Plasmon-Enhanced Single Photon Emission from a Nanoassembled

Mar 20, 2009 - Hot spots between diamond and gold nanoparticles provide an efficient near-field coupling, despite the mismatch in size and shape. Our ...
0 downloads 11 Views 948KB Size
NANO LETTERS

Plasmon-Enhanced Single Photon Emission from a Nanoassembled Metal-Diamond Hybrid Structure at Room Temperature

2009 Vol. 9, No. 4 1694-1698

Stefan Schietinger,* Michael Barth, Thomas Aichele, and Oliver Benson Nano-Optics, Institute of Physics, Humboldt-UniVersita¨t zu Berlin, HausVogteiplatz 5-7, D-10117 Berlin, Germany Received February 5, 2009; Revised Manuscript Received March 12, 2009

ABSTRACT In this Letter we present the controlled coupling of a single nitrogen vacancy center to a plasmonic structure. With the help of an atomic force microscope, a single nanodiamond containing a single nitrogen vacancy center and two gold nanospheres are assembled step-by-step. We show that both the excitation rate and the radiative decay rate of the color center are enhanced by about 1 order of magnitude, while the single photon character of the emission is maintained. Hot spots between diamond and gold nanoparticles provide an efficient near-field coupling, despite the mismatch in size and shape. Our approach provides hybrid systems as important building blocks for novel nanophotonic light sources in advanced plasmonic devices stable even at room temperature.

The controlled coupling of single quantum emitters to metal nanoantennas is one of the most promising yet challenging tasks in current plasmonic research.1-5 To exploit the quantumcharacteroftheemissionincombinedphotonic-plasmonic devices, a stable emitter-antenna system is required. The emission properties of quantum emitters, such as atoms, molecules, or quantum dots, depend not only on their intrinsic structure but also on the local electromagnetic environment.6 Modifications of their emission lifetime, spectral distribution, and spatial emission pattern have been demonstrated near planar interfaces,7 through resonant coupling to optical microcavities,8,9 and, more recently, through coupling to metal nanostructures.1-5,10-12 The excitation of plasmon resonances in these structures, which act as optical nanoantennas, leads to highly localized photon fields and therefore to an enhancement of the excitation as well as of the radiative and nonradiative decay rates of nearby emitters. The delicate interplay between these different processes has been disentangled in elegant single molecule studies, employing scanning near-field probes as controllable nanoantennas.1,2,5 In these experiments the focus is on obtaining control over the distance between emitter and antenna in coupled systems. Here, we use scanning probe manipulation for precise nanoassembly14-16 of stable photonic and plasmonic nanoelements. As a single photon source we employ nitrogen vacancy (NV) color centers in diamond,17-19 which represent one of the most stable solid-state quantum emitter systems, * Corresponding author, [email protected]. 10.1021/nl900384c CCC: $40.75 Published on Web 03/20/2009

 2009 American Chemical Society

operating even at room temperature. For the first time, we demonstrate that individual diamond nanocrystals, containing single nitrogen vacancy color centers, can be integrated as single photon emitters in a plasmonic structure assembled from one or more gold nanoparticles via manipulation with an atomic force microscope.14,16 In our setup the deterministic coupling between a diamond nanocrystal (20-35 nm height) and spherical gold nanoparticles (60 nm diameter) is achieved through controlled manipulation with an atomic force microscope (AFM) on top of a homemade inverted confocal microscope (Figure 1b). This allows us to simultaneously perform optical measurements and nanoscale manipulation (translation and rotation) on the particles. On the excitation side we use a frequency-doubled Nd:YAG laser (λexc ) 532 nm, Coherent) and a frequency-doubled, picosecond-pulsed amplified diode laser (λexc ) 531 nm, 100 ps pulse width, 2.5 MHz repetition rate, PicoQuant) for continuous wave and pulsed excitation, respectively. The polarization direction of the excitation is controlled by a λ/2 waveplate. An oil immersion objective (60×/1.4 NA, Olympus) is used for focusing the excitation light onto the sample and for collection of the fluorescence emission. On the detection side, after passing a 550 nm longpass filter and a 50 µm pinhole the emission light can either be monitored by an EMCCD camera (Andor), spectrally dispersed and imaged by a spectrograph (Princeton Instruments; Andor) or be analyzed in a Hanbury-Brown and Twiss correlator (containing two single photon counting modules,

