Continuous-Wave ODMR: Principles & Applications

• CW-ODMR is a spectroscopy technique that continuously uses optical pumping and microwave driving to detect spin transitions through measurable changes in fluorescence.
• It leverages optical spin initialization and microwave-induced state redistribution to analyze spin Hamiltonians, resonance splitting, and spectral line shapes.
• The method supports various instrument architectures—from coplanar waveguides to high-frequency setups—for applications in room-temperature magnetometry and engineered AC sensing.

Continuous-wave optically detected magnetic resonance (CW-ODMR) is a magnetic-resonance spectroscopy modality in which optical excitation and microwave driving are applied continuously and resonance is read out as a microwave-induced change in an optical observable, most commonly fluorescence or photoluminescence. In the canonical nitrogen-vacancy (NV) implementation, continuous green illumination polarizes the spin into a bright state while a swept microwave field drives ground-state transitions, producing dips in the detected optical signal; in other material classes the same CW logic appears as photoluminescence-detected magnetic resonance (PLDMR) or related wavelength-resolved optical detection (Kollarics et al., 2020, Babashah et al., 2022, Negyedi et al., 2016, 2206.13636).

1. Definition and measurement logic

CW-ODMR is defined operationally by simultaneous continuous optical pumping and continuous microwave excitation during data acquisition. In the NV-center case, the experiment is “truly a CW optical pump + CW microwave drive” even when low-frequency modulation is superposed for phase-sensitive detection; microwave chopping or frequency modulation changes the readout protocol, not the continuous-wave character of the spin driving itself (Kollarics et al., 2020). In the broader ODMR literature, the same continuous-wave logic is described as recording the optical emission or absorption as a function of microwave frequency while the spin system is continuously optically initialized and read out (Babashah et al., 2022).

The optical observable is system dependent. For NV centers, resonance appears as a drop in photoluminescence when microwaves transfer population out of the brighter $m_s=0$ state into the darker $m_s=\pm1$ manifold (Dokai et al., 4 Jun 2026). In molecular and semiconductor PLDMR, the microwave field redistributes populations among spin sublevels of long-lived excitonic or pair states, and the measured quantity is the microwave-induced change in photoluminescence, often denoted $\Delta \mathrm{PL}$ or $\Delta \mathrm{PL}/\mathrm{PL}$ (Negyedi et al., 2016, 2206.13636). The same general ODMR condition was summarized as relying on optical spin initialization and spin-state-dependent optical properties, including fluorescence, phosphorescence, absorption, or photocurrent (Babashah et al., 2022).

CW-ODMR is distinct from pulsed ODMR. In the pulsed case, coherent sequences such as Rabi, Ramsey, or echo are used; in the CW case, the resonance spectrum is obtained directly from the steady-state optical response under simultaneous optical pumping and microwave driving. This distinction is not merely terminological. A continuous optical readout with continuously detected fluorescence is not, by itself, sufficient to classify an experiment as CW-ODMR unless the microwave field is tuned to the spin transition that is being optically read out (Wolfe et al., 2015).

2. Spin Hamiltonians, resonances, and optical contrast

For ensemble NV centers at low bias field, a widely used reduced Hamiltonian is

$$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$

with transition frequencies

$f_{i\pm}=D\pm\gamma B_i,$

where $D=2.87$ GHz, $\gamma = 28$ MHz/mT, and $B_i=\mathbf B\cdot \mathbf e_i$ is the projection of the magnetic field onto NV orientation $i$ (Dubey et al., 25 Apr 2025). In a more general near-zero-field NV description including strain, the Hamiltonian is written as

$m_s=\pm1$0

with $m_s=\pm1$1, $m_s=\pm1$2, and $m_s=\pm1$3 (Dokai et al., 4 Jun 2026).

