Nanoscale dipolar fields in artificial spin ice probed by scanning NV magnetometry
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Materials for Quantum Technology |
| Published | 2026-09-01 |
| Authors | Ephraim Spindler, Vinayak Shantaram Bhat, Elke Neu, Mathias Weiler, M. Benjamin Jungfleisch |
| DOI | 10.1088/2633-4356/aea8b7 |
Scanning magnetometry with one nitrogen-vacancy center found nine ice-rule violations in the 1000 nm lattice sample and one in the 910 nm sample. The comparison used near-coercive reverse pulses within scanned regions of about 10 µm by 10 µm. Micromagnetic calculations then reproduced type I, II, and III vertex fields, with residuals typically within ±1 mT.
Ice-rule response and lattice spacing
Section titled “Ice-rule response and lattice spacing”Each square-lattice vertex contains four nanomagnets with Ising-like orientations. Type I and II vertices obey the two-in, two-out ice rule. Type III and IV vertices violate that rule and act as monopole-like excitations.
| Sample | Lattice constant | Reverse pulse | Vertex state | Violations in scan |
|---|---|---|---|---|
| S-A | 1000 nm | −38.3 mT | Type I domains separated by type II chains | 9 |
| S-B | 910 nm | −36.3 mT | Type I domains separated by type II chains | 1 |
Both samples were first saturated along x at ≥60 mT. Near the coercive pulse, type I vertices dominated. Type II chains defined boundaries between domains, and type III violations appeared at their endpoints. The authors attribute the larger violation count in S-A to weaker coupling at its larger lattice constant. Independent nanomagnet reversals in that sample permit more ice-rule defects.
Separate pulses bracketing the coercive field showed S-A retaining a uniform type II state at −35.8 and −40.3 mT. S-B developed local type I regions at −33.9 and −38.5 mT. These results connect lattice spacing with both field-driven transitions and defect formation.
Sample fabrication and magnetic preparation
Section titled “Sample fabrication and magnetic preparation”The arrays were patterned with a 100 keV electron-beam lithography system. Electron-beam evaporation deposited a 25 nm Ni81Fe19 film with a 3 nm Al cap. Each nanomagnet was approximately 110 nm wide and 810 nm long.
All magnets were rotated by 45° to permit full saturation along x. The patterned regions measured 25 µm by 100 µm. Before each measurement pulse, the authors saturated the structures along x at ≥60 mT.
Pulses lasted a few seconds near each coercive field and were followed by a low remnant field. The samples shared nanomagnet dimensions, so lattice constant was the controlled geometric difference.
Quenching imaging
Section titled “Quenching imaging”Contact-mode quenching imaging used a negative z bias of approximately −8.2 mT, as later determined by modeling. The ASI stray field added vectorially to that bias. North-pole vertices had higher fluorescence counts because their off-axis field reduced quenching.
Fields typically above 5 mT produced enough quenching for fast all-optical identification. Quenching images required 10 ms to 30 ms per pixel. A representative spectrum at 15.1 mT showed about 20% less off-resonant fluorescence than the 6.3 mT spectrum. Quenching therefore provides a rapid qualitative state map rather than a vector field measurement.
Vector-resolved cODMR
Section titled “Vector-resolved cODMR”Quantitative cODMR scans used a lifted probe height of 100 nm to preserve contrast. The stationary ProteusQ probe used a (100)-oriented diamond nanopillar with an NV center at its apex. The expected NV depth was 9 ± 4 nm after 6 keV nitrogen implantation.
A 515 nm laser with NA 0.7 excited the center and collected its red fluorescence. The two resonance frequencies, f− and f+, were fitted with a double Lorentzian. Using D = 2.87 GHz and γ = 28 GHz T−1, the authors extracted axial and transverse fields relative to the NV axis.
Representative spectra used total field magnitudes of 6.3 mT at 39° and 15.1 mT at 60° relative to the NV axis. At 4.0 mT off-axis, the fitted axial and transverse sensitivities were 3.4 and 28.9 µT/√Hz. At 13.0 mT, they were 4.6 and 12.2 µT/√Hz.
The resonance frequencies produced separate two-dimensional maps of field magnitude, axial component, and transverse component. These maps support direct comparison with simulated stray fields rather than inference from image contrast alone.
Micromagnetic fitting
Section titled “Micromagnetic fitting”The authors modeled individual vertices in Mumax3 through Aithericon. The fitting loss combined squared deviations in the axial and transverse fields. It optimized the external field, NV stand-off distance, NV orientation, effective saturation magnetization, and lateral alignment.