Figure 1. Optical properties and nanoscale manipulation of diamond nanocrystals and gold nanoparticles. (a) Fluorescence spectrum of a gold nanoparticle (orange) and of a single NV center in a diamond nanocrystal (black). The latter exhibits two zero phonon lines at 575 nm (ZPL1) and 638 nm (ZPL2), attributed to the NV0 and NV- charged states, respectively. The inset shows the corresponding g(2)(τ) autocorrelation function, clearly demonstrating the single photon character of the emission. (b) Experimental setup for simultaneous farfield optical detection and near-field manipulation of individual diamond nanocrystals (NCs) and gold nanoparticles (NPs). Correlation measurements are performed with a Hanbury-Brown and Twiss (HBT) setup. (c) Emission intensity of a single diamond nanocrystal as a function of the polarization of the excitation light for two different nanodiamond orientations (blue and red dots, indicated by corresponding AFM images, same color scale as in (d). The solid lines represent modified cos2(θ) fit functions, which also take saturation effects into account. (d) AFM images of different gold-diamond configurations. The gold sphere is moved around the diamond nanocrystal, which stays fixed in its orientation (indicated by red arrows, marking a characteristic lobe), thus testing the optimal configuration for emission enhancement.

Perkin-Elmer, and a time correlator, PicoQuant). Additional spectral filtering is applied (except for measurements of the fluorescence spectra) using a 630 nm long-pass filter and a 750 nm short-pass filter to filter out the main part of the gold fluorescence and residual infrared light from the excitation light. An AFM (JPK Instruments) in tapping mode is used for imaging the sample, while manipulation of the particles is performed in contact mode. In a preliminary step the sample is prepared by spin-coating an aqueous solution of diamond nanocrystals (Microdiamant) and gold nanoparticles (60 nm, BBI) onto a glass coverslip. The diamond nanocrystals show typical heights of 20-35 nm when imaged with the AFM, while their lateral dimensions can be more than twice as large. Approximately 1% of these nanocrystals contain a single NV center and can be selected by making a confocal scan of the sample. Before starting the assembly, the luminescence properties of all employed particles (gold and diamond) are investigated. The emission from the gold spheres is shown in Figure 1a and is attributed to a radiative decay of plasmons originating from interband transitions in the metal.20 The plasmon resonance of our gold particles lies at 540 nm. In Figure 1a only a tail of the emission is visible due to the filtering at 550 nm. A representative fluorescence spectrum of a NV center in a diamond nanocrystal is also shown in Figure 1a. The broad emission band features two zero phonon lines at 575 and 638 nm, which are associated with the neutral (NV0) Nano Lett., Vol. 9, No. 4, 2009

and negatively charged (NV-) state of the NV center, respectively.19 Autocorrelation measurements (inset in Figure 1a) clearly show the single photon nature of the emission, proving that both zero phonon lines originate from one and the same NV center which undergoes charging/decharging processes during illumination.21 As the transition dipole moment is oriented perpendicular to the symmetry axis of the NV center, a dependence of the fluorescence intensity on the polarization angle of the excitation field is induced.22 This anisotropy can by exploited to determine the orientation of the diamond nanocrystal, as is demonstrated in Figure 1c. The nanodiamond is rotated using the AFM tip and the fluorescence intensity is monitored as a function of the excitation polarization angle for two specific nanodiamond orientations. The observed shift of the intensity minima/maxima clearly corresponds to the rotation angle of the nanocrystal. In a next step, we start our nanoassembly by bringing a nearby gold nanoparticle in direct contact with the diamond nanocrystal. As the exact position of the NV center inside the nanodiamond and therefore the distance to the gold sphere is unknown, the optimal arrangement has to be explored experimentally by moving the gold sphere around the diamond nanocrystal and measuring the optical response. A sequence of three successive steps is shown in Figure 1d. Note that only the gold sphere is moved while the diamond nanocrystal (same as in Figure 1c) stays fixed in its 1695

