Multimodal Super‐Resolution Imaging of Nitrogen‐Vacancy Centers via High‐Index‐Induced Structured Illumination Microscopy and Optically Detected Magnetic Resonance Spectrometry
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Advanced Science |
| Published | 2026-03-03 |
| Authors | Kyu Ri Choi, Mohammed Zia Jalaludeen, Samuel Begumya, Yan Qiu Du, Dong Hee Park |
| DOI | 10.1002/advs.202522495 |
The method resolved two NV ensemble features separated by approximately 98 nm inside bulk diamond, improving resolution by more than twofold over diffraction-limited imaging. High-index speckle illumination and optically detected magnetic resonance (ODMR) modulation supplied complementary structured signals during blind-SIM reconstruction. The acquisition also produced local ODMR spectra and magnetic-field vectors for emitters about 100 µm beneath the diamond surface.
Sub-100 nm imaging in bulk diamond
Section titled “Sub-100 nm imaging in bulk diamond”Two emission peaks separated by approximately 300 nm were unresolved in the diffraction-limited image but distinct after reconstruction. Other regions resolved features at approximately 170 nm and approximately 150 nm. The approximately 98 nm pair provided the strongest demonstration of sub-100 nm performance. Closely spaced ensembles merged in diffraction-limited images when their separation fell below approximately 360 nm. This result exceeded the twofold gain cited for conventional linear SIM.
The reconstructed distances are optical estimates rather than direct measurements of buried emitter positions. Separate nanodiamond measurements supported reconstruction fidelity because scanning electron microscopy confirmed their emitter positions. The same reconstructed emitters showed ODMR contrast under microwave excitation, which supported their identification as NV ensembles.
Creating a buried NV ensemble
Section titled “Creating a buried NV ensemble”Femtosecond irradiation at 780 nm created NV centers in type 1b HPHT diamond. A confocal microscope with a tunable infrared pulsed laser performed the writing under ambient conditions. Volumetric Z-stack fluorescence imaging located the emission layer approximately 100 µm below the surface. Characterization used 514 nm continuous-wave excitation and collected fluorescence from 600 to 800 nm through a 40×, 0.95 NA objective.
Laser-generated GR1 vacancies recombined with nearby substitutional nitrogen inside the focal volume. The authors report that this route avoided high-temperature annealing and minimized collateral damage and graphitization. Each written square measured 200 × 200 nm². Paired squares had nominal center-to-center separations of 250 nm, 300 nm, and 400 nm. The corresponding NV-forming regions spanned approximately 50 to 450 nm, 100 to 500 nm, and 200 to 600 nm. Stochastic NV generation produced local spacing smaller than the written-square geometry implied.
How diamond extends the illumination bandwidth
Section titled “How diamond extends the illumination bandwidth”The authors’ model treats incident speckle as compressed when it enters diamond with refractive index n ≈ 2.4 from air with n = 1. FDTD simulations predicted finer real-space features in diamond than in air. The simulated Fourier spectra extended beyond the free-space cutoff k₀ to an in-diamond cutoff of nk₀. This higher spatial-frequency content supplied the sub-diffraction modulation needed for structured illumination.
The experiments inferred the excitation structure from fluorescence modulation rather than mapping the compressed field directly. Direct field mapping at the buried interface was not available, so the high-k structure remained model-supported and indirectly encoded.
Synchronizing speckle and ODMR acquisition
Section titled “Synchronizing speckle and ODMR acquisition”The negatively charged nitrogen-vacancy center is a spin-1 system with m_s = 0 and m_s = ±1 ground-state sublevels. The zero-field splitting between m_s = 0 and m_s = ±1 was 2.87 GHz. Optical excitation produced about 20% less fluorescence from the m_s = ±1 states than from m_s = 0.
A 532 nm continuous-wave laser was modulated by an acousto-optic modulator, and its first-order beam entered a multimode fiber. Mechanical stretching and a motorized fan generated successive speckle patterns. Patterns reached the diamond through a 60×, 0.9 NA objective. A loop antenna applied radio-frequency fields while a nearby magnet supplied the static field. The RF frequency step was 10 MHz.
A pulse blaster synchronized the laser, RF source, and sCMOS camera. For each static speckle pattern, the camera recorded fluorescence across a full RF sweep under RF-on and RF-off conditions. A 600 nm long-pass filter rejected excitation light before detection.
Reconstruction and acquisition cost
Section titled “Reconstruction and acquisition cost”Blind-SIM iteratively estimated both the effective illumination and the sample distribution from low-resolution frames. The MATLAB implementation minimized a cost function comparing measured images with predicted data. The full dataset contained 31,320 raw frames. It comprised 261 RF frequency points, 20 frames per point, and six repetitions. Each point used ten RF-on and ten RF-off frames.
