High‐Temperature Magnetic Field Mapping Using a Single‐Crystal Diamond MEMS Resonator Array
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Advanced Science |
| Published | 2026-09-16 |
| Authors | Zilong Zhang, Keyun Gu, Satoshi Koizumi, Masaya Toda, Meiyong Liao |
| DOI | 10.1002/advs.77779 |
An 8 × 8 single-crystal diamond MEMS array mapped magnetic fields at 573 K. The single-pixel response was 3.5 ms, and frequency stability stayed at the parts-per-billion level. Magnetic noise density was 0.43 nT Hz⁻¹ᐟ² at 300 K and 5.45 nT Hz⁻¹ᐟ² at 573 K. The array produced two- and three-dimensional maps at a 160 µm sampling interval.
Each FeGa/Ti/diamond cantilever acted as an independently calibrated magnetic pixel. A magnetic field changed FeGa’s Young’s modulus through the ΔE effect, which shifted the cantilever’s mechanical resonance. The 3.5 ms value describes one pixel under pulsed excitation. It does not describe full-frame acquisition, which used sequential laser Doppler vibrometry.
Sensor architecture and calibration
Section titled “Sensor architecture and calibration”The resonant element was a 90 nm FeGa magnetic film on a 10 nm Ti adhesion layer and an SCD base. HRTEM identified clean interfaces, a disordered interfacial layer of about 0.5 nm, FeGa (220) orientation, and diamond (111) facets. EDS mapping showed uniform Fe, Ga, Ti, and C distributions. The authors suggest that oriented FeGa could improve soft magnetic properties, but they did not isolate that contribution.
Before imaging, each pixel was calibrated under a known field to obtain its sensitivity S_ij. The local field was calculated as H_ij = Δf_ij / S_ij. This pixel-level calibration accounted for sensitivity variation among resonators. For imaging at 573 K, the authors recalibrated every pixel at 573 K and used those temperature-specific values.
Diamond resonator fabrication and inspection
Section titled “Diamond resonator fabrication and inspection”Fabrication used smart-cut and ion-implantation-assisted release to form monolithic SCD-on-SCD resonators. The active diamond was an epilayer grown on an HPHT SCD substrate. Carbon implantation used 180 keV and a dose of 10¹⁶ cm⁻². MPCVD growth used 500 sccm hydrogen, 0.5% methane in hydrogen, 1000 W microwave power, 840 °C, and 3 h.
Implantation formed an approximately 200 nm graphite-like sacrificial layer. A 150 nm Al mask guided oxygen ICP etching. Hot H₂SO₄ and HNO₃ removed the mask and released the resonators. Annealing ran at 1100 °C for 3 h under a vacuum below 10⁻⁷ Pa and improved Q. RF magnetron sputtering deposited FeGa/Ti under argon at 10 sccm, 1 Pa, 100 W, and 300 K.
Three-dimensional laser profilometry showed slight upward curvature after sacrificial-layer release. Resonance was measured with laser Doppler vibrometry, a PZT actuator, a lock-in amplifier, and copper coils for applied fields. Measurements ran below 10⁻³ Pa. The optical system used a 633 nm He-Ne laser below 1 mW.
Resonant and magnetic response
Section titled “Resonant and magnetic response”After FeGa/Ti deposition, Q factors remained approximately 4000 to 6000. Resonance frequencies varied by about 5% across the array. The resonators gave linear vibration amplitude versus RF excitation, while actuation voltage had negligible effect on resonance frequency.
At 300 K, an applied field shifted the resonance peak toward lower frequency. One sensor had a sensitivity of 12.7 Hz/mT, and array sensitivities ranged from 7.2 to 12.7 Hz/mT. At 0.5 mT, the 3 dB bandwidth was 100 Hz. The relation τ ≅ 0.35/f_3dB gave the reported 3.5 ms response.
At 573 K, the maximum measured sensitivity reached 28.58 Hz/mT. The applied field left Q largely unchanged. Q decreased with temperature, which the authors attributed to increased intrinsic damping. They attributed the fast response to FeGa magnetization, SCD mechanical stability, and magnetoelastic coupling through Ti.
Elevated-temperature noise and stability
Section titled “Elevated-temperature noise and stability”Frequency noise was measured by tracking resonance with a phase-locked loop for 300 s. Data transfer was 899.5 Sa s⁻¹, and the target bandwidth was 1 Hz. An FFT of the tracked frequency gave the magnetic noise density after division by the measured magnetic sensitivity.
| Measure | 300 K | 573 K | Basis |
|---|---|---|---|
| Magnetic noise density | 0.43 nT Hz⁻¹ᐟ² | 5.45 nT Hz⁻¹ᐟ² | 100 µm resonator, 1 Hz target bandwidth |
| Minimum frequency fluctuation | 7.96 × 10⁻⁹ | 2.13 × 10⁻⁷ | Allan deviation |
| Allan-derived minimum field | 244.3 nT at 4.41 s averaging | 2919.3 nT at 5.04 s averaging | Allan deviation |
These two noise measures answer different questions. Spectral density describes short-term frequency noise in a 1 Hz bandwidth. The Allan deviation result estimates quasi-static or DC field resolution at a stated averaging time. Low-frequency drift, temperature changes, and long-term stability affect the Allan-derived value.
