Spin transport and magnetic proximity effect in CoFeB/normal metal/Pt trilayers

This study demonstrates that inserting an interlayer (Al, Cr, or Ta) in CoFeB/Pt trilayers suppresses the magnetic proximity effect-induced ferromagnetism in the Pt layer, thereby significantly reducing the total damping and highlighting the critical role of the proximity effect in spin transport parameter extraction.

The Gist

The Big Picture: A Traffic Jam of Spins

Imagine you are trying to run a very efficient delivery service. In the world of electronics, the "packages" being delivered are not boxes, but tiny bits of magnetic energy called spins.

The scientists in this paper are studying a specific delivery route made of three layers of metal:

1. The Warehouse (CoFeB): A magnetic material that generates the spins.
2. The Highway (Pt - Platinum): A non-magnetic metal that the spins travel through.
3. The Interlayer (The "Buffer"): A thin sheet of material (like Aluminum, Chromium, or Tantalum) that can be placed between the Warehouse and the Highway.

The Problem:
When the Warehouse (CoFeB) is directly touching the Highway (Pt), the delivery system gets clogged. The "traffic" (damping) slows down significantly. In physics terms, this is called high damping. The spins lose their energy too quickly, which is bad for making fast, efficient computer memory or processors.

The scientists wanted to know: Why is the traffic so bad? Is it because the Highway is bad, or is the Warehouse messing with the Highway?

The Mystery: The "Ghost" Magnet

The researchers suspected a phenomenon called the Magnetic Proximity Effect (MPE).

Think of the Platinum (Pt) highway as a calm, non-magnetic river. When the magnetic Warehouse (CoFeB) sits right on top of it, it's like a powerful magnet hovering over the water. Even though the water isn't supposed to be magnetic, the magnet's influence "haunts" the top few inches of the river, turning that water into a temporary, weak magnet.

This "ghost magnet" (the magnetized Platinum) creates extra friction. It grabs the spinning packages and slows them down, causing the high damping the scientists observed.

The Experiment: Building a Wall

To test this theory, the scientists built a wall between the Warehouse and the Highway. They inserted a thin layer of different materials (Aluminum, Chromium, or Tantalum) between the CoFeB and the Pt.

The Result:
As soon as they put up this wall, the traffic jam disappeared! The damping dropped significantly, regardless of what material they used for the wall.

The Analogy:
Imagine the "ghost magnet" in the Platinum is a sticky trap.

• Without the wall: The Warehouse is right next to the trap. The spins get stuck immediately.
• With the wall: The wall acts as a shield. It stops the Warehouse's magnetic influence from reaching the Platinum. The Platinum returns to being a calm, non-magnetic river, and the spins flow smoothly again.

The Proof: Taking a "Magnetic X-Ray"

How did they know the Platinum was actually turning into a magnet? They used a special tool called HHG-TMOKE.

Think of this as a super-powered, element-specific X-ray.

• Regular X-rays show bones.
• This special X-ray can look at a specific element (like Platinum) and ask, "Are you magnetic right now?"

They shone this light on the samples:

1. CoFeB/Pt (No wall): The X-ray screamed, "Yes! The Platinum is magnetic!" (It showed a signal at the Platinum energy level).
2. CoFeB/Al/Pt (With wall): The X-ray said, "Nope. The Platinum is just normal metal."

This confirmed that the "ghost magnet" was real and that the wall successfully killed it.

The Twist: Why the Math Was Confusing

Here is the tricky part. When scientists try to model how spins move, they use a number called the Spin Diffusion Length. You can think of this as "how far a spin can run before it gets tired."

• The Old Model: When they looked at the "clogged" system (CoFeB/Pt), the math suggested the spins could only run about 1 nanometer before getting tired.
• The New Reality: When they added the wall and fixed the clog, the math showed the spins could run 2.5 nanometers.