At zero or weak field, the $m_s=\pm1$4 and $m_s=\pm1$5 states are often reorganized into the bright and dark superpositions

$m_s=\pm1$6

which are directly relevant for CW-ODMR-based AC sensing schemes and for transverse-field formulations (Dokai et al., 4 Jun 2026, Okaniwa et al., 2023). In zero field, the $m_s=\pm1$7 transitions are degenerate in the simplest NV picture; finite field lifts that degeneracy and produces Zeeman splitting, which is the basis of DC magnetometry and orientation-sensitive spectral analysis (Kollarics et al., 2020).

The optical contrast is ultimately determined by the microwave-induced change in steady-state population of the brighter spin state. One fluorescence model expresses the mean photon number as

$m_s=\pm1$8

where $m_s=\pm1$9 is the occupation probability of $\Delta \mathrm{PL}$0, $\Delta \mathrm{PL}$1, and $\Delta \mathrm{PL}$2 (Dokai et al., 4 Jun 2026). In chopped-microwave NV experiments, a standard interpretation consistent with the reported acquisition protocol is that the ODMR signal is the difference between photoluminescence with microwave off and on, while the normalized contrast is the corresponding ratio to the microwave-off fluorescence (Kollarics et al., 2020).

In other material systems the Hamiltonian and readout basis differ, but the resonance logic remains analogous. In singlet-fission molecular crystals, for example, CW-ODMR detects transitions within triplet and quintet manifolds through their effect on emissive $\Delta \mathrm{PL}$3TT-related photoluminescence (2206.13636). In ensemble nanodiamond thermometry, the measured CW-ODMR feature can deviate strongly from a single Lorentzian because it is the convolution of many single-NV responses with distributed zero-field splitting and strain parameters (Kato et al., 15 May 2026).

3. Instrument architectures and detection modes

CW-ODMR has been implemented with both resonant and non-resonant microwave structures. A broadband NV spectrometer based on a coplanar waveguide (CPW) rather than a cavity uses a continuous-wave 532 nm laser, an HP83751B microwave signal generator, broadband amplification, and a $\Delta \mathrm{PL}$4-terminated CPW on which the diamond is placed directly. In that system, microwave frequency sweeps are detected by chopping the microwave output with the TTL output of a lock-in amplifier, while fixed-frequency field sweeps use microwave frequency modulation to generate derivative-like lineshapes analogous to field-modulated ESR (Kollarics et al., 2020). The same work reports CPW gaps of $\Delta \mathrm{PL}$5, separation of $\Delta \mathrm{PL}$6, total width $\Delta \mathrm{PL}$7, and a calculated microwave magnetic-field conversion of $\Delta \mathrm{PL}$8, compared with $\Delta \mathrm{PL}$9 for a cylindrical TE$\Delta \mathrm{PL}/\mathrm{PL}$0 resonator (Kollarics et al., 2020).

A distinct cavity-based CW-PLDMR implementation was developed for single-walled carbon nanotubes. It uses unmodulated continuous optical excitation, square-wave chopping of the microwave irradiation at typically $\Delta \mathrm{PL}/\mathrm{PL}$1 kHz, a home-built TE$\Delta \mathrm{PL}/\mathrm{PL}$2 cylindrical cavity around $\Delta \mathrm{PL}/\mathrm{PL}$3 GHz, a Horiba JY iHR320 spectrograph, and a liquid-nitrogen-cooled InGaAs detector. That instrument was designed for tunable visible excitation and wavelength-resolved near-infrared detection, and it simultaneously records DC photoluminescence and lock-in-detected $\Delta \mathrm{PL}/\mathrm{PL}$4 (Negyedi et al., 2016). Its reported practical sensitivity reaches $\Delta \mathrm{PL}/\mathrm{PL}$5 at 1 kHz chopping frequency and 1 s time constant, with spectrograph resolution of $\Delta \mathrm{PL}/\mathrm{PL}$6 nm (Negyedi et al., 2016).

