Fluorescent Nanodiamond Layers

• Fluorescent nanodiamond layers are engineered NV-containing structures that preserve and tune photoluminescence, spin, and charge responses.
• They are fabricated via diverse methods such as covalent attachment, electrostatic self-assembly, electrospinning, and CVD growth, each offering unique trade-offs in coverage and performance.
• These layers underpin applications in quantum imaging, biointerfaces, and chemical sensing by balancing optical brightness and analyte accessibility.

Fluorescent nanodiamond (FND) layers are organized forms of NV-containing nanodiamonds deployed as surface-bound arrays, submonolayer coatings, electrospun composite mats, microwave-plasma CVD-grown nanodiamond films, or shell-engineered core–shell particles. In this literature, the unifying function of an FND layer is to preserve or tune the photoluminescence, spin, or charge-state response of the nitrogen-vacancy center while enabling integration with technologically relevant substrates, biointerfaces, or chemically active microenvironments. Reported implementations span covalently assembled nanoparticle arrays, electrostatically self-assembled coatings for magnetic imaging, hydrogenated layers for all-optical voltage and ion readout, PLGA nanofiber mats for cell-compatible sensing, CVD-grown heavy-nitrogen-shell nanodiamond layers with long $T_1$, and double-layer silica architectures that combine radical generation with relaxometric detection (Kianinia et al., 2016, Price et al., 2018, Chea et al., 5 Aug 2025, Voorhoeve et al., 15 Feb 2026, Su et al., 26 Dec 2025, Prooth et al., 12 May 2026).

1. Physical basis and architectural scope

The dominant optically active defect in these systems is the NV center. In FND-polymer nanofiber mats, the NV spin Hamiltonian is written as

$H = D\,S_z^2 + \gamma_e\,B\cdot S,$

with zero-field splitting $D \simeq 2.870\ \mathrm{GHz}$ and electron gyromagnetic ratio $\gamma_e \simeq 28\ \mathrm{GHz/T}$; under 532 nm excitation, the same platform shows broadband photoluminescence extending to $\sim 800\ \mathrm{nm}$ with a zero-phonon line at 637 nm (Price et al., 2018). Self-assembled FND coatings for quantum imaging likewise show the NV$^-$ zero-phonon line at 637 nm and a broad phonon sideband above 650 nm (Chea et al., 5 Aug 2025).

Within the cited work, “layer” does not denote a single morphology. Kianinia et al. reported lithographically defined arrays of carboxylated 35 nm FNDs covalently attached to amine-functionalized EBID carbon seeds on planar and non-planar substrates (Kianinia et al., 2016). Chea et al. reported dense and homogeneous electrostatically self-assembled coatings on Si and quartz (Chea et al., 5 Aug 2025). Other work used hydrogenated FNDs to form submonolayers of loosely packed particles on quartz or ITO, enabling charge-state-based voltage and ion concentration imaging (Voorhoeve et al., 15 Feb 2026). Electrospun PLGA/fND nanofiber mats constitute a three-dimensional fibrous layer rather than a planar particulate coating (Price et al., 2018). A distinct category is the CVD-grown nanodiamond layer grown by heterogeneous nucleation and terminated by a heavy-nitrogen shell-doping step (Prooth et al., 12 May 2026). At the particle level, Su et al. described a double-layer silica architecture in which an inner dense silica shell preserves NV properties and an outer mesoporous shell mediates catalytic radical chemistry (Su et al., 26 Dec 2025).

A recurrent source of confusion is to treat all FND layers as interchangeable replacements for bulk diamond. The reports instead describe architecture-specific trade-offs among optical brightness, aggregation, spin relaxation, surface accessibility, and substrate compatibility. This suggests that “FND layer” is best understood as a family of engineered geometries rather than a single materials platform.

