Strongly enhanced photon collection from diamond defect centres under micro-fabricated integrated solid immersion lenses
J. P. Hadden, J. P. Harrison, A. C. Stanley-Clarke, L. Marseglia, Y-L. D. Ho, B. R. Patton, J. L. O'Brien, J. G. Rarity
aa r X i v : . [ qu a n t - ph ] J un Strongly enhanced photon collection from diamond defect centres undermicro-fabricated integrated solid immersion lenses
J. P. Hadden, ∗ J. P. Harrison, A. C. Stanley-Clarke, L. Marseglia,Y-L. D. Ho, B. R. Patton, J. L. O’Brien, and J. G. Rarity
Centre for Quantum Photonics, H. H. Wills Physics Laboratory & Department of Electrical and Electronic Engineering,University of Bristol, Merchant Venturers Building, Woodland Road, Bristol, BS8 1UB, UK (Dated: October 28, 2018)The efficiency of collecting photons from optically active defect centres in bulk diamond is greatlyreduced by refraction and reflection at the diamond-air interface. We report on the fabrication andmeasurement of a geometrical solution to the problem; integrated solid immersion lenses (SILs)etched directly into the surface of diamond. An increase of a factor of 10 was observed in the satu-rated count-rate from a single negatively charged nitrogen-vacancy (NV − ) within a 5 µ m diameterSIL compared with NV − s under a planar surface in the same crystal. A factor of 3 reduction inbackground emission was also observed due to the reduced excitation volume with a SIL present.Such a system is potentially scalable and easily adaptable to other defect centres in bulk diamond. The ability to address single defect centres in dia-mond using confocal microscopy allows optical accessto these single ‘atom like’ systems trapped within amacro-scale solid. The negatively charged nitrogen-vacancy centre (NV − ) is of particular interest forapplications such as single photon generation[1, 2],nanoscale magnetometery[3], and fundamental investi-gations of spin interactions and entanglement at roomtemperature[4–6]. Other defect centres that exhibitsingle photon emission have also been identified (e.g.the nickel-related ‘NE8’[7], the silicon-vacancy[8], andchromium related centres[9]), but the search continues forother defect centres with spin properties like those of theNV − centre[10]. The high refractive index of diamondcauses refraction of the emitted light at the diamond-airinterface, reducing the possible angular collection of amicroscope objective. Thus the NV − photon collectionefficiency is severely reduced. This is a problem regard-less of the application, or of the particular defect centre ofinterest. Here we report on the fabrication and measure-ment of hemispherical integrated solid immersion lenses(SILs) etched directly into the surface of diamond. Thesestructures eliminate surface refraction, thus increasingthe numerical aperture (NA) of the microscope system.This allows a substantial increase in the resolution andbackground rejection of our system, along with a strongenhancement in NV − photon collection efficiency. More-over, this geometrical solution can easily be applied toother defect centres in bulk diamond which emit at dif-ferent wavelengths[11].The photon collection efficiency from NV − centres indiamond has previously been improved by using NV − centres located within nanocrystals small enough that thecentres effectively emit into free space[2, 12], or nanopho-tonic structures such as nanowires which guide emissiontowards collection[13]. Photon collection is increased bya factor of up to about 5 in the former case and 10 inthe latter. However with the NV − centres positioned soclose to the surface, local strain, impurities and other surface effects have been shown to degrade the stability,and spin coherence time of the NV − centre[14, 15], so asolution which improves the photon collection efficiencyfrom NV − centres in bulk diamond is desirable. By etch-ing hemispherical SILs into the surface of the diamond wecan increase photon collection efficiency without requir-ing the centre to be close to the surface. Rays traced froma defect centre located at the origin of a hemispherical di-amond surface are normal to the surface at all points onthe hemisphere. Therefore no refraction occurs and thetheoretical