Diamond photonics platform enabled by femtosecond laser writing

Diamond is a promising platform for sensing and quantum processing owing to the remarkable properties of the nitrogen-vacancy (NV) impurity. The electrons of the NV center, largely localized at the vacancy site, combine to form a spin triplet, which can be polarized with 532 nm laser light, even at room temperature. The NV’s states are isolated from environmental perturbations making their spin coherence comparable to trapped ions. An important breakthrough would be in connecting, using waveguides, multiple diamond NVs together optically. However, still lacking is an efficient photonic fabrication method for diamond akin to the photolithographic methods that have revolutionized silicon photonics. Here, we report the first demonstration of three dimensional buried optical waveguides in diamond, inscribed by focused femtosecond high repetition rate laser pulses. Within the waveguides, high quality NV properties are observed, making them promising for integrated magnetometer or quantum information systems on a diamond chip.

for photonics applications. Using an angled plasma etching process, suspended triangular nanobeam waveguides were recently demonstrated 14,17 with losses < 10 dB/cm at visible wavelengths however for each design, a custom Faraday cage is required. Waveguides have also been demonstrated in diamond using isotropic etching 18 and ion beam writing 19 but as with other methods, optical circuits are restricted to 2D geometries.
We propose a disruptive technology using femtosecond laser writing to realize a 3D photonics toolkit for diamond. In this method, focused ultrashort pulses are nonlinearly absorbed in the bulk of a transparent material, leading to a localized modification 20 . In glasses this modification is a refractive index increase which enables 3D waveguide writing [20][21][22][23][24][25][26][27][28] but in crystals, the lattice is damaged yielding a decrease in refractive index 22 . Here, we laser-write closely spaced parallel lines in diamond to confine a guided optical mode between them. The high repetition rate of 500 kHz was found to reduce the formation of highly absorptive graphite, enabling optical waveguiding. Crucially, we find high quality NV center properties within the guiding region. Optical waveguides formed by directly focused femtosecond laser pulses can potentially be patterned in three dimensions to optically link or address NVs, to enable efficient excitation and collection, and make quantum information systems and spin-based sensing a reality.

Results
The effect of repetition rate on bulk modification of diamond. The initial focus of our study was to produce uniform modification lines below the surface of diamond, which could be used to form optical waveguides. At 500-kHz repetition rate, we found that an average power of 50 mW and a scan speed of 0.5 mm/s along < 110 > crystallographic directions produced uniform and reproducible modifications in the bulk of diamond. As shown in the optical microscope image in Fig. 1(a), the cross section of the laser-written track was approximately 5 μ m transversely and 22 μ m vertically. The significant vertical elongation of the line is attributed to the large spherical aberration caused by the mismatch between the index matching oil (n = 1.5) and diamond (n = 2.4). This asymmetry in the modification could be corrected using adaptive optics 29 . (a) Transverse optical microscope image of single laser-induced track written with 500-kHz repetition rate, 50-mW average power and 0.5-mm/s scan speed. (b) μ Raman spectra (532-nm excitation wavelength) at four different vertical positions inside the modification. 'Out' refers to a spectrum taken outside the track. The spectra have been normalized to the diamond peak to show the change in the relative intensity of the G-peak inside the structure. (c) μ Raman spectra (normalized to the G-peak) in the center of modification tracks at repetition rates of 5 kHz, 25 kHz and 500 kHz, with pulse energy held constant (800 nJ) to produce a similar size modification at each repetition rate. (d) Tauc plot for diamond with tracks written over the entire sample at 50-μm depth and 20-μ m line separation for 5-kHz and 500-kHz repetition rates. It is considered that the absorption in the visible region is only due to the modification tracks, with the rest of the sample being transparent. To better understand the structure of the femtosecond laser-written lines, μ Raman spectroscopy was performed. As a reference, the pristine diamond 30 has a characteristic Raman peak centered at 1332 cm −1 with a typical FWHM of around 2.3 cm −1 . For the above mentioned laser processing conditions, μ Raman characterization ( Fig. 1(b)) revealed that within the modification, there is a reduction of the intensity of the peak at 1332 cm −1 to 15% of the original intensity along with an increase of its width by about 2 cm −1 , evidencing increased disorder in the diamond lattice. Moreover, at least two new bands appear: the G-peak at 1575 cm −1 and the D-peak at 1360 cm −1 , showing a transformation of the sp 3 bonding of diamond into sp 2 bonding. In our case, the widths of D and G peaks (greater than 100 cm −1 ) and the intensity ratio between them (I(D)/I(G) less and close to 1) indicate that these sp 2 clusters are mainly in an amorphous carbon phase rather than graphite 31 . This is in contrast to previous work with 1-kHz repetition rate Ti:Sapphire femtosecond lasers 32 , which demonstrated the formation of micrographitic lines in diamond, which are not desirable for photonic devices due to the strong absorption.