2. Assembly on solid supports

The most spatially resolved substrate-bound architecture in the cited literature is the EBID-directed array. Kianinia et al. formed amorphous-carbon disks of approximately 90 nm diameter and approximately 20 nm height by parking a defocused 15 keV, 300 pA electron beam for 30 s in naphthalene vapor. After NH$_3$ RIE at 6 Pa and 100 W for 45 s, the seed surfaces were converted to a high density of $-$NH$_2$ groups, and commercial 35 nm $-$COOH-terminated FNDs were coupled from water using EDC at a fixed molar ratio $H = D\,S_z^2 + \gamma_e\,B\cdot S,$0. With an FND concentration of $H = D\,S_z^2 + \gamma_e\,B\cdot S,$1 and 6 h immersion at room temperature under gentle agitation, more than 92% of seeds captured at least one FND and no off-seed FNDs were observed, corresponding to 100% lateral selectivity (Kianinia et al., 2016). The same report gives a minimal feature size of about 90 nm and a routine pitch of $H = D\,S_z^2 + \gamma_e\,B\cdot S,$2.

A chemically simpler route is electrostatic self-assembly. In Chea et al., Si or quartz chips of $H = D\,S_z^2 + \gamma_e\,B\cdot S,$3 were cleaned by sonication in acetone, ethanol, and DI water, followed by UV–ozone treatment. A positively charged PAH layer was formed by 5 min vertical immersion in $H = D\,S_z^2 + \gamma_e\,B\cdot S,$4–$H = D\,S_z^2 + \gamma_e\,B\cdot S,$5 PAH, producing a surface zeta potential $H = D\,S_z^2 + \gamma_e\,B\cdot S,$6. Negatively charged commercial HPHT FNDs with 120 nm hydrodynamic diameter were then assembled from aqueous suspension by varying concentration $H = D\,S_z^2 + \gamma_e\,B\cdot S,$7 from $H = D\,S_z^2 + \gamma_e\,B\cdot S,$8 to $H = D\,S_z^2 + \gamma_e\,B\cdot S,$9, immersion time $D \simeq 2.870\ \mathrm{GHz}$0 from $D \simeq 2.870\ \mathrm{GHz}$1 to $D \simeq 2.870\ \mathrm{GHz}$2, and pH between 6.4 and 4.0 (Chea et al., 5 Aug 2025). The reported optimum for maximizing single-particle density while minimizing aggregates was $D \simeq 2.870\ \mathrm{GHz}$3, $D \simeq 2.870\ \mathrm{GHz}$4, immersion time $D \simeq 2.870\ \mathrm{GHz}$5, and pure DI water as the low-ionic-strength medium. Under those conditions, AFM gave a surface coverage $D \simeq 2.870\ \mathrm{GHz}$6, corresponding to a particle density of approximately $D \simeq 2.870\ \mathrm{GHz}$7, while confocal PL exceeded 20 kcps on 91.4% of the surface and 10 kcps on 99.9% of the surface (Chea et al., 5 Aug 2025).

The adsorption kinetics in this self-assembled system were described by a random sequential adsorption model,

$D \simeq 2.870\ \mathrm{GHz}$8

and, in the diffusion-limited regime, by $D \simeq 2.870\ \mathrm{GHz}$9, matching the empirical $\gamma_e \simeq 28\ \mathrm{GHz/T}$0 scaling of the single-particle count (Chea et al., 5 Aug 2025). A key practical point is that increasing concentration does not monotonically improve layer quality: sparse coverage was observed at $\gamma_e \simeq 28\ \mathrm{GHz/T}$1, the single-particle count peaked at $\gamma_e \simeq 28\ \mathrm{GHz/T}$2, and the overall density decreased at $\gamma_e \simeq 28\ \mathrm{GHz/T}$3 because of aggregation in suspension. Likewise, lowering pH reduced the magnitude of the negative FND zeta potential and initially increased surface loading, but below pH 4.5 the colloidal stability collapsed and coverage fell to zero (Chea et al., 5 Aug 2025).

A related but electrochemically distinct layer was formed from hydrogenated sub-30 nm FNDs. After electron irradiation, argon annealing, and oxidation, FND-Oxy powders were annealed in forming gas at $\gamma_e \simeq 28\ \mathrm{GHz/T}$4 for 1 h to yield FND-Hyd with surface C–H and CH$\gamma_e \simeq 28\ \mathrm{GHz/T}$5 groups and zeta potentials greater than $\gamma_e \simeq 28\ \mathrm{GHz/T}$6, compared with approximately $\gamma_e \simeq 28\ \mathrm{GHz/T}$7 for FND-Oxy. Quartz or ITO-coated coverslips were cleaned, immersed in $\gamma_e \simeq 28\ \mathrm{GHz/T}$8 FND-Hyd suspension for 10 min, rinsed, and dried, yielding submonolayers of loosely packed FNDs. For wide-field voltage imaging on ITO, $\gamma_e \simeq 28\ \mathrm{GHz/T}$9 suspensions were spin-cast at 1000 rpm for 10 s and 4000 rpm for 30 s, then baked at $\sim 800\ \mathrm{nm}$0 for 10 min (Voorhoeve et al., 15 Feb 2026).

