Boosting nonlinear frequency-conversion efficiencies in hybrid metal–dielectric nanostructures generally requires the enhancement of optical fields that interact constructively with nonlinear dielectrics. Inevitably for localized surface plasmons, spectra subject to this enhancement tend to span narrowly. As a result, because of the spectral mismatch of resonant modes at frequencies participating in nonlinear optical processes, strong nonlinear signal generations endure the disadvantage of rapid degradations. Here, we experimentally design a multiband enhanced second-harmonic generation platform of three-dimensional metal–dielectric-metal nanocavities that consist of thin ZnO films integrated with silver mushroom arrays. Varying geometric parameters, we demonstrate that the introduction of ZnO materials in intracavity regions enables us to modulate fundamental-frequency-related resonant modes, resulting in strong coupling induced plasmon hybridization between localized and propagating surface plasmons. Meanwhile, ZnO materials can also serve as an efficient nonlinear dielectric, which provides a potential to obtain a well-defined coherent interplay between hybridized resonant modes and nonlinear susceptibilities of dielectric materials at multi-frequency. Finally, not only is the conversion efficiency of ZnO materials increased by almost two orders of magnitude with respect to hybrid un-pattered systems at several wavelengths over a 100-nm spectral range but also a hybrid plasmon-light coupling scheme in three-dimensional nanostructures can be developed.
As an essential research topic in the area of nonlinear frequency conversion, second-harmonic generation (SHG) signifies a second-order nonlinear optical effect that generates photons with twice the fundamental frequency (FF) of pumping lasers.1 Particularly, due to its wide applications to labeling-free ultrasensitive in situ characterization technique and high-resolution microscopy,2–4 SHG in plasmonic nanostructures have attracted extensive attention for decades. Within these nanostructures, strong nonlinear optical effects can be frequently achieved by matching localized surface plasmon resonances (LSPRs) with the frequency of fundamental excitation, second-harmonic (SH) emission, or both.4–10 Inevitably for LSPRs, spectra subject to this enhancement tend to span narrowly. Therefore, due to the spectral mismatch of resonant modes at frequencies participating in nonlinear optical processes, efficient nonlinear signal generations endure the disadvantage of rapid degradations. Under this circumstance, special attention has been paid to satisfying critical frequency-matching conditions, which severely limit the use of localized SHG over a wide operating bandwidth.
Notably, due to high losses (from Ohmic losses) and inherently low nonlinear coefficients of traditional plasmonic materials, e.g., gold, silver, and aluminum,11 the focus on nonlinear frequency conversion has switched from plasmonic nanodevices to hybrid structured systems consisting of metallic and dielectric parts.12–19 In these systems, impressively high field confinements, extremely low losses, and intrinsically large nonlinear susceptibilities are integrated. Moreover, the coherent interplay between hybrid components, e.g., plasmonic-photonic coupling and plasmonic-excitonic state, leads to the strong reshaping of second-order polar responses.12–14,19 Inspired by these pioneering progresses, here, we propose a concept of robust nanostructures with the high frequency tunability in the configuration of hybrid metal–dielectric–metal (MDM) nanocavities. Because of the existence of dielectric layers, an ultrathin gap that contributes to the extreme confinement of light into subwavelength mode volumes appears.17 However, the requirement of high-valued nonlinear conversion efficiencies in planar MDM nanocavities occasionally encounters inevitable hindrances, especially for the cancellation between SH sources from two sides of the gap, referred to as “SHG silencing.”8,20 Very recently, processing plasmonic materials in three dimensions offers a potential to spatially reshape the distribution of local-field energies in which electromagnetic-field enhancements are not exclusively confined to the interface or the substrate but primarily redistributed in air. This processing provides additional degrees of freedom to modulate nonlinear scattering processes in a highly sensitive way.17,21–24
In this work, we experimentally design a multiband plasmon-enhanced SHG platform of three-dimensional MDM (3D-MDM) nanocavities that consist of thin ZnO films integrated with silver mushroom arrays. Studies among 3D-MDM nanocavities with various geometry parameters reveal that ZnO materials in intracavity regions can not only modify mode couplings near fundamental wavelengths but also serve as an efficient nonlinear dielectric for the possibility to allow amplifying SHG signals at several wavelengths simultaneously. This possible amplification is intimately related to (1) the strong coupling induced plasmon hybridization between localized and propagating surface plasmons and (2) the well-defined symmetry (orientation) matching between hybridized resonant modes and nonlinear susceptibilities of dielectric materials over a wide operating bandwidth. Finally, the results of 3D finite-difference time-domain (3D-FDTD) simulations agree excellently with experimental data.
