1. Introduction

Metallic architectures sustaining plasmonic resonances have been widely considered as promising objects for manipulating electromagnetic field or boosting optical processes because of their instinct virtue of localized field enhancement [1–6 ]. Especially, when the plasmonic metal nanostructures contain sub-10 nm gaps, such narrow gaps are capable of focusing free-space light into sub-diffraction-limited regions and thereby enable extremely extensive field enhancements [6,7 ]. This remarkable merit makes the gap-plasmon nanostructures exceptional for enhancing light-matter interaction on the nanoscale, and various photonic and optoelectronic functional devices have been cultivated based on them [2,8–11 ].

Previous studies about gap-plasmon nanostructures and their corresponding applications exhibit two trends. One is that the majority of the involved materials for constructing sub-10 nm gap-plasmon architectures are gold and silver, whose fabrication and integration processes could be easily realized by using e-beam lithography techniques and chemical depositon methods [2,8,12–15 ]. As a result, the optical working wavelengths of such gap-plasmon nanostructures always lie in the visible range due to their intrinsic location of plasmon resonant wavelengths, leaving other spectral regions, such as ultraviolet (UV) and near-UV regions, unexplored. However, compared to the visible one, light in the UV region has higher energy, shorter wavelength, and smaller scale of diffraction limit. Therefore, it is urgent to find out an UV-enhanced gap-plasmon nanostructure which would enable us to gain some new applications of plasmonic materials in UV light sources, sensors, and optical memories with high storage density [16–18 ].

The other trend is that the main utilization of gap-plasmon nanostructure is serving as substrates for optical enhancements, like Raman scattering of molecules or photoluminescence of nanoemitters [8,13,19–21 ]. Beside the success in this kind of application, the third-order and second-order optical nonlinearities of these materials are also significant due to their strong localized field enhancement [22,23 ]. Recently, many works have been done in these gap-plasmon structrures, including third-order harmonic spectroscopy, four-wave mixing (4WM), second-harmonic generation (SHG), and sum frequency generation(SFG) [15,24–27 ]. But most were carried out using gold and silver materials in the visible and near infrared wavelengths, seldom was taken in UV or near-UV region.

Taken all the aforementioned situations into consideration, one can find that an alternative material which has low-cost and can be built into nanostructures so as to possess excellent UV optical nonlinearity is in need for the development of gap-plasmon nanoscience. Fortunately, aluminum (Al) has been demonstrated as a promising material for plasmonic nanostructure and device fabrication since aluminum nanostructures show great tunability of plasmonic energy extending to the UV regime [28,29 ], as well as stronger nonlinear photoluminescence compared to gold and silver ones [30], which makes aluminum nanostructures applicable in various kinds of UV-enhanced devices such as CMOS compativble devices [31], plasmonic nanoantennas [32], etc. While seldom work was done to investigate their third-order nonlinear optical susceptibility (χ ⁽³⁾). So, herein, by using an optimized anodic aluminum oxide (AAO) template growing method and sputtering technique, we fabricated Al nanorod array film (NRAF) which exhibited large third-order optical absorption as well as nonlinear optical susceptibility in a wide spectral range from 360 nm to 900 nm, according to our measurements performed by Z-scan technique . Especially, by comparing with Au or Ag NRAF, near-UV enhanced nonlinear optical properties can be obtained significantly in Al-NRAF, which implies that Al-NRAFs are good plasmonic nanostructure candidates in the fields of nonlinear optics, photonics and optoelectronics in a broad wavelength range extending to near-UV regions.

2. Experimental

2.1 Preparation of the nanorod array film

The Au, Ag and Al nanorod array film were fabricated on porous AAO templates. The templates were prepared with an optimized two-step anodization process [24]. Prior to the anodization, the aluminum sheets (purity 99.999%) were degreased in acetone and electro polished at 0 °C under a constant voltage of 16 V for 4 min in a mixture of HClO₄ and C₂H₅OH. In the first anodization step, the treated aluminum sheet was exposed to a H₂C₂O₄ solution (0.3 M) under constant voltage of 40 V at ~4 °C for 3 h. The alumina layer produced in this process was then removed by wet chemical etching in a mixture of phosphoric acid (0.15 M) and chromic acid (0.60 M) at 60 °C for 1 h. In the second step, the etched alumina sheet was anodized with 10 wt% H₃PO₄ aqueous solution. The growth voltage was maintained at 40 V for 3 minutes. Thereafter the voltage was quickly raised to 80 V for another 2 hours. During these processes the preparation temperature was controlled around 15 °C. Finally, the Au, Ag, and Al nanorod array films were sputtered (Q150T ES, Quorum) onto the as-grown AAO templates [24].

