🔬 mesoscale physics

High-Q, high-power spin Hall nano-oscillators with thick NiFe layers enabled by SiC substrates

This study demonstrates that utilizing high thermal conductivity SiC substrates enables spin Hall nano-oscillators with thick NiFe layers to achieve record-high simultaneous output power and coherence, overcoming the thermal limitations that typically degrade performance in such devices.

Original authors: Maha Khademi, Akash Kumar, Martina Ahlberg, Artem Litvinenko, Xiaotian Zhao, Saroj P. Dash, Johan Åkerman

Published 2026-10-27

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Maha Khademi, Akash Kumar, Martina Ahlberg, Artem Litvinenko, Xiaotian Zhao, Saroj P. Dash, Johan Åkerman

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where tiny electronic components can generate their own steady, rhythmic pulses of energy, acting as the heartbeat for future computers that think more like the human brain than like traditional calculators. These devices, known as spin Hall nano-oscillators, are microscopic engines that convert electricity into microwave signals. They work by sending a current through a stack of thin metal layers, where the flow of electrons creates a twisting force that makes the magnetism in one layer spin in a continuous, synchronized circle. This spinning motion generates a radio wave that can be used for everything from processing information to controlling quantum systems. However, for these oscillators to be truly useful in real-world technology, they need to be both powerful and incredibly stable, producing a clean signal without wavering. The challenge has been that making them stronger often makes them hotter, and that heat causes the signal to become messy and unreliable, limiting how much power they can handle.

Researchers have long suspected that making the magnetic layer in these devices thicker would solve some of these problems, as a thicker layer could carry more energy and produce a clearer signal. Yet, this idea hit a wall: the extra thickness required much higher electrical currents to keep the magnetism spinning, and those currents generated so much heat that the devices would overheat and fail before they could reach their full potential. It was a classic engineering bottleneck where the solution to one problem created a worse one. To break through this barrier, a team of scientists turned their attention to the material the device sits on. Instead of the standard materials used in labs, they tested a specific type of ceramic called silicon carbide, known for its ability to pull heat away from hot spots much faster than other materials.

The team built a series of these nano-oscillators using a stack of tungsten and nickel-iron, placing them on three different types of substrates: aluminum oxide, silicon, and silicon carbide. They systematically varied the thickness of the magnetic nickel-iron layer, testing versions that were five nanometers thick all the way up to ten nanometers. By running electrical currents through these devices and measuring the resulting signals, they discovered that the silicon carbide substrate acted as a highly efficient heat sink. While the devices on the other materials began to overheat and lose their stability at lower currents, the silicon carbide versions stayed cool enough to handle much higher power levels. This thermal advantage allowed the researchers to push the devices into a high-performance zone that was previously inaccessible, where the spinning magnetic mode could expand and stabilize.

The results were striking. The best-performing device, built with a seven-nanometer-thick magnetic layer on a silicon carbide base, produced an average output power of 13.1 pW and maintained a quality factor of 23,000. In the world of these single, self-running oscillators, this combination of power and signal clarity is a record, surpassing all previous reports. The silicon carbide substrate did not just help the device run hotter; it allowed the device to run better. The team found that while they could even sustain oscillations in layers as thick as ten nanometers on silicon carbide, the seven-nanometer version offered the perfect balance, delivering the strongest and most coherent signal. This work demonstrates that by simply changing the foundation the device sits on, engineers can overcome the heat limitations that have held back the development of these powerful, coherent microwave sources, opening the door to more robust networks of oscillators for next-generation computing.

Technical Summary: High-Q, High-Power Spin Hall Nano-Oscillators Enabled by SiC Substrates

Problem Statement
Spin Hall nano-oscillators (SHNOs) are promising nanoscale microwave sources for applications in neuromorphic computing and Ising machines. A primary route to improving their performance—specifically output power and coherence—is increasing the thickness of the ferromagnetic (FM) layer (e.g., NiFe). Thicker FM layers offer larger magnetic mode volumes, reduced interfacial damping, and stronger spin-wave coupling. However, this strategy faces a critical bottleneck: thicker layers require higher drive currents to compensate for damping, leading to significant Joule heating. This thermal rise degrades device performance by reducing saturation magnetization, increasing damping, suppressing the anisotropic magnetoresistance (AMR) readout signal, and triggering magnetic instabilities. Existing substrates, such as sapphire (Al $_2$ O $_3$ ) and high-resistivity silicon (Si), lack the necessary combination of high thermal conductivity and electrical resistivity to manage the heat generated in thick-FM SHNOs at room temperature, limiting the accessible current range and preventing the realization of high-power, high-coherence operation.