In fact, when we reduce the repetition rate from 500 kHz to 5 kHz, we observe a sharper G-peak with a slight displacement to higher wavenumber ( Fig. 1(c)), which implies a greater concentration of nanocrystalline graphite clusters 33 . We also found that the second order peaks at 2700 cm −1 (2 D peak) and 2900 cm −1 (D + G peak) appear for 5 and 25 kHz, but not for 500 kHz, evidencing increased graphitization at these lower repetition rates. Compared to the laser repetition rate, we found that the pulse energy and scan speed played a lesser role in the amount of graphitization.
Bulk optical waveguiding in diamond. Motivated by the μ Raman results at 500-kHz repetition rate, we attempted to form waveguides in diamond using the type II method to produce an optical waveguide between two closely inscribed modification lines 22 . Figure 2(a) shows the transverse optical microscope image of the pair of laser-inscribed lines written 50 μ m below the surface using the same laser processing conditions as the single line in Fig. 1(a). For 13-μ m line separation, we produced for the first time to our knowledge, a buried optical waveguide in diamond using femtosecond laser writing. While the waveguide showed single mode behavior when scanning the input launch fiber transversely, when scanned vertically, modes could be excited at three different depths: the lowest loss mode was centered between the modification lines, as shown in Fig. 2(a). A mode could also be excited a few microns above and below this position (not shown in Fig. 2(a)). The mode field diameter (MFD) of the central and lowest loss mode is 10 μ m × 11 μ m at 635-nm wavelength with an insertion loss of 14 dB using butt-coupled single mode fibers (including a coupling loss of 2.8 dB/facet, Fresnel reflection loss of 0.3 dB/facet and propagation loss of 16 dB/cm). We also found optical waveguiding with similar losses in the near infrared (808 nm and 1550 nm). We attribute the propagation loss mainly to scattering loss due to the overlap of the optical mode with the modification tracks. The propagation loss would be much higher should graphite be present in the laser-written lines. By evaluating the attenuation of the optical mode 34 we estimate a waveguide damping loss of greater than 10 4 dB/cm for type II modification tracks containing graphite. Further optimization of laser processing conditions to yield less overlap of the optical mode with the modification lines is expected to reduce the moderately high damping losses reported here.
At 532 nm, the wavelength for incoherent excitation of NV centers in diamond, the MFD was 9 μ m × 9 μ m and similar losses were found compared to 635-nm wavelength. By using an end-fire free-space coupling setup with an adjustable polarizer at the input, we found that the type II waveguide from Fig. 2(a) supported only the TM mode. Similar polarization-dependent behavior has been reported in type II waveguides in other crystals such as lithium niobate 22 , KDP 35 , and Ti:Sapphire 36 and can be attributed to the elliptical shape of the modification, which results in a different stress-induced refractive index distribution for TE and TM polarizations 22 .
