Native NIR-emitting single colour centres in CVD diamond

Single-photon sources are a fundamental element for developing quantum technologies, and sources based on colour centres in diamonds are among the most promising candidates. The well-known nitrogen vacancy centres are characterized by several limitations, and thus few other defects have recently been considered. In the present work, we characterize, in detail, native efficient single colour centres emitting in the near infra-red (λ = 740–780 nm) in both standard IIa single-crystal and electronic-grade polycrystalline commercial chemical vapour deposited (CVD) diamond samples. In the former case, a high-temperature (T > 1000 °C) annealing process in vacuum is necessary to induce the formation/activation of luminescent centres with good emission properties, while in the latter case the annealing process has marginally beneficial effects on the number and performance of native centres in commercially available samples. Although displaying significant variability in several photo-physical properties (emission wavelength, emission rate instabilities, saturation behaviours), these centres generally display appealing photophysical properties for applications as single photon sources: short lifetimes (0.7–3 ns), high emission rates (∼50–500 × 103 photons s−1) and strongly (>95%) polarized light. The native centres are tentatively attributed to impurities incorporated in the diamond crystal during the CVD growth of high-quality type-IIa samples, and offer promising perspectives in diamond-based photonics.

Recently, the activation of NIR centres upon high-temperature (T > 1000°C) thermal annealing in forming gas (4% H 2 in Ar) in pristine (i.e. unimplanted) single-crystal diamonds was investigated, both in terms of creation yield and spectral emission properties [53].
It is worth noting that, in contrast to what was observed in the NV − case, most of the above-mentioned centres frequently display zero-phonon emission lines which are dispersed on a relatively large spectral range (i.e. ∼740-780 nm). This indicates that different structural configurations are possible for the same impurity and/or that the centres are extremely sensitive to local strain fields. Therefore, an unequivocal attribution of their spectral features to specific defect complexes requires particular care: to this scope, complementary techniques such as vibrionic [54] and stress-dependent [55] photoluminescence (PL) spectrometry measurements, EPR [56][57][58] and XANES [59] can provide useful information. As a matter of fact, several works reported promising NIR-emitting single colour centres in diamond with tentative (if any) attributions [53,[60][61][62].
In the present work we report on the observation and optical characterization in commercial chemical vapour deposited (CVD) diamond samples of native NIR-emitting (λ = 740-780 nm) single colour centres with promising photo-physical properties (linewidth, emission rate, photo-stability and polarization). While being characterized by similar properties to those reported in previous works (i.e. high brightness, narrow emission in the λ = 740-780 nm spectral range, ∼ns radiative lifetimes, linear polarization), the centres were not formed upon ion implantation or intentional doping during CVD growth, but directly observed in as-grown polycrystalline samples characterized by the highest degree of purity, while a high-temperature annealing procedure was necessary to activate/form the centres in IIa single-crystal samples characterized by a lower degree of purity. Despite a still unclear attribution, the availability of native NIR centres with good emission properties in commercial CVD samples offers numerous opportunities in diamond-based photonics.

Experimental setup
In the present study, five 3 × 3 × 0.3 mm 3 single-crystal CVD diamond samples produced by Element Six in different lots were employed. The samples are classified as type IIa, their nominal substitutional nitrogen and boron concentrations being <1 ppm and <0.05 ppm, respectively. The crystal orientation is <100> and they are optically polished on both of their two larger faces. A 5 × 5 × 0.2 mm 3 polycrystalline CVD diamond sample produced by Element Six was also investigated. This sample is classified as type IIa and is referred to as 'detector grade' by the producer, with nominal substitutional nitrogen and boron concentrations of <50 ppb and <1 ppb, respectively. The average microcrystal size was in the 20-50 μm range.
All of the above-mentioned samples were thermally annealed in vacuum (p < 10 −6 mbar) at a temperature of 1450°C for 1 h. After the high-temperature annealing, surface graphitization was removed by oxidizing the sample in air for 30 min at a temperature of 400°C, followed by a 30 min exposure to an oxygen plasma (30 W radiofrequency power, 20 sccm oxygen flux, p = 2.5 × 10 −2 mbar).
