1 Introduction

A number of color centers (light-emitting defects) in diamond, such as nitrogen-vacancy (NV) [1, 2], silicon-vacancy (SiV) [3–5], nickel (NE8) [6] and chromium-related optical defects [7] are the subject of intensive research due to the prospects of using them as the single photon emitters in quantum information technologies [8]. While NV center is the most studied of those luminescent defects, other centers may have particular optical features more appropriate for these applications. For instance, SiV centers demonstrate much narrower emission spectrum (zero-phonon line—ZPL) than that of the NV centers and much shorter lifetime [3, 9]. The increase in PL collection efficiency is important issue that can be achieved, among other approaches, by surface micro- or nanostructuring of local emitters in the form of diamond pillars [8, 10]. The concentration of light within the pillar due to the total internal reflection results in strongly directional PL emission to be collected with appropriate microscope optics.

Typically, a top–bottom approach is used to fabricate the photoemitters in two steps by ion implantation, followed by reactive ion etching with an appropriate mask [10, 11]. The fabrication process would be greatly simplified if one could avoid the implantation and etching processes to directly grow pillars using in situ Si (or N) doping. Recently, Singh et al. [12] have produced a patterned nanodiamond array with fluorescent SiV centers using pre-growth local seeding of nanodiamonds on SiO2 substrate by scanning probe ‘dip-pen’ lithography followed by microwave plasma chemical vapor (MPCVD) diamond deposition. The obtained submicron CVD diamond particles demonstrated SiV photoluminescence (PL) as a result of the substrate etching by atomic hydrogen and Si incorporation in the growing diamond from the formed SiH4 precursor. However, the diamond particles have uncontrolled shape; they are principally polycrystalline, so the SiV PL yield is expected to be less effective than that of single crystal (SC) diamond photoemitters. Here, we demonstrate a new bottom-up approach to grow epitaxially the regular SC as well as microcrystalline diamond microstubs with imbedded SiV centers by MPCVD through a-Si mask with holes controlling the stub shape.

2 Experimental

The flowchart for the diamond stub fabrication is depicted in Fig. 1. First, a commercial SC diamond substrate [the plate of synthetic Ib type diamond produced by high-pressure–high-temperature (HPHT) technique] with 4 × 4 × 0.6 mm dimension and (100) orientation polished to the roughness of R a ~ 1 nm was coated with 600-nm-thick amorphous Si (a-Si) layer, by microwave plasma CVD deposition using SiH4–H2 gas mixture under the following conditions: 300 °C substrate temperature, 20/480 sccm SiH4/H2 flow rate, 40 Torr pressure, 3 h deposition time.

Fig. 1
figure 1

Fabrication schematic of diamond microemitters with SiV fluorescent centers

Full size image

Then, an array of through holes was drilled in the a-Si mask by ablation with second harmonic of Ti: sapphire femtosecond laser (Spectra-Physics, τ = 110 fs, λ = 400 nm). The laser beam was focused with a quartz lens to a spot diameter of ~1 µm. Local laser fluence on the surface was about 0.3 J/cm2. The 300 pulses in one spot were required to form one through hole in the a-Si mask to make possible a further diamond epitaxy on the opened substrate. Periodic hole arrays (10 µm distance between the holes) have been produced by consequent laser irradiation.

As a next step, diamond deposition was performed in a microwave plasma CVD system ARDIS-100 (Optosystems Ltd.) operated at 2.45 GHz frequency in CH4–H2 gas mixture under the following conditions: 120 Torr gas pressure, 20 sccm CH4 flow rate, 480 sccm H2 flow rate, 750 °C substrate temperature, 2.37 kW microwave power, 25 min deposition time. The diamond growth started epitaxially from the HPHT diamond substrate being confined by the hole sidewalls, and thus, no preliminary seeding process was required. After the growth, the Si mask was etched in HF+HNO3+CH3COOH acid mixture to leave the array of diamond stubs.

Raman and PL spectra for the individual stubs were studied using LabRAM HR800 spectrometer (HORIBA-Jobin Yvon) in confocal mode with excitation at 488 nm wavelength (Ar+ laser) in a spot of 1.5 μm diameter at incident laser power of 35 mW. Surface morphology was studied by scanning electron microscope JSM 6510LV (JEOL).

Surface distribution of SiV centers and the PL lifetime were investigated with a homemade luminescence microscope in a confocal configuration connected to Hanbury–Brown–Twiss interferometer (Fig. 2). A CW diode laser (Coherent Inc., 532 nm, 300 mW) coupled to an optical fiber was used for PL excitation. The laser beam was directed by dichroic mirror through an oil immersion objective (Nikon Apo TIRF 100x/1.49) and focused on the sample surface. The luminescence radiation was collected in back-side geometry. Red light and narrow band (730–750 nm)-passing filters were used for cutting off the reflected laser radiation and transmitting PL emission. The PL emission was collected with the avalanche photodiode (PerkinElmer SPCM-AQRH-14-F) via a single mode fiber. A piezoelectric translation stage was applied for PL mapping of the sample surface. The estimated spatial resolution of PL probe is ca. 300 nm. For PL lifetime measuring, the luminescence microscope was modified just by replacing the pump CW laser with the pulsed diode laser (PICO Quant, 532 nm, 60 ps, 200 mW, 80 MHz).