NA of collection is increased by a value equalto the refractive index ( n d = 2.42) of diamond[16]. Thisallows a significant increase in the collection efficiency.Other advantages related to the increased NA of a SILare an imaging magnification (by a factor of n d later-ally) and a reduced (by n d longitudinally) laser excita-tion volume[16]. The latter both increases the resolu-tion and boosts the effective power density for a giveninput laser power. SILs (also known as numerical aper-ture increasing lenses) have previously been used to boostphoton collection from quantum dots[11, 17] and singlemolecules in anthracene crystals[18]. Both implementa-tions used millimetre scale SILs placed on the surface ofthe sample. These are less scalable and also incur lossesand aberrations caused by SIL-sample mismatch, surfacereflections and gaps between the surfaces. The use of amicro-scale integrated SIL overcomes these problems. In-tegrated micro-lenses have previously been fabricated indiamond using inductively-coupled plasma etching (andother etching techniques), however they have not beenused to improve collection of defect centre emission[19].Using a finite-difference time-domain (FDTD) method,we simulated the collection efficiency (into a microscopeobjective with 0.9 NA) for a dipole located: 2.5 µ m belowa planar diamond surface; at the focal point of a 2.5 µ mradius hemisphere; and at the focal point of a 2.5 µ m ra-dius hemisphere surrounded by a 2 µ m wide trench. Thelatter case is included because to etch the entire surfaceof the sample to the base of the SIL, as in the ideal case, FIG. 1: Representation of electromagnetic field intensity inthe YZ plane for a single frequency calculated from FDTDsimulations for: (a) dipole 2.5 µ m beneath a planar diamondsurface, (b) dipole at origin of 2.5 µ m radius hemisphere, (c)dipole at origin of 2.5 µ m radius hemisphere with 2 µ m widetrench. would be too time consuming. The size of the trenchis chosen such that the entire numerical aperture of thecollection lens is utilised. It can be clearly seen froma cross-section of the simulated electromagnetic field in-tensity shown in Fig. 1 that the SIL increases the fieldintensity crossing the diamond-air interface.The collection efficiencies were calculated for wave-lengths in the range of 600-800 nm (covering the major-ity of the NV − emission spectrum) and then an averagewas taken. The collection efficiencies calculated in thisway were 5.6%, 29.8%, and 28.6% respectively. The firsttwo cases are consistent with efficiencies calculated bypurely analytic methods[20]. In other words, we expect a ∼ µ m alongany of the three Cartesian axes the collection efficiencyshould remain above 20% (c.f. 28.6% at the focus). Inother words the SILs are relatively tolerant to lateral orlongitudinal placement error. It should also be notedthat the magnification effect of the SILs means that a 1 µ m lateral displacement in the diamond corresponds toa measured ‘image’ distance (in a confocal scan) of 2.42 µ m, placing it at the edge of the lens in a confocal image.SILs were fabricated in a polycrystalline type IIa CVD FIG. 2: (a) FIB image showing 7 SILs and location grid lines.(b) Confocal image of the same area with SILs labelled. Thebright line is a diamond crystal grain boundary. diamond sample (Element Six) measuring ∼ × − centres in the sample, and a re-gion was chosen where the density was low enough toresolve single NV − s. A 30 keV gallium focused ion beamsystem (FEI Strata FIB-201) was used to fabricate theSILs. Hemispheres were approximated by etching con-centric rings of increasing depth and diameter. This wasachieved by varying the beam current and dwell time andwater was used throughout to assist the etching process.A FIB image showing 7 of the 14 SILs fabricated alongwith FIB etched grid lines is shown in Fig. 2(a), and aconfocal image of the same area with the SILs labelled isshown in Fig. 2(b).Optical characterisation was performed using a laser FIG. 3: (a) Confocal image of SIL K. The boundary of the SIL is marked in white. Note that the colour scale has been chosento maximise the visibility of the SIL and surroundings - the intensity of the bright spot