Highly nitrogen-vacancy doped diamond nanostructures fabricated by ion implantation and optimum annealing

In diamond, many impurities and their vacancy-related centers exhibit bright luminescence in a wide wavelength range^1–5 and have thus been of interest as single-photon emitters, optically accessible quantum bits, and magnetic-field sensors.^6–8 To obtain higher emission efficiency and sensitivity, diamond photonic nanostructures have also been fabricated to enhance electromagnetic interactions between impurity-related centers and optical fields confined in the nanostructures.^9,10 Most of the quantum optical applications using single or a few photons require the formation of single or low-density impurity centers in diamond. In contrast, this study focuses on nanostructured diamond with a high concentration of impurity centers, which is needed to achieve a diamond color center laser available as a tunable visible light source.^11,12 It has been recently proposed that a diamond color center laser can also work as a sensitive magnetometer, thanks to a large ensemble of an impurity center and its intense optical output.¹³ Since a high concentration of impurity centers yields a sufficient optical gain for lasing and a high sensitivity for magnetic fields, we should take an approach to impurity doping different from that for previous quantum optical devices. In addition, fine fabrication of diamond optical nanocavities is required for strong optical confinement. However, it is still challenging to prepare air-suspended single-crystal diamond membranes, which are platforms for fabricating planar nanocavities. One of the developed methods is to combine mechanical thinning and long-time oxygen reactive ion etching (RIE) of diamond substrates. However, with this method, it is difficult to fabricate large-area membranes with a uniform thickness because of a limited parallelism of the starting diamond substrates.¹⁴ Two other methods are the heteroepitaxial growth of thin diamond films on semiconductor substrates with an intermediate catalyst¹⁵ and anisotropic diamond etching with a specially designed metal cage or unique mask deposition.^16,17 Although these pioneering methods can successfully fabricate high-quality membranes, in this study, we took an approach that uses high-dose ion implantation,^18,19 which has several advantageous characteristics. First, it is based on a common fabrication procedure that does not require additional crystal growth and special metal cages. Second, it enables to fabricate large single-crystal diamond membranes with a uniform thickness because the thickness is precisely determined by ion implantation, not by substrate parallelism, which is preferable for simultaneous fabrication of many devices in a sample. Finally, the ion implantation can create a dense ensemble of impurity-related emission centers in diamond membranes.

An outline of the fabrication process in this study is shown in Fig. 1(a). Accelerated ions are implanted into a single-crystal diamond substrate with a high native nitrogen concentration to create an intermediate amorphous carbon (a-C) layer in the substrate. The thicknesses of the top diamond membrane and the a-C layer can be controlled by the ion energy and dose amount, which results in membrane thickness that is uniform over the entire sample area. At the same time, since the ion implantation creates numerous vacancies, a dense ensemble of nitrogen-vacancy (NV) centers can be created in the diamond membrane after thermal annealing. However, the ion implantation severely damages the crystal in the fabricated membrane, so the annealing condition should be well optimized to recover the crystal quality and obtain the largest possible emission intensity from the NV centers. To find the optimum annealing condition in this fabrication method, we performed a systematic analysis to characterize the diamond membranes and kinetics of the NV centers. In the diamond membrane supported by a-C, fine diamond nanowires that are potentially available for various photonic applications^20,21 were patterned by electron beam (EB) lithography with a metal mask and reactive ion etching. The a-C layer was removed by boiling acid to fabricate air-suspended diamond nanowires at the top of the substrate.²² Finally, the nanowires were transferred to another substrate, which enables us to measure photoluminescence (PL) from only the thin diamond nanowires without substrate emission.

