Skip to content

Nanodiamond Spin Probes for Sensing Reactive Intermediates in Heterogeneous Photocatalysis

FieldValue
JournalAdvanced Functional Materials
Published2026-07-09
AuthorsBradley T. Flinn, Robert E. Munday, E. Stephen Davies, Andrei N. Khlobystov, Melissa L. Mather
DOI10.1002/adfm.76935

NV centres in fluorescent nanodiamond provided a local readout of reactive oxygen species from heterogeneous photocatalysts. The Fenton assay reached about 100 nM sensitivity. Lower superoxide response also tracked greater epoxide selectivity in potassium-doped graphitic carbon nitride.

Spin-active molecules reduced NV photoluminescence contrast by depolarizing nearby NV spins. The sensors were carboxylated FND particles with an average diameter of about 100 nm and more than 1,000 NV centres per particle.

The authors deposited 5 µL of a 0.1 mg mL−1 FND suspension on isopropanol-cleaned glass slides. Each sample dried for more than 4 hours at 60 °C. An inverted microscope supplied 550 nm excitation at about 1 mW cm−2 through a 60× objective with a numerical aperture of 1.49.

Optically detected magnetic resonance used microwaves at 2.87 GHz to saturate the NV transition. A comparison frequency of 2.77 GHz provided an off-resonance reference. The authors recorded photoluminescence with an sCMOS camera over a maximum field of 200 × 200 µm. The sample temperature did not exceed 28 °C during standard measurements.

A lower difference between on-resonance and off-resonance photoluminescence indicated greater depolarization. This contrast change served as the quantitative measure of spin-active species near the nanodiamond.

Fenton reaction establishes radical sensitivity

Section titled “Fenton reaction establishes radical sensitivity”

The Fenton reaction provided a controlled source of hydroxyl radicals. Added hydrogen peroxide oxidized Fe²⁺ to Fe³⁺ while generating •OH in water. The colour changed immediately from pale green to orange-brown as Fe³⁺ formed.

Iron controls produced only a small contrast decrease before radical generation. Fe³⁺ caused a slightly larger response than Fe²⁺, consistent with its greater number of unpaired electrons. The hydroxyl-radical response was much larger and was not masked by iron.

The authors varied Fe²⁺ and H₂O₂ concentrations from 1 nM to 10 mM. The resulting method sensitivity was of the order of 100 nM. DMPO was not required to observe the bulk Fenton radical signal.

Adding 10 mM DMPO produced a similar concentration trend. However, the spin population then contained iron species, •OH, and DMPOX formed by DMPO oxidation. The agreement still supported direct hydroxyl-radical detection in the homogeneous reaction.

Earlier FND relaxometry work detected •OH during H₂O₂ photolysis under UVC irradiation. That method required more complex equipment, longer experiments, and produced larger errors. The ODMR protocol used a common optical microscope and microwave hardware.

The method then tested commercial P25 TiO₂ and 1.8 w% potassium-doped graphitic carbon nitride, abbreviated 1.8K-g-CN. P25 contains TiO₂ nanoparticles with an average diameter of about 25 nm. The carbon nitride was prepared as micron-sized flakes.

The catalysts and FNDs were drop cast separately onto glass slides. The samples contained aqueous H₂O₂ and 100 mM DMPO. TiO₂ was illuminated with 370 nm UV light. K-g-CN was illuminated with 456 nm blue light.

The catalyst samples had different dark contrast levels. TiO₂ produced no intrinsic ODMR response and retained 100% residual contrast. In contrast, 1.8K-g-CN retained 47% residual contrast. EPR and XPS associated this response with catalyst radicals, including an EPR line at g = 2.004.

Unlike the Fenton reaction, photocatalysis produced an observable contrast reduction only when DMPO was present. The authors attribute this difference to surface-generated radicals. These radicals have short lifetimes and can recombine with the catalyst before entering solution.

H₂O₂ concentrations ranged from 1 nM to 100 mM. The detection onset was approximately 1 µM for TiO₂ and 100 µM for 1.8K-g-CN. Control measurements without a catalyst produced no measurable contrast change under either illumination condition.

Two factors could explain the different onsets. First, NV response scales with the inverse cube of source distance. TiO₂ nanoparticles can pack closer to FNDs than micron-sized carbon nitride flakes. Second, carbon nitride has a typical valence-band position of about 1.4 V versus NHE. This prevents H₂O₂ oxidation under irradiation and changes the balance between radical generation and back reactions.

EPR provided complementary identification of the spin adducts detected by the nanodiamond probes. The EPR measurements identified different products in aqueous and nonaqueous photocatalysis.

Photocatalysis conditionAssigned EPR featureCoupling constantsg-value
TiO₂ or 1.8K-g-CN with aqueous H₂O₂ and DMPODMPO-OH quartetaN ≈ aβH = 1.5 mT2.006
1.8K-g-CN with DMPO in acetonitrileDMPO-OOHaN = 1.35 mT and aβH = 1.09 mT2.008

The aqueous DMPO-OH spectrum appeared under illumination for both catalysts. This supported assignment of the spin-active products as hydroxyl-radical adducts.