At a 50 ms frame period and 20 frames per second, full acquisition took approximately 26.1 minutes, excluding minor hardware overhead. Only 100 to 300 frames were required for reconstruction. Toggling RF near identified resonances could reduce acquisition to approximately 5 to 15 seconds. The reported excitation power was approximately 60 mW.
A speckle-only reconstruction used the same frame count at a fixed off-resonant RF frequency. It separated emitters less robustly than the combined speckle and ODMR reconstruction. This matched comparison supports a reconstruction contribution from ODMR modulation. The reconstruction ran on an NVIDIA GeForce RTX 5090 and 44-core CPU for 300 × 300 pixel images.
Magnetic field mapping
Section titled “Magnetic field mapping”A representative spectrum contained eight peaks, corresponding to four NV axis orientations. The wide-field image was divided into tiles, and pixel intensities across the stack formed local ODMR spectra. The demonstration used four tiles, and each spectrum contained six peaks from three NV orientations. Hyperfine splitting depends on the magnetic-field projection along each NV axis.
The outermost resonances had the largest Zeeman splitting and were assumed to come from axes most closely aligned with the field. The analysis solved the NV Hamiltonian for field strength and projection angle using lower and upper resonance frequencies. A numerical solver matched calculated and measured eigenfrequencies, while least-squares fitting and outlier rejection constrained the inversion. The inversion selected an optimum path within a trusted radius in field-parameter space. Regions with insufficient signal-to-noise ratio were excluded from the final map.
The retrieved vectors were uniform, consistent with magnet field lines distributed uniformly over several tens of millimeters. Repeated measurements across laser-written sites gave relative field-strength uncertainty below approximately 6% and angular precision within 1 to 3°. The authors describe this vector retrieval as a proof of concept.
Limits and engineering implications
Section titled “Limits and engineering implications”Localization-based methods can offer comparable nominal resolution, but the paper says they usually require sparse, independently blinking emitters. That requirement limits their fit for dense subsurface NV ensembles in bulk diamond. STED and SMLM can require scanning or point-by-point excitation, whereas the demonstrated method uses wide-field speckle acquisition. Nonlinear SIM can improve resolution in principle, but high intensities and specialized photoswitchable fluorophores do not fit dense NV ensembles.
The method obtains its high-k effect from the diamond host instead of an added engineered substrate. The authors state that moderate excitation intensity helps preserve NV spin coherence and reduce laser-induced heating. Imaging about 100 µm deep with a 0.9 NA air objective introduced spherical aberration. Blind-SIM and ODMR modulation helped retain reconstruction quality under those conditions. Adaptive optics or immersion objectives could reduce the aberration, but the paper presents them as future options. Lock-in camera detection could later improve magnetic sensitivity and precision.
Direct nanoscale characterization of the buried emitter distribution was not available through scanning electron microscopy or atomic force microscopy. The high-k excitation field was also not mapped directly, leaving near-field probes or a thin fluorescent layer as future validation methods. For material selection, the relevant properties are bulk optical access, a controlled buried NV layer, and compatibility with the 532 nm microscope.
Materials and support
Section titled “Materials and support”The imaging application calls for a diamond plate with optical access to a laser-written NV layer about 100 µm below one surface. The diamond materials catalog lists single-crystal and polycrystalline MPCVD plates and wafers up to 125 mm, with thickness from 0.1 to 500 µm. The paper used type 1b HPHT diamond, so nitrogen content needs confirmation before substitution; engineering support can review the writing depth and optical geometry.
View Original Abstract
Structured illumination microscopy (SIM) allows for optical imaging with spatial resolution that surpasses the diffraction limit. To further enhance this super-resolution capability, we present a multimodal approach that integrates SIM with optically detected magnetic resonance (ODMR) of nitrogen-vacancy (NV) centers in diamond-a platform considered one of the most promising platforms for quantum sensing and computing. Both theoretical modeling and experimental results reveal that diamond’s high refractive index supports finer structured illumination patterns, thereby accessing higher spatial frequencies and yielding improved resolution. This combined strategy not only enhances NV center localization precision and achieves sub-100-nm resolution but also facilitates spin coherence analysis. Leveraging the high refractive index of diamond in SIM thus offers a powerful pathway toward advancing quantum applications at the nanoscale.
Single Crystal Cvd Diamond Plate
View Specs →
Boron Doped Single Crystal Diamond Plate
View Specs →