First, we fabricated single-crystal diamond membranes with different thicknesses and performed multiple analyses to characterize them. To create a sacrificial a-C layer for air suspension, accelerated helium (He) ions were implanted into 0.3-mm-thick type-Ib single-crystal diamond substrates with a (001) polished surface (Sumitomo Electric Corp.). The substrate area was 3 × 3 mm². Based on Monte Carlo simulation and the critical vacancy concentration for amorphization, the acceleration voltage and ion dose were set to 160 kV and 4 × 10¹⁶ cm⁻² for 400-nm-thick membranes used in the optical characterization and to 50 kV and 2 × 10¹⁶ cm⁻² for 150-nm-thick membranes used for fine nanowire fabrication, respectively. Metal nails fixing the diamond substrates made shadows that blocked the ion irradiation so that unimplanted regions were obtained along with ion-implanted regions, which enable their optical comparison within the same fabrication run. The sample temperature was increased during ion irradiation, but less than 100 °C. The ion-implanted substrates were annealed for 4 h at a pressure of 600 Torr with an argon flow. The annealing temperature (T_A) ranged from 1000 °C to 1350 °C. Figure 1(b) is a cross-sectional transmission electron microscope (TEM) image of an ion-implanted substrate after thermal annealing at T_A = 1150 °C. Here, the ion acceleration voltage was 160 kV. In this image, four distinguishable regions are found: an a-C layer for surface protection, a sample top layer, a low-contrast middle layer, and the diamond substrate. The a-C protection layer was deposited by a sputtering method on only the sample used for TEM observation. The top and low-contrast middle layers are 400-nm and 150-nm thick, respectively. In Fig. 1(b), the simulated depth profiles of the ion-induced vacancy concentration are also shown.²³ An upper boundary between the top and low-contrast middle layers was created at the depth where the vacancy concentration reaches to the critical value of ∼8 × 10²² cm⁻³.¹⁹ A lower boundary is found at the depth larger than that expected by simulation, which might cause deeper ion penetration in the lower-density middle layer. During annealing, slight sharpening of the layer boundaries was observed, but there was no large change in the layered structure. As seen in the inset of Fig. 1(b), an optical microscope image for the sample shows no interference fringe in the entire ion-implanted region, which confirms high thickness uniformity of the top diamond layer. Figure 1(c) shows selective area electron diffraction (SAED) images along the [110] zone axis at three regions where the probing area is approximately 60 nm in diameter. The SAED patterns of the top layer and the substrate show ordered spots corresponding to the crystalline diamond structure, and the spots are identical, which indicates that the top layer maintains crystalline diamond with an orientation identical to that of the substrate. The absence of a grain feature in the TEM image suggests that the top layer is a single crystal. In contrast, the SAED of the low-contrast middle layer shows a center spot and weak halos, indicating the absence of long-ranged periodical structures. The innermost halo is slightly weighted to the horizontal position. This diffraction pattern indicates that the middle layer is a-C embedding graphite, whose honeycomb plane is weakly aligned parallel to [001].^24,25 Figure 1(d) shows electron energy loss spectroscopy (EELS) spectra near the carbon K-edge absorption.²⁶ At the capping a-C and the low-contrast middle layer, the EELS spectra exhibit a peak at 285 eV corresponding to carbon sp² bonds, which is further evidence that the middle layer consists of a-C. However, the complete disappearance of the peak at 285 eV from both the top layer and substrate confirms that they are crystalline diamond with no significant sp² bonds. The multiple peaks observed above 290 eV originate from carbon sp³ bonds. From these TEM, SAED, and EELS results, we can conclude that the top and low-contrast middle layers are single-crystal diamond and a-C, respectively. Although there is no significant difference in the TEM images and EELS spectra between the top diamond layer and the substrate, their optical properties should be investigated with sensitive luminescence spectroscopy, as discussed later.

The concentrations of nitrogen in pristine and fabricated samples are estimated to be 1.4 × 10¹⁹ cm⁻³ by secondary ion mass spectroscopy (SIMS), as shown in Fig. 1(e). The corresponding nitrogen ratio is as high as 80 ppm, which results from the high-pressure and high-temperature (HPHT) synthesis of type-Ib diamond. The depth distributions in the nitrogen concentration are flat across the top, middle a-C, and substrate regions, which verifies the uniformity of the nitrogen concentration in the three layers and that there was no change in its concentration during the fabrication.