In our experiments and simulations, we utilize an array of silver mushrooms featured with centrosymmetry-breaking units in 3D space separated by a small gap. Figure 1(a) schematically displays the hybrid structured system that will be studied by utilizing nanoimprint lithography combined with a two-step deposition process including electro-chemical deposition (ECD) and atomic-layer deposition (ALD).21,25 Within this system, the array of silver mushrooms is grown on the flat gold substrate (labeled as SMA). Then, during the ALD processing, a conformal ZnO film with 100 nm thickness is created. The thin ZnO film covers the entire surface of SMA, which we called a hybrid ZnO-silver mushroom array (labeled as HZSMA). Therefore, nonplanar MDM nanocavities between adjacent pilei can be obtained. Coordinates are chosen such that mushrooms lie on the x–y plane. Measurements of SHG are performed by focusing a tunable Ti:sapphire laser onto samples from the air side (z direction) at a 45°-incident angle, while reflection-scattered signals with twice the frequency of fundamental waves are collected. Here, the utilization of oblique incident beams can reduce the 3D symmetry of nanostructures, inducing effective second-order nonlinear processes.23,26 In Fig. 1(b), the side-view scanning electron microscopy (SEM) image depicts that silver mushrooms are made of stipes and pilei with the height of stipes and the diameter of pilei equal to 220 nm and 300 nm, respectively. At ∼10 nm, gap sizes (g) among adjacent pilei can be tuned via optimizing the ECD deposition time. The arrangement of mushrooms with 450-nm period along both the x and y axes (px = py = p) exhibits a high degree of homogeneity, leading to the achievement of uniform and reproducible responses [Fig. 1(c)]. Furthermore, the cross section of HZSMA shown in Fig. 1(d) illustrates the hybrid structured system that consists of thin ZnO films integrated with SMA. Additionally, we perform energy-dispersive spectroscopy analyses to display the change of the weight ratio of elements silver (Ag), gold (Au), zinc (Zn), oxygen (O), carbon (C), nitrogen (N), and silicon (Si) for different sampled regions, i.e., Regions I, II, and III, denoting the region across the convex surface of mushrooms, the region among adjacent pilei, and the pileus-substrate cavity region, respectively [see the insets in Figs. 1(b)–1(d)]. In particular, the existence of elements Zn and O for Region I–III reveals the evolution of HZSMA and verifies the presence of homogeneous ZnO materials that cover the entire surface of mushrooms including interpileus gaps and pileus-substrate cavities.
As the first step of our experimental procedure, we introduce the reflection configuration of SH spectroscopy for estimating the samples’ nonlinearity activity. As shown in Fig. 2(a), 130 fs laser pulses tunable from 710 nm to 900 nm (Mira 900, Coherent, Inc.) under the 45°-incident angle are focused on samples using a lens L1 (f = 100 mm). The laser power can be continuously controlled by a neutral density filter combined with the half-wave plate-Glan–Taylor prism pair. A long-pass filter at 700 nm can also be used to filter the possible spurious photoluminescence in the laser line. Then, samples are mounted on an x–y sample stage at the focus of incident beams, and scattered SH signals in the 45°-reflection direction are collected by a lens L2 (f = 100 mm). Note that these incident/reflection angles are defined as those between incident/collection directions and normal directions of the sample stage. The distance between lens L3 (f = 200 mm) and L2 is much larger than the focus length of L3 (nontelecentric imaging system). Finally, collected signals are focused by L3, filtered by a short-pass filter at 550 nm, and measured using an electron multiplying charge coupled device (EMCCD) camera (Newton 970) attached to a spectrometer (Andor SR550) with different integral time. Expectedly, the emission peak shifts from 390 nm to 440 nm by sweeping the excitation wavelength (λex) between 780 nm and 880 nm [Fig. 2(b)]. Meanwhile, the measured signal intensities increase quadratically with the increasing excitation power ranging from 10 mW to 80 mW [Fig. 2(c)]. Within this range, SHG outputs remain stable, suggesting that no significant damage is caused by the incident laser radiation to our samples. Experimental results agree satisfactorily with the essential characteristics of SHG.27,28 Afterward, measurements are also taken on SMA and ZnO films with 100 nm thickness deposited either on the un-patterned gold substrate or on the un-patterned silicon substrate (labeled as ZnO@Au and ZnO@Si). As shown in Fig. 2(d), the emission signal of HZSMA appears much stronger than that of SMA, ZnO@Au, and ZnO@Si [Fig. 2(d)]. Next, we conduct a comparative study of λex-dependent SHG performances among HZSMA, SMA, and ZnO@Au [Fig. 2(e)]. This study demonstrates that, at ∼820 nm and 860 nm, λex-dependent peak values reach a maximum for HZSMA and SMA, respectively. Moreover, for SMA, when λex sweeps into the region from 820 nm to 780 nm, more than one order of magnitude reduction in the maximum intensity can be observed. By comparison, the λex-dependent degradation of nonlinear signal generations (SH energy loss) for HZSMA ranges from 20% to 40% over a wide operating bandwidth. Also, the inherent weakness and the frequency-related insensitiveness for SH responses in ZnO@Au are observed. It is worth noting that in λex-dependent measurements, each SH response has been maximized by optimizing the optical path for keeping the impinging power as a constant and ensuring the same configuration of incident beams such as the position and the size of incident-beam spots.
