Bright optical centre in diamond with narrow, highly polarised and nearly phonon-free fluorescence at room temperature

The spectral properties of the L1 centres are firstly described in the following. Figure 2(a) shows the background corrected fluorescence of a typical L1 centre under 532 nm excitation at room temperature. The fluorescence spectrum has a very narrow emission line at 581.7 nm (2.134 eV). A weak and characteristic vibronic sideband is found, showing a broad peak around 600 nm together with some features between 640 and 650 nm (with energies of 196.4, 209.1 and 219.2 meV). The FWHM of the main 581.7 nm emission line was determined to be less than 1.4 nm. The resolution of the home-built spectrometer is in the order of 1 nm, thus the linewidth may be even narrower. Remarkably, about 55% of the intensity is found in the main emission line. The brightness of the L1 centres varies from one centre to the other. The brightest ones do reach emission rates of about 7 × 10⁵ counts s⁻¹ at saturation (at about 3 mW with an objective with NA = 0.95), much higher than the NV centres. The saturation curve of a typical L1 centre is shown in figure 2(b). The single emitter nature of the defect is proved by the second-order photon autocorrelation function plotted in figure 2(c) which goes to zero at zero time delay. The g⁽²⁾ function has been background corrected and normalised. From the fit of the dip, a lifetime of about 1.7 ns is estimated for the excited state. Furthermore, the bunching shoulders indicate the presence of a long-lived shelving state of about 700 ns. At this stage it is not possible to conclude whether the main emission line is the ZPL of the defect. The spectral study of several centres reveals a spread in the position of the emission line. This spread extends over more than 20 nm, between 577 and 605 nm, with a maximum at about 582 nm (figure 3(a)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It is believed that this spread is due to stress or environmental changes and that all the centres observed here are of the same nature. Note that for SiV or Cr centres, the ZPL can shift for several nm due to the same reasons [32, 33]. A spread in the brightness of the centres is also observed, which can vary of about one order of magnitude. No correlation between the line position and the brightness of the centres could be found.

The vibronic sideband emission of the L1 centres is weak compared to the main emission line, even at room temperature, indicating a weak electron-phonon coupling. With respect to the spread of the emission line, the sidebands of several centres have been more precisely analysed. To this purpose, the sidebands have been plotted (in figure 3(b)) not in wavelength scale but as the energy difference with respect to the position of the main emission line (and normalised to its intensity). It can be seen at first sight that the sidebands of the L1 centres are similar. They all present a broad peak at about 60 meV (dotted line) as well as narrower features at higher energy, between 190 and 220 meV (shadowed area). These are attributed to local vibrational modes (LVM), since the phonon density of states has a sharp high-energy cutoff at 165 meV and decreases strongly below 70 meV [34].

A deeper analysis of these LVM may help understanding the structure of the L1 centre and what kind of atoms are involved. Furthermore, additional lines of higher intensity can be seen in the sideband region of some centres (indicated by the stars). At this stage it is not possible to conclude whether these features are electronic transitions or vibronic features. Further measurements at low temperature are required for clarification.

An additional property shared by all the L1 centres is a strong polarisation, measured both in absorption and in emission. For absorption measurements (figure 4(a)), a half-wave plate has been used to rotate the polarisation of the laser excitation, whereas a linear polarisation analyser has been placed in the detection path for the measurements in emission (not shown). Figure 4(a) shows the (background-corrected) polarisation diagrams in absorption of two typical L1 centres, representative of all the measured centres. The absorption maxima of a single centre are found either along the [110] direction or the [−110] direction. Similar polarisation diagrams are obtained in emission. It can further be seen that no absorption takes place for the polarisation perpendicular to the one of maximum absorption. The L1 centres behave therefore as single dipoles in absorption. A two-fold symmetry in the projected (001) plane is found and the fluorescence is maximum, or zero, along the 〈110〉 projected axes. No L1 centres having maxima along the [100] or the [010] directions could be found. The distribution of the directions of maximum absorption of more than 100 L1 centres is plotted in figure 4(b). This distribution indicates that the dipoles of the L1 centres are oriented along the 〈111〉 axes (figure 4(c)). If they would have been oriented along the 〈110〉 axes (as it will be seen later for the 546 nm centres), additional absorption maxima of lower intensity would have been found along the [100] and [010] directions, which is not the case. Moreover, the knowledge of the dipole orientation is of importance in what concerns the integration or coupling of single defect centres within photonic structures such as cavities or pillars. The narrow emission line at room temperature, the short lifetime, the brightness and the highly polarised emission are all important properties for the use of the L1 centre as a single-photon source. Although ZPL emission peaks have already been reported in the same spectral region as for the L1 centres, on the basis of the previously mentioned features and of the vibronic sideband this centre does not correspond to any known natural or artificial diamond defect line in the literature.