A single-photon-sensitive confocal NIR-PL microscopy system was developed in the present work, as schematically shown in figure 1. The centres are excited with both continuous and pulsed laser light at λ = 690 nm emitted by a solid-state laser source, coupled into a singlemode optical fibre and expanded by a 4 × objective. The pulsed light excitation is performed at 80 MHz repetition rate with 100 ps pulse width; subsequently, the beam is directed to a dichroic mirror (λ > 700 nm) which reflects it to an objective (100×, NA = 0.9, air; or alternatively 100 × O2, NA = 1.3, oil) focusing on the sample. The sample is mounted on a remotely controlled three-axis piezo-electric stage, allowing positioning in a 100 × 100 μm 2 area range with micrometric accuracy. The induced luminescence beam is collected together with the scattered excitation beam by the same focusing objective and then filtered through the dichroic mirror and a subsequent filters set (λ > 730 nm), thus allowing a suitable (>10 12 ) attenuation of the λ = 690 nm excitation component of the beam. The filtered beam is then focused with an achromatic doublet and coupled into a graded-index multimode optical fibre, acting not only as an optical connection to the detectors, but also as the pinhole aperture for the confocal system. The fibre leads to an integrated 50:50 beam-splitter (BS) whose outputs are connected to two photon-counters based on Si-single-photon-avalanche photo-diodes (SPADs) operating in Geiger mode. This experimental configuration reproduces a 'Hanbury Brown and Twiss' interferometer [7], that allows the identification of the presence of a single emitter via the measurement of the second-order autocorrelation function g (2) (t) from the coincidence counts between the two detectors. In the coincidence circuit the difference between the arrival times of single photons at the two detectors is measured by means of a time-to-amplitude converter whose output feeds a multi-channel analyser recording a histogram of the above-mentioned time intervals. Simultaneously, the total counts detected at the two SPADs are recorded by a digital counter, allowing a measurement of the total luminescence intensity for each pixel.
The collection of PL spectra from single centres was carried with a single-grating monochromator with 1600 grooves mm −1 blazed at 600 nm connected to one of the abovementioned SPADs.
A complementary setup was developed for polarization measurements both on the absorbed and emitted radiation from single centres, where the above-mentioned dichroic mirror was replaced by a 50:50 BS and the polarization of the excitation beam was selected using a combination of a half-waveplate (λ = 800 nm) and a Glan-Taylor polarizer. For the detection of polarized PL a film-polarizer was placed before the achromatic doublet. Care was taken in subtracting a slight rotation of the pump light polarization induced by the 50:50 BS as it was verified that all polarizers did not introduce a measurable beam shift during their rotation. The polarization dependence of the reflection and transmission of the BS were suitably taken into account 7 .

Annealing effect and spectral properties
In single-crystal samples, thermal annealing was generally found to have the effect of inducing the formation of isolated luminescent centres in all samples under analysis.
As an example, figures 2(a) and (b) report confocal PL intensity maps obtained from a depth of ∼5 μm below the surface in the same region of a single-crystal sample before and after 1450°C thermal annealing, respectively. In figure 2(a), one isolated bright spot (circled in red in the map) was found to be a single emitter by means of autocorrelation statistics measurements. As shown in figure 2(b), after thermal annealing at 1450°C, the concentration of centres per surface area in the single-crystal samples increased significantly, with an average value of ∼1-10 centres per 50 × 50 μm 2 . Due to the improvement of the emission properties (emission rate, stability) upon annealing, after thermal processing it was possible to identify and characterize a larger set of centres, with a direct advantage for the statistical relevance of the acquired data. The depth distribution in the location of the centres was found to be uniform within the range of depths probed by the confocal system (∼50 μm).
The emission properties of the isolated centre before thermal annealing (see figure 2(a)) are reported in figure 3. Experimental data are characterized by poor statistics due to the low emission rates (∼10 4 photons s −1 ) and severe photo-bleaching of the centres. Figure 3(a) shows the PL emission of the centre at room temperature: a single zero phonon line (ZPL) emission at λ ≅ 744 nm was observed, with two phonon sidebands. Figure 3(b) shows the relevant secondorder autocorrelation chronograms, which were fitted with the following function that describes the temporal evolution of a three-level system [63]: where β 1 and (β 2 , a) account for the anti-bunching and bunching features of the chronogram, respectively. Note that the background due to diffused luminescence and Raman emission was subtracted from experimental chronograms according to the procedure described in [7] and the temporal response of the detectors was taken into account by convoluting the function in equation (1) with a Gaussian function with FWHM = 0.7 ns (as directly measured with a ps pulsed laser). A value of (β 1 ) −1 = (0.77 ± 0.07) ns was obtained, together with non-negligible bunching parameters: (β 2 ) −1 = (0.033 ± 0.004) ns and a = (0.76 ± 0.04). The spectral position of the ZPL emission after thermal annealing (see figure 2(b)) is subjected to dispersion in the 740−780 nm range, as shown for three typical PL spectra reported in figure 4(a). No pronounced phonon sidebands were observed after thermal annealing, while partial overlap with the first-order Raman line was possible, depending on the ZPL position. It is worth noting that the spectral resolution of the spectrometer poses a Δλ = 5 nm upper limit to the ZPL emission width at room temperature. Due to time constraints, a limited number of centres (i.e. ∼50) was characterized. Nonetheless, within the limitation imposed by statistics, from the ZPL position histogram reported in figure 4(b) it can be noted that (i) the distribution of the ZPL positions is not uniform in the above-mentioned spectral range but rather multimodal, with larger populations for λ ≅ 750 nm, λ ≅ 755 nm and λ ≅ 765 nm; (ii) the distribution does not vary significantly from sample to sample. In our interpretation, the observation of a multi-modal distribution of the ZPL spectral positions indicates that a discrete set of different  (1)).