2.2 Instrumentation and measurements

Scanning electron microscopy (SEM) images were obtained by using a Zeiss Auriga-39-34 operated at an accelerating voltage of 5.0 kV. The absorption spectrum was recorded by an ultraviolet-visible-near-infrared (UV-VIS-NIR) spectrophotometer (PerkinElmer Lambda 950). The optical nonlinearity results were acquired with a Z-scan setup by using a Ti:Sapphire based femtosecond laser system. The laser output from 700 nm to 900 nm was obtained directly from a Mira 900 laser (Coherent). Laser in the range of 500-700 nm and 360-500 nm was generated by an APE OPO system and a harmonic generator, respectively. The photoluminescence measurements were taken by using a CCD camera (Newton 970) through a monochromator with a focus length of 500 mm.

2.3 Numerical simulations

The electric field distributions of Al-NRAF were conducted by using finite-difference time-domain (FDTD) method with a commercially available FDTD simulation software package from Lumerical Solutions. In the calculations the Al nanorod was modeled as core-shell structure composed of a dielectric Al₂O₃ core with 20 nm diameter and 105 nm height. The thickness of the Al shell was set to be 15 nm. The bottom of the nanorod was attached to a dielectric matrix with ordered honeycomb-like air pores inside. The diameter of the air pore was set as 70 nm, and the centre-to-centre distance of adjacent pores was 100 nm. The refraction index of Al₂O₃ was 1.76, and we used periodic boundary conditions in x and y directions while perfectly matched layers in z direction.

3. Results and discussions

Figures 1(a) and 1(b) show the top-view and cross-section SEM images of the Al-NRAF, respectively. One can see that the nanorods look uniformly and the rod diameter is about 50 nm, and gaps between neighboring rods are around 7 nm wide. According to the fact that strong localized field enhancement and large third-order optical nonlinearity usually appear in samples of high particle density and high percolation threshold [33], it is reasonable to expect large third-order optical nonlinearities in this kind of Al-NRAF. It is also worth noticing that the existence of the naturally generated alumina outside the Al particles will further enhance the optical nonlinearities of the sample [28].

 

Fig. 1 SEM images of Al-NRAF. (a) Top-view SEM image of Al-NRAF. The average diameter of Al nanorods is about 50 nm and the average interparticle distance is around 7 nm. (b) Cross-section SEM image of Al-NRAF.

Download Full Size | PDF

Figure 2 is the optical absorption spectra of Au-NRAF, Ag-NRAF, and Al-NRAF, and the inset is the corresponding transmission spectra. For Au-NRAF and Ag-NRAF, strong surface plasmon induced absorption peaks can be observed around 600 nm and 480 nm, respectively. While for Al-NRAF, the absorption grows monotonously as the wavelength goes from NIR to UV region. This feature of Al-NRAF can be resulted from two reasons. On one hand, the resonant absorption band may locate in the deep violet region which is beyond our measurement ability; on the other hand, the possible existence of alumina which forms outside the sputtered Al layer will influence the performance of plasmon resonant behavior of Al. For all of the three samples, the absorption decreases as the wavelength increases at the longer wavelength side, which means that all the samples are semi-continuous [34,35 ].

 

Fig. 2 Absorption spectra of Au-NRAF, Ag-NRAF, and Al-NRAF, respectively. The corresponding transmission spectra are shown in the inset. The surface plasmon induced absorption peaks For Au-NRAF and Ag-NRAF are located around 600 nm and 480 nm, respectively. While for Al-NRAF, the absorption monotonously grows as the wavelength goes from NIR to UV region.

