Combining self-assembly technology and the thin-film annealing method, annealed Ag nano-islands (Ag NIs) with SiO2 microsphere (MS) arrays were prepared as surface-enhanced Raman scattering (SERS) substrate in this paper. A “top-binding” model was established to analyze the effect of SiO2 MS at different annealing temperatures and different Ag film thickness. The corresponding comparative analysis with and without MS templates have been carried on. We simulated the electric field enhancement, coupling enhancement spectrum and the optical focus function of SiO2 MS. Characterization and experiments showed that the use of a SiO2 MS array as a dewetting template can well regulate the size and morphology of Ag NIs. Among them, SiO2 MS-Ag-20 can achieve the most uniform particle size at an annealing temperature of 873.15 K. According to the UV-Vis absorption spectrum, it had an active SERS analytical enhancement factor (AEF) of ∼3.91×105 and good SERS reproducibility (relative standard deviation, RSD = 0.113) for R6G when the wavelength was 532 nm.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Surface-enhanced Raman scattering (SERS) has been widely used in the fields of chemistry, physics, optics, and materials science since its discovery in 1974 . The enhancement mechanism in SERS technology comes from the electromagnetic field enhancement (EM) caused by the localized surface plasmons excited on the surface of the metal nanostructure, and the chemical enhancement (CM) formed by the interaction between biomolecules and metals. The contribution of EM is dominant . In EM enhancement, highly enhanced electric field can be produced at the nanogap between metal nanostructures due to the localized surface plasmon resonance (LSPR) [3,4].
As a kind of template that is easy to prepare and relatively cheap, microspheres (MS) have been widely used in the research of SERS, which can obtain a large-area uniform periodic structure through a self-assembly technology. Chen et al.  prepared a three-dimensional self-assembled polystyrene (PS) microsphere/ultra-thin silver film composite structure with an excellent SERS performance (enhancement factor (EF) ≈ 108) and uniformity (relative standard deviation (RSD) ≈ 8%). Lei et al.  used PS microsphere array as a template, removed the template and surface silicon after depositing the gold film, and then performed secondary coating to obtain an elevated bowtie nano-antenna array. A super high SERS activity was obtained by near-field enhancement of the gap cavity. Chen et al.  used the PS template removal method to prepare periodic nanoarrays and silver film-coupled nano-antennas, and studied the influence of film coupling on the optical performance of the antenna. The RSD for detecting the intensity of the SERS spectrum of R6G (rhodamine 6G) was lower than 10%, and the enhancement factor reached 107. Zhao et al.  transferred monolayer AuNPs into a nanocavity prepared with PS microspheres as a template to construct Au particle-in-hemispherical honeycomb nanoarray (PIHHN) as an ultrasensitive and spatially reproducible SERS substrate, the capacity of detection for R6G in an optimal PIHHN substrate was as low as a concentration of 10−15 mol/L, and the RSD of signal deviation was no more than 5.6%. Hu et al.  reported a new hierarchical AAO (HAAO) template with the hexagonally ordered unit cells and the radially distributed nanochannels formed by integrating the self-assembled PS microsphere template into the AAO fabrication process and rationalized in terms of mechanical stress and electric-field induced oxide dissolution, and R6G can be detected for concentration down to 10−10 mol/L. Zhao et al.  prepared a gold-plated glass nanostripe array by reactive ion etching and sputter deposition on a glass slide covered with a single layer of PS colloid, indicating the quantifiable triphenylphosphine (TPP) detection in a large concentration range. A SERS substrate with a nanogap array was manufactured with etching an assembled PS ball array, with EF of 2×106 . Farcau and Astilean  performed a scanning confocal SERS microscopy to investigate the spatial (lateral) variations of the SERS enhancements on gold films over nanospheres (AuFoN) substrates. Results proved that only a small fraction of the total exposed metalic surface provided the major part of the enhancement in this honeycomb structure. Niu et al.  prepared a uniform thin film assembled by magnetic noble metal composite microspheres with high sensitivity and reproducible SERS performance, and the RSD value of the signal intensity of the main Raman peak was still lower than 0.2 at a low concentration of 10−8 mol/L. However, for PS composite SERS substrates, it is difficult to work well in some high temperature applications.