To obtain the highest possible NV center emission and concentration, we looked for an optimum annealing temperature while measuring luminescence from the top diamond layer. Low-temperature cathodoluminescence (CL) spectroscopy with a low-energy electron beam was performed with an electron acceleration voltage of 5 kV at which the electron beam excites only the top 400-nm-thick diamond. Figures 2(a) and 2(b) show wideband CL spectra of an unimplanted and implanted diamond region with different annealing temperatures. The magnified CL spectra at a wavelength of around 570 nm are shown in Figs. 2(c) and 2(d) to focus on the nitrogen-related emission peaks. In Figs. 2(a) and 2(c), at T_A = 1150 °C, the unimplanted diamond exhibits a large sharp peak at 505.2 nm accompanied by a broad phonon sideband and a weak peak at 576.1 nm overlaying the phonon sideband. Although the annealing temperature caused the peak wavelengths to fluctuate slightly, they can correspond to zero phonon lines (ZPLs) from H3 centers at 505.8 nm and neutral NV (NV⁰) centers at 576.7 nm.^1,27 The H3 center is known as an emission center formed by a vacancy together with two nitrogen atoms (NVN). As discussed later, the fluctuation of the emission peak wavelength was not observed in a diamond nanowire released from the substrate. This indicates that the wavelength fluctuation might come from non-uniform strain in the top diamond layer supported by the a-C layer, which results in mechanical bending of air-suspended nanowires, as shown in Fig. 4(a). At T_A = 1000 °C, the CL intensity of the NV⁰ peak is one order of magnitude larger than that of H3. However, as the annealing temperature increases, the CL intensity of the H3 peak becomes larger than that of the NV⁰ peak, and then, it is hard to find the weak NV⁰ peak due to the intense phonon sideband of H3. At T_A = 1350 °C, the entire CL intensity clearly decreases. In contrast, Figs. 2(b) and 2(d) show that the CL peak of NV⁰ in the ion-implanted diamond is dominant at all annealing temperatures. At T_A = 1350 °C, the H3 peak grows largely and an additional peak appears at 416.6 nm, which indicates the N3 center originating from multiple nitrogen atoms.²⁷ 

To summarize the behavior of H3 and NV⁰ centers, the CL peak intensities of ZPLs from H3 and NV⁰ centers as a function of the annealing temperature are shown in Fig. 3. Here, the peak intensity excludes the broad background from phonon sidebands. In the unimplanted diamond shown in Fig. 3(a), the CL intensity of H3 peaks at T_A = 1150 °C, while the intensity of NV⁰ decreases monotonically as the annealing temperature increases. At T_A > 1150 °C, the CL intensity of H3 is larger than that of NV⁰. However, the ion-implanted region exhibits different behavior, as shown in Fig. 3(b). The CL intensity of NV⁰ is always larger than that of H3 and gradually increases as the annealing temperature increases, while the intensity of NV⁰ seems to saturate at T_A > 1150 °C. Here, note that the CL intensity of H3 increases largely at T_A > 1250 °C, and finally, the intensities of H3 and NV⁰ become comparable at T_A = 1350 °C. In nitrogen-implanted type-IIa substrates, the NV center emission is saturated at T_A > 950 °C,²⁸ which might cause the consumption of most mobile vacancies. Furthermore, a similar crossover of the emission intensities between NV and H3 centers has been observed at T_A ∼ 1400 °C.²⁹ Figure 3(c) shows the CL intensity ratio for H3 and NV⁰ centers. In the unimplanted region after annealing at T_A > 1150 °C, the H3 center becomes dominant. In contrast, the CL from the ion-implanted region is always dominated by the NV⁰ center although the H3 center is prominent at the high annealing temperature. If the emission quantum efficiency is stable within this annealing temperature range, the CL intensity is proportional to the concentration of the emission centers. Figure 3 shows that the optimal annealing temperature to obtain the largest emission from NV⁰ centers is 1150 °C, which is expected to provide the highest NV concentration in the ion-implanted diamond layer while suppressing H3 center creation. This high optimum annealing temperature is preferable for the ion-implanted diamond because high temperature annealing can reduce the number of various vacancies working as nonradiative centers quenching the NV emission.³⁰ 

