like water and certain phase change materials (e.g., Germanium Telluride), exhibit anomalous behavior, switching between fragile and strong liquid characteristics. This is often indicative of an underlying liquid-to-liquid phase transformation. At an inflection point, many properties can diverge, signaling a high-order phase transition.
For phase change materials, this transition is practically important:
- Fragile behavior at high temperatures: Promotes viscous flow, enabling fast crystallization speeds for memory writing.
- Strong behavior at low temperatures: Leads to very high activation energy, ensuring excellent data retention and preventing "drift" (changes in conductivity over time) in memory cells, which would otherwise make it difficult to distinguish between 0s and 1s.
This unique combination of behaviors allows PCMs to achieve both fast switching and long-term data stability.
Takeaways
- Phase‑change materials switch between amorphous and crystalline states, causing large changes in optical reflectivity and electrical conductivity, which made rewritable optical discs possible and now powers modern phase‑change memory devices.
- Threshold switching allows a highly insulating amorphous PCM to become conductive under a voltage pulse, providing the high current needed for Joule heating and crystallization without atomic rearrangement.
- Intel’s 3D XPoint (Optane) stacks GST‑like PCMs with selector devices to prevent sneak‑path currents, delivering high density but struggling with cost, competition, and ecosystem challenges.
- In photonics, the huge refractive index contrast of PCMs enables reconfigurable optical elements, multi‑level optical memories, and in‑memory analog computing for AI by performing matrix multiplication with light.
- PCMs show fragile‑liquid behavior at high temperatures for fast crystallization and strong‑liquid behavior at low temperatures for long‑term data retention, a duality that also informs glass viscosity and processing considerations.
Frequently Asked Questions
What is threshold switching in phase‑change memory and how does it enable writing?
Threshold switching is a purely electronic transition that makes an amorphous PCM suddenly conductive when a voltage exceeds a critical value. This temporary conductivity allows a large current to flow, generating Joule heating that crystallizes the material for the "set" operation, while the reverse pulse resets it.
How do phase‑change materials enable optical analog computing for AI workloads?
Phase‑change materials enable optical analog computing by using their switchable transmittance to encode matrix weights in a photonic circuit. Light carrying input vectors passes through PCM cells, and the varying transmission multiplies the signals; a single photodetector then sums the products, performing fast, in‑memory matrix multiplication for AI tasks.
Who is MIT OpenCourseWare on YouTube?
MIT OpenCourseWare is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.
Does this page include the full transcript of the video?
Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.
of whether glass is
liquid or a solid is complex, depending heavily on temperature and the timescale of observation. While intuitively glass seems solid, its rheological behavior can blur the lines.
Helpful resources related to this video
If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.
Links may be affiliate links. We only include resources that are genuinely relevant to the topic.