Annealing noble metal films to obtain a variety of nanostructures has been one of the common methods to obtain high-performance SERS substrates. Wang et al.  successfully prepared Ag NIs/Al2O3/Ag structure through annealing and self-encapsulation technology, with EF of 3.9×108 for R6G. Quan et al.  reported a strategy to investigate the correlation between the dewetting temperature of metal film and SERS intensity. Sun et al.  prepared a three-dimensional structure based on large-area flexible carbon fiber cloth decorated by Ag nanoparticles (Ag NPs-CFC) by high temperature annealing, indicating that the analytical enhancement factor (AEF) reached to 2.4×1012 and a detection limit was as low as 1.0×10−14 mol/L. It is potential to achieve SERS substrate with excellent Raman enhancement performance using combining self-assembly microspheres technology and annealing method, as well as to keep working well in some high temperature applications.
In this paper, we deposited silver films of different thicknesses on the SiO2 MS arrays, and then annealed the samples for one hour at different temperatures. The silver film finally formed Ag NIs with uniform size and distribution through the template dewetting process. Their properties are investigated in detail.
2. Preparation and characterization
2.1 Materials and instruments
The silicon wafer is P-type (100) single-throwing hydrogen peroxide type (1∼10 Ω·cm), the thickness of the oxide layer is 300 ± 10 nm (Zhejiang Lijing Optoelectronics Technology Co., Ltd.), surface hydroxylated SiO2 MS with a diameter of 600 nm (Zhongke Leiming Bio Medical Nanotechnology Company), Rhodamine 6G (Shanghai Aladdin Co., Ltd.). The equipments for samples preparation are ultra-high vacuum organic/inorganic thermal evaporation coating system (QX-300), and CVD (chemical vapor deposition) system (Anhui Beiyike Equipment Technology Co., Ltd.). Raman data are recorded by a confocal microscope Raman spectrometer (Horiba's LabRAM HR Evolution), with the excitation wavelength of 532 nm, the power of 1.5 mW, and the integration time of 2 s. The two-dimensional morphology of samples was characterized using Quattro field emission environmental scanning electron microscope (SEM).
The preparation process of the substrate was divided into three main steps, as shown in Fig. 1. The first step was the preparation of structured templates: the self-assembly of the SiO2 MS array was relatively easy. Due to huge volume, the hydroxylation of the MS and the larger Hamaker constant of SiO2 than that of H2O, the SiO2 MS will spontaneously aggregate in water solvents. Therefore, we directly adopted the method to drip SiO2 MS emulsion on the clean wafer and dry it in a vacuum oven, to prepare a self-assembled MS array. The layer of MS array can be controlled by the MS concentration.
The second step was to prepare the silver film by a thermal evaporation apparatus. During the evaporation process, the evaporation speed was controlled at 1·Å·S−1 to ensure the thickness uniformity of the silver film. In order to avoid the oxidation effect brought by the coating process, the working pressure of the evaporation chamber was controlled below 3×10−5 Pa. In this paper, silver films with thicknesses of 50 nm, 30 nm and 20 nm were deposited on the substrate, and SiO2 MS-Ag-50, SiO2 MS-Ag-30 and SiO2 MS-Ag-20 were obtained respectively. For comparison, we also prepared SiO2 wafer-Ag-50, SiO2 wafer-Ag-30 and SiO2 wafer-Ag-20 on blank silicon wafers.
The third step was to anneal the sample. Put the prepared substrate coated with silver film into a quartz tube furnace for annealing treatment at different temperatures. In order to ensure the temperature and accuracy of the temperature during the annealing process, we put the sample in the center of the heating zone of the tube furnace, and put pipe plugs at both ends of the heating zone to prevent the sample from being oxidized in the tube during the process of pumping and ventilation. Or the sample will be blown away because the air velocity was too high. Before annealing, vacuumize the entire system to a target pressure of 2.5×10° Pa in the tube. Then during the entire annealing process, a mixture of hydrogen (reducing agent) and argon (inert gas) was continuously introduced to ensure that the metal film was not oxidized. The flow rate of hydrogen and argon was 40 sccm and 120 sccm, respectively. The entire thermal annealing process was divided into four stages: the temperature at the first stage was from room temperature to 473.15 K. In order to effectively avoid the instability in the early heating process of the tube furnace, the heating rate was set to 10 K/min. At the second stage it was from 473.15 K to the target temperature with the heating rate of 15 K/min. At the third stage the temperature was kept at the target temperature for one hour to achieve the dewetting process of the silver film. Finally, the high-temperature system was shut down, and the sample was taken out after the tube furnace temperature has naturally cooled to room temperature.