To quantitatively describe these discrepancies, we define the conversion efficiency as , where denote the SH and FF average powers.23,29 In our experiments, the average power of FF beams () can be obtained in front of samples (after the focus of incident beams) via using a power meter. Also, the average power of SH signals () can be estimated by considering transmission coefficients of optical components and quantum efficiencies of the EMCCD detector. Consequently, for a 50-mW average power of fundamental waves at 780 nm, conversion efficiencies of nominally 2.6 × 10−12, 3.2 × 10−14, 3.2 × 10−14, and 9.8 × 10−15 are obtained for HZSMA, SMA, ZnO@Au, and ZnO@Si, respectively. To illustrate a fair comparison, we use as the nonlinear coefficient,29 indicating that it does not depend on the pumping power used for SHG. With this analysis, the values of ξ for HZSMA, SMA, ZnO@Au, and ZnO@Si are corrected by 5.2 × 10−11, 3.2 × 10−13, 3.2 × 10−13, and 1.96 × 10−13 (W−1), respectively. More importantly, the comparison between HZSMA and SMA (ZnO@Au and ZnO@Si) yields a factor of ∼163 (163, 265), revealing that the external efficiency of HZSMA reaches nearly two-order magnitude higher than that of references. Similarly, the values of ξ for HZSMA and SMA at 820 nm and 860 nm are given by 9.2 × 10−11 (W−1) and 2.5 × 10−11 (W−1), respectively. Then, a comparison of ξ demonstrates that a fourfold reduction in the maximum external efficiency between HZSMA and SMA can be attained. Experimental results mentioned above exhibit that SH responses in fully metallic nanostructures satisfy the critical frequency-matching condition, while our hybrid structured system allows a broad tunability of generated SH signals featured with frequency-insensitive enhancements. Moreover, due to the relative uniform λex-dependent intensity distribution in both HZSMA and ZnO@Au, we can reasonably deduce that observed SH signals in such a hybrid structured system primarily originate from the nonlinear response of the ZnO material itself. It should be pointed out that, for more accurately checking the role of ZnO materials played in hybrid structured systems, alternative dielectric materials with the same index, but with the negligible nonlinear response, should be further considered. However, due to the limitation of the index matching in the measured wavelength range combined with the compatibility of ALD processes, the replacement of alternative dielectric materials remains uncertain to a large extent such as the distribution of index contrast in the intracavity region and the variation of gap sizes among adjacent pilei. This uncertainty may complicate the comparison of linear and nonlinear characterizations among these samples. Nevertheless, we can still qualitatively deduce that the maximum external efficiency of the hybrid structure that consists of alternative dielectric materials may resemble that of SMA. At the same time, the λex-dependent intensity distribution in such a structure may experience a spectral shift when compared to that in SMA.