physical conditions (defect structure or charge state, local strain) determines different PL emission properties.
With the scope of investigating the role of vacancy-related defects in the formation of the observed centres, a single-crystal sample was irradiated with 15.6 MeVC 5+ ions prior to thermal annealing. The implantation was performed at the scanning ion microbeam line of LABEC laboratories (Florence) [64,65] at fluences ranging from 1 × 10 13 cm −2 to 1 × 10 16 cm −2 over areas with sizes ranging from 50 × 50 μm 2 to 750 × 750 μm 2 . No significant differences were found in the distribution and PL emission properties of single centres located in the implanted portions of the irradiated sample with respect to both the unimplanted portions and to the unirradiated samples. Therefore, it is reasonable to conclude that vacancies do not play any significant role in the formation of the colour centres under investigation: this observation is compatible with what has been reported in [53]. If the limited diffusivity of foreign atoms in diamond is considered [66,67], the significant increase in colour centres after annealing can hardly be related to the migration of impurities over significant distances, but rather to their local rearrangement and/or to the removal of nearby PL-quenching defects. In [53], the presence of hydrogen in the atmosphere during the 1000°C annealing was found to have a significant correlation with the formation of NIR-emitting centres with similar emission properties to those reported in the present work. Although hydrogen was never intentionally introduced in the annealing environment in the present work, the effect of trace H impurities in the process chamber cannot be ruled out in principle. Moreover, it is worth noting that in the present work the higher (1450°C) annealing temperature could have more effectively contributed to the dissociation of molecular hydrogen into atomic hydrogen.
In the poly-crystalline sample, native NIR-emitting single centres with good emission properties (i.e. count rates up to 1 × 10 6 photons s −1 and absence of bunching behaviour in ∼90% of observed emitters) were found even before thermal annealing (see figure 5). This observation is compatible with the higher purity of this sample, i.e. in this case thermal annealing is not necessary to remove nearby PL-quenching defects. Moreover, it is worth noting that the observation of the centres in as-grown poly-crystalline samples might be related to the effect of a higher hydrogen concentration at the grain boundaries [68]. The centres were found to be often located in the proximity of dislocations and grain boundaries. Thermal annealing at 1450°C was found to only have a slightly positive effect on the photo-physical properties of the centres.

Emission properties
PL instability features such as photo-bleaching and photo-blinking affect many (>50%) of the single centres even after thermal annealing. This effect is particularly evident in standard type-IIa single-crystal samples with respect to the 'detector-grade' poly-crystalline sample. Similarly to what was reported for the spectral dispersion in the ZPL emission, different emissioninstability features are observed from centre to centre: in some cases sudden random jumps in emission rates take place, while in others clear evidence of photo-blinking is observable. An example of the former phenomenon is shown in figure 6(a), in which the time evolution of the emission rate of a single centre in one of the single-crystal samples is reported for a laser pump power of ∼1.2 mW.
