Negative temperature coefficient of Gilbert damping in magnetic bilayers

This paper reports a counterintuitive negative temperature coefficient of Gilbert damping in Py/Nd bilayers, where damping decreases with rising temperature due to thermally induced dynamic separation of interfacial and bulk magnetization, a phenomenon that can be tuned via capping layer thickness to enhance spintronic device performance.

The Gist

Imagine a spinning top. In the world of magnets, this top is a tiny magnetic particle. When you nudge it, it wobbles and spins before eventually settling down. How fast it settles is determined by something called Gilbert damping. Think of damping as the "friction" or "air resistance" that slows the spin down.

In most materials, if you heat them up, this friction gets worse. It's like trying to spin a top in hot, thick soup; the heat makes the atoms jittery, creating more chaos and resistance, so the top stops spinning faster. This is the standard rule for almost all magnetic metals.

The Surprise Discovery
The researchers in this paper found a magnetic "trick" that breaks this rule. They created a sandwich made of two layers: a magnetic layer called Permalloy (Py) and a non-magnetic layer called Neodymium (Nd).

When they heated this specific sandwich, something strange happened: the friction actually went down. Instead of the top slowing down faster in the heat, it kept spinning for longer. The "damping" coefficient had a negative temperature coefficient, meaning heat made the system less resistant to motion, which is the opposite of what usually happens.

The "Crowded Dance Floor" Analogy
To understand why, imagine the magnetic atoms as dancers on a floor.

1. The Normal Case (Pure Metal): In a regular metal, the dancers are all holding hands tightly. When you heat the room (increase temperature), everyone starts shaking and jumping around wildly. This chaos makes it hard for the group to move in sync, so they stop dancing (relax) very quickly. More heat = more friction.
2. The Special Case (The Py/Nd Sandwich): In this experiment, the researchers added a "spin pump" effect at the boundary where the two layers meet. This is like having a very strict bouncer at the edge of the dance floor who tries to pull the dancers out of sync to stop them.
   • At low temperatures: The dancers are calm. The bouncer is very effective, pulling on the dancers at the edge and creating a lot of friction. The whole group stops quickly.
   • At high temperatures: The dancers start shaking and jumping wildly on their own. Because they are so jittery, they start to let go of each other's hands near the edge. The connection between the dancers at the edge and the dancers in the middle gets weak.
   • The Result: The "bouncer" (the spin pump) can no longer grab the dancers effectively because the edge dancers are too chaotic and disconnected from the group. The friction at the edge disappears, and the whole group spins more freely.

How They Proved It
The team used two methods to confirm this:

• Computer Simulations: They built a virtual model of these atomic dancers and watched them spin at different temperatures. The computer showed that as the temperature rose, the connection between the surface and the bulk (the middle) broke down, reducing the friction.
• Real Experiments: They used ultra-fast laser pulses to heat up real samples of this magnetic sandwich. By measuring how the magnetism wobbled and settled, they confirmed that the damping decreased as the sample got hotter, matching their computer predictions.

Why It Matters (According to the Paper)
The paper explains that this effect happens specifically because the "spin pumping" (the bouncer) is very strong at the interface, but the heat causes the surface atoms to become so chaotic that they disconnect from the bulk.

The researchers note that this is a new way to control how magnetic devices behave. Since many devices (like computer memory) get hot when they work, being able to engineer materials where heat reduces friction could help make these devices switch faster or use less energy. They also mention that other rare-earth metals might do the same thing, offering a new playground for designing better magnetic tools.

Technical Summary

Technical Summary: Negative Temperature Coefficient of Gilbert Damping in Magnetic Bilayers

Problem Statement
The Gilbert damping factor is a critical parameter governing the switching speed and energy dissipation in spintronic devices, such as magnetic random access memory (MRAM) and heat-assisted magnetic recording (HAMR). In simple metals and bulk ferromagnetic films, intrinsic Gilbert damping typically increases with temperature due to enhanced thermal spin fluctuations, often diverging near the Curie temperature. While extrinsic contributions like surface scattering and spin pumping are known to influence damping in nanoscale bilayers, the temperature dependence of these effects over a wide range (including near the Curie temperature) remains underexplored. Specifically, the behavior of damping in ferromagnetic/non-magnetic bilayers where spin pumping is significant has not been fully characterized regarding its temperature coefficient.