Fig. 2
figure 2

Optical scheme for PL mapping with fluorescence microscope

Full size image

3 Results and discussion

SEM image of the periodic hole massive in the Si mask is shown in Fig. 3a. The holes are not perfectly round because of nonuniform intensity distribution in the laser beam. The diamond stubs revealed after the mask removal are seen in Fig. 3b. The height of the cylindrical part of stubs is determined by the mask thickness. The top surface of the stubs is composed mostly of (100) oriented crystallites. The stubs shape resembles a milk mushroom on 600 nm stipe with a dip of ≈1 μm (Fig. 3b).

Fig. 3
figure 3

SEM images of the laser-drilled holes in Si mask (a) and diamond stubs grown through the holes (b)

Full size image

Since the diamond surface below the Si mask has been also subjected to a certain laser ablation, a cavity was formed on the substrate with the maximum depth in the hole center. The diamond nucleates and grows from the inclined sidewalls of the cavity as well from its bottom, forming a dip in the growing stub. The stub’s height is 2.2 μm as measured by optical profilometer (NewView 5000, Zygo Corp.), that is significantly larger than the mask thickness. The stub diameter, ≈5 µm, exceeds the hole diameter as the lateral diamond growth starts from the moment when the stub height becomes larger than the mask thickness. The growth would result in shorter, almost vertical stubs, when stopped earlier. Therefore, the microstub’s aspect ratio height/diameter was 0.43. The growing stubs are automatically doped with Si since the silicon mask is subjected to continuous etching by atomic hydrogen from MW plasma [13], and then, the volatile SiHx species are transported to the diamond surface to result in Si doping of the stubs. Thus, the Si mask served simultaneously both as the template and the doping source for diamond.

Figure 4 shows Raman and PL spectra taken from one of the stubs and from undoped substrate area outside the stub. The diamond Raman peak at 1,332.8 cm−1 with full-width at half-maximum (FWHM) of 6.6 cm−1 is seen for the stub.

Fig. 4
figure 4

PL and Raman (inset) spectra for CVD diamond stub (red lines) and HPHT diamond substrate surface between the stubs (blue lines) after the Si mask removal

Full size image

Other features for this spectrum are G broad band at ≈ 1,500 cm−1 of sp2-carbon and a weak peak at 1,152 cm−1 from trans-polyacetylene [14], while the strong narrow diamond peak at 1,332.8 cm−1, without any sign of other carbon phases, characterizes the substrate. The appearance of non-diamond carbon can be related to (1) the presence of grain boundaries; (2) too high local concentration of Si in gas phase within the hole in the mask, resulting in enhanced defect abundance in the growing diamond.

A strong emission from SiV center with zero-phonon line (ZPL) at wavelength λ = 738.6 nm is observed in PL spectra for the Si-doped stubs (Fig. 4). The SiV PL intensity is ≈70 times higher than that for the Raman diamond peak. For an ensemble of SiV centers, the width (FWHM) of SiV ZPL as small as 6 nm is found which is typical for Si-doped CVD diamonds [15].

The fluorescence image of an area with the stub array (Fig. 5a) reveals the strong SiV PL within each stub, the shape and dimensions of the PL-active zone being identical to those for the stubs. Moreover, emission is weaker from the central part of the stubs due to thinner and/or more defective material there, as seen again at the SEM image. The brightest PL comes from (100) facets of diamond grains. The area outside the stubs contributes very low signal, as expected.

Fig. 5
figure 5

Pseudo-colored fluorescence image of the stubs array (a); PL decay kinetics (b)

Full size image

From time-resolved measurements, we determined SiV PL decay time τ of 0.76 ± 0.15 ns (Fig. 5b), that is in agreement with literature data for different diamond materials and doping methods, including ion implantation (τ = 1.2 ns) in type IIa natural diamond films [3] and doping of nanodiamond particles (τ = 1.1 ns) [4] and epitaxial diamond films (τ = 1.3 ns) [15] in course of CVD process. To reduce SiV concentration for producing single photon emitter, a bilayer mask with holes can be used with a very thin Si layer on a thicker, neutral material containing no Si, or, as alternative, the doping from SiH4 added in the plasma can be employed [16].

Finally, using the same approach, the diamond Si-doped microstubs of 1.8 μm in diameter and aspect ratio of 0.4 have been grown on polycrystalline diamond surface (Fig. 6). The potential advantage of this substrate material is its availability in large size. The polycrystalline film was synthesized in the same CVD system [17]. We used unpolished relatively smooth fine-grained nucleation side of a free-standing 8 × 8 mm2 400-μm-thick diamond film covered with 0.7-μm-thick Si mask. The SiV PL intensity for the stubs was 64 times higher than that for the Raman diamond peak in this case; however, the PL contrast (stub/substrate) was low, ≈2, because of a strong background SiV emission from the substrate. The nucleation side of the substrate was strongly contaminated with silicon due to Si substrate etching by atomic hydrogen in microwave plasma in the course of diamond wafer growth. Nevertheless, the produced regular stubs, still with aspect ratio <1, having the uniform diameter clearly demonstrate the potential capability to grow the structures with pillars using the proposed approach and the holes with higher aspect ratio.

Fig. 6
figure 6

SEM image of diamond stubs grown on polycrystalline CVD diamond substrate

Full size image

4 Conclusions

We have demonstrated the bottom-up approach to produce diamond photoemitters in the form of short pillars (stubs) with SiV color centers strongly emitting at 738 nm wavelength. The fabrication process includes the CVD diamond homoepitaxial growth through vertical channels formed in Si mask on a diamond substrate, the mask also serving as the Si source for diamond doping. Arrays of the microsized stubs with bright PL can be easily obtained. Other silicon-containing materials, such as SiO2, seem to be appropriate as the mask in this scheme [18, 19].