Apart from a slight overall decrease in the count rate, which is attributed to instrumental features (i.e. a slight drift of the sample stage), the system is randomly oscillating among three distinct emission rate regimes, highlighted by the blue dashed lines and labeled as (1)-(3). This kind of process was already observed in both Ni/Si-related [37] and Cr-based [46] single centres in diamond, and in the present work (as much as in [62]) is tentatively attributed to the presence of neighbouring defects with unstable charge state that activate/deactivate non-radiative decay paths via resonant energy transfer, although this explanation of the observed process cannot be considered as conclusive. Figure 6(b) shows the second-order autocorrelation chronograms measured from the same centre in time windows characterized by each of the above-mentioned emission rate regimes. Experimental data were interpolated with equation (1), as for the beforeannealing case. The correlation between bunching/anti-bunching behaviours of the autocorrelation functions and the relevant emission regimes is not straightforward, thus making their interpretation difficult: as indicated by the fitting parameters reported in table 1 (i) for the lowest emission rate (labelled as (1) in figure 6(a)), a very strong bunching behaviour is observed; (ii) for the intermediate emission rate (labeled as (2) in figure 6(a)), the bunching behaviour is fairly negligible and finally (iii) the highest emission rate (labeled as (3) in figure 6(a)) is associated with a non-negligible bunching behaviour. Although only tentatively, the above-mentioned behaviour can be attributed to quantum efficiency fluctuations induced by interaction with surrounding defects [69] affected by reversible changes in their charge state [37]: these effects induce 'on-off' periods of the colour centres of the order of tens of ns, as resulting from the fitting of the 'bunching' components of the second-order autocorrelation functions ( figure 6(b)). Moreover, as a partial support of this interpretation, it is worth noting that, although this kind of measurement was not possible with  (1)) are reported as red lines.  (1)) of the second-order autocorrelation functions reported in figure 6(b) in correspondence of different emission rate regimes (see figure 6(a)).
Emission rate (10 3 cps) (β 1 ) −1 (ns) (β 2 ) −1 (ns) a ∼85 2.25 ± 0.08 (4.68 ± 0.07) × 10 −2 1.420 ± 0.014 ∼135 (8.27 ± 0.18) × 10 −1 (2.34 ± 0.07) × 10 −2 (3.5 ± 0.6) × 10 −2 ∼190 1.00 ± 0.02 (3.1 ± 0.5) × 10 −2 (8.02 ± 0.08) × 10 −1 our experimental setup, random 'jumps' among different emission regimes were previously associated to spectral shifts in ZPL emission [46,62]. Centres subjected to the above-mentioned instabilities can bleach (i.e. 'jump' into a state with extremely low emission rate) over indefinitely long times. Usually they can be retrieved by exposing them for ∼10-20 s to a λ = 532 nm CW excitation with ∼3 mW power.   distributions associated with the above-mentioned mean values (continuous red lines). In particular, in the case of the former centre (figures 7(a) and (c)), the emission rate distribution is fully compatible with Poissonian statistics: assuming that the distortion effects on photon statistics due to SPAD dead-time (i.e. 50 ns) are negligible at the measured count rate [72], we conclude that the emission rate of the first colour centre is constant in time. On the other hand, in the latter case (figures 7(b) and (d)) a significant discrepancy from a Poissonian distribution is clearly observable, suggesting that the photon emission rate fluctuates in time. For comparison, a Poissonian distribution of a colour centre that would have the same intensity but with a constant photon emission rate is superimposed in figures 7(c) and (d). As mentioned above, a possible interpretation of this phenomenon may reside in the quenching mechanisms between the centres and surrounding defects, which determine 'on-off' periods on a different time-scale with respect to what is observed in figure 6 (i.e. ∼10 2 μs). Second-order autocorrelation measurements were carried out as a function of excitation power in centres unaffected by emission instabilities, with the purpose of estimating their excited-state lifetimes. A degree of variability was also observed from centre to centre in this respect. As shown in figures 8(a) and (d), centres with ZPL emission centred at different wavelengths (λ ZPL = 755 nm and λ ZPL = 772 nm, respectively) exhibit distinct saturation behaviours in the 'emission rate versus pump power' trend, as resulting from the fitting of the experimental data with the following expression: This indicates either that distinct non-radiative decay paths are present in different centres or that they are characterized by significantly different absorption cross-sections. The secondorder autocorrelation function was also measured at different powers from the same centres (figures 8(b) and (e)) and from the linear fitting of the '1/β 1 versus pump power' trends it was possible to estimate the centre's lifetime [39,60,61]. As shown in figures 8(c) and (f), the two centres exhibit different lifetimes, i.e. (0.7 ± 0.2) ns and (2.6 ± 0.2) ns respectively. It is worth noting that, although displaying some dispersion, all lifetime values are comprised between 0.7 ns and 3 ns, over a population of ∼20 probed centres.