The annealing temperature dependence of the CL intensities results from the reaction among a nitrogen atom, vacancy, NV, and H3 center. First, the reaction in the unimplanted diamond is discussed. Since it is suggested that a vacancy can be mobile at T_A > 600 °C due to the relatively low diffusion energy of 2.3 eV,¹ a mobile vacancy V^* meets a substitutional nitrogen atom N to forms a NV center even at low temperature,³¹ which is expressed by

However, in the case of no ion irradiation, the NV creation can be limited because the vacancy concentration [V] is insufficient to maintain the creation of NV centers, while the dissociation of NV centers, which is the reverse process of Eq. (1), is possible by thermal activation.^28,29,32 In contrast, the nitrogen concentration [N] in the type-Ib diamond substrate is high enough to create A-aggregates (N–N) and H3 centers (N–V–N).³³ At high temperature, a NV center can be mobile by the diffusion of a substitutional nitrogen atom assisted by mobile vacancies, which might be accompanied by the temporal dissociation of the NV center. If the mobile NV center (N–V)^* meets another nitrogen atom, an H3 center is created through the A-aggregate formation under a high [N] and [N] > [V],^31,34 which is expressed by

The total free energy of the H3 center is lower than that of the NV center.³¹ This H3 creation results in a decrease in the CL intensity for NV⁰ and its increase for H3 at annealing temperatures ranging from 1000 °C to 1150 °C, as shown in Fig. 3(a). However, at higher annealing temperature, an H3 center can be dissociated to an A-aggregate (N–N) and vacancy, which is a reverse process of the last reaction in Eq. (2). This dissociation is experimentally observed as the CL intensity reduction of H3 at T_A > 1250 °C. It is expected that some of the free vacancies from the dissociated H3 centers are reused to create new NV and H3 centers;³³ however, the vacancy concentration could be insufficient to recover their concentrations due to anneal out of the vacancies,³⁰ and then, the A-aggregate accumulates.³⁴ In the ion-implanted region, numerous vacancies are introduced into type-Ib diamond by the high-dose ion implantation. Figures 1(b) and 1(e) show that the vacancy concentration [V] ∼ 10²² cm⁻³ is three orders of magnitude larger than the nitrogen concentration [N] ∼ 10¹⁹ cm⁻³, which satisfies [N] ≪ [V]. In this case, the creation of an NV center in Eq. (1) proceeds well by the existing rich nitrogen and ion-created numerous vacancies during annealing, which is consistent with the CL intensity increase for NV⁰, as shown in Fig. 3(b). At high temperature, some of the NV center can be dissociated and transformed to an H3 center through the reaction in Eq. (2). This reaction is observed as a rapid increase in the CL intensity of H3 at T_A > 1250 °C, and then, the consumption of the NV center for the H3 creation can cause the saturation of the CL intensity of NV⁰.