In order to avoid random errors, we prepared five samples for one kind of SERS substrates in each batch of experiments. During the annealing process, all samples are placed in the middle of the tube furnace. Raman intensities are averaged data from several times measurements.
We performed SEM characterization on samples of SiO2 MS-Ag-50, SiO2 MS-Ag-30 and SiO2 MS-Ag-20 annealed at 673.15 K, 773.15 K and 873.15 K for one hour, respectively. The choice of annealing temperature was based on Gadkari's work . The result was shown in Fig. 2.
When the thickness of the silver film was 50 nm, most of the silver film was well preserved at the annealing temperature of 673.15 K (Fig. 2(a1)), indicating that the thickness of the silver film was too large, which made the formation of holes too difficult. This temperature was not enough to provide enough energy to make the silver atoms fully diffused. At the annealing temperature of 773.15 K, pores began to form in a large area (Fig. 2(a2)). At the annealing temperature of 873.15 K, the pores grew in a large area, and the silver film underwent an incomplete dewetting process, which has become numerous worm-shaped silver nano-strips (Fig. 2(a3)). When the thickness of the silver film was 30 nm, at the annealing temperature of 673.15 K, the dewetting of the silver film was almost completed, and many independent Ag NIs have been formed (Fig. 2(b1)). However, the particle size of the NIs was not uniform. The distribution of unseparated worm-shaped silver nanostructures and the NIs were relatively random. As the annealing temperature increased to 773.15 K and 873.15 K, the uniformity of the Ag NIs was not significantly improved (Figs. 2(b2, b3)). When the thickness of the silver film was 20 nm, the average particle size of the NIs showed good uniformity at all three annealing temperatures, and the NIs were almost all distributed in the center of the self-assembled SiO2 MS (Figs. 2(c1, c2, c3)). In order to further analyze the influence of annealing temperature and film thickness on the appearance of NIs, we calculated the particle size and gap of NIs under different conditions, shown in Fig. 3.
The deposited silver film with a thickness of 50 nm did not form a NI structure, so the particle size and gap value can’t be counted. As shown in Fig. 3(a), when the thickness of the silver film was 30 nm, with the increase of the annealing temperature, the average particle size was 0.31 µm, 0.33 µm and 0.36 µm at 673.15 K, 773.15 K and 873.15 K, respectively. The average gap has also become an increasing trend, respectively 0.21 µm, 0.24 µm and 0.26 µm. At the same time, the RSDdiameter reflected the uniformity of the size of NIs. The corresponding uniformity increased with the increase of annealing temperature.
As shown in Fig. 3(b), when the thickness of the deposited silver film was 20 nm, with the increase of the annealing temperature, the particle size distribution still showed an increasing trend. The average particle size was 0.24 µm, 0.26 µm, 0.27 µm at 673.15 K, 773.15 K and 873.15 K, respectively. The average gap was 0.26 µm, 0.31 µm and 0.32 µm, respectively. Through calculated RSDdiameter, we got that the size uniformity of NIs at the three annealing temperatures was relatively high, and the RSD of SiO2 MS-Ag-20 at the annealing temperature of 873.15 K was less than 10%. In order to show the uniformity of the substrate more clearly, we performed a lateral SEM characterization of the sample, as shown in Fig. 4(a). Ag NIs were mainly distributed on the top of the SiO2 MS, shown in Fig. 4(b). The distribution of Ag NIs basically followed the distribution of self-assembled SiO2 MS, presenting a large-area uniform NI array.
3. Results and discussion
3.1 Template dewetting
In the process of template dewetting of the metal film, only when the film thickness and the surface topography parameters of the template conform to a certain semi-quantitative relationship, the binding function of the template can be well realized [18–21].
The brief process of the dewetting of the silver film on SiO2 MS was shown in Fig. 5(a). Due to the high shape retention, the deposited metal film will have the same morphological characteristics as the template. For the silver film on the surface of the MS, the density of silver atoms on the upper film was less than that of the lower film. This will cause some nano-holes to be generated spontaneously on the surface of the film to maintain the isotropy of the silver film. Compared with the silver film deposited on the SiO2 wafer, the crown-shaped silver film requires a lower temperature to achieve the same degree of dewetting. From Fig. 5(b), we can see Ag NIs conformed to the “top-binding” due to the role of the MS template, which was different from those common annealing templates such as nanopore and nanowire in which the Ag NIs conformed to the “gap-binding” [19,22].