We further confirmed that 500 kHz offers a better regime for type II modification of diamond by lowering the repetition rate of the laser to 5 kHz. Over a wide range of laser processing parameters tested (200 nJ to 1 μ J pulse energy, track separations of 10 to 18 μ m, scan speeds of 0.01 to 10 mm/s), we could not obtain waveguides with well-confined modes or reasonably low insertion loss. Waveguiding may be impeded by increased graphitization at lower repetition rates ( Fig. 1(c)) resulting in higher absorption of the launched visible light, supported by absorption measurements on samples with tracks written over the entire 5 mm × 5 mm area (20-μ m spacing between lines at 50-μ m depth). As shown in Fig. 1(d), for visible wavelengths, the absorption is slightly higher for lower repetition rate processing. We also found that the optical bandgap was reduced for 5-kHz processing, which is expected when graphitic inclusions are present 37 .
To achieve single mode guiding, we wrote a second vertically offset type II modification ( Fig. 2(b)). This four-line modification supported a single mode at 635-nm wavelength with a MFD of 9 μ m × 10 μ m, an insertion loss of 19 dB and a coupling loss of 2.4 dB/facet. Similar to the case of the two-line type II waveguide, the four-line structure supported only the TM mode. Although each of the four lines shown in Fig. 2(b) was written with the same laser processing conditions, the vertical elongation of the lines increases dramatically with depth due to increased spherical aberration, which could be avoided by employing a spatial light modulator to impose a phase profile that is equal and opposite to the aberration introduced by the refractive index mismatch 38 . This would allow more flexible and precise 3D patterning of the barrier zones of the type II waveguide, enabling a more symmetric four-line structure or even an annular type modification for confining light.
μ Raman measurements in diamond are sensitive to the presence of stress in the material. A shift of the diamond peak to higher (lower) wavenumber is associated with a compressive (tensile) stress. In Fig. 3(a) we present a map showing the peak shift near two laser written tracks. It can be seen that inside the guiding region the stress is mainly compressive, with a shift of about + 1.5 cm −1 compared to pristine diamond. A similar shift in the diamond peak is observed at the center of the four-line type II structure (Fig. 3(b)). However, previous studies have shown that compressive stress in diamond results in a decrease of the refractive index 39 . Therefore, the optical waveguiding we observe in the type II structures may be due to increased polarizability 40 or simply from a reduced refractive index in the modification lines, which serve as barriers for confining the optical mode. Figure 4(b) plots the width of diamond Raman peak within the guiding region of the two-line type II waveguide. Within experimental uncertainty (1 cm −1 ), the width of the crystal peak in the guiding region is the same as its value in pristine diamond (2.3 cm −1 ). This is evidence that the crystalline structure in the guiding region is preserved, even if it is under compressive stress as shown in Fig. 3. NV center properties within waveguides. The type II modification is beneficial for quantum information and magnetometry applications relying on NV centers, since the waveguide mode propagates in the undamaged region between the laser-inscribed lines. Photoluminescence (PL) characterization using a confocal microscope revealed a reduction in the intensity of the NV's zero phonon line (ZPL) transition within the modification lines (Fig. 5(a)). However, in the region between the modification lines where the optical mode propagates, the PL spectrum appears to be the same as that of pristine diamond. To give a quantitative estimate we compare the ratios of the integrated intensity of the NV − center ZPL to the 1 st order Raman of the pristine diamond and waveguide spectra, which are 0.728 ± 0.006 and 0.733 ± 0.006, respectively. This suggests there is not a statistically significant change in the level of NV − centers within the waveguide region.
By launching 532-nm light into the waveguide and collecting the light using a spectrometer, we found a similar PL spectrum as that obtained by confocal microscopy (Fig. 5(b)). As the waveguide only supported the TM mode, the difference in the PL recorded from the TM and the TE configuration (shadowed in Fig. 5(b)) is the actual contribution of the guiding region to the PL, compared to the bulk material excited around the guide (Fig. 5(b) inset). These measurements provide strong evidence that the PL of the NV centers excited with the 532-nm laser inside the guiding region can be effectively collected and guided using the fabricated waveguide.   Within the modification lines, a bright signal was observed between 730 nm and 800 nm, which is attributed to the GR1 color center (740 nm) associated with neutral vacancies 41 created by laser writing. It is possible that other vacancy complexes and interstitial defects are also induced causing spectral features similar to the radiation B Band of diamond which has undergone ion implantation and annealing 42 .