Next, we attempted to fabricate diamond nanowires in a thin membrane. Diamond nanowires have often been fabricated by vertical deep RIE of a diamond substrate containing NV centers.^20,35 These standing nanowires work as efficient photon emitters because they have a higher extraction efficiency of the NV emission to the vertical direction. However, their sidewall fluctuates because of a long time exposure to reactive oxygen plasma. In this study, we used EB lithography and oxygen RIE for lateral patterning of fine diamond nanowires in a 150-nm-thick diamond membrane annealed at the optimized T_A = 1150 °C. Since the length and width of the nanowires are defined by EB lithography and short RIE of less than 5 min, fine nanowires can be achieved with a high aspect ratio, which means a long nanowire with a small cross section. Figure 4(a) shows a diamond comb fabricated with a patterned metal mask. As shown in Figs. 4(b) and 4(c), the comb is suspended in air after the removal of the sacrificial a-C layer by a H₂SO₄–HNO₃ mixture boiling at 400 °C. The side etching depth was ∼500 nm to fix the comb by its large base. To fabricate a larger air-suspended membrane, the side etching depth can reach over several micrometers by increasing the temperature of the boiling acid. The nanowires were snapped and released from the comb by scratching. A free nanowire was picked up with a tungsten-tip probe attached to a three-axis piezo-manipulator and transferred to a platinum-deposited Si (Pt/Si) substrate that has high thermal conduction and no luminescence in our wavelength range of interest. Figure 4(d) shows a nanowire placed on the Pt/Si substrate. The length and width of this nanowire are 10 µm and 770 nm, respectively. Although one end of the nanowire is not smooth because of the mechanical breaking, this process successfully fabricates finely shaped nanowires with a high aspect ratio, which is difficult with previous methods using vertical deep etching.²⁰ In Fig. 4(d), a large nanowire was selected for easy manipulation. However, narrower nanowires as small as 100 nm can be fabricated by reducing the acceleration voltage of He ions and our fine EB lithography.

We measured PL from the transferred nanowires by low-temperature micro-PL spectroscopy using a confocal setup with a low numerical-aperture (NA = 0.4) objective. The wavelength of the continuous-wave (cw) laser used for the time-integrated PL measurement was 532 nm. In contrast to CL spectroscopy, PL spectroscopy with the green laser detects negatively charged NV (NV⁻) centers by resonant excitation in addition to neutral centers (NV⁰).³⁶ Since the Pt/Si substrate has no PL, only the PL from a nanowire can be detected. The PL spectra of the nanowire in Fig. 4(d) are shown in Fig. 5(a). Two clear PL peaks are found at 575.8 nm and 638.5 nm, which are in good agreement with those from the NV⁰ and NV⁻ centers in diamond.²⁷ Furthermore, no change in the peak wavelength is observed as the excited position changes. As shown in Figs. 5(c) and 5(d), the PLs of NV⁰ and NV⁻ centers are almost uniformly distributed in the nanowire, while the end part of the nanowire exhibits larger PL of NV⁻. For a comparison with a diamond substrate often used in diamond photonic devices, Fig. 6(a) shows PL spectra of the nanowire and a nitrogen-doped type-IIa diamond substrate grown by chemical vapor deposition (CVD), where nitrogen ions were implanted with an acceleration voltage of 100 kV and a dose of 1.0 × 10¹² cm⁻², and annealing was performed at 950 °C for 3 h in argon.²⁸ In the nanowire, the PL peak of NV centers is broadened by 5.0 nm (3.7 THz in frequency), which has been observed in the ion-implanted type-Ib diamond substrates possibly due to residual damage and local strain.³⁷ This linewidth is at least 50 times larger than the linewidth of <0.1 nm (<70 GHz, spectrometer resolution limit) in the nitrogen-doped CVD diamond, but the broad peaks and phonon sidebands are preferable for wider gain bandwidth and tunability of a color center laser. The PL peak intensity ratios between NV⁻ and NV⁰ (NV⁻/NV⁰) are 0.34 for the nanowire and 0.42 for the CVD diamond substrate, which indicates no large difference in their charge states. To examine the PL dynamics, we measured the PL decays of NV⁻ in the nanowire and the nitrogen-doped CVD diamond substrate. Here, the samples were excited by pulsed laser light at a wavelength of 518 nm. The pulse width and repetition frequency were ∼0.3 ns and 10 MHz, respectively. As shown in Fig. 6(b), although the PL does not exhibit single exponential decay, the PL lifetime, defined as when the normalized PL intensity becomes e⁻¹, is around 1.3 ns for the nanowire, which is 1/3 of the lifetime of 3.9 ns for the CVD diamond substrate. This PL decay acceleration indicates that nonradiative recombination is enhanced in the nanowires. Even though the optimum annealing was performed, the crystal damage caused by ion implantation still remains and works as severe nonradiative recombination centers. Here, the radiative recombination (spontaneous emission) lifetime of NV⁻ is assumed to be invariant. No plasmonic enhancement in the PL decay was observed on the Pt/Si substrate because the PL decay was identical even in a diamond membrane placed on a Si substrate.