is about 120 kcps. (b) Optical spectrumrecorded at the point of highest intensity in (a). The sharp drops in spectral intensity at 630 nm and 700 nm are caused bythe spectral filter and not by chromatism in the SIL. (c) Second order intensity correlation function recorded at the same pointas in (b). (d) Comparison of the total count rate for the single NV − under SIL K with that for a single NV − under a planarsurface as a function of excitation power. The background in each case is measured from a point close to the NV − centre. scanning confocal microscope system built in-house. Theoutput from a frequency-doubled Nd:YAG laser (CoboltSamba) was focused onto the sample using a 0.9 NA mi-croscope objective (Nikon). Laser-induced fluorescencewas collected by the same objective and focused ontothe 8.9 µ m core of an optical fibre (serving as the con-focal aperture). A 532 nm long-pass filter was used toreject stray laser light. A band-pass filter centred at 675nm with a 67 nm transmission band was used to blockfirst and second order Raman scatter. It was tilted toshift the transmission band to 630-700 nm so that theNV − zero-phonon emission at 637 nm was transmitted.The fluorescence photons were then directed either to aspectrograph (Shamrock 303) fitted with a CCD camera(Newton 920) to record optical spectra, or to a Hanbury-Brown and Twiss detection scheme to record the secondorder intensity correlation function ( g (2) ( τ )).Of the 14 SILs etched, 2 fell on grain boundaries andwere not investigated further. Of the remaining 12, 5contained single NV − centres. One of these 5 actuallyhad two centres that could be resolved and analysed sep-arately. Only data from the ‘best’ of the single NV − containing SILs (denoted ‘SIL K’) is presented in fullhere. The enhancement of the other SILs is commentedon briefly.A confocal scan of a 20 × µ m area of the samplein the region of SIL K is shown in Fig. 3(a). A highintensity region is evident near the origin of the SIL. Thisregion is 1 µ m away from the SIL’s origin in the confocalscan, corresponding to 0.41 µ m in ‘real’ space. The brightfeature at the top is a grain boundary.The optical spectrum recorded at this point (Fig.3(b)) displays the characteristic NV − emission profile,with zero-phonon line around 637 nm and broad phononassisted side-band at longer wavelengths. The anti-bunching dip in the second order intensity correlationfunction (Fig. 3(c)) clearly indicates that the emissionarises from a single centre. The data has been normalisedand corrected for background as described by Beveratos et al. [2].To quantify any enhancement in collection efficiency,photon count-rate as a function of laser power wasrecorded for the NV − in SIL K and for other single NV − centres in the same diamond grain under an unmodi-fied planar surface. A comparison of the saturated in-tensities (Fig. 3(d)) indicates an enhancement of ∼ − spectrum is be-ing collected. In principle then, optimisation of the filterswould result in an absolute count rate of 500 kcps.The 3 other SILs containing a single NV − centre hadenhancement factors of: 10, 8, and 8. For the SIL withtwo separate single NV − s, the enhancement factors were6 and 3.6.As well as improving the angular collection efficiency,the SILs also modify the profile of the excitation beam.Elimination of refraction in the excitation beam meansthat it can be focused to a smaller volume, thus increasedresolution is to be expected. Assuming that the excita-tion point spread function is Gaussian in shape, we canestimate the excitation spot size from the full width halfmaximum (FWHM) of a line scan across an NV − cen-tre. Figure 4 shows a comparison between the line scanacross a planar NV − centre and across the one underSIL K. The FWHM is 360 ±