Optically detected magnetic resonance (ODMR) characterization of the NV centers' electron spin within the buried diamond waveguides showed hyperfine structure ( Fig. 6(a,b)) with linewidths of Δ = 1.7 ± 0.4 MHz, comparable to those measured from pristine diamond (Δ = 1.8 ± 0.3 MHz) for the same measurement parameters. These linewidths provide a lower bound on the pure dephasing time (T 2 * = 1/(π Δ ) of the electron spin ensemble where T 2 * = 0.2 μ s for the case of the NV centers in the waveguide. The linewidths recorded in these measurements are broader than those reported for similar grades of diamond (0.2-1 MHz 43 ) though this could be caused by power broadening 44 and imperfect static field alignment 45 . Further, the lifetime of the excited state transition (11.0 ± 1.5 ns) was the same for NV centers within the waveguide and pristine diamond (Fig. 6(c)). These measurements provide evidence that the electronic structure of the NV centers was not severely affected by femtosecond laser inscription. Direct measurement of the spin coherence time of NV centers and the application of dynamical decoupling techniques will give a more complete picture of the effect of the laser writing process on the NV center and its coherence properties.

Discussion
The reduction in graphitization with MHz-repetition rates compared to kHz rates may be due to the higher temperatures driven by the higher pulse delivery rate. Jerng et al. observed that above growth temperatures of 1100°C, amorphous carbon was produced instead of nanocrystalline graphite films 46 . In agreement with characterization by μ Raman spectroscopy, absorption and waveguide transmission, we have found that the electrical conductivity of lines written with the femtosecond laser is increased at lower repetition rates 47 . Future characterization including electro-chemical etching of the laser-written lines will seek to provide more insight into the modifications induced within the MHz repetition rate regime exploring other exposure parameters such as beam shaping with the SLM, pulse duration and wavelength.
Working at room temperature, magnetometry devices based on NV ensembles 48 would benefit from waveguides for efficient collection and routing of the fluorescence signal. Even more compelling is a diamond waveguide device with integrated Bragg reflectors at green and infrared wavelengths, to enable an integrated cavity enhanced magnetometer 49 . Bragg gratings waveguides have been demonstrated in glass by periodically modulating the intensity of the femtosecond laser pulse train during waveguide writing 50,51 , and a similar approach will be applied to diamond. Diamond waveguides with Bragg reflection functionality would enable a significant enhancement of the magnetic-field sensitivity since the cavities increase the absorption path length by a factor proportional to their finesse 52 . Diamond has shown itself as an important material for quantum photonics, being a host to defect centers with atom-like properties having long-lived spin quantum states and well-defined optical transitions. The 3D photonics toolkit that we have developed demonstrating high quality NV properties while allowing easy interfacing to standard optical fibers may help diamond reach its full potential for quantum technologies.

Methods
The femtosecond laser used for waveguide writing in diamond was a regeneratively amplified Yb:KGW system (Pharos, Light Conversion) with 230-fs pulse duration, 515-nm wavelength (frequency doubled), focused with a 1.25-NA oil immersion lens (RMS100X-O 100 × Olympus Plan Achromat Oil Immersion Objective, 100× oil immersion, Olympus). Employing such a high NA allows for a smaller focal volume, to minimize the writing power and avoid self-focusing (critical power 2 MW at 515 nm), which lead to vertically elongated waveguide cross sections and non-reproducible results.