Finally, the concentration of the NV center in the nanowire was approximately estimated by comparing the PL intensities and PL decay rates of the NV center in the nanowire and the CVD diamond substrate. The PL intensity is expressed by I = βηN, where β, η, and N are the emission collection efficiency, quantum yield of the PL from the NV center, and areal concentration of the NV center, respectively. Here, β is assumed to be constant because there is no photonic structure to largely modify the emission extraction. η is defined by η = Γ_R/(Γ_R + Γ_NR) = Γ_R/Γ_PL, where Γ_R and Γ_NR are the radiative and nonradiative decay rates of the NV center, respectively. Γ_PL is the PL decay rate expressed by Γ_R + Γ_NR. Γ_R should be constant in both the nanowire and the CVD diamond substrate when there is no plasmonic enhancement. However, the difference in Γ_NR appears as the change in Γ_PL, as shown in Fig. 6(b). Based on these relations for the nanowire and CVD substrate, the concentration ratio of the NV center can be estimated from the ratios of the PL intensity and its decay with the relation N_NW/N_CVD = (I_NW/I_CVD)(Γ_NW/Γ_CVD). Γ_NW/Γ_CVD is estimated to be 3 from the result shown in Fig. 6(b), and I_NW/I_CVD is calculated to be ∼3 from the PL spectra integrated in the available detection range from 560 nm to 800 nm as shown in Fig. 6(a), where the NV⁰, NV⁻, and their phonon sidebands are taken account. Finally, N_NW/N_CVD ∼ 9 is derived, which means that the NV concentration is almost one order of magnitude higher than that in the CVD diamond substrate. Since the typical concentration of our nitrogen-doped CVD diamond substrate is N_CVD ∼ 1 × 10¹¹ cm⁻², which is calculated from a NV creation yield of ∼10%³⁸ and an approximate comparison with the emission intensity of a single NV center, N_NW is on the order of 10¹² cm⁻². Assuming a uniform depth distribution of the NV centers in the nanowire, the approximate volume concentration and the creation yield of the NV centers are calculated to be 7 × 10¹⁶ cm⁻³ (0.4 ppm) and 0.5%, respectively, from the native nitrogen concentration measured by SIMS.

In summary, we fabricated single-crystal diamond thin membranes by He ion implantation, which enabled us to obtain a high doping concentration of NV centers. Systematic CL spectroscopy revealed the kinetics of nitrogen atoms and vacancies during high temperature annealing, and we found the annealing temperature at which the largest luminescence intensity of NV centers can be obtained while suppressing the transformation from NV to H3 centers. Air-suspended fine diamond nanowires with a high aspect ratio were also successfully fabricated in a thin diamond membrane and transferred to a Pt/Si substrate. They exhibit intense PL from NV centers with the concentration as high as 7 × 10¹⁶ cm⁻³. These light-emitting diamond nanowires can be used to form laser nanocavities or to couple with photonic crystal slabs.^39,40 Although reducing of the nonradiative process is a remaining issue for higher emission efficiency, the diamond nanowires and planar nanostructures containing a dense ensemble of NV centers can be key materials for achieving diamond color center lasers and magnetic-field sensing devices on various photonic platforms.

We thank Hiroo Omi, Makoto Takamura, Masanobu Hiroki, and Kazuhide Kumakura for assistance with the thermal annealing and Takehiko Tawara and Yoshitaka Taniyasu for support with the oxygen reactive ion etching and cathodoluminescence machines, respectively. This work is part of a joint research project supported by the National Institute of Information and Communications Technology (NICT).