3. Composite and in situ grown layer systems

Electrospun FND layers embed nanodiamonds in a fibrous polymer scaffold. In the PLGA platform, a pellet obtained from a commercial aqueous suspension of 100 nm FNDs with more than $\sim 800\ \mathrm{nm}$1 NV$\sim 800\ \mathrm{nm}$2 centers per particle was dried under N$\sim 800\ \mathrm{nm}$3, resuspended in methanol, diluted with chloroform to a 2:1 CF:MeOH mixture, and combined with 80 mg PLGA in a final volume of 1 mL to yield an 8% w/v polymer solution. After stirring and 1 h sonication, the viscous suspension was electrospun at $\sim 800\ \mathrm{nm}$4 through a stainless-steel needle at typically $\sim 800\ \mathrm{nm}$5–$\sim 800\ \mathrm{nm}$6 with a 10–15 cm needle-to-collector distance (Price et al., 2018). Mats of tens of microns thickness were obtained by adjusting spin duration from 30 min to 2 h. Image analysis gave a mean diameter of 365 nm for control fibers and 306 nm for fND-loaded fibers; after 1 h spinning, mat thickness was typically 10–30 $\sim 800\ \mathrm{nm}$7m. In the optimized CF:MeOH (2:1) formulation, fluorescence micrographs showed no large aggregates above $\sim 800\ \mathrm{nm}$8m and more than 90% of FNDs were spaced less than $\sim 800\ \mathrm{nm}$9m apart along well-formed fibers (Price et al., 2018).

Microwave-plasma CVD provides a fundamentally different route in which the nanodiamond layer is grown rather than deposited. Reported growths used a 2.45 GHz microwave reactor at 120 mbar and 1 kW with total gas flow $^-$0, CH$^-$1 flow of $^-$2, substrate temperatures between 700 and $^-$3, and growth durations of 9–15 min to obtain approximately 50–200 nm average particle heights (Prooth et al., 12 May 2026). Substrate preparation used Si(100) or sapphire wafers, optionally patterned with 200 nm PMMA holes on a 1–2 $^-$4m square grid or seeded with 5 nm commercial nanodiamonds. The growth profile consisted of an H$^-$5 ramping step, an H$^-$6+CH$^-$7 growth step, and a 30 s H$^-$8+CH$^-$9+N$_3$0 doping step for heavy-nitrogen shell formation (Prooth et al., 12 May 2026). The temperature dependence showed the expected competition between nucleation density and vertical growth rate: lower temperature produced smaller particles and higher nucleation density, whereas higher temperature reduced nucleation density and increased vertical growth.

The shell-doping step is central to the CVD-grown layer concept. During the final 30 s of growth, $_3$1 was introduced, corresponding to a chamber concentration of approximately 1.25% and requiring about 10 s to saturate. For faster shell saturation, CH$_3$2 could be pulsed to $_3$3 for 2 s before returning to $_3$4, reducing CH$_3$5 steady-state time from roughly 30 s to roughly 2 s (Prooth et al., 12 May 2026). The resulting $_3$6-doped shell thicknesses were on the order of 5–15 nm. Secondary nucleation frequently occurred on $_3$7 facets during the nitrogen pulse, producing small satellite crystallites and occasional twinning; the report states that shorter $_3$8 pulses and lower substrate temperature yield more uniform shells (Prooth et al., 12 May 2026).

A further layered architecture exists at the single-particle level rather than as a substrate coating. In the double-layer silica design, 40 nm carboxylated FNDs were first coated with a dense silica shell via TEOS hydrolysis and condensation in ethanol with NH$_3$9OH, producing a standard inner-shell thickness of $-$0 and a tunable range from $-$1 to $-$2. A second TEOS/CTAB step created an outer mesoporous shell approximately 20–30 nm thick, giving core–shell particles with overall diameter $-$3 and zeta potential around $-$4 after template removal (Su et al., 26 Dec 2025). This architecture is not a planar FND film, but it is nonetheless a layered FND system in the strict structural sense.