To identify the mechanism that governs these high-valued enhancement factors, we measure the reflectance spectra of HZSMA and SMA using a linearly p-polarized white light source, respectively [Figs. 3(a) and 3(b)]. Both of them feature the same period, stipe height, diameter, and g value but different refractive indexes (n) in the intracavity region, i.e., n = 1.0 for air in SMA and wavelength-dependent n ranging from 2.1 to 1.9 for ZnO thin films30 in HZSMA. Then, four dips appear in both samples including fundamental-frequency-related (FF-related) resonances such as dips D1 and D2. The spectral position of maximum SH peak values agrees satisfactorily with one half that of FF-related resonances, as demonstrated in Figs. 2 and 3, indicating that FF-related plasmon-driven enhancements play a significant role in the amplification of SH emission powers. The dominant role of plasmonic resonances played in enhanced SH responses can also be regarded as a criterion to exclude the influence of the wavelength-dependent fluctuation in fundamental waves on signal characterizations (see the supplementary material for detailed discussions). Moreover, the present work depends on our previous understanding that SHG signals in these 3D nanostructures can be boosted by (a) interpileus-gap resonances, (b) pileus-substrate cavity modes, or the plasmon hybridization between (a) and (b).21 Armed with this demonstration, we can reasonably deduce that dips D4, D3, and D1 originate from the plasmon hybridization of both LSPR-related interpileus-gap resonances and pileus-substrate cavity modes, and the dip D5 stems from the plasmon hybridization between collective surface modes and pileus-substrate cavity modes. To provide more physical insights, we carry out 3D-FDTD simulations (FDTD Solutions 8.7, Lumerical Solutions, Vancouver) to demonstrate local-field distributions and electromagnetic-field vectors of several plasmonic resonances. The dimension of the modeled system is chosen to match the previously fabricated structure in which the conformal ZnO thin films cover on the entire surface of SMA including the air–silver interface, interpileus-gap, and pileus-substrate cavity regions (see Fig. S1 and the supplementary material for 3D-FDTD simulated methods). Optical constants of gold, silver, and ZnO are taken from Refs. 30 and 31. For SMA, electromagnetic fields at D1 are primarily concentrated near the end point of mushroom umbrellas and redistributed in the intracavity region, exhibiting the plasmon hybridization between interpileus-gap resonances and pileus-substrate cavity modes [Fig. 3(c)]. For HZSMA, due to the introduction of nonlinear dielectrics, linear optical properties have changed [Fig. 3(b)]. However, four dips still remain, including the FF-related dip D2. Notably, the local-field distribution at D2 is primarily modulated by collective surface modes propagating on the convex surface of mushrooms in the array along the x axis, featured with umbrella-related “hot spots” [Fig. 3(d)]. In contrast, the near-field distribution at D3′ resembles that at D1, indicating the plasmon hybridization of both interpileus-gap and pileus-substrate resonances. Differing from the electromagnetic-field distribution at D1, plasmonic-resonance-driven enhancements of pileus-substrate cavity modes at D3′ have been severely suppressed [Fig. 3(e)]. It is worth noting that ZnO is chosen as the nonlinear dielectric in our experiments because its Wurtzite structure is the characteristic of a C6v symmetry with the nonvanishing element17 ideally for z-polarized fields across the air–silver interface. Consequently, a well-defined symmetry (orientation) matching between local modes at D2 and nonlinear susceptibilities of ZnO materials can be attained in the region of the convex surface of mushrooms, contributing to the generation of strong nonlinear currents.
With the aid of the previous demonstration, coupled local-field energies in hybrid structured systems are primarily concentrated near the end point of mushroom umbrellas and redistributed either on the convex surface of mushrooms or in the pileus-substrate cavity region. Furthermore, because a large fraction of the spatial region with local fields is exposed to the environment, the index sensitivity of LSPR-based sensors is increased by lifting metal nanoparticles above the substrate in 3D space.32 Accordingly, we carry on n-dependent simulations by varying n ranging from 1.00 to 2.50 to identify the evolution of several plasmonic resonances [Fig. 4(a)]. See the supplementary material for n-dependent simulated methods. Consequently, as n increases from 1.00 to 1.70, the dip D1 experiences a redshift toward the near-infrared region where the detection efficiency varies poorly in our instruments; the dip D3 remains almost unchanged; dips D4 and D5 exhibit a similar redshift fashion and continue to approach each other, resulting in the generation of a new hybridized resonant dip D4′. Particularly, when n increases from 1.70 to 2.10, a clear anti-crossing behavior can be observed, indicating that near n = 1.70 modes between D4′ and D3 are strongly coupled and forming the occurrence of new hybrid polaritonic states (D3′ and D2). Furthermore, we conduct a comparative study between experimental and simulated results (see Fig. S2 in the supplementary material). This study reveals a qualitative agreement for the monitor of several plasmonic resonances, including FF-related dips D2 and