We also found that there were also some small-sized Ag NIs distributed in the lower part of the SiO2 MS, which were not swallowed by the larger Ag NIs at the top. This phenomenon can be solved by modifying the diameter of the microspheres or increasing the annealing temperature. Because the binding function can be perfectly realized only if the parameters of the template and the film thickness meet a certain relationship. As for temperature, when it increases, the surface diffusion rate of silver atoms increases exponentially. This can provide kinetic energy for the atom to overcome gravity and move upward.
In order to visually analyze the binding effect of the MS template on the Ag NIs, we prepared SiO2 wafer-Ag-20 for a reference experiment. The annealing temperature was 873.15 K and 1073.15 K, respectively, and the time was 1 h.
Shown in Fig. 6(a), the NIs obtained in SiO2 wafer-Ag-20 had a large size span and there were more super small NIs. The average diameter was 216.1 nm and the RSDdiameter was ∼36%. Compared to Fig. 6(b), the advantages of template annealing can be seen. The particle size of the “top-binding” Ag NIs was more uniform and the distribution was relatively regular. Shown in Figs. 6(c), 6(d), when the annealing temperature was further increased to 1073.15 K, the surface diffusion of silver atoms became more intense, and the NIs can swallow each other, thus breaking through the “top-binding”. The biggest diameter can exceed 1500 nm (Fig. 6(d)). At an annealing temperature of 1073.15 K, the template function failed.
3.1.1 Influence of the wavelength and size of Ag NI
Through the Young's relationship , the surface energy values of Ag and SiO2 can be fixed, and the height of the NI was 0.772 times the diameter (the contact angle was about 123°) . This guided us in setting up the simulation model. When the Ag NIs grew on the spherical surface, the relationship between θ and the height of the Ag NI was relatively complicated. For NI of any arbitrary size, we have made the following calculations, shown in Fig. 7.
We set θ = 123°, R was set to 300 nm, and r was set to 130 nm, then we can calculate that the height of the NI exposed to the air was 0.88 times the diameter. This indicated that the NIs grown on the spherical surface were closer to the spherical shape than the flat-grown NIs, which also reflected the modulation of the MS template.
FDTD solutions was used to simulate the spatial distribution of the electric field intensity (Figs. 11(a1∼a7, b1∼b7)). We used plane waves of different wavelengths (0.3 µm∼0.8 µm) to investigate the electric field intensity distribution of the system on two different structures (SiO2 wafer & SiO2 MS). Based on SEM images, the size of Ag NI was 0.26 µm. In our experiment, the surface of the SiO2 MS was hydroxylated due to the process, and the refractive index was greater than that of ordinary SiO2, which was n=1.65. The index of silver was set to the default Palik (0∼2 µm) in the FDTD material library. The surrounding medium was set to air and the perfect matching layer (PML) was used as the boundary condition. The number of layers was 16, the accuracy of FDTD was level 2, and the accuracy of meshing was 2 nm.
Shown in Figs. 8(c), 8(d), the maximum electric field intensity values of the two systems changed similarly with wavelengths and Ag NI sizes. When the radius of the Ag NI was 0.13 µm, the maximum electric field of SiO2 MS-Ag NI appeared at the incident wavelength of ∼400 nm, and the maximum electric field of SiO2 wafer-Ag NI appeared at the incident wavelength of ∼532 nm. When the incident wavelength was 532 nm, within a given range (0.07 µm∼0.17 µm), the electric field intensity of the two systems both showed an approximately decreasing trend as the radius of the NI increased.
3.1.2 Far field
Normally, the reflection and scattering between the cells at close distances (sample SiO2 MS-Ag NI) usually re-excite the LSPR to enhance the electric field. The far-field simulations on the cell structures of the two systems were performed to verify this phenomenon, and the results were shown in Fig. 9.
The reflection and scattering power of the dipole SiO2 MS-Ag NI on the xy plane was greater. The intensity in the x-axis direction (incident polarization direction) was nearly 5 times that of SiO2 wafer-Ag NI. Figure 9(b) showed the mutual excitation effect between the cells. The coupling effect of SiO2 MS-Ag NIs was significantly stronger than that of SiO2 wafer-Ag NIs, and both exhibited wavelength sensitivity.
3.1.3 SiO2 MS enhancement property
A near-field simulation on SiO2 MS was performed using FDTD. The monitor and polarization settings were shown in Fig. 10(a). The incident wavelength was 532 nm.