Figure 2. Experimental realization and numerical simulation of a gold-diamond hybrid structure. (a) AFM images of a single diamond nanocrystal (left), to which one (middle) or two (right) gold nanoparticles are coupled. (b) Corresponding numerical simulations of the intensity enhancement of the excitation light, which is linearly polarized along the x axis. Upper row: schematic representation of the particle configuration. Middle row: x-y cross section. Lower row: x-z cross section. The insets in the middle and lower row indicate the respective directions of propagation and polarization of the impinging light. The field intensity is normalized to the value at the center of the bare diamond nanocrystal and displayed in a logarithmic color scale. (c) Theoretical enhancement of the excitation rate γexc and of the radiative decay rate γrad as a function of the dipole position within the diamond nanocrystal, coupled to one (blue) or two (red) gold spheres: upper part, γexc at 532 nm; lower part, γrad at 680 nm (solid lines) and 600 nm (dashed lines), corresponding to the emission maximum of the NV- and NV0 state, respectively. The dipole is oriented parallel to the x axis and moved along the central line in the nanodiamond as indicated in the inset. All rates are normalized to the corresponding rates of the bare diamond nanocrystal, γexc,0 and γrad,0, respectively.

orientation due to the stronger adhesion to the surface. The final position is a trade-off between the optimal orientation and the optimal distance of the NV center’s dipole to the gold sphere.23 Subsequently, further assembly can be performed by positioning other gold nanoparticles in the same manner. We now investigate in detail the optical properties of two specific configurations as shown in Figure 2a, namely, a diamond nanocrystal with one gold sphere attached (configuration A, middle panel) and the same nanocrystal sandwiched between two gold spheres (configuration B, right panel). For this experiment, a smaller nanodiamond is chosen (see left panel in Figure 2a) to allow shorter distances between opposing gold spheres. Three-dimensional finitedifference time-domain calculations are performed with the FDTD Solutions software package (Lumerical Solutions) to estimate the plasmonic enhancement effects in these configurations (Figure 2b,c). The diamond nanocrystal is modeled as a truncated four-sided pyramid. Realistic material parameters are used, and the influence of the glass substrate as well as the excitation with a high-NA objective is taken into account. The mesh size is set to 1 nm in the vicinity of the nanoparticle system. Simulations of the excitation field predict a strong field enhancement only for polarization along the x axis (Figure 2b). In this case, pronounced hot spots are formed at the contact points of the diamond nanocrystal and the gold spheres due to the steep jump in refractive index. These hot spots give rise to an enhanced electric field inside the nanodiamond, even if the latter is considerably smaller than the gold spheres. Simulations with a classical dipole as emitter predict a strong enhancement of the radiative decay rate γrad only if the dipole is orientated along the x axis. In 1696

Figure 2c the corresponding values are shown as a function of the dipole position inside the diamond nanocrystal. The experimental results are presented in Figure 3. A strong modification of the optical properties of the NV center in the presence of the gold nanoparticles is observed. The emission from the NV0 state is significantly more enhanced than that from the NV- state (see also Figure 2c) as it is closer to the plasmon resonance at 540 nm. Therefore, the emission spectrum is drastically changed at its shortwavelength side (Figure 3a). However, to suppress the fluorescence of the gold and therefore reduce background in the single photon emission, we filter out the light emitted below 630 nm. As predicted by the simulation, a pronounced polarization dependence of the excitation rate is induced (Figure 3b), leading to an intensity contrast of at least a factor of 12 between polarization directions parallel and perpendicular to the x axis. In the latter case, the emission mainly stems from fluorescence of the gold particle(s) (see Figure 1a and Figure 3a). From time-resolved measurements (Figure 3c) we obtain an increase of the excited-state decay rate (1/τdec) by factors of 7.5 and 9.5 for configurations A and B, respectively. To check whether this enhancement is due to radiative or nonradiative processes, power-dependent measurements were performed (Figure 3d). The data were fitted with a saturation model of the form P ) ξσΦPexc/(1 + σΦτradPexc), where σ is the absorption cross section, τrad is the radiative lifetime, Φ is the internal quantum efficiency, and ξ is the total collection efficiency of our setup. Since the simulation results show a negligible influence of the gold spheres on the collection efficiency, it is reasonable to assume that ξ does not change between different measurements. Nano Lett., Vol. 9, No. 4, 2009