Zoom In Zoom Out Reset image size

Besides their high linear polarisation, several L1 centres show a blinking behaviour. Such centres, if integrated in nano-diamonds, are of interest as fluorescent markers for stochastic methods of super resolution microscopy techniques such as photo-activated localisation microscopy (PALM) [35] or stochastic optical reconstruction microscopy (STORM) [36]. In this view it is important to characterise the duty cycles and blinking dynamics of the centres. For a use in STORM/PALM techniques, the fluorescent markers need to be ON only a small period of time with respect to the time they are OFF. The sub-diffraction nanometer accuracy is obtained by reconstructing an overall image from a series of single images in which only a fraction of the centres is turned ON. In the scan of figure 1, four L1 centres having different blinking dynamics can be found. The fluorescence time trace of a chosen blinking single L1 centre (dotted circle) is recorded for different laser powers (from 200 to 1000 μW) for 2 min each, and plotted in figure 5(a). The higher the excitation, the faster the blinking of the L1 centre is driven.

Zoom In Zoom Out Reset image size

The resulting durations of the ON and OFF states obtained from the time traces have then been sorted in a statistical distribution (not shown) and further fitted by Gaussian functions. Note that the intensity of the OFF state corresponds to the background fluorescence coming from the diamond. A threshold for the fluorescence intensity is then chosen to discriminate between the states in the trace. The distribution of the time spent in the bright ON state decreases exponentially, as well as for the time spent in the OFF state. As seen in figure 5(b), the bright to dark ratio (duty cycle) decreases as the laser power is increased. It is of advantage for STORM that the smallest duty cycle (about 10%) happens when the centres are the brightest. This would lead to smaller acquisition times for the single images and thus faster elaboration of the high-resolution reconstructed image. By analogy with previous work on NV^-/0 photo-induced blinking of NV centres [37], the 'ionising' and 'recharging' rates of the L1 centre are plotted in figure 5(c), as a function of the laser power. Both curves are fitted using a second order polynomial function with fitting parameters found to be a = 0 and b = 2 × 10⁻⁶ for the ionisation rate curve, and a = 1.5 × 10⁻⁴ and b = 1.3 × 10⁻⁷ for the recharging rate curve. Both rates possess a term of quadratic increase with the laser power, characteristic of two-photon processes in the ionisation and recharging processes. The purely quadratic power dependence of the ionising rate suggests that the ground state of the L1 centre lies more than 2.13 eV below the conduction band, following a similar reasoning applied to NV centres in [37].

The cause of the blinking behaviour is not clear at this stage of the investigation. Possible explanations are a transition in a dark state or a change of the charge state. In some cases, the L1 centres become irreversibly dark, thus a change of the charge state is more likely. These behaviours are probably due to different local environments of the L1 centres, as it can also be observed for other optical centres in diamond such as the NV centre and the SiV centre [38]. Further investigations of the stability/blinking behaviour of the L1 centres are planned, when donor and/or acceptor atoms are co-implanted to the L1 centres. Finally, we have investigated whether the L1 centre is ESR active in the MW spectral ranges of NV and ST1 centres (0.94–1.04 GHz, 1.22–1.32 GHz and 2.82–2.92 GHz) but no resonances could be found in the optical fluorescence.

