Crystallisation kinetics of supercooled liquid palladium

This study employs classical molecular dynamics simulations to characterize the crystallisation kinetics of supercooled liquid palladium, revealing diffusion-limited growth and a homogeneous nucleation maximum near $0.5 T_{\mathrm{m}}$ that aligns with time-resolved X-ray diffraction experiments and indicates that homogeneous nucleation governs the achievable supercooling in rapidly quenched Pd thin films.

The Gist

Imagine you have a pot of molten metal, specifically Palladium (a shiny, silver-white metal). If you let it cool down slowly, it naturally wants to turn back into a solid crystal, like water turning into ice. But what if you could cool it down so incredibly fast that it doesn't have time to organize itself? Instead of becoming a crystal, it gets "frozen" in a messy, disordered state, turning into a metal glass.

This paper is a detective story about how fast Palladium tries to turn back into a crystal when it's supercooled, and whether we can cool it fast enough to stop it.

The Two Competing Forces

Think of the liquid metal as a crowded dance floor.

1. The Desire to Organize (Thermodynamics): As the metal gets colder, the atoms "want" to line up in neat rows (crystallize) because it's more stable. The colder it gets, the stronger this urge becomes.
2. The Lack of Energy (Kinetics): However, as it gets colder, the atoms get sluggish. They move slower and slower, like people in a thick, sticky syrup. They can't find their way to the neat rows fast enough.

The battle between "wanting to organize" and "being too slow to move" decides if the metal becomes a crystal or a glass.

The Experiment: A Digital Time Machine

The researchers couldn't watch individual atoms move in real-time with a microscope because it happens too fast (in billionths of a second). Instead, they built a massive digital simulation (a "movie" made of 1.37 million atoms) to watch what happens.

They also did a real-world experiment using a super-powerful X-ray laser (like a high-speed camera) to zap thin films of Palladium, melt them, and watch them cool down.

What They Discovered

1. The "Speed Limit" of Atoms
They found that as the metal cools, the atoms move slower and slower in a very predictable way. It's like a car slowing down as it drives up a hill; the steeper the hill (colder the temperature), the slower the car goes. They calculated exactly how much energy is needed for an atom to take a step.

2. The "Seed" Problem (Nucleation)
To turn into a crystal, the liquid needs a "seed" to start growing.

• The Finding: The metal is incredibly good at making these seeds. Even when it's very cold, tiny crystal seeds pop up spontaneously everywhere at once.
• The Analogy: Imagine trying to keep a room full of people from forming a conga line. In most materials, you might be able to stop them. In Palladium, the moment the music stops (cooling starts), people immediately start linking arms. The researchers found that the "perfect" temperature for these seeds to form is about half the metal's melting point. At this temperature, the urge to organize is strong, but the atoms are still moving fast enough to link up.

3. The Growth Speed
Once a seed forms, it grows rapidly.

• The Finding: The crystal front moves at speeds of several meters per second.
• The Mechanism: The researchers tested two theories on how it grows.
  • Theory A (Collision-Limited): Atoms crash into the crystal and stick instantly, like rain hitting a windshield.
  • Theory B (Diffusion-Limited): Atoms have to wiggle and shuffle through the liquid to find a spot to stick, like people trying to find a seat in a crowded theater.
  • The Verdict: The data showed that Theory B is correct. The atoms have to shuffle around to find their place. The "crash and stick" theory predicted the metal would grow 100 times faster than it actually did.

4. The "Glass" Goal
The ultimate goal of this research was to see if we could cool Palladium fast enough to turn it into a glass (vitrification) rather than a crystal.

• The Result: To stop the crystals from forming, you need to cool the metal at a rate of 10 trillion degrees per second (10¹³ K/s).
• The Reality Check: The real-world experiment they did cooled the metal at about 500 billion degrees per second (5×10¹¹ K/s).
• The Conclusion: The real-world cooling was too slow. The metal simply didn't have enough time to avoid crystallizing. The "seeds" formed and grew before the metal could freeze into a glass.

The Big Picture

This paper tells us that pure Palladium is a "bad citizen" when it comes to making metal glass. It is too eager to turn back into a crystal. Even with the fastest cooling techniques currently available, the atoms organize themselves too quickly.

The researchers used their super-computer simulations to predict exactly when and where the crystals would start forming, and their predictions matched perfectly with the real-world X-ray laser experiments. This confirms that in these thin films, the crystals are forming from scratch (homogeneous nucleation) rather than starting from dirt or container walls (heterogeneous nucleation).

In short: You can't easily turn pure Palladium into a glass because it's just too good at organizing itself. To do it, you would need to cool it down faster than nature currently allows in their experiments.