Methodology
The authors fabricated W(5 nm)/NiFe( $t_{\text{NiFe}}$ )/Al $_2$ O $_3$ (4 nm) nanoconstriction (NC) SHNOs with varying NiFe thicknesses ( $t_{\text{NiFe}}$ = 5, 7, 8, 9, and 10 nm) on three different substrates: c-axis Al $_2$ O $_3$ , intrinsic high-resistivity Si, and high-purity 4H SiC. The devices featured constriction widths ranging from 20 to 200 nm.

Thermal Characterization: The team employed COMSOL Multiphysics simulations to model current distribution, Joule heating, and temperature profiles. They experimentally verified thermal behavior by measuring the current-dependent resistance ( $R$ vs. $I_{\text{dc}}^2$ ) to identify the onset of thermal runaway.
Electrical and Magnetic Measurements: Auto-oscillation dynamics were characterized via power spectral density (PSD) maps under applied magnetic fields. Key metrics included integrated output power, linewidth ( $\Delta f$ ), and quality factor ( $Q$ ). Spin-torque ferromagnetic resonance (ST-FMR) was used to extract effective magnetization ( $\mu_0 M_{\text{eff}}$ ) and Gilbert damping ( $\alpha_{\text{eff}}$ ).
Systematic Comparison: Performance was systematically compared across substrates and NiFe thicknesses to isolate the effects of thermal management and current redistribution.

Key Contributions and Results

Superior Thermal Management of SiC: The study demonstrates that SiC substrates, with a room-temperature thermal conductivity of $\approx$ 450 W/m·K (compared to 35 W/m·K for Al $_2$ O $_3$ and 130 W/m·K for Si), enable substantially more efficient heat spreading. Simulations and resistance measurements confirm that SiC devices remain significantly cooler under high drive currents, delaying thermal runaway and allowing access to higher current regimes.
Record Performance for Single Free-Running SHNOs: The optimal device configuration was identified as a 200 nm wide NC with a 7 nm NiFe layer on a SiC substrate. This device achieved:
- An average integrated output power of 13.1 pW.
- A mean quality factor of $Q = 23,000$ (with individual peaks reaching $Q \approx 45,000$ and linewidths as low as 200 kHz).
- This represents the highest simultaneous output power and coherence reported for a single free-running SHNO, surpassing previous records which typically showed powers of 1–9 pW with $Q$ factors below 1,000.
Enabling Thick-FM Operation: While the optimal performance occurred at 7 nm, the superior thermal properties of SiC allowed stable auto-oscillations to be sustained in NiFe layers as thick as 10 nm. In contrast, devices on Al $_2$ O $_3$ and Si showed degraded performance (broader linewidths and lower power) when the NiFe thickness was increased from 5 nm to 7 nm, as they could not reach the necessary high-current "expanding mode" regime without thermal instability.
Substrate-Dependent Mode Evolution: The efficient heat dissipation in SiC devices allowed the auto-oscillation mode to expand from the constriction edge into the interior of the device at high currents. This "expanding mode" regime is crucial for maximizing output power and minimizing linewidth, a state that was inaccessible in conventional substrates due to premature thermal limitations.

Significance
The paper establishes thermal substrate engineering, specifically the use of SiC, as a viable and effective pathway to realizing high-coherence, thick-ferromagnet spintronic oscillators. By overcoming the Joule heating bottleneck, SiC substrates enable the simultaneous achievement of high output power and high spectral coherence in single free-running devices. The authors note that this capability is essential for scalable networks of coupled oscillators, where robust mutual synchronization requires large total drive currents and efficient heat spreading. Furthermore, the compatibility of SiC with Si-based microelectronics and its emerging role as a platform for spin-defect quantum devices suggests potential for integrating nanoscale microwave-field sources with quantum systems. The results provide a foundational step toward next-generation microwave sources and unconventional computing hardware.

Technical Summary: High-Q, High-Power Spin Hall Nano-Oscillators Enabled by SiC Substrates

More like this