The second kind of centre which was found together with the L1 centre within the implanted samples is denoted here as the L2 centre. Note that, in the first sample, the L1 centres were found in majority (90%–95%) with respect to the L2 centres (5%–10%) whereas an opposite ratio was found in the second implanted sample. The L2 centre presents different properties than the L1 centre. Firstly, the fluorescence spectrum of a typical L2 centre is plotted in figure 6(a) (with background correction). It can be seen that, at room temperature, the emission consists of a narrow ZPL at 545.9 nm (2.274 eV) and of a broad phonon sideband extending up to 800 nm. The FWHM of the ZPL is 1.4 nm and about 2% of the light is emitted in the ZPL. The observation of several L2 centres revealed that the position of the ZPL is spread, in the range 540–547 nm. The brightness of the L2 centres is changing from one centre to the other and most of the L2 centres show a blinking behaviour, as illustrated in figure 6(b). The single emitter nature of the L2 centres is demonstrated by the g⁽²⁾ function plotted in figure 6(c), taken at an excitation power of 200 μW. To fit these data, we assumed a three level model. From this approach, an excited state lifetime of about 9 ns has been extracted. At higher excitation power (1.66 mW) bunching shoulders appear (inset of figure 6(c)), revealing a long-lived shelving state of about 700 ns.

Zoom In Zoom Out Reset image size

Polarisation-dependent absorption measurements have also been realised with the L2 centres. Four different possible orientations are observed. The highest brightness is observed when the polarisation is along the 〈110〉 axes, while the lower brightness centres are observed for polarisation along the 〈100〉 axes. For symmetry reasons, it can therefore be deduced from these measurements that the absorption dipoles are oriented along the 〈110〉 axes. There are six equivalent positions for 〈110〉 oriented defects, which however reduce to four, as measured here, when projected onto the (001) plane.

The L2 centres were further subjected to continuous ODMR measurement schemes. This revealed two positive lines around 0.99 and 1.27 GHz with linewidths of 1–4 MHz (figure 7(a)). For these lines, contrasts of up to 6% and 20% were found, respectively. No intrinsic hyperfine splitting could be found, indicating that the atoms involved in the defect do not have a nuclear spin.

Zoom In Zoom Out Reset image size

The kinetics of the shelving state observed in the second-order photon correlation function at high power has been investigated, with and without the employment of microwave π-pulses at 1.26 GHz. In the decay curves of the fluorescence (figure 7(b)), double exponential decays are observed, suggesting that the spin triplet is the shelving state, and not the ground state. A possible level scheme for this defect, with the corresponding measured lifetimes, is depicted in figure 7(c), as inspired from [24]. An additional presumed fourth level (dotted line) is however suggested here to cope with the discrepancy of the g⁽²⁾ power sweep measurements (not shown) when a three level system is assumed. This additional level may furthermore account for the decrease of fluorescence when the excitation is increased (observed in the saturation curve, grey hollow circles in figure 6(b)).

The positive and relatively high contrast of the ODMR spectra is due to the longer lifetime (∼2500 ns) of the spin triplet sublevel $| 0\rangle$ than the lifetime of the $| -\rangle$ sublevel (∼700 ns) and of the $| +\rangle$ sublevel (∼100 ns). This makes an optical pumping of the $| 0\rangle$ sublevel possible as well as a coherent manipulation of the associated spin. In the context of quantum technologies in the solid state based on the coupling between electron and nuclear spins, the spin-free ground state of the L2 centre is of advantage compared to the triplet ground state of the NV^- centre, for which the presence of the electron spin causes decoherence to the nuclear spin. In the case of the L2 centre, the spin manipulation and readout can be achieved within the metastable excited spin state.

Rabi oscillations have then been performed to illustrate the coherent control between the triplet spin states; both between the $| 0\rangle$ and $| +\rangle$ states (1.27 GHz) and between the $| 0\rangle$ and $| -\rangle$ states (0.99 GHz). Two schemes have been applied: with constant laser illumination (figure 7(d) left) and with pulsed light (figure 7(d) right). As seen in the two figures, the amplitude of the oscillations and the decay envelope can be well fitted using the model and time constants presented in figure 7(c).