Download Full Size | PDF

Figures 3(a) and 3(b) are typical third-order optical nonlinear absorption and nonlinear refraction curves of Al-NRAF obtained at 810 nm with the peak laser irradiance I ₀ of 1.1 × 10¹² W/m², while the insets are corresponding typical Z-scan curves for Au-NRAF and Ag-NRAF. The data are extracted from the open- and close-aperture Z-scan data, respectively. The nonlinear absorption results of the sample was calculated according to the theoretical expression between transmittance T and sample position z that is given as $T(z)={\displaystyle \sum _{m=0}^{\infty }{[-q_o]}^m/{\left(1+z^2/z_o^2\right)}^m{(m+1)}^{\raisebox{1ex}{$3$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}}$, where z ₀ is Rayleigh range which equals to nπω ₀ ²/λ, ω ₀ is the waist radius at focus [36]. The q ₀ value is extracted to be −1.3 by fitting the open-aperture Z-scan data with the expression of T(z). The expression between q ₀ and the nonlinear absorption coefficient β is presented as $q_0=I_0\cdot L_{\text{eff}}\cdot \beta$, where I ₀ is the peak irradiance of the laser beam at focus and L _eff is an effective thickness of the sample, so the value of β is calculated to be −6.9 × 10⁻⁵ m/W, and the corresponding value of Imχ ⁽³⁾ is −8.2 × 10⁻⁶ esu.

 

Fig. 3 (a) Open-aperture Z-scan data of Al-NRAF, the insets are typical open-aperture Z-scan data for Au-NRAF and Ag-NRAF at 810 nm. (b) Close-aperture Z-scan transmittance normalized by dividing the open-aperture one of Al-NRAF, the insets are typical curves for Au-NRAF and Ag-NRAF, respectively. The solid line is a fitting curve and the symbols are experimental data (λ _ex = 810 nm, I ₀ = 1.1 × 10¹² W/m²). (c) and (d) show the imaginary part of the third-order nonlinear optical susceptibility (Imχ ⁽³⁾) and figure of merit (Imχ ⁽³⁾/α) of Al-NRAF, versus excitation wavelength (I ₀ = 1.1 × 10¹² W/m²), the data of Au-NRAF and partial data of Ag-NRAF are also demonstrated for comparison. (e) and (f) are simulated field distribution of Al-NRAF under 810 nm and 400 nm laser excitation, respectively (obtained 15 nm below the top of the nanorod).

Download Full Size | PDF

In the nonlinear refraction results calculation, T(z) can be theoretically written as $T(z)=1_0+\frac{4\Delta {\Phi }_0\times ({z/z}_{\text{0}})}{\left\{\left[{\left(z/z_0\right)}^2+9\right]\left[{\left({z/z}_0\right)}^2+1\right]\right\}}$ [36]. The value of ΔΦ ₀ is extracted to be −0.9 by fitting the nonlinear refraction data. The relation betweenΔΦ ₀ and nonlinear refraction coefficient γ is expressed as $\Delta {\Phi }_0=k\cdot I_0\cdot L_{\text{eff}}\cdot \gamma$, where k is the wave number, so that the value of γ is calculated to be −6.5 × 10⁻¹² m²/W, and the corresponding value of Reχ ⁽³⁾ is −5.1 × 10⁻⁶ esu. The total third-order optical susceptibility is obtained to be 9.7 × 10⁻⁶ esu. In addition, that value of AAO template is on the order of 10⁻¹⁰ esu [37], which can be neglected compared to the Al-NRAF.

Figures 3(c) and 3(d) are the excitation wavelength dependent imaginary part of the third-order nonlinear optical susceptibility (-Imχ ⁽³⁾) and the figure of merit (-Imχ ⁽³⁾/α) of Al-NRAF, Au-NRAF, and Ag-NRAF, respectively. Since the light scattering is strong at short wavelengths which hinders the measurement of nonlinear refraction, the wavelength dependent results of real part of χ ⁽³⁾ are not given here. For Al-NRAF, the value of -Imχ ⁽³⁾ increases from 5.3 × 10⁻⁶ esu to 2.7 × 10⁻⁵ esu and the value of -Imχ ⁽³⁾/α grows from 1.0 × 10⁻¹¹ esu·cm to 2.5 × 10⁻¹¹ esu•cm, respectively, as excitation wavelength decreases from 900 nm to 360 nm. All these values are extremely large compared to those reported in many other articles [38–43 ], indicating quite outstanding nonlinear optical performance of this NRAF structure. Besides, the large values of -Imχ ⁽³⁾ for the Al-NRAF can be achieved in the full measurement wavelength range, which can be attributed to strong local field enhancement induced by the intense interaction between Al on the nanorods. Furthermore, it is worth noticing that the maximum -Imχ ⁽³⁾ as well as the -Imχ ⁽³⁾/α were obtained at the shortest wavelength measured (further measurements at much shorter wavelength was limited by our laser output wavelength). The existence of strong electric field enhancements in the Al-NRAF can be revealed by the FDTD simulations, as shown in Figs. 3(e) and 3(f) where the excitation wavelengths are set as 810 nm (e) and 400 nm (f), respectively. After comparing the field distributions of different incident light at the same color scale, one can find that the field enhancements at shorter wavelength is much more intensive, which is not only in accordance with the absorption behavior of Al-NRAF, but also further suggests that the Al-NRAF sample may exhibit excellent performance in the blue and near-UV regions rather than in visible region.