The scientists realized the math was lying. The Platinum didn't actually change its physical ability to let spins run. Instead, the "ghost magnet" (the MPE) was acting like a hidden speed bump that the standard math didn't account for.

When they tried to fix the math by just changing the "speed limit" (resistivity) or the "road width" (spin diffusion length) of the Platinum, they couldn't explain the data. They only got the numbers to match when they admitted: "Okay, the top layer of the Platinum is actually a magnet, and the rest is normal."

The Conclusion

1. The Culprit: The high energy loss (damping) in CoFeB/Pt systems isn't just because Platinum is a bad conductor; it's because the magnetic CoFeB turns the top of the Platinum into a magnet (MPE), creating extra friction.
2. The Fix: Putting a thin spacer layer between them stops this magnetic influence, making the system much more efficient.
3. The Lesson: When scientists try to measure how well these materials work, they have to be careful. If they don't account for the "ghost magnet" effect, they will get the wrong numbers for how fast spins can travel.

In short: The researchers found that a magnetic layer can "infect" its neighbor, slowing everything down. By putting a simple barrier in between, they stopped the infection and restored the flow, proving that sometimes, the best way to fix a traffic jam is to build a wall.

Technical Summary

1. Problem Statement

The study addresses a significant discrepancy in the field of spintronics regarding the spin pumping effect in Ferromagnet/Normal Metal (FM/NM) bilayers, specifically CoFeB/Pt.

• The Phenomenon: When a ferromagnet (FM) is excited by microwaves, it pumps a spin current into an adjacent non-magnetic (NM) layer, increasing the magnetic damping ($\alpha$). In CoFeB/Pt systems, this damping is unusually high.
• The Discrepancy: Previous studies showed a lack of correlation between measured inverse spin Hall effect (ISHE) voltages and the effective spin mixing conductance ($G^{\uparrow\downarrow}$) extracted from damping data in CoFeB/Pt compared to CoFeB/Ta.
• The Hypothesis: The authors hypothesize that the Magnetic Proximity Effect (MPE)—where the Pt layer acquires a static magnetic polarization due to direct contact with the ferromagnet—is the primary cause of the enhanced damping. However, the impact of MPE on parameter extraction in standard spin transport models is not well understood. Specifically, it is unclear if standard models can accurately describe systems where only a few atomic layers of the NM are magnetized.

2. Methodology

The researchers employed a combination of experimental characterization and phenomenological modeling:

• Sample Fabrication: They grew polycrystalline CoFeB(11 nm) / X(t) / Pt(3 nm) trilayers on Si/SiO$_2$ substrates via RF sputtering.
  • Variable X (Spacer): Aluminum (Al), Chromium (Cr), and Tantalum (Ta) were used as interlayers with varying thicknesses ($t$).
  • Control Samples: CoFeB/Pt (no spacer) and CoFeB/Al/Pt were compared. A CoFeB/Al/Ta system served as a reference without MPE.
• Damping Measurements: Ferromagnetic Resonance (FMR) linewidths were measured using a Vector Network Analyzer (VNA) setup to extract the damping parameter ($\alpha$) as a function of frequency.
• Direct MPE Verification: To confirm the presence of magnetic order in Pt, the team used Transverse Magneto-Optical Kerr Effect (TMOKE) spectroscopy with High-Harmonic Generation (HHG) in the Extreme Ultraviolet (XUV) range. This technique is element-sensitive, allowing them to probe the magnetic asymmetry specifically at the Pt $N_{7}$ absorption edge ($\sim$71.2 eV).
• Theoretical Modeling: They utilized a spin relaxation model (based on the work of Boone et al.) to simulate the damping. The model accounts for:
  • Spin mixing conductance ($G^{\uparrow\downarrow}$).
  • Spin diffusion lengths ($\lambda_{spin}$) of the layers.
  • Resistivity and spin backflow effects.
  • They tested two scenarios: (1) a homogeneously magnetized Pt film and (2) a film where only the interface (approx. 2 monolayers) is magnetized.