For large-area NV work, a planar ring antenna was designed specifically for room-temperature ODMR. It has a resonance frequency around $\Delta \mathrm{PL}/\mathrm{PL}$7 GHz, bandwidth of $\Delta \mathrm{PL}/\mathrm{PL}$8 MHz, measured bandwidths of $\Delta \mathrm{PL}/\mathrm{PL}$9 MHz with diamond and $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$0 MHz without diamond, and a $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$1-mm-diameter center hole with fairly uniform microwave magnetic field in the central region (Sasaki et al., 2016). The design targets the practical requirements of broadband operation, spatial uniformity, and optical access for imaging.

At the opposite extreme of frequency and field, single-NV CW-ODMR has been demonstrated at $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$2 GHz and $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$3 T using quasioptics and a corrugated waveguide integrated into a confocal microscope inside a $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$4 T superconducting magnet. That system uses a continuous-wave 532 nm laser, confocal fluorescence detection with avalanche photodiodes, and field-swept ODMR at fixed microwave frequency; the resonance was assigned to the $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$5 transition and fitted with a Gaussian (Stepanov et al., 2015).

Cryogenic CW-ODMR in constrained environments has also been realized in a variable temperature insert. In that setup, a continuous-wave 532 nm laser with typical output power of 100 mW excites an NV ensemble through an optical path of about 190 cm, while a Windfreak SynthHD v2, a fast microwave switch, and a custom CPW provide frequency-stepped microwave excitation. The laser remains continuously on while the microwaves are gated on and off; fluorescence is sampled in synchronized $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$6 windows within $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$7 microwave-off and $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$8 microwave-on sub-intervals, repeated $$\hat{\mathcal{H}}^{i}=D (\hat{S}_z^i)^2 +\gamma B_{i}\hat{S}^i_z,$$9 times per frequency point (Tong et al., 4 Dec 2025).

4. Spectral structure, overlap, and data analysis

In NV ensembles, the number and arrangement of CW-ODMR dips depend strongly on the field orientation relative to the four NV classes. For a $f_{i\pm}=D\pm\gamma B_i,$0-cut diamond, the four unit vectors are

$f_{i\pm}=D\pm\gamma B_i,$1

$f_{i\pm}=D\pm\gamma B_i,$2

and the projections satisfy

$f_{i\pm}=D\pm\gamma B_i,$3

At low bias field, this symmetry produces a taxonomy from one to eight observable dips. Along $f_{i\pm}=D\pm\gamma B_i,$4, all four $f_{i\pm}=D\pm\gamma B_i,$5 are equal and only two dips remain; along $f_{i\pm}=D\pm\gamma B_i,$6, one pair has zero projection and three dips appear; along generic directions, all four projections are distinct and the full eight-dip structure is resolved (Dubey et al., 25 Apr 2025). The practical consequence is that low-field ensemble CW-ODMR often contains overlap-induced ambiguities in vector-field reconstruction.

Ensemble line shapes can also deviate systematically from conventional Lorentzian or Voigt fits. In fluorescent nanodiamond ensembles, the local feature near the center of the ODMR dip can exhibit a “small peak inside a dip,” arising from convolution over distributed zero-field splitting and strain. Starting from an ensemble model with Lorentzian distributions in $f_{i\pm}=D\pm\gamma B_i,$7, $f_{i\pm}=D\pm\gamma B_i,$8, and $f_{i\pm}=D\pm\gamma B_i,$9, the local spectral feature near $D=2.87$0 is approximated by

$D=2.87$1

with parameters determined from the derivatives of the full ensemble spectrum at $D=2.87$2 (Kato et al., 15 May 2026). In the reported experiments, conventional Lorentzian and Voigt fits had $D=2.87$3 below $D=2.87$4 within 2866–2872 MHz, whereas the dip–peak model yielded $D=2.87$5; the same work reported an improvement in resonance-frequency precision by approximately a factor of $D=2.87$6 under identical acquisition conditions and optimal performance near about $D=2.87$7 ODMR contrast (Kato et al., 15 May 2026).