The local electric field enhancement inside the SiO2-MS was caused by the focusing effect of the MS and the phenomenon of light interference, shown in Fig. 10(b). Figure 10(c) was the field enhancement spectrum of two different materials of SiO2 and PS. For non-metallic materials, the refractive index does not change with wavelength. We take nPS=1.59. The results showed that the field enhancement effect of MS varied with excitation. The enhancement effect of SiO2-MS in the visible light band was stronger than that of PS MS. We guessed this may be related to the higher refractive index.
Figures 11(a1, a2) showed the propagation state of the incident wave on the SiO2 MS plane and the reflected wave from the system at a certain point in time. Figures 11(a3, a4) were the case of a blank SiO2 wafer. The optical focusing function of MS can be seen. MS can also refocus and enhance the reflected field of the system. All NIs attached to the surface of the MS will be excited as shown in Fig. 11(b).
3.1.4 UV-Vis absorption
Figure 12(a) showed the absorption curve of SiO2 MS-Ag-20 in the air and solution; the corresponding Raman signals irradiated by 633 nm and 532 nm were also shown.
The absorption spectrum of the R6G solution had a peak at ∼526 nm, which was conducive to the resonance enhancement in the chemical enhancement mechanism. The comparison of the absorption of SiO2 MS-Ag-20, R6G and SiO2 substrate were shown in Fig. 12(b). The absorption of R6G on SiO2 MS-Ag-20 samples was shown in Fig. 12(c). We can see a wider absorption region around 532 nm.
3.2 Raman measurements
In Raman measurements, an objective lens was a “10× visible” microscope objective lens with a numerical aperture NA=0.25, a working distance WD of 10.6 mm, and a spot diameter of 2.59 µm. The number of grating lines was 600 l/mm. The spectral scanning range was 500∼2000cm−1. The concentration of R6G was 10−6 mol/L. The mapping area was a random 50×50 µm2, the sampling interval was 10 µm, and the number of detection points was 6×6 = 36. The calculated average Raman signals on four samples were shown in Fig. 13(a).
Figure 13(b) was the mapping results of SiO2 MS-Ag-50 at an annealing temperature of 873.15 K, at 1650 cm−1. The calculated RSD was 35.4%, meaning a poor uniformity. In order to compare the detection uniformity of these three substrates more intuitively, we set the ordinate to the same. The RSD of SiO2 MS-Ag-30, MS-Ag-20 and SiO2 wafer-Ag-20 was 19.2%, 11.3% and 30.1%, respectively, shown in Figs. 13(c), 13(d), 13(e). Detection uniformity was orderly sorted by SiO2 MS-Ag-20 > SiO2 MS-Ag-30 > SiO2 wafer-Ag-20. Compared with gold films over nanospheres , flexible carbon fiber cloth decorated by Ag nanoparticles , metal-graphene oxide nanostructured film , our sample SiO2 MS-Ag-20 has better uniformity, using a low cost and simple preparation method.
The averaged intensity of SiO2 wafer-Ag-20 at 1650 cm−1 was stronger than the other three groups of substrates. We thought that this phenomenon can be explained as: during the dewetting process of the silver film deposited on the SiO2 wafer, SiO2 wafer-Ag-20 has smaller Ag NIs. According to our simulation, the smaller Ag NIs will be excited to form a stronger electric field, and the gap of some Ag NIs may be small, and strong “hot spots” may be generated. The limit of detection (LOD) of the four samples was explored, as shown in Fig. 14.
We used the intensity of the characteristic peak at 1650 cm−1 as the IRS to calculate AEF. The calculation rule of AEF is as follows:
In this paper, a self-assembled SiO2 MS array was used as the annealing template to successfully fabricate a highly uniform SiO2 MS-Ag NIs composite structure SERS substrate. The results showed that both the thickness of the silver film and the annealing temperature can affect the degree of dewetting, thereby affecting the size and appearance of the NIs. SiO2 MS-Ag-20 can achieve best SERS detection uniformity (RSD = 11.3%) and nice enhancement performance (AEF = 3.91×105) at an annealing temperature of 873.15 K. In the next step, we will use SiO2 MS with different parameters to conduct experiments and expand the selection range of silver films to further explore the semi-quantitative relationship in MS template dewetting.
Central University Basic Research Fund of China (CQU2018CDHB1A07); National Outstanding Youth Foundation of China (cstc2019jcyjjqX0018); National Natural Science Foundation of China (61875024).
We would like to thank Dr. Gong Xiangnan at Analytical and Testing Centre of Chongqing University for his help in Raman measurement.
The authors declare no conflicts of interest.