The first pass relaxed the magnetization without an external field. It assumed a 178 nm NV-to-surface distance, comprising a 75 nm magnetic stand-off, 100 nm lift, and 3 nm Al cap. The type I example had a mean absolute field deviation of 0.162 mT.
Relaxing the magnetization with the fitted external field reproduced small tilts near the rounded nanomagnet ends. The deviation fell to 0.119 mT, about 27% lower than the first pass. Those tilts changed the ASI stray field from a nearly symmetric pattern.
A 5 nm grid for the magnetic stand-off distance gave dMSO = 65 ± 5 nm. A ±10° orientation grid gave φ = 277 ± 1° and θ = 54.75 ± 1°. This revealed a 7° azimuthal misalignment from the nominal φ = 270°.
The final fitted local external field was µ0(Hext,x, Hext,y, Hext,z) = (5.8, 3.7, −8.7) mT. The fitted effective saturation magnetization was 586 ± 42 kA m−1. Most of its uncertainty came from the stand-off distance.
Accuracy, reference points, and device implications
Section titled “Accuracy, reference points, and device implications”The final maps agreed across type I, II, and III vertices. Experimental and simulated field magnitude, axial component, and transverse component differed by no more than about ±1 mT in most locations. The authors attribute the remaining residuals to structural imperfections, mechanical drift, and the planar data model. The probe follows sample topography, whereas the analysis treats each scan as a plane.
The effective saturation magnetization was below the cited Ni81Fe19 literature value of 860 kA m−1. The authors suggest that a conical cross-section reduces magnetic volume. An AFM-derived cross-section supports that interpretation. The fitted value therefore describes the fabricated nanomagnets, including geometry, rather than ideal bulk Permalloy.
The NV measurement operated under ambient conditions and required neither synchrotron access nor ultra-high vacuum. Unlike MFM, the NV probe did not use a magnetic tip that could perturb the sample. The extended interaction volume of an MFM tip also complicated its extraction of local field information.
Quenching imaging enables rapid counting of ice-rule violations over larger areas. cODMR converts selected vertices into quantitative tests of the stray-field model. The authors link lattice spacing to dipolar coupling and the resulting defect formation in these arrays.
The nine-versus-one comparison remains specific to the tested geometry and pulse conditions. It does not establish a universal relation between spacing and defect density. Local field maps can still validate fabricated arrays before their stray fields are used in magnonic or spintronic designs.
The fitted tilts reveal weak local fields below the nanomagnets’ switching threshold. The method therefore quantifies dipolar interactions while also exposing fabrication-dependent changes in magnetic volume.
Materials and support
Section titled “Materials and support”The measurement requires a single-crystal diamond probe containing a shallow NV center for scanning Ni81Fe19 artificial spin ice. 6CCVD’s diamond materials catalog includes single-crystal diamond grown by MPCVD, matching the base material family, but NV implantation, nanopillar geometry, and crystallographic orientation require separate qualification. The custom fabrication and engineering support service can review those requirements before a sensing format is selected.
View Original Abstract
Abstract We investigate dipolar coupling fields in two square-lattice artificial spin ice (ASI) systems with different lattice constants using scanning probe microscopy based on a single nitrogen-vacancy (NV) center in diamond. This technique offers nanoscale spatial resolution, operates under ambient conditions, and provides quantitative stray field measurements, making it uniquely suited for studying nanoscale magnetic textures. Our approach combines fluorescence quenching imaging and continuous-wave optically detected magnetic resonance (cODMR). A comparison of the two ASI samples, which differ in their lattice constants of 1000 nm and 910 nm respectively, reveals differences in the appearance of ice-rule violations—deviations from the lowest energy configuration in ASI vertices. We attribute these variations to varying coupling strengths dictated by the lattice constant. From the cODMR data, we extract both axial and transverse components of the local magnetic field relative to the NV axis. Micromagnetic modeling of these measurements allows for an iterative determination of the external magnetic field orientation, the detection of subtle magnetization tilts induced by weak external fields (well below the nanomagnets’ switching threshold), and an estimation of the effective saturation magnetization, thereby accounting for deviations in nanomagnet dimensions. These findings provide crucial insights into the tunable magnetic interactions in ASI, paving the way for the design of advanced magnonic and spintronic devices.
Products
View Specs →
Single Crystal Cvd Diamond Plate
View Specs →