30 nm in the planar caseand 289 ± ± λ ex / NA = 90 nm for oursystem with excitation wavelength λ ex = 532 nm and anincreased NA of 2.18[16]. The reduced performance islikely to arise from aberrations within our system causedby the NV − centre not being precisely at the origin ofthe SIL.Improved resolution is important because a smaller ex-citation volume decreases the proportion of unwantedbackground fluorescence collected along with the desiredNV − fluorescence. This is clearly visible in (Fig. 3(d))where the signal to background ratio is much higher withthe SIL, and in fact the absolute value of backgroundis 3 times lower at equivalent pump powers. A smallerexcitation volume also means that the power density isincreased (for a given laser power) and less power is re-quired to saturate an NV − centre. SIL K obviously con-tains a very well positioned NV − centre and we do seevariation in these enhancements from SIL to SIL. We as-cribe this not only to NV − position, but also to variationof dipole orientation, since the crystal orientation is un-known in this polycrystalline sample. We are currentlystudying both of these effects.This demonstration of strongly enhanced photon col-lection efficiency from NV − centres using integrated solidimmersion lenses is a step towards efficient single photonsources as well as efficient optical spin read-out in com-pact micro-structured devices. The enhancement com- -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2-0.20.00.20.40.60.81.0 Sil K Line-scan Planar Line-scan SIL K Gaussian Fit SIL K Inferred Gaussian Fit Planar Gaussian FitSIL K Image FWHM - 289–2nmSIL K Inferred FWHM - 119–1nmPlanar FWHM - 360–30nm P l ana r N V C oun t R a t e ( c p s ) X position ( m)
FIG. 4: Comparison of confocal line scan of the single NV − under SIL K and from a single NV − under a planar surfaceboth fitted with a normalised Gaussian function. The inferred‘real’ fit of SIL K taking into account the magnification of theSIL is also plotted. The full width half maximums (FWHM)of the two fits, and the inferred ‘real’ FWHM of SIL K’s areshown with their associated uncertainties. pares favourably to those reported from nanocrystal andnanowire devices[12, 13], with the advantage that NV − centres are not located close to the surface where theymight suffer from surface-induced decoherence. Sincethe enhancement is wavelength independent when thedipole is at the origin of the SIL it may be used withother defect centres in bulk diamond. Such a solutionis robust and potentially scalable, since SILs could befabricated over existing defect centres after characterisa-tion. Refinements in surface symmetry, SIL placementand smoothness should allow us to achieve even greaterenhancement as our fabrication technique is perfected.This work was supported by EPSRC, European UnionSixth Framework Program project EQUIND IST-034368,NanoSci-ERA project NEDQIT and the LeverhulmeTrust. JGR and JLOB are both supported by individ-ual European research council fellowships and by RoyalSociety Wolfson research merit awards. ∗ [email protected] [1] C. Kurtsiefer, S. Mayer, P. Zarda, and H. Weinfurter,Phys. rev. lett. , 290 (2000).[2] A. Beveratos et al., Eur. Phys. J. D , 191 (2002).[3] G. Balasubramanian et al., Nature , 648 (2008).[4] F. Jelezko, T. Gaebel, I. Popa, M. Domhan, A. Gruber,and J. Wrachtrup, Phys. Rev. Lett. , 1 (2004).[5] R. Hanson, F. Mendoza, R. Epstein, and D. Awschalom,Phys. Rev. Lett. , 1 (2006).[6] P. Neumann et al., Science , 1326 (2008).[7] T. Gaebel, I. Popa, A. Gruber, M. Domhan, F. Jelezko,and J. Wrachtrup, New J. Phys. , 98 (2004).[8] C. Wang, C. Kurtsiefer, H. Weinfurter, and B. Burchard,Jour. Phys. B , 37 (2006). [9] I. Aharonovich, S. Castelletto, D. A. Simpson, A. D.Greentree, and S. Prawer, Phys. Rev. A (2010).[10] J. R. Weber et al., Proc. Natl. Acad. Sci. USA. , 8513(2010).[11] Z. Liu et al., Appl. Phys. Lett. , 071905 (2005).[12] E. Ampem-Lassen, D. A. Simpson, B. C. Gibson, S. Trp-kovski, F. M. Hossain, S. T. Huntington, K. Ganesan,L. C. Hollenberg, and S. Prawer, Optics Express ,11287 (2009).[13] T. M. Babinec et al., Nat. Nano. , 195 (2010).[14] C. Bradac et al., Nat. Nano. , 345 (2010). [15] J. R. Rabeau et al., Nano lett. , 3433 (2007).[16] T. R. Corle and G. S. Kino, Confocal Scanning Opti-cal Microscopy and Related Imaging Systems (AcademicPress,, San Diego, 1996).[17] V. Zwiller et al., Appl. Phys. Lett. , 2476 (2001).[18] J.-B. Trebbia, H. Ruf, P. Tamarat, and B. Lounis, Opt.Express , 23986 (2009).[19] C. Lee, E. Gu, M. Dawson, I. Friel, and G. Scarsbrook,Diam. and Relat. Mater. , 1292 (2008).[20] W. Barnes et al., Eur. Phys. J. D18