The ODMR spectra of the L2 centres are very similar to the ones reported for the ST1 centre [24] (zero-field splitting parameters D = 1.135 GHz and E = 139 MHz) even if the ST1 centre shows a third resonance peak at 278 MHz. The optical spectra at room temperature are slightly shifted: the ZPL of the ST1 centre is at 550 nm and it is weaker than the one observed for the L2 centre. However, the ST1 and L2 centres show the same excited state lifetime of about 9 ns. The ST1 centre (the nature of which is still unknown) was found in an ultrapure single-crystal HPHT diamond, after the fabrication of nanowires [24]² . The wires had been etched from the diamond surface by reactive ion etching using inductively coupled oxygen plasma. Afterwards, the sample had been annealed in vacuum at 850 °C for several hours. It was suggested by the authors that oxygen could be a potential constituent. Indeed, the ODMR measurements of single ST1 centres did not show any hyperfine coupling and ¹⁶O has no nuclear spin. In the case of the L2 centre observed in our sample, we also did not find any hyperfine coupling. It is therefore very likely that the L2 centre and the ST1 centre are the same defect, possibly with oxygen involved in it. The shift of the ZPLs may be explained by lattice deformation or by the presence of another impurity species.

Both the L1 and L2 centres have been created simultaneously twice, in two different samples, using the same implantation conditions and cathode source. It is important to note that only the implanted faces (front face of sample 1 and front and back faces of sample 2) show the presence of the L1 and L2 centres. No centre can be found on the unimplanted backface of sample 1, which excludes that the centres may have been induced solely by thermal annealing. However, the (probably) bi-atomic molecular ion used for implantation does not allow an unambiguous identification of their nature. The magnetic field of the 90° bending magnet of the accelerator was set up to select only the ions having a mass/charge ratio of 27 amu/elementary charge, from a cathode made of a mixture of graphite and boron nitride. Note that even if it has not been intentionally introduced in the cathode, a significant ¹⁶O^- peak was observed in the mass spectrum revealing the presence of oxygen. The implanted ions were therefore expected to consist most likely of (¹³C¹⁴N)⁻, (¹²C¹⁵N)⁻ and (¹¹B¹⁶O)⁻ molecules [39, 40], the relative amount of BO^- being nevertheless difficult to estimate. However, (¹²C¹⁵N)⁻ or (¹³C¹⁴N)⁻ molecules are routinely used to produce NV centres and, surprisingly, no NV centres have been found within the implanted samples. This fact, together with the absence of hyperfine coupling within the ODMR spectra of the L2 centres, does support the (¹¹B¹⁶O)⁻ hypothesis for the L2 centres. Besides, a control implantation with an aluminium source (in case of a non-negligible Al content in the boron/graphite cathode) which also has a mass of 27 amu has been done and neither the L1 centres nor the L2 centres have been produced. Note that a control implantation of ¹¹B^- ions also did not lead to the creation of these centres. Furthermore, the unintentional implantation of ions of neighbouring masses 26 or 28 amu is not excluded, but is unlikely because it would have led mainly to the creation of NV centres (¹²C¹⁴N) or SiV centres (²⁸Si), which were not observed. The different L1/L2 ratios observed between sample 1 and sample 2 are not yet understood, however different oxidation degrees of the B-C-N cathode (at the moment of the implantation) may play a role. Regarding the high-energy triple feature of the LVM of the L1 centres, the blinking of some of the L1 and L2 centres, the fact that they were created together and the absence of NV centres, further experimental and theoretical studies related to the possible involvement of hydrogen in the L1 and L2 centres would be of interest, since the hydrogen content and homogeneity may vary significantly between different CVD diamonds. For this purpose, NV centres could be co-implanted and be used as local hydrogen sensors: it has recently been shown that hydrogen diffusion leads to significant passivation of NV centres, making them optically dark [41, 42]. Note that a third sample (element six, electronic grade) has been implanted with the mass 27 amu, using another B-C-N cathode source and that the L1 and L2 centres could not be found. Ab initio calculations of the LVM observed in the fluorescence spectra of the L1 centres, together with the different polarisation properties may help finding the nature and structure of the L1 and L2 centres.

Finally, it would be of interest to study the application potential of the L2 centre as a new colour fluorescent marker for sub-diffraction STED microscopy [43]. The STED technique requires fluorescent markers having broad and spectrally shifted absorption and emission. This is typically the case for the NV centre which, together with an exceptional photo-stability, can be imaged down to less than 10 nm using this technique [26, 44]. The fluorescence of the L2 centre is similar and blue-shifted by about 90 nm with respect to the NV centre. A STED setup dedicated to the NV centres (532 nm excitation and 775 nm depletion typically) is expected to be also well suited to the L2 centres.