Skip to content

Nanodiamond Sensing of Stray Fields during Domain Reversal

FieldValue
JournalNano Letters
Published2026-09-11
AuthorsKin On Ho, Hamza A. Abudayyeh, Andrew Perez, Hui‐Ping Chang, Liang Chang
DOI10.1021/acs.nanolett.6c03535

Ensemble nitrogen-vacancy centers in 140 nm nanodiamonds resolved three different local stray-field responses during reversal of a CoFeB strip. The sensor at the edge registered a positive field jump of about 0.6 mT. A sensor just outside the strip registered a negative jump greater than 1 mT. A sensor near the center showed no discernible change. These results show that a domain flip does not shift every nearby field in the same direction.

The device stack consisted of a substrate and four layers. These were 100 nm SiO2, 10 nm Ta, 1.2 nm Co20Fe60B20, and approximately 1 nm MgO. Interfacial coupling produced perpendicular magnetic anisotropy. The CoFeB strip was approximately 5 µm wide and 83 µm long. Notches helped nucleate domain walls.

An external field was applied perpendicular to the film. It was swept between approximately −35 mT and +30 mT, depending on direction. Polar magneto-optical Kerr effect microscopy gave a coercive field of 21.0 ± 0.4 mT. MOKE detected the perpendicular magnetization change but could not quantify magnetization or the local stray field.

The 140 nm nanodiamonds were drop-cast from a solution with a density of 10 µg/mL. A Pt wire about 70 µm in diameter served as the microwave antenna. Its center was about 80 µm from the strip center. Laser power was approximately 11 µW, and microwave power was +24 dBm. Near the coercive field of approximately 20 mT, the nanodiamond sensitivity was approximately 20 µT/√Hz.

ND3, a particle far from the strip with negligible sample stray field, calibrated the applied field. This particle also confirmed that the measured field change was local to the CoFeB strip.

Full-spectrum fitting for incomplete ODMR data

Section titled “Full-spectrum fitting for incomplete ODMR data”

Conventional analysis usually fits multiple Lorentzian peaks and uses only resonance frequencies for field reconstruction. That approach discards contrast and linewidth information. It also typically requires three NV orientations to determine a unique field vector, along with broadband microwave excitation.

The authors instead fit the entire optically detected magnetic resonance spectrum with the NV spin Hamiltonian and a phenomenological lineshape. The model summed four NV orientations in a local frame. Random particle orientation removed the need to specify the crystal axes in the laboratory. The known applied field direction remained perpendicular to the sample plane.

The complete model had ten parameters. These covered two zero-field splitting parameters, D and E. They also covered three magnetic field parameters, three Lorentzian parameters, and two effective contrast modulations. Calibrations fixed the zero-field splitting, intrinsic linewidth, and modulation terms. Five parameters then remained free: field magnitude, polar angle, azimuthal angle, baseline, and contrast.

This fit retained weak and overlapping resonances. It could extract the local field when peaks became nearly degenerate at low field. At low field, the spectrum collapsed toward two broad resonances. The fit also remained usable when resonances moved outside the accessible window in the field regime above 15 mT. The authors attribute this stability to constraints imposed by the Hamiltonian and calibrated lineshape.

Three local responses at the switching field

Section titled “Three local responses at the switching field”

Five nanodiamonds at different positions were investigated. The CoFeB strip was first saturated under a large field of one polarity. Optically detected magnetic resonance spectra were then recorded while the field swept toward the opposite polarity.

ND1 sat at the strip edge. Its resonance frequency increased abruptly near a coercive field of 22.9 ± 0.3 mT. The full-spectrum fit converted this shift into a magnetic-field change of approximately 0.6 mT. The jump appeared during sweeps in both directions. The paper identifies this reversal with the independent MOKE measurement.

ND2 sat just outside the strip. Its frequency decreased during domain reversal, indicating a local field decrease. The measured decrease exceeded 1 mT. This sign was opposite to the expected increase when magnetization aligns with the applied field.

ND5 sat near the strip center. Its spectra showed no discernible frequency jump during either sweep direction. The paper reports no numerical zero field change for this response. Together, ND1, ND2, and ND5 demonstrate strong position dependence in the reversal-induced field change.

Edge fields reproduced by micromagnetic simulation

Section titled “Edge fields reproduced by micromagnetic simulation”

For a thin film with perpendicular magnetic anisotropy, stray fields are generated mainly at the strip boundaries. The analysis used a semi-infinite thin film with 2w much greater than thickness t. In this limit, the in-plane component Bx was largest at the edge. The out-of-plane component Bz reached its extrema at x = ±w ± z. Both components fall rapidly with distance from the edge.