Technical Summary

Technical Summary: Crystallisation Kinetics of Supercooled Liquid Palladium

Problem Statement
Understanding the crystallisation kinetics of supercooled liquids (SCLs) is a fundamental challenge in materials science, particularly for pure metals which exhibit exceptionally high nucleation and growth rates, making vitrification difficult. While the competition between thermodynamic driving forces and atomic mobility dictates whether a liquid crystallises or forms a glass, the specific mechanisms governing rapid crystal growth in pure metals remain unresolved. A central open question is whether growth is limited by atomic diffusion (thermally activated) or by atomic collisions (constrained by the speed of sound). Furthermore, distinguishing between homogeneous and heterogeneous nucleation in experimental settings is challenging, as heterogeneous sites often dominate. This study focuses on pure palladium (Pd), a representative face-centred cubic (FCC) metal and a primary constituent of many metallic glasses, to quantify these kinetics and validate them against ultrafast experimental data.

Methodology
The authors employed classical molecular dynamics (MD) simulations using the LAMMPS code and an embedded-atom method (EAM) potential specifically developed for Pd. To ensure statistical robustness and minimize finite-size effects, the primary simulations utilized a system of 1,372,000 atoms, while a complementary set of 256,000-atom systems was used for constructing Time-Temperature-Transformation (TTT) diagrams via ensemble averaging.

The simulation protocol involved:

1. Melting an initial FCC configuration at 2500 K to erase thermal history.
2. Instantaneous quenching to temperatures ranging from 0.38 to 0.65 $T_m$ (700–1150 K).
3. Monitoring the evolution of microstructure using OVITO software with Polyhedral Template Matching (PTM) and Grain Segmentation algorithms to distinguish ordered grains from disordered liquid.
4. Calculating self-diffusion coefficients from mean-squared displacement (MSD) and determining nucleation rates ($J$) and growth rates ($U$) from the temporal evolution of grain density and size.

These simulations were directly compared with time-resolved X-ray diffraction (XRD) data obtained from optically molten Pd thin films at the European XFEL. The experimental setup involved femtosecond laser excitation followed by X-ray probing, achieving cooling rates of approximately $5 \times 10^{11}$ K/s.

Key Results

• Self-Diffusion: The self-diffusion coefficient ($D$) of liquid Pd follows Arrhenius behavior over the investigated temperature range, with an activation energy of 467(6) meV/atom and a pre-exponential factor of $5.8(4) \times 10^{-8}$ m²/s. No crossover indicative of a liquid-liquid phase transition was observed.
• Nucleation Kinetics: The steady-state homogeneous nucleation rate exhibits a maximum of approximately $4 \times 10^{35}$ m⁻³ s⁻¹ near $0.5 T_m$. The data were successfully fitted within the framework of Classical Nucleation Theory (CNT), yielding a liquid-crystal surface energy of $\gamma \approx 0.17(1)$ J/m².
• Growth Mechanism: Crystal growth velocities were found to be on the order of meters per second (1–8 m/s). The temperature dependence and magnitude of these rates align closely with the diffusion-limited Wilson–Frenkel model. In contrast, the collision-limited growth model overestimated the velocities by nearly two orders of magnitude, indicating that growth is thermally activated rather than collision-limited.
• TTT Diagram and Critical Cooling Rate: A TTT diagram constructed from the simulations reveals a "nose" near $0.5 T_m$ at approximately 100 ps. This corresponds to a critical cooling rate for vitrification of $\sim 10^{13}$ K/s.
• Experimental Validation: The MD-predicted crystallisation onset times and temperatures for a cooling rate of $5 \times 10^{11}$ K/s match the experimental XRD observations. The experimentally observed supercooling reached approximately $0.65 T_m$, consistent with the simulation predictions.

Significance and Claims
The paper establishes that for rapidly quenched pure Pd thin films, the achievable supercooling is governed by homogeneous nucleation rather than heterogeneous nucleation. The agreement between the MD simulations and the XFEL pump-probe experiments validates the use of large-scale MD for studying ultrafast crystallisation kinetics.

The study concludes that:

1. Pure FCC metals like Pd are extremely difficult to vitrify because their critical cooling rate ($\sim 10^{13}$ K/s) far exceeds the rates achievable in the studied thin-film experiments ($5 \times 10^{11}$ K/s).
2. The crystallisation front propagation in supercooled Pd is diffusion-limited, supporting the Wilson–Frenkel mechanism over the collision-limited regime.
3. The results provide a consistent physical framework for interpreting experimental data on metallic glass formation, confirming that the limitations observed in experiments are intrinsic to the material's homogeneous nucleation properties rather than experimental artifacts or insufficient quenching techniques alone.