Methodology
The study employs a dual approach combining atomistic spin dynamics (ASD) simulations and experimental measurements:

1. Atomistic Simulations: The authors utilized a classical spin model based on a 3D Heisenberg spin Hamiltonian to simulate magnetic relaxation in Permalloy (Py)/Neodymium (Nd) bilayers. The simulations covered temperatures from 20 K to 800 K. The model incorporated a stochastic Landau–Lifshitz-Gilbert (sLLG) equation to account for thermal fluctuations. Crucially, the simulations distinguished between bulk damping and interface damping, modeling the spin pumping effect by increasing the intrinsic damping at the Py/Nd interface based on the Nd layer thickness ($d_{NM}$) and the spin diffusion length ($\lambda_{sdl}$).
2. Experimental Measurements: Time-resolved magneto-optical Kerr effect (TR-MOKE) measurements were conducted on Py/Nd bilayers. To vary the effective temperature during relaxation without changing the ambient environment, ultrafast laser pulses with variable pump fluence were used to induce ultrafast demagnetization and elevate the electronic temperature. The resulting magnetization precession was fitted to extract the relaxation time ($\tau$) and precession frequency ($f$), allowing the calculation of the effective Gilbert damping ($\alpha_{eff}$).

Key Results
The study reveals a counter-intuitive phenomenon in Py/Nd bilayers:

• Negative Temperature Coefficient: Contrary to the behavior of pure Py films (where damping increases with temperature), Py/Nd bilayers exhibit a decrease in effective Gilbert damping as temperature rises. In simulations for a Py(6 nm)/Nd(32 nm) bilayer, $\alpha_{eff}$ dropped from approximately 0.082 at 20 K to 0.064 at 800 K. Experimental TR-MOKE data confirmed this trend, showing a continuous decrease in damping with increasing pump fluence (and thus increasing electronic temperature).
• Thickness Dependence: The effect is strongly dependent on the thickness of the Nd capping layer. Thicker Nd layers (8 nm to 32 nm) exhibit the negative temperature coefficient, while thinner layers (e.g., 1 nm) or pure Py films show the conventional positive temperature coefficient.
• Control Material Comparison: Comparative measurements on Pt/Py and Pt$_{0.68}$Nd$_{0.32}$/Py bilayers showed a positive temperature coefficient. This is attributed to the short spin diffusion lengths in these materials, which limit the spin accumulation and interfacial damping enhancement, keeping bulk and interface damping values close.

Mechanism of Action
The paper attributes the negative temperature coefficient to a "dynamic separation" of interfacial and bulk magnetization dynamics:

1. Spin Pumping Enhancement: In Py/Nd bilayers, the interface exhibits significantly enhanced damping due to strong spin pumping.
2. Thermal Decoupling: As temperature increases, thermal fluctuations become more pronounced, particularly at the surface/interface where exchange bonds are reduced (lower coordination).
3. Reduced Effective Damping: These enhanced surface fluctuations cause the interfacial magnetization to decouple dynamically from the bulk magnetization during the relaxation process. This partial decoupling reduces the efficiency of the spin-pumping-induced damping enhancement, leading to a net decrease in the effective Gilbert damping observed at the macroscopic level.

Significance and Claims
The authors claim to present a new spintronic effect where the dynamic properties of nanoscale materials can be engineered to have a negative temperature coefficient of Gilbert damping. This finding challenges the conventional expectation that damping universally increases with temperature in magnetic systems.

The significance lies in the ability to control the temperature dependence of damping by varying the non-magnetic capping layer thickness and material choice. The authors suggest this could be utilized to modify the dynamic properties of nanoscale devices for enhanced energy efficiency or improved switching dynamics, particularly in scenarios where devices operate at elevated temperatures. The study provides a framework for understanding how interfacial spin accumulation and thermal fluctuations interact to dictate macroscopic damping behavior in bilayer systems.