To demonstrate the novel nonlinear optical properties of Al-NRAF at blue and near-UV wavelength regions, comparative data of Au–NRAF and Ag-NRAF are also given in Figs. 3(c) and 3(d) (the morphologies of Ag and Au NRAF can be found in [26]). For Au-NRAF, the trend of the wavelength dependent -Imχ ⁽³⁾ agrees well with its absorption spectrum. At wavelengths longer than 510 nm, Au-NRAF has larger values of -Imχ ⁽³⁾ than Al-NRAF, while the result is opposite when the wavelength is shorter than 510 nm. When it comes to Ag-NRAF, since photochromism always happens at near-UV and resonance wavelengths, the data in 360-720 nm region were not stable and then they were not given here [44]. Since different samples hold different linear absorptions, figure of merit would be a good parameter for comparison between different samples. From Fig. 3(d) one can see that between 360 nm and 720 nm, Al-NRAF shows much better performance than Au-NRAF. Therefore, we conclude that Al-NRAF is an excellent plasmonics nanostructure candidate for blue and near-UV applications.

Except for the comparison of the third-order optical nonlinearities, PL enhancement experiments of CdSe/ZnS quantum dots (QDs) were also performed on different substrates. Figure 4(a) is a typical PL spectrum of CdSe/ZnS QDs excited by a 400 nm laser. Figures 4(b)-4(e) are normalized PL intensity of QDs on the substrates glass, Au-NRAF, Ag-NRAF, and Al-NRAF, respectively. For better comparison, all the PL measurements were performed under the same average pump power, and the data were normalized to the one obtained at 440 nm. To avoid PL quenching of QDs on metallic structure surface, TiO₂ layer with thickness of several nanometers was prepared between the QDs and the metallic surfaces by sputtering Ti metal at room temperature in a mixture atmosphere of Ar and O₂ (base vacuum = 1.3 × 10⁻³ Pa, partial pressure ratio = 0.5:0.5, total pressure = 5.3 Pa). One can see that the PL intensity of QDs on glass and Au-NRAF substrates did not change much as the excitation wavelength varied from 440 nm to 380 nm. While the PL intensity drops greatly as the excitation wavelength decreases on Ag-NRAF substrate, and grows drastically as the wavelength decreases for Al-NRAF substrate, respectively. This indicates that Al- NRAF is the best substrate to enhance the PL intensities at blue and near-UV wavelengths.

 

Fig. 4 Excitation wavelength dependent peak PL intensities of CdSe/ZnS QDs on different substrates. (a) Typical PL spectrum of CdSe/ZnS QDs. (b)-(e) are excitation wavelength dependent peak PL intensities of QDs on glass, Au NRAF, Ag NRAF, and Al NRAF substrates, respectively, normalized to the PL intensity at 460 nm for better comparison.

Download Full Size | PDF

4. Conclusion

Regularly distributed Al-NRAF has been deposited on an AAO template by using sputtering method. The SEM images show that the average diameter of the nanorods is about 50 nm and the width of the gaps between adjacent rods is about 7 nm, respectively. Large third-order optical susceptibilities with -Imχ ⁽³⁾ of from 5.3 × 10⁻⁶ to 2.7 × 10⁻⁵ esu, and -Imχ ⁽³⁾/α of from 1.0 × 10⁻¹¹ to 2.5 × 10⁻¹¹ esu•cm in 900 nm to 360 nm wavelength region were obtained, and the maximum appeared at the shortest measured wavelength. Comparisons of the image part of third-order optical susceptibility, figure of merit, and PL enhancements between Au-NRAF, Ag-NRAF, and Al-NRAF samples demonstrated that Al-NRAF has the best nonlinear optical properties at the blue and near-UV wavelength regions. FDTD simulations, as well as the SEM morphologies and the absorption results suggested that the large third-order nonlinearity came from the large localized field enhancement. The results obtained here show that Al-based gap-plasmonic nanostructures have large third-order optical nonlinearities and strong localized field enhancement at blue and near-UV wavelengths, and can therefore be applied in many kinds of UV-enhanced substrates and devices, like nonlinear plasmonics, SERS substrates and solar cells.