The authors tested this model with OOMMF simulations. They solved the Landau-Lifshitz-Gilbert equation using 2 × 2 × 1 nm³ cells. The simulation used saturation magnetization Ms = 1.3 × 10⁶ A/m and damping constant α = 0.008. Exchange stiffness was A = 1.8 × 10⁻¹¹ J/m. The strip center was at x = 0 µm, with its edge at x = 2.5 µm.

The before-reversal and after-reversal fields were calculated at Hc − δ and Hc + δ. A standoff of z = 70 nm represented half the average nanodiamond size. Matching the measured 0.6 mT change for ND1 gave a most probable lateral distance. This distance was about 740 nm from the sample edge. The greater than 1 mT change for ND2 was consistent with a closer position of about 300 nm.

During reversal, the domain-associated stray field changes direction across the coercive field. Its contribution relative to the applied field differs at the edge, just outside, and near the center. The position estimates were model results, not direct optical measurements. They depend on the assumed 70 nm standoff and selected domain configurations.

The authors also note that strong gradients require small particles. They judged the 140 nm nanodiamond suitable for capturing this local field.

NV scanning probe microscopy, or NV-SPM, offers excellent spatial resolution but can perturb the native magnetic texture with its tip. Probe-to-probe differences in NV properties, standoff, and geometry require careful calibration. NV widefield microscopy, or NV-WFM, covers a large field of view. Its resolution is limited by optical diffraction to about 500 nm.

Nanodiamond sensing sits between these approaches in resolution and experimental handling. Its spatial scale is set mainly by particle size. Commercial nanodiamonds as small as approximately 30 nm were available, while this experiment used 140 nm particles. The particles can be deposited directly on devices and moved between setups without specialized bulk-diamond tip handling. The authors describe this as a simple, low-cost, and adaptable approach.

Prior NV-SPM work at a standoff of approximately 120 nm distinguished Bloch and Néel domain walls. Those studies predominantly examined static magnetic configurations. Earlier nanodiamond studies covered domain wall oscillations, ferromagnetic resonance, and spin waves. The present experiment followed a field-driven reversal and resolved local changes at several positions.

The comparison also defines practical limits. Nanodiamond particles still require screening for suitable properties. Standoff and placement matter because the local field changes sharply near an edge. The authors state that high-quality particles could support pulsed protocols for improved sensitivity.

Limits and implications for device engineering

Section titled “Limits and implications for device engineering”

The work is a proof of principle on one patterned CoFeB strip. Five particles were examined in the main text, while ND3 and ND4 data appear in the Supporting Information. The central result is not a general field jump for every site. It is a set of measured, position-dependent responses to one reversal process.

The field extraction depends on calibrated Hamiltonian and lineshape parameters. Limited microwave bandwidth can hide some NV resonances in the field regime above 15 mT. The position estimates from simulation assume a 70 nm standoff and selected domain configurations. These conditions limit how directly the inferred distances transfer to other devices.

For device engineering, the method measures short-range magnetic fields at the scale relevant to patterned magnetic elements. The authors identify domain-based neuromorphic architectures as an application where these fields can be confused with electrically generated fields. Distinguishing them is necessary to control crosstalk and support scalable operation.

Nanodiamond sensing is especially relevant when strips narrow to hundreds or tens of nanometers. NV-WFM remains diffraction limited near 500 nm, while three-dimensional geometries can complicate NV-SPM. Particle size, placement, and standoff therefore belong in the sensor design. For a device engineer, the minimum useful particle diameter, allowable standoff, and placement tolerance are part of the measurement specification.

View Original Abstract

Quantitative measurement of nanoscale stray magnetic fields during domain reversal is important for optimizing magnetic memory and logic devices. Yet, achieving high spatial resolution with minimally invasive probes remains challenging. Here we demonstrate nitrogen vacancy centers in nanodiamonds (NDs) as local sensors in a CoFeB strip. During magnetic-field-driven reversal, the ND sensors resolve three distinct local responses depending on their positions relative to the strip: a positive frequency jump at the edge, an unexpected negative jump just outside the strip, and a negligible change near the center. These contrasting signals are quantitatively explained by magnetostatic boundary fields from micromagnetic simulations. We further implement a fully constrained fitting procedure for optically detected magnetic resonance spectra, enabling robust field extraction even when resonances are partially resolved. Together, these results extend ND magnetometry to quantitatively characterize heterogeneous stray fields during magnetic domain reversal in thin films and spintronic devices.