4. Morphology, optical response, and spin-state characterization

Morphological characterization across the literature relies on SEM, AFM, DLS, and fluorescence mapping. In the CVD-grown layers, SEM imaging was carried out with Zeiss 450 FEGSEM (Gemini 2) or FEI Quanta instruments at nominal 2–3 nm resolution, with image analysis in ImageJ using Gaussian blur, thresholding, watershed, and particle analysis; the reported SEM datasets showed sub-100 nm and even sub-50 nm crystals with clean morphology (Prooth et al., 12 May 2026). In self-assembled coatings for quantum imaging, AFM scans of $-$5 with 1024×1024 pixels and a 55 nm height threshold were used to separate single-particle platelets from aggregates, with typical FND $-$6-heights of 40–80 nm (Chea et al., 5 Aug 2025). In the hydrogenated sub-30 nm layers, AFM showed aggregates of approximately 100–200 nm lateral extent with $-$7-heights not exceeding 25 nm, while DLS indicated hydrodynamic diameters increasing from approximately 30 nm for FND-Oxy to approximately 50 nm for FND-Hyd (Voorhoeve et al., 15 Feb 2026).

Photoluminescence served both as a functional signal and, in the CVD work, as a size proxy. For shell-doped CVD FNDs, confocal PL maps of $-$8 with 100×100 pixels were processed in Python/SciPy, and the total PL intensity was related to particle diameter through

$-$9

implying

$_2$0

with $_2$1, $_2$2 at 50 $_2$3W, and $_2$4 (Prooth et al., 12 May 2026). In that study, PL-derived size distributions agreed well with SEM histograms, including at 1 mW excitation for enhanced detection of particles smaller than 100 nm. In the PLGA nanofiber mats, normalized emission spectra from drop-cast beads and embedded fibers overlaid closely, and no significant quenching by PLGA was reported (Price et al., 2018).

Spin-state characterization is equally architecture dependent. In the CVD-grown shell-doped FNDs, inversion-recovery measurements used

$_2$5

with $_2$6, while Hahn-echo data were fit to $_2$7 with $_2$8 (Prooth et al., 12 May 2026). Typical $_2$9 values for shell-doped CVD FNDs were 650–1035 $-$0s, the mean longitudinal coherence time was reported as 800 $-$1s, and the maximum exceeded 1.8 ms, close to bulk theoretical values for particles averaging around 60 nm in size (Prooth et al., 12 May 2026). By contrast, the same shell-doped particles showed $-$2s, compared with approximately 2–3 $-$3s in reference HPHT powder (Prooth et al., 12 May 2026). The paper explicitly models the relaxation rates as

$-$4

and

$-$5

arguing that the combination of high $-$6 and short $-$7 indicates suppression of low-frequency magnetic noise but persistent high-frequency surface magnetic or charge noise (Prooth et al., 12 May 2026).

Other FND layers show different spin responses. In the PLGA mats, the spin-polarization recovery followed

$-$8

with measured $-$9s for drop-cast fNDs and $H = D\,S_z^2 + \gamma_e\,B\cdot S,$00s for embedded fibers, indicating approximately twofold shortening upon embedding, attributed to changed surface chemistry, polymer proximity, and mechanical strain (Price et al., 2018). In the double-layer silica particles, the baseline relaxometry value was $H = D\,S_z^2 + \gamma_e\,B\cdot S,$01s for ensemble silica-FNDs and $H = D\,S_z^2 + \gamma_e\,B\cdot S,$02s for raw single-particle FNDs, while the dense inner shell preserved $H = D\,S_z^2 + \gamma_e\,B\cdot S,$03–$H = D\,S_z^2 + \gamma_e\,B\cdot S,$04s for near-surface NV centers (Su et al., 26 Dec 2025). A common misconception is therefore that improvements in one spin metric necessarily generalize across all layered architectures; the published data do not support that simplification.