D3′. The relative shift trend in dips D3′ and D2 can be attributed to fabrication imperfections such as the inevitable variation of gap sizes among adjacent pilei. As depicted in Fig. 4(b), a schematic of the resonances’ interaction demonstrates that two interacting channels are characterized with the hybridization between interpileus-gap resonances and collective surface modes propagating on the convex surface (labeled as gap-surface modes, highlighted in red boxes and red lines) and the hybridization of both interpileus-gap resonances and pileus-substrate cavity modes (labeled as gap-cavity modes, highlighted in blue boxes and blue lines). When two channels are strongly coupled, an avoided crossing with a Rabi splitting can be distinctively observed. This avoided crossing stands for a signature of the strong coupling, leading to the coherent energy transfer and substantial energy repartitioning among modes,12,33–35 i.e., D3′ and D2. Consequently, the charge distribution of D2 (highlighted in the black box) reflects a dipolar interaction with all neighbors, which resembles that of D3, while the dominance of the FF-related resonance (D2) switches from LSPR-related dependences (D3) to collective-resonance-driven enhancements (D4′). An increase of local-field energy densities across the silver–dielectric interface, which originates from the extensively spill-out of coupled local-field energies from LSPR-related interpileus-gap regions, can be attained. As a result, a high degree of the spatial overlap between hybridized resonant modes near fundamental wavelengths and nonlinear susceptibilities of ZnO materials on the convex surface of mushrooms may result in the generation of strong nonlinear currents, contributing to the improvement of conversion efficiencies. The dominant role of gap-surface modes played in multiband enhanced SH responses can be further confirmed by conducting g-dependent SHG measurements for HZSMA (see Fig. S3 and the supplementary material for experimental results and detailed discussions). Notably, due to the dispersion of ZnO thin films, the FF-related resonant dip D2 may experience a redshift ranging from 783 nm to 833 nm as n increases from 1.90 to 2.10. Additionally, the near-field interaction of these n-dependent resonances behaves very similarly [see the insets highlighted in the black box of Fig. 4(b)], revealing that the reshaping of improved local-field energy densities across the air–silver interface can be attained over a wide operating bandwidth. This reshaping provides a potential to obtain the well-defined symmetry (orientation) matching between hybridized resonant modes and nonlinear susceptibilities of dielectric materials at several wavelengths simultaneously and further contributes to enhanced SH signals at multi-frequency.
In summary, we have experimentally designed a multiband plasmon-enhanced SHG platform by constructing hybrid structured systems with SMA coupled to thin ZnO films. Geometry-dependent measurements demonstrate that, on the proposed platform, the conversion efficiency of ZnO materials reaches up to two-order magnitude more highly than that of several references at several wavelengths over a 100-nm spectral range. The n-dependent characteristics reveal that, due to the introduction of ZnO materials, the high-index contrast in the intracavity region facilitates the tuning of FF-related resonances at different spectral positions. Additionally, this contrast results in the strong coupling induced plasmon hybridization between LSPR-related plasmonic fields and out-of-plane propagating surface modes, leading to the formation of Rabi-splitting type hybridized resonances and the redistribution of coupled local-field energies. Meanwhile, ZnO materials can also be regarded as an efficient nonlinear dielectric, which provides the possibility to obtain the well-defined coherent interplay between hybridized resonant modes and nonlinear susceptibilities of dielectric materials at multi-frequency. This coherent interaction allows amplifying nonlinear optical responses of dielectric materials over a wide operating bandwidth. Hopefully, our work will provide insights for understanding the governing mechanism of plasmon-enhanced nonlinear optical processes in hybrid structured systems and benefit ultra-low-threshold, efficient, and reproducible nonlinear wavelength conversion for quantum-optical and nonlinear metamaterial applications.
See the supplementary material for the discussion regarding the influence of the wavelength-dependent fluctuation in fundamental waves on signal characterizations, 3D-FDTD simulation technique, comparative study between experimental and simulated results, and g-dependent SHG measurements for HZSMA.
The authors would like to thank Professor Junbo Han, Professor Jinfeng Zhu, Professor Shengli Huang, Dr. Zongwei Ma, Dr. Jie Zheng, Dr. Jingyu Wang, and Dr. Renxian Gao for experimental assistance and helpful discussion. This work was supported by the National Natural Science Foundation of China (Grant Nos. 12004121, 91850119, and 21673192), the Ministry of Science and Technology of the People’s Republic of China (Grant Nos. 2016YFA0200601 and 2017YFA0204902), the Natural Science Foundation of Jiangxi Province (Grant No. 20192ACB20032), and the Scientific Research Funds of Huaqiao University (Grant No. 605-50X19028).
The data that support the findings of this study are available from the corresponding author upon reasonable request.