3. Key Contributions and Results

A. Direct Confirmation of MPE

Using HHG-TMOKE, the authors provided direct, element-specific evidence of MPE:

• CoFeB/Pt: A clear magnetic asymmetry signal was observed at the Pt $N_{7}$ edge, confirming that the Pt layer possesses an induced ferromagnetic moment.
• CoFeB/Al/Pt & CoFeB/MgO: No magnetic signal was detected at the Pt edge. The Al interlayer successfully suppressed the MPE, proving that direct contact is required for the effect.

B. Damping Reduction via Interlayers

• Observation: Introducing a thin interlayer (even 1 nm of Al, Cr, or Ta) between CoFeB and Pt caused a drastic reduction in the total damping parameter $\alpha$.
  • CoFeB/Pt: $\alpha \approx 10.5 \times 10^{-3}$
  • CoFeB/Al(1)/Pt: $\alpha \approx 6.6 \times 10^{-3}$
• Material Independence: The reduction occurred regardless of the spacer material (Al, Cr, or Ta), indicating that the effect is driven by the suppression of the MPE rather than changes in interface transparency or spin diffusion properties of the spacer itself.
• Control: In CoFeB/Al/Ta systems (where no MPE exists), the damping remained low and constant regardless of the Al thickness, confirming that the high damping in CoFeB/Pt is indeed an MPE artifact.

C. Modeling and Parameter Extraction

The authors attempted to reproduce the experimental damping data using the spin relaxation model:

• Effective Spin Diffusion Length: To fit the high damping of the CoFeB/Pt bilayer, the model required an artificially reduced spin diffusion length for Pt ($\lambda_{spin} \approx 0.9$ nm). When the spacer was introduced (suppressing MPE), the model required a larger $\lambda_{spin} \approx 2.5$ nm to fit the lower damping.
• Limitation of Standard Models: The authors tested whether a standard model could explain the high damping by assuming only a 2-monolayer magnetized Pt interface (the physical reality of MPE) while keeping the rest of the Pt non-magnetic.
  • They found that simply reducing the spin diffusion length or adjusting resistivity of the thin magnetized layer was insufficient to match the experimental damping values.
  • Even increasing the spin mixing conductance ($G^{\uparrow\downarrow}$) by five orders of magnitude failed to fully explain the data without assuming the entire Pt film was magnetized.
• Conclusion on Modeling: The phenomenological spin transport model, which treats the NM layer as a homogeneous bulk material, fails to accurately describe systems with MPE unless it assumes the entire NM layer is magnetized. This suggests that current models cannot fully decouple the MPE contribution from standard spin pumping contributions.

4. Significance

• Resolution of Discrepancies: The study clarifies why spin transport parameters (like $G^{\uparrow\downarrow}$ and $\lambda_{spin}$) extracted from CoFeB/Pt systems often differ from other FM/NM systems. The "extra" damping is largely due to the MPE, not just spin pumping efficiency.
• Impact on Literature: The wide range of reported spin diffusion lengths for Pt in literature may be attributed to the varying presence or absence of MPE in different sample preparations.
• Modeling Limitations: The work highlights a critical gap in theoretical spin transport models. Standard models struggle to account for the complex physics of a partially magnetized interface (MPE) versus a bulk magnetized layer. This implies that extracting intrinsic spin transport parameters from FM/Pt bilayers without spacer layers may yield erroneous values.
• Experimental Validation: The use of HHG-TMOKE provides a robust method for directly verifying the existence of MPE, moving beyond indirect inference from damping measurements.

In summary, the paper demonstrates that the Magnetic Proximity Effect is the dominant factor enhancing damping in CoFeB/Pt systems. While interlayers effectively suppress this effect, current spin transport models are insufficient to accurately describe the physics of the partially magnetized interface, leading to potential misinterpretations of spin transport parameters.