5. Quantum sensing modalities

Magnetic imaging with FND layers has been demonstrated in both electrospun and self-assembled formats. In the PLGA nanofiber system, wide-field continuous-wave ODMR was performed with 532 nm excitation, a 1.49 NA 60× TIRF objective, and microwave sweeps from 2.77 to 2.97 GHz in 2 MHz steps delivered through a 0.125 mm Cu wire. Ten sweeps were recorded in approximately 2 s, yielding a spectrum with a contrast dip near 2.87 GHz, full width at half maximum $H = D\,S_z^2 + \gamma_e\,B\cdot S,$05, and contrast $H = D\,S_z^2 + \gamma_e\,B\cdot S,$06–$H = D\,S_z^2 + \gamma_e\,B\cdot S,$07 (Price et al., 2018). A two-point CW-ODMR protocol provided dynamic range up to approximately 1 mT before saturation, 20 ms acquisition for two images, a 50 Hz update rate, a minimum detectable field of approximately $H = D\,S_z^2 + \gamma_e\,B\cdot S,$08, and an estimated sensitivity of $H = D\,S_z^2 + \gamma_e\,B\cdot S,$09 for a single-ND region of interest (Price et al., 2018).

Self-assembled dense coatings on quartz were used for magnetic field and magnetic noise imaging. In the ODMR analysis, the effective local field was written as

$H = D\,S_z^2 + \gamma_e\,B\cdot S,$10

with $H = D\,S_z^2 + \gamma_e\,B\cdot S,$11 (Chea et al., 5 Aug 2025). Without Fe$H = D\,S_z^2 + \gamma_e\,B\cdot S,$12O$H = D\,S_z^2 + \gamma_e\,B\cdot S,$13, the linewidth was $H = D\,S_z^2 + \gamma_e\,B\cdot S,$14, corresponding to $H = D\,S_z^2 + \gamma_e\,B\cdot S,$15; with Fe$H = D\,S_z^2 + \gamma_e\,B\cdot S,$16O$H = D\,S_z^2 + \gamma_e\,B\cdot S,$17, $H = D\,S_z^2 + \gamma_e\,B\cdot S,$18 and $H = D\,S_z^2 + \gamma_e\,B\cdot S,$19. The ROI histogram peak shifted from approximately 0.5 G to approximately 1.0 G on the relative scale used in the figure, and field maps resolved $H = D\,S_z^2 + \gamma_e\,B\cdot S,$20 up to 3.5 G around aggregates (Chea et al., 5 Aug 2025). The same platform performed $H = D\,S_z^2 + \gamma_e\,B\cdot S,$21 relaxometry using single-exponential fits to $H = D\,S_z^2 + \gamma_e\,B\cdot S,$22, with $H = D\,S_z^2 + \gamma_e\,B\cdot S,$23s without Fe$H = D\,S_z^2 + \gamma_e\,B\cdot S,$24O$H = D\,S_z^2 + \gamma_e\,B\cdot S,$25, $H = D\,S_z^2 + \gamma_e\,B\cdot S,$26–$H = D\,S_z^2 + \gamma_e\,B\cdot S,$27s with Fe$H = D\,S_z^2 + \gamma_e\,B\cdot S,$28O$H = D\,S_z^2 + \gamma_e\,B\cdot S,$29, and local regions under strong superparamagnetic noise dropping below 10 $H = D\,S_z^2 + \gamma_e\,B\cdot S,$30s, which was below the experimental resolution (Chea et al., 5 Aug 2025).

A distinct sensing mode is all-optical voltage and ion concentration imaging based on NV charge-state modulation. In sub-30 nm hydrogenated FND layers, surface hydrogenation drove near-surface NV centers toward the non-fluorescent NV$H = D\,S_z^2 + \gamma_e\,B\cdot S,$31 state through negative electron affinity, while UV–ozone restored positive electron affinity and reversed NV$H = D\,S_z^2 + \gamma_e\,B\cdot S,$32 back to NV$H = D\,S_z^2 + \gamma_e\,B\cdot S,$33 with approximately 70% PL recovery (Voorhoeve et al., 15 Feb 2026). In aqueous electrochemical cells containing 0.17 M NaCl, the voltage-dependent PL change was defined as