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
1. M. Fleischmann, P. J. Hendra, and A. J. Mcquillan, “Raman spectra of pyridine adsorbed at a silver electrode,” Chem. Phys. Lett. 26(2), 163–166 (1974). [CrossRef]
2. S. Y. Ding, J. Yi, J. F. Li, R. Bin, D. Y. Wu, R. Panneerselvam, and Z. Q. Tian, “Nanostructure-based plasmon-enhanced Raman spectroscopy for surface analysis of materials,” Nat. Rev. Mater. 1(6), 16021 (2016). [CrossRef]
3. L. Tong, H. Xu, and M. Käll, “Nanogaps for SERS applications,” MRS Bull. 39(2), 163–168 (2014). [CrossRef]
4. Y. Guo, J. Yu, C. Li, Z. Li, J. Pan, A. Liu, and C. Zhang, “SERS substrate based on the flexible hybrid of polydimethylsiloxane and silver colloid decorated with silver nanoparticles,” Opt. Express 26(17), 21784–21796 (2018). [CrossRef]
5. J. F. Chen, T. Li, and J. Zhang, “Raman enhancement properties of a high uniformity PS microsphere-Ag nanoparticle substrate,” Opt. Mater. Express 10(12), 3215–3225 (2020). [CrossRef]
6. L. Feng, R. P. Ma, Y. D. Wang, D. R. Xu, D. Y. Xiao, L. X. Liu, and N. Lu, “Silver-coated elevated bowtie nanoantenna arrays: Improving the near-field enhancement of gap cavities for highly active SERS,” Nano Res. 8(11), 3715–3724 (2015). [CrossRef]
7. J. F. Chen, C. Zhang, J. Zhang, and Y. Zhu, “Raman enhancement of large-area silver grating arrays based on self-assembled polystyrene microspheres,” Opt. Mater. Express 11(4), 1234–1246 (2021). [CrossRef]
8. W. Zhao, Y. Zhang, J. Yang, J. Li, Y. Feng, M. Quan, Z. Yang, and S. Xiao, “Synergistic plasmon resonance coupling and light capture in ordered nanoarrays as ultrasensitive and reproducible SERS substrates,” Nanoscale 12(35), 18056–18066 (2020). [CrossRef]
9. Y. Hu, X. Wang, M. Zhang, S. Wang, S. Li, and G. Chen, “A hierarchical anodic aluminum oxide template,” Nano Lett. 21(1), 250–257 (2021). [CrossRef]
10. Q. Zhao, G. Liu, H. Zhang, and W. Cai, “Ordered gold-coated glass nano-sting array with large density tips as highly SERS-active chips for detection of trace organophosphorous toxicant,” Nanotechnology 31(41), 415301 (2020). [CrossRef]
11. L. Xia, Z. Yang, S. Yin, W. Guo, S. Li, W. Xie, D. Huang, Q. Deng, H. Shi, and H. Cui, “Surface enhanced Raman scattering substrate with metallic nanogap array fabricated by etching the assembled polystyrene spheres array,” Opt. Express 21(9), 11349–11355 (2013). [CrossRef]
12. C. Farcau and S. Astilean, “Mapping the SERS efficiency and hot-spots localization on gold film over nanospheres substrates,” J. Phys. Chem. C 114(27), 11717–11722 (2010). [CrossRef]
13. C. Niu, B. Zou, Y. Qiang, L. Cheng, H. Zheng, and S. Zhou, “Highly sensitive and reproducible SERS performance from uniform film assembled by magnetic noble metal composite microspheres,” Langmuir 32(3), 858–863 (2016). [CrossRef]
14. Z. K. Wang, J. M. Quan, C. Zhang, Y. Zhu, and J. Zhang, “AgNIs/Al2O3/Ag as SERS substrates using a self-encapsulation technology,” Opt. Express 28(21), 31993–32001 (2020). [CrossRef]
15. J. Quan, J. Zhang, X. Qi, J. Li, N. Wang, and Y. Zhu, “A study on the correlation between the dewetting temperature of Ag film and SERS intensity,” Sci. Rep. 7(1), 14771 (2017). [CrossRef]
16. N. Sun, Z. Wang, M. Jiang, and J. Zhang, “Flexible carbon fiber cloth decorated by Ag nanoparticles for high Raman enhancement,” Opt. Mater. Express 11(4), 1321–1333 (2021). [CrossRef]
17. P. R. Gadkari, A. P. Warren, R. M. Todi, R. V. Petrova, and K. R. Coffey