Figure 3. Optical characterization of a gold-diamond hybrid structure. All measurements were performed on the three configurations shown in Figure 2a, i.e., bare diamond, diamond with one gold sphere (configuration A), and diamond sandwiched between two gold spheres (configuration B). (a) Fluorescence spectrum of the bare diamond (black, scaled by a factor of 6) and of configuration A under excitation parallel (blue) as well as perpendicular (orange) to the x axis at the same excitation intensity. Corresponding measurements on configuration B showed similar results and are omitted for clarity. The dashed lines indicate the lower and upper edge of the spectral detection window used in all subsequent measurements (see text). (b) Emission intensity of configuration A as a function of the polarization angle of the excitation light, corrected for the background emission from the gold. Corresponding measurements on configuration B showed similar results and are omitted for clarity. (c) Fluorescence time traces of the bare diamond (black), configuration A (blue), and configuration B (red). An additional time trace for configuration B, recorded after applying a laser-melting procedure, is also shown (green). (d) Emission intensity of the bare diamond (black), configuration A (blue), and configuration B (red) as a function of the excitation power, corrected for the background emission from the gold. (e, f) Normalized autocorrelation measurements on bare diamond (e, f, black), configuration A (e, blue), and configuration B (f, red). No background was subtracted here.

Under strong excitation, the maximal number of emitted photons is only restricted by the radiative lifetime and the formula reduces to P ) ξ/τrad. With this, we deduce an increase of the radiative decay rate γrad by a factor of 5.8 and 8.9 for configuration A and B, respectively, which corresponds to quantum efficiencies Φ ) τdec/τrad of 0.78 and 0.93 (for the bare NV center a value of 0.99 is assumed).21 Obviously, in our configuration the first gold sphere induces a noticeable nonradiative decay channel (thus reducing the quantum efficiency), while the second gold sphere predominantly enhances the radiative decay rate (thus partly restoring the original quantum efficiency). This can be attributed to different distances of the gold particles to the NV center and thus a different impact of fluorescence quenching effects.1,2 Note that the enhancement of γrad by nearly an order of magnitude is, to our knowledge, the highest achieved so far for NV centers in diamond and is equivalent to an increase of the maximum single photon rate by the same factor. These enhancement factors are remarkably higher than that reported previously.1 This can be explained by the much higher index of refraction of diamond surrounding the emitting dipole.23 Under weak excitation, P reduces to P ) ξσΦPexc and it is also possible to calculated the enhancement of the excitation rate γexc ∝ σPexc from the slope of the powerdependent measurements in Figure 3d at weak excitation intensities, where the fluorescence is linearly dependent on the excitation intensity. A 12- and 14-fold enhancement rate can be deduced for configurations A and B, respectively. For other diamond nanocrystals an increase of up to a factor Nano Lett., Vol. 9, No. 4, 2009

of 18 was observed (data not shown). We want to mention that this is the overall enhancement of γexc and stems from an enhanced Pexc because of local field effects as well as a higher σ because of plasmonic amplification of the excitation transition. The effect of the second gold particle is not as strong as that predicted by the FDTD calculation as the simulation just represents an idealized case of a dipole oriented along and located on the axis depicted in Figure 2c. While both, γexc and γrad are thus strongly enhanced, the single photon character of the emission is still preserved, as can be seen from corresponding measurements of the autocorrelation function g(2)(τ) (Figure 3e and Figure 3f). The fact that g(2)(0) > 0 is mainly due to background emission from the glass substrate for the case of the bare diamond, while for configurations A and B the background predominantly stems from the fluorescence of the gold spheres. This does not represent a fundamental limitation for the application as a single photon source, since the decay time of the gold nanoparticles is much faster (