$H = D\,S_z^2 + \gamma_e\,B\cdot S,$34

FND-Hyd exhibited a monotonic increase of +42% at $H = D\,S_z^2 + \gamma_e\,B\cdot S,$35 and a decrease of $H = D\,S_z^2 + \gamma_e\,B\cdot S,$36 at $H = D\,S_z^2 + \gamma_e\,B\cdot S,$37, whereas FND-Oxy showed less than 5% variation (Voorhoeve et al., 15 Feb 2026). In the linear regime $H = D\,S_z^2 + \gamma_e\,B\cdot S,$38, the slope was approximately $H = D\,S_z^2 + \gamma_e\,B\cdot S,$39, and the best-performing aggregate yielded a shot-noise-limited voltage sensitivity of $H = D\,S_z^2 + \gamma_e\,B\cdot S,$40 (Voorhoeve et al., 15 Feb 2026). The initial PL response occurred within $H = D\,S_z^2 + \gamma_e\,B\cdot S,$41, limited by the camera, and the effective voltage-imaging resolution was reported as approximately 400 nm for individual aggregates (Voorhoeve et al., 15 Feb 2026).

The same hydrogenated layers were used for ionic imaging. Under $H = D\,S_z^2 + \gamma_e\,B\cdot S,$42 bias in 1.7 M NaCl between platinum microelectrodes separated by 160 $H = D\,S_z^2 + \gamma_e\,B\cdot S,$43m, simulations based on the Nernst–Planck equations gave local $H = D\,S_z^2 + \gamma_e\,B\cdot S,$44 up to about 7 mM near the electrodes, and the spatial PL maps reproduced the calculated concentration profile with sub-10 $H = D\,S_z^2 + \gamma_e\,B\cdot S,$45m fidelity (Voorhoeve et al., 15 Feb 2026). The concentration sensitivity was defined as

$H = D\,S_z^2 + \gamma_e\,B\cdot S,$46

with a mean value of approximately $H = D\,S_z^2 + \gamma_e\,B\cdot S,$47 and peaks at $H = D\,S_z^2 + \gamma_e\,B\cdot S,$48 for certain aggregates in the detailed analysis (Voorhoeve et al., 15 Feb 2026). The abstract of the same report summarizes the sensitivity as up to $H = D\,S_z^2 + \gamma_e\,B\cdot S,$49 per millimolar NaCl (Voorhoeve et al., 15 Feb 2026). This suggests that the apparent sensitivity is geometry- and aggregate-dependent, rather than a single invariant material constant.

The double-layer silica architecture extends FND layers into catalytic sensing. Su et al. expressed the NV relaxometry response in the presence of hydroxyl radicals as

$H = D\,S_z^2 + \gamma_e\,B\cdot S,$50

with $H = D\,S_z^2 + \gamma_e\,B\cdot S,$51 mapped to local radical concentration (Su et al., 26 Dec 2025). Gd(DTPA)$H = D\,S_z^2 + \gamma_e\,B\cdot S,$52 loading over 100 aM to 100 fM gave $H = D\,S_z^2 + \gamma_e\,B\cdot S,$53–$H = D\,S_z^2 + \gamma_e\,B\cdot S,$54 at 100 aM and $H = D\,S_z^2 + \gamma_e\,B\cdot S,$55–$H = D\,S_z^2 + \gamma_e\,B\cdot S,$56 at 100 fM, corresponding through Monte Carlo analysis to hydroxyl concentrations from approximately 0.1 M to several mol/L in the mesoporous shell (Su et al., 26 Dec 2025). The empirical calibration

$H = D\,S_z^2 + \gamma_e\,B\cdot S,$57

enabled real-time readout directly from $H = D\,S_z^2 + \gamma_e\,B\cdot S,$58 measurements. Among 177 individual MS-silica-FNDs challenged with 100 aM Gd(DTPA)$H = D\,S_z^2 + \gamma_e\,B\cdot S,$59, more than 85% showed positive $H = D\,S_z^2 + \gamma_e\,B\cdot S,$60 and $H = D\,S_z^2 + \gamma_e\,B\cdot S,$61 (Su et al., 26 Dec 2025).