The polarization of the incident laser was perpendicular to the scan direction. The repetition rate of the laser was variable from 1 MHz to single pulse. Computer-controlled, 3-axis motion stages (ABL-1000, Aerotech) interfaced by CAD-based software (ScaBase, Altechna) with an integrated acousto-optic modulator (AOM) were used to translate the sample relative to the laser to form the desired photonic structures.
Polished 5 mm × 5 mm × 0.5 mm synthetic single-crystal diamond samples (type II, optical grade with nitrogen impurities 100 ppb) were acquired from MB Optics. Laser-inscribed structures were characterized for their morphology using white-light optical microscopy in transmission mode with 10× and 40× magnification objectives (Eclipse ME600, Nikon). We also fabricated waveguides in electronic grade single-crystal diamond (nitrogen impurities 5 ppb, dimensions 3 mm × 3 mm × 0.2 mm) from MB optics which revealed similar propagation losses and mode profiles.
For waveguide transmission measurements, high resolution 3-axis manual positioners (Nanomax MAX313D, Thorlabs) were used. The four-axis central waveguide manipulator (MicroBlock MBT401D, Thorlabs) enabled transverse displacement between sets of diamond waveguides. Light sources at 808 nm (S1FC808, Thorlabs), 635 nm (TLS001-635, Thorlabs) and 532 nm (4301-010, Uniphase) were coupled to the waveguides using the appropriate Thorlabs single-mode fibers for each of the visible wavelengths (460HP for 532 nm, SM600 for 635 nm, 780HP for 808 nm). To test the polarization dependence of the waveguide transmission, free-space coupling was used with 10× (5721-H-B, Newport) and 60× (5721-H-B, Newport) lenses at the input and output, respectively. At the output, light was coupled to an optical power meter (818-SL, Newport) to measure the power transmitted through the waveguide. To measure the near-field waveguide mode profile, a 60× asphere (5721-H-B, Newport) was used to image the light to a beam profiler (SP620U, Spiricon).
Micro-Raman spectra were recorded using a Labram Aramis Jobin Yvon Horiba microRaman system with a DPSS laser source of 532 nm and equipped with a confocal microscope and an air-cooled CCD. A 50 × (100× ) objective was used to focus the laser on the sample as well as to collect the Raman signal, with a spatial resolution of about 1 micron. A wavenumber accuracy of about 1 cm −1 can be achieved with a 1800 line/mm grating.
To observe the characteristic fluorescence from the negatively charged NV center at 637 nm (and the phonon sidebands), 532-nm light from a 1-W frequency-doubled Nd:YVO 4 laser (Verdi, Coherent) was coupled with and collected with the free space optics described above and detected with a spectrometer (Ocean Optics model HR2000). A notch filter in the green was used to attenuate the pump wavelength.
For confocal photoluminescence measurements, nitrogen-vacancy defects were excited with a DPPS 532-nm laser (CL532-500-L, CrystaLaser) focused on to the sample with a 0.55-NA objective (100× Plan Apo SL Infinity Corrected Objective, Mitutoyo). Photoluminescence was collected through the same objective, filtered from the excitation light using a dichroic beamsplitter (ZT 532 RDC, Chroma) and long-pass filters (ET 555 LP Chroma, FELH 0650 Thorlabs) and focused into a single mode fibre which provided the confocal aperture. Photon counting of the filtered light was performed using an avalanche photodiode (SPQR-14, Perkin-Elmer). Optically detected magnetic resonance measurements were performed by monitoring the fluorescence rate while scanning the frequency of a microwave field driven through a 20 μ m copper wire on the surface of the sample generated using a commercial microwave source (Agilent ESG E4433B) amplified by a high power broad band amplifier (Minicircuits ZHL-16W-43+ ). A static magnetic field was applied using a permanent magnet. The field was ~90 G for the measurements in Fig. 6(a,b). Lifetime measurements were performed using a supercontinuum source for excitation filtered at 532 nm (Fianium Whitelase WL-SC400-4), and a time correlated single photon counting board (Timeharp 260).