PL measurements were also performed under pulsed excitation, so that the lifetime was directly extrapolated by fitting the relevant chronograms. Figure 9 shows the results of secondorder autocorrelation measurements carried in pulsed excitation conditions from a single centre in one of the single-crystal samples after thermal annealing.
As shown in figure 9(a), the spacing between the peaks of the g (2) (t) function corresponds to the period of the pulsed excitation, i.e. 12.5 ns, while at null time delay the observed weak peak is due to unfiltered first-order Raman emission. The zoomed-in autocorrelation chronogram reported in semi-logarithmic scale in figure 9(b) exhibits a single-exponential decay behaviour. From the fitting of the above-mentioned exponential decay, a lifetime of (2.05 ± 0.05) ns is obtained, in good agreement with typical values obtained from the abovementioned power-dependent measurements.

Polarization measurements
Polarization measurements from different centres were performed both in absorption and reflection. In the former case, the full (i.e. over all polarizations) PL emission intensity of the centre in the λ > 730 nm spectral range was measured as a function of the excitation light polarization. In the latter case the PL emission intensity was measured at different polarizations while keeping the excitation polarization in the direction that maximizes the PL yield. Despite the above-mentioned variability in several photo-physical characteristics, all observed centres consistently displayed a dipole behaviour both in absorption and emission, with the polarization axis aligned with the <110> or <111> crystallographic direction. A high degree of polarization (>95%, basically limited by the instrumental sensitivity of the setup) was observed in all centres under investigation, a significant property for quantum technology applications. In figures 10(a)-(d) the absorption and emission polarization diagrams for the two abovementioned centres emitting respectively at λ ZPL = 755 nm and λ ZPL = 772 nm are reported. As shown in figure 10, in the former case (λ ZPL = 755 nm: figures 10(a) and (b)) the excitation and emission polarizations are parallel, while in the latter (λ ZPL = 772 nm: figures 10(c) and (d)) they are mutually orthogonal. These mutual orientations were systematically observed in all centres under investigation.

Conclusions
We reported the observation and photo-physical characterization of NIR-emitting single colour centres in both optical-grade single-crystal and electronic-grade poly-crystalline commercial CVD diamond samples. In the former case, a high-temperature (T > 1000°C) annealing process in vacuum was necessary to induce the formation/activation of luminescent centres with good emission properties, while in the latter case the annealing process had marginal effects on the number and performance of native centres in commercially available samples. While no definite hypotheses can be drawn about the attribution of the centres to specific defect complexes, our observations indicate that they are based on native impurities and/or structural defects in highquality commercial CVD samples, with the possible exception of light elements present in trace concentrations in the annealing chamber.
The properties of the centres can be summarized as follows.
• Their PL emission is characterized by narrow ZPL lines, with low background and no measurable phonon sidebands; such a property represents a fundamental advantage with respect to NV − centres as single-photon emitters. • The ZPL positions are dispersed over a relatively broad spectral range, with a multi-modal statistical distribution; this could, in principle, represent a limitation for quantum technology applications, nevertheless this issue can be addressed when the attribution of these centres is fully clarified. • While the typical emission rate of the centres is high compared to 'standard' NV − centres [70,71], different centres display a variety of emission instabilities; the centres unaffected by blinking behaviours display Poissonian statistics. • The saturation behaviour as a function of excitation power also displays significant variability; in this case, significant improvements in addressing this issue will be possible once the attribution of the centres is clarified. • All centres are characterized by lifetime values between 0.7 ns and 3 ns and their emission is linearly polarized.
It is worth noting that the observed centres have similar properties to those of previously reported emitters obtained in diamond by direct ion implantation or intentional doping during CVD growth and characterized by different tentative attributions. While a definitive defect attribution cannot be presented in the present work, the obtained results indicate that (i) the structure of the observed defects is related to native impurities incorporated during high-purity (i.e. 'electronic grade') CVD growth; (ii) single substitutional vacancies do not have a direct influence on the defect formation; and (iii) hydrogen might play a role in the formation of the defect. The photophysical properties of the centres indicate that more common defects (such as the NV centre or other nitrogen-related complexes) play a 'quenching' role in reducing the quantum efficiency of these centres via resonant energy transfer in samples characterized by lower purity standards.
The observation in commercially available 'standard type IIa' and 'electronic-grade' CVD diamonds of efficient native NIR colour centres emitting fully polarized light in a convenient spectral range, with no need for specific sample processing (i.e. doping from vapour phase or by ion implantation), offers appealing perspectives in the fields of diamond-based single-photon emitters and quantum optics.