6. Robustness, biointerfaces, scalability, and limitations

Mechanical and chemical robustness are decisive for any practical FND layer. In the covalent EBID/EDC arrays, Bransonic ultra-sonication at 185 W and 40 kHz produced no measurable detachment for up to 3 h, and after 12 h approximately 90% of the original FNDs remained bound (Kianinia et al., 2016). Because the attachment chemistry is based on amide formation between $H = D\,S_z^2 + \gamma_e\,B\cdot S,$62COOH-terminated FNDs and $H = D\,S_z^2 + \gamma_e\,B\cdot S,$63NH$H = D\,S_z^2 + \gamma_e\,B\cdot S,$64-rich EBID carbon seeds, the resulting arrays tolerate subsequent wet processing and were used for on-chip ODMR mapping with a 30 $H = D\,S_z^2 + \gamma_e\,B\cdot S,$65m copper microwire over the array, resolving local fields from 0 to 3 mT across more than 90 individual pixels (Kianinia et al., 2016).

Biological interfacing is best documented for the PLGA/fND nanofiber mats. Neural stem cells at passages 1–3 were seeded at $H = D\,S_z^2 + \gamma_e\,B\cdot S,$66 on poly-ornithine/laminin-coated mats, expanded for 2–3 days, and differentiated for 7 days. Viability was $H = D\,S_z^2 + \gamma_e\,B\cdot S,$67 on fibers versus approximately 88% on glass; pyknotic nuclei were 4% on fibers versus 2.5% on glass. Lineage-marker expression was approximately 45% GFAP-positive astrocytes on fibers versus approximately 43% on glass, 27% Tuj-1-positive neurons versus 22%, and 19% MBP-positive oligodendrocytes versus 14% (Price et al., 2018). After 21 days, neuronal networks were confirmed by Fluo-5F AM calcium imaging, and ODMR spectra were acquired from single FNDs under live neurons and from small cell aggregates (Price et al., 2018). These data support the narrower claim that FND-containing nanofiber layers can sustain prolonged culture while retaining quantum readout capability.

Scalability differs sharply among fabrication routes. Solution-based self-assembly requires no vacuum or CVD reactor and is explicitly described as compatible with wafer-scale or roll-to-roll coating on Si, glass, quartz, flexible polymers, and bioscaffolds (Chea et al., 5 Aug 2025). The CVD route, although more equipment-intensive, includes explicit scale-up strategies: sapphire or Si with a thin Mo sacrificial layer permits clean lift-off of large-area FND films; photolithography can define centimeter-scale arrays of nucleation sites at 1–10 $H = D\,S_z^2 + \gamma_e\,B\cdot S,$68m pitch; spin-coating or spray-seeding over 4-inch wafers yields more than $H = D\,S_z^2 + \gamma_e\,B\cdot S,$69 individual nanodiamonds per wafer; and wet etching of Mo or thin Ni interlayers in HCl or HNO$H = D\,S_z^2 + \gamma_e\,B\cdot S,$70 releases intact films transferable to glass, PDMS, or microfluidic chips (Prooth et al., 12 May 2026). The same study notes that Mo does not incorporate into diamond and survives 100 h H$H = D\,S_z^2 + \gamma_e\,B\cdot S,$71 plasma (Prooth et al., 12 May 2026).

The limitations reported across the literature are specific and nontrivial. Dense coverage is not automatically advantageous, because self-assembly at excessively high suspension concentration or too low pH promotes aggregation or colloidal collapse (Chea et al., 5 Aug 2025). Embedding FNDs in PLGA shortens $H = D\,S_z^2 + \gamma_e\,B\cdot S,$72 relative to drop-cast particles (Price et al., 2018). Long $H = D\,S_z^2 + \gamma_e\,B\cdot S,$73 in CVD-grown shell-doped nanodiamond does not imply long $H = D\,S_z^2 + \gamma_e\,B\cdot S,$74, because the two timescales couple to different spectral components of the noise bath (Prooth et al., 12 May 2026). Hydrogenated all-optical layers still exhibit heterogeneity because aggregates vary in sensitivity, monolayer coverage is challenging, and the long-term stability of hydrogenation under aqueous bias merits further study (Voorhoeve et al., 15 Feb 2026). In the silica-engineered particles, the stated advantage is precisely a trade-off: the inner dense shell passivates surface spin noise and preserves NV-center $H = D\,S_z^2 + \gamma_e\,B\cdot S,$75, while the outer mesoporous shell preserves pore-mediated access for water and radicals (Su et al., 26 Dec 2025). A plausible implication is that future FND-layer design will continue to balance passivation against analyte accessibility rather than maximizing either parameter in isolation.