The FeGa/Ti/SCD platform showed a low temperature coefficient of frequency, but the paper gave no numerical value. Repeated heating and cooling tests were not performed on the present array. Prior work on individual SCD resonators showed stable repeated cycling, but the authors state that array-level thermal robustness still needs testing.
Spatial field reconstruction
Section titled “Spatial field reconstruction”A 3 × 3 × 8 mm³ permanent magnet began 10 mm above the resonator plane. A motorized stage moved it along z in 1 mm steps. At 300 K, the authors reconstructed field maps over 1120 × 1120 µm². They used three source positions: z0, z0 + 1 mm, and z0 + 2 mm. Field intensity decreased and broadened as the source moved farther from the array.
The array also produced a fixed-position magnet image at 573 K using temperature-matched calibration. A 90° in-plane rotation of the magnet changed the reconstructed pattern while its center and source distance stayed fixed.
Laser spot repositioning took about 100 ms between adjacent sensors. A complete array dataset required about 6.6 s, and reconstruction with visualization took about 1 min. Origin’s Data Boundary contouring used a smoothing parameter of 0.05 and a 100-fold total-points increase. These operations changed the display only, not measured pixel values.
The 160 µm pitch limits the demonstrated sampling interval. It does not establish the intrinsic resolution of the sensing mechanism. The millimeter-scale magnet and its separation also broadened the measured field.
Comparison and engineering limits
Section titled “Comparison and engineering limits”Table 1 compares the 3.5 ms pixel response with 100 ms for Hall-bar microscopy. It lists 6 nm for scanning NV center microscopy and 0.5 µm for scanning SQUID microscopy. The 160 µm value is only an array sampling interval, so the paper does not claim nanoscale mapping.
The authors position the device as a high-temperature ΔE-effect sensor. It does not have the highest room-temperature sensitivity or the lowest reported detection limit in their comparison. SQUIDs can reach fT-level detection, but they require cryogenic cooling. The diamond array instead keeps magnetic sensing operational at 573 K.
For device selection, the array is relevant when local magnetic mapping must continue at elevated temperature. Each pixel needs calibration at the operating temperature, and the present optical readout limits throughput. The maps remain proof-of-concept images rather than fully optimized quantitative measurements.
Accuracy can be affected by sensitivity variation, resonance-fit uncertainty, baseline drift, alignment, and stage positioning. Thermal drift during long sequential acquisitions was not separately quantified. The authors call for repeatability analysis, drift correction, parallel or multiplexed electrical readout, pixel addressing, and crosstalk control.
Materials and support
Section titled “Materials and support”The paper needs MPCVD-grown single-crystal diamond mechanical resonators with Ti adhesion and FeGa magnetostrictive coatings. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal diamond in 0.1 to 500 micrometer thicknesses, plates and wafers up to 125 mm, and Ti metallization. For custom device preparation, engineering support can assess the fabrication steps needed beyond the catalog materials.
View Original Abstract
ABSTRACT Magnetic imaging has long faced fundamental trade‐offs among speed, sensitivity, and spatial resolution, and its performance typically deteriorates at elevated temperatures. Here, we present a single‐crystal diamond (SCD) microelectromechanical system (MEMS) resonator array for spatial magnetic‐field mapping at elevated temperatures. An 8 × 8 cantilevers array enables pitch‐limited spatial magnetic‐field mapping with millisecond single‐pixel temporal response and nT Hz 1/2 magnetic noise level at 573 K, while maintaining frequency stability at parts‐per‐billion precision. These capabilities exhibit competitive performance in sensing metrics, enabling two‐ and three‐dimensional magnetic‐field reconstruction and demonstrate elevated‐temperature magnetic‐field mapping up to 573 K. These results demonstrate the potential of SCD MEMS resonator array for spatial magnetic‐field sensing under elevated‐temperature conditions.
Electronic Grade Intrinsic Diamond Epitaxial Wafer H Terminated Superior 100 Mosfet Ready Substrate
View Specs →Products
View Specs →
Single Crystal Cvd Diamond Plate
View Specs →References we cover
Section titled “References we cover”- Progress in semiconductor diamond photodetectors and MEMS sensors
- Effect of Deep-Defects Excitation on Mechanical Energy Dissipation of Single-Crystal Diamond
- Optimization of a Diamond Nitrogen Vacancy Centre Magnetometer for Sensing of Biological Signals
- Scanned probe imaging of nanoscale magnetism at cryogenic temperatures with a single-spin quantum sensor