In acetonitrile, the observed DMPO-OOH quartet supported superoxide formation. The DMPO-OOH signal remained stable for more than 45 minutes in acetonitrile. No DMPO-OH analogue was detected above noise under those conditions.

EPR has high sensitivity but poor spatial resolution for this application. The nanodiamond readout supplied the local response, while EPR established the adduct identity.

Superoxide response tracks product selectivity

Section titled “Superoxide response tracks product selectivity”

The authors then examined potassium-doped graphitic carbon nitrides with loadings from 0 to 6 w%. Acetonitrile supplied the nonaqueous medium, dissolved O₂ supplied the oxidant, and DMPO trapped the photogenerated radicals.

Samples were irradiated at 456 nm for 10 minutes before recording the second ODMR spectrum. This allowed the DMPO-OOH adduct to accumulate. Each NV result averaged two independent photocatalytic runs.

A weaker light-induced contrast reduction corresponded to less superoxide formation. These catalysts also produced greater epoxide selectivity during alkene oxidation. The relationship was clearest for 3.3K-g-CN.

The proposed mechanism assigns epoxide formation to hydrogen peroxide, which is spin silent under the reported conditions. Superoxide promotes oxidative cleavage toward the aldehyde product. Increasing potassium loading is associated with greater hydrogen peroxide formation and reduced superoxide production.

The authors tested this correlation with trans-anethole under batch conditions. Each reaction contained 10 mg catalyst, 0.4 mmol trans-anethole, and 4 mL acetonitrile. Irradiation continued for 4 hours at 456 nm with continuous stirring. The irradiance was 1.24 W cm−2 at a lamp distance of 2 cm. Gas chromatography and NMR measured conversion, epoxide yield, and aldehyde yield.

Optical detection remains a constraint for photocatalysts that emit visible light. Mobile carbon nitride flakes partially obscured the diamond photoluminescence and produced data artefacts. Background subtraction can separate red-region catalyst emission from the NV response, but it does not remove every artefact.

Inherently paramagnetic catalysts also reduce the initial ODMR contrast. The 47% residual contrast measured for 1.8K-g-CN lowered its signal-to-noise ratio and observable spin-noise sensitivity. Dark catalyst measurements are therefore needed to separate intrinsic contrast loss from light-induced depolarization.

The residual contrast saturated in the millimolar H₂O₂ regime. The authors suggest spin-trap quenching or saturation around the FND surface. Deeper implanted NV centres in the 100 nm particles may remain insensitive to surrounding solution spins. The reported sensing range for superoxide detection was about 20 nm.

The supplied manuscript was posted as a preprint on 15 February 2026. It states that the data may be preliminary and that the work had not been peer reviewed.

The method supports a local screening workflow for heterogeneous photocatalysts. Illumination-induced NV contrast loss indicates spin-active chemistry near the catalyst surface. EPR can then identify whether the trapped species correspond to hydroxyl radicals or superoxide.

For selective alkene epoxidation, the reported correlation favors catalysts with a weaker superoxide signal. The 3.3K-g-CN material showed the clearest combination of low radical response and high epoxide selectivity. This offers a selection criterion based on radical formation rather than product analysis alone.

The experiments do not yet map regional activity across a catalyst surface. The authors present nanoscale surface sensing as a future use of the method. The present result is a measured contrast change caused by illumination, combined with EPR identification of the radical adduct.

View Original Abstract

ABSTRACT Heterogeneous photocatalysts are promising materials for turning solar energy into chemical fuels, but their widespread application is hampered by low efficiency and selectivity. To facilitate future catalyst design, the evaluation of reactive intermediates, such as reactive oxygen species (ROS), formed during photocatalysis is vital. Challenges arise in ROS detection due to their high reactivity and interactions with the catalyst surface. Techniques such as electron paramagnetic resonance (EPR) often rely on bulk‐sensing apparatus that are highly sensitive but have poor spatial resolution. Here, a fast in situ method for evaluating the performance of photocatalysts in ROS formation is developed, utilizing nitrogen‐vacancy (NV) centers in fluorescent nanodiamond (FND) particles as sensors. This work demonstrates the ability to detect ROS (e.g., • OH, O 2 •- ) in both the absence and presence of spin‐trap adducts (e.g., 5,5‐Dimethyl‐1‐pyrroline N‐oxide, DMPO), the latter allowing for further ROS identification by EPR. An exploration of various photocatalysts demonstrates the technique’s ability to probe ROS generation during photocatalysis. Correlation with bulk photocatalytic activity highlights the method’s suitability for rapidly evaluating photocatalyst efficiency. This methodology shows promise for sensing local, nanoscale photocatalytic activity on catalyst surfaces and can act to expedite the discovery of next‐generation catalysts.