When spectra are noisy or sparse, nonparametric analysis can outperform standard fitting. A clustering algorithm based on two-stage K-means-like grouping first partitions data by photoluminescence level into $D=2.87$8 rows and then clusters the lowest row into $D=2.87$9 resonance columns. On synthetic and experimental ODMR data, this method was reported to achieve about $\gamma = 28$0 better accuracy, $\gamma = 28$1 better resolution, or about $\gamma = 28$2 fewer data points than standard statistical fitting, while remaining usable even in regimes where conventional fitting failed to converge in over $\gamma = 28$3 of spectra (Stone et al., 2024).

A further analytical complication is that the modulation-frequency dependence of cwODMR does not behave like conventional ESR absorption. In the Kaplan-Solomon-Mott-type intermediate-pair model, the observables are

$\gamma = 28$4

and the lock-in-detected in-phase and out-of-phase components depend on the full singlet–triplet kinetics under square-wave microwave modulation (Lee et al., 2012). That analysis showed that a large number of quantitatively different models cannot be differentiated, that the sign of cwODMR can depend on recombination, dissociation, intersystem crossing, pair generation, modulation frequency, microwave power, and temperature, and that radiative and non-radiative recombination cannot be distinguished from signal sign alone (Lee et al., 2012). A common misconception is therefore that enhancement or quenching in CW-ODMR directly identifies the dominant microscopic pathway; the modulation-frequency analysis does not support that simplification.

5. Extended operating regimes and engineered CW-ODMR sensing

CW-ODMR has been extended beyond static-field spectroscopy into resonant AC magnetometry. A room-temperature NV method without pulse sequences and without an externally applied DC magnetic field uses the strain-split upper states

$\gamma = 28$5

with energies $\gamma = 28$6 and $\gamma = 28$7, so that the $\gamma = 28$8 splitting is $\gamma = 28$9 in the MHz range (Saijo et al., 2018). Under a resonant AC field, the GHz CW-ODMR spectrum splits according to

$B_i=\mathbf B\cdot \mathbf e_i$0

and the demonstrated sensitivity was

$B_i=\mathbf B\cdot \mathbf e_i$1

at room temperature (Saijo et al., 2018).

Frequency-tunable AC sensing has been pursued by engineering dressed states within CW-ODMR. One route uses two RF tones to create RF double-dressed states, making the detectable target frequency tunable according to

$B_i=\mathbf B\cdot \mathbf e_i$2

with an estimated bandwidth of $B_i=\mathbf B\cdot \mathbf e_i$3 MHz and optimal sensitivity around

$B_i=\mathbf B\cdot \mathbf e_i$4

near $B_i=\mathbf B\cdot \mathbf e_i$5 MHz (Okaniwa et al., 2023). A later microwave-dressed proposal replaces RF dressing with GHz microwave dressing and derives the resonance condition

$B_i=\mathbf B\cdot \mathbf e_i$6

predicting tunable AC detection frequencies up to the order of $B_i=\mathbf B\cdot \mathbf e_i$7 MHz. In the corresponding numerical model, an ideal-readout sensitivity of $B_i=\mathbf B\cdot \mathbf e_i$8 and a realistic single-NV sensitivity of about $B_i=\mathbf B\cdot \mathbf e_i$9 were reported, with sensitivity remaining roughly constant as $i$0 increased to $i$1 MHz (Dokai et al., 4 Jun 2026).

Broadband CW-ODMR is also important in purely spectroscopic settings. The CPW-based optical–microwave pump–probe platform demonstrated NV ODMR not only around the zero-field splitting near $i$2 GHz but also in a field-swept experiment at fixed microwave frequency of $i$3 GHz, with derivative Lorentzian lineshapes produced by microwave frequency modulation and a reported modulation sensitivity of $i$4 (Kollarics et al., 2020). At much higher field, single-NV CW-ODMR at $i$5 GHz and $i$6 T established that optical spin readout remains viable in the high-frequency/high-field regime, with stable fluorescence up to $i$7 T and an inferred NV $i$8-factor between $i$9 and $m_s=\pm1$00 (Stepanov et al., 2015).

6. Boundary cases, material breadth, and conceptual limits

CW-ODMR is not restricted to NV centers. Tunable-laser, wavelength-resolved PLDMR has been implemented for single-walled carbon nanotubes with visible excitation from 560–900 nm and near-infrared detection from 1000–2000 nm, explicitly to separate species-dependent optical transitions in heterogeneous nanotube ensembles (Negyedi et al., 2016). Broadband CW-ODMR/PLDMR has also resolved triplet and quintet resonances in the singlet-fission crystal TES TIPS-TT, including zero-field parameters $m_s=\pm1$01 MHz, $m_s=\pm1$02 MHz, $m_s=\pm1$03 MHz, and $m_s=\pm1$04 MHz, with microwave amplitude modulation at 200 Hz and optical readout of $m_s=\pm1$05 at 5 K (2206.13636). These examples show that the defining structure of CW-ODMR is methodological rather than material-specific: continuous optical pumping, continuous or quasi-continuous microwave driving, and optically detected resonance-induced redistribution of spin populations.

At the same time, several adjacent techniques are often conflated with CW-ODMR and should be separated conceptually. In nanodiamond detection of ferromagnetic dynamics on yttrium iron garnet, continuous optical pumping and optical readout are used, but the microwave field is tuned to excite the ferromagnet rather than the NV spin transition; the authors explicitly described the method as “distinct from commonly used ODMR techniques” (Wolfe et al., 2015). A plausible implication is that continuous optical NV fluorescence readout is not sufficient to classify a method as CW-ODMR unless the spin transition being optically read out is itself microwave driven.

A second boundary case is microscale reflectance-detected ODMR in a $m_s=\pm1$06Te quantum well. There the experiment uses pulsed 13.8 GHz microwaves, pulsed optical acquisition, and differential “MW ON”/“MW OFF” reflectance spectroscopy while sweeping magnetic field; the optical observable is the microwave-induced change in excitonic Zeeman splitting rather than fluorescence intensity (Dydniański et al., 2024). The work is ODMR in function and terminology, but not textbook simultaneous continuous-wave optical and microwave excitation. This suggests that the ODMR label spans a broader family of optically detected magnetic-resonance measurements, within which CW-ODMR denotes the steady-state continuous-drive subset.

Finally, the breadth of CW-ODMR instrumentation has made it a general platform for sensing and spectroscopy rather than a single standardized experiment. Open-source control environments such as Qudi were specifically extended to speed up CW-ODMR acquisition, relax instrument requirements, and support ensemble measurements with analog photodetectors, random microwave sweep orders, and real-time fitting (Babashah et al., 2022). The resulting experimental landscape includes room-temperature magnetometry, cryogenic magnetometry in variable temperature inserts, wavelength-resolved spectroscopy in the near infrared, broadband microwave delivery on CPWs and planar antennas, and high-field/high-frequency single-defect ODMR (Babashah et al., 2022, Tong et al., 4 Dec 2025).

CW-ODMR is therefore best understood as a family of steady-state optical magnetic-resonance methods unified by continuous optical pumping, frequency-domain microwave interrogation, and optical detection of resonance-induced state redistribution. Its technical challenges are correspondingly diverse: resonance overlap in NV ensembles, model mismatch in ensemble line shapes, ambiguity of sign in spin-dependent recombination systems, microwave-bandwidth constraints, optical-throughput limitations, and the need to distinguish true CW-ODMR from related continuous-readout but off-resonant or differentially gated optical sensing modalities (Dubey et al., 25 Apr 2025, Kato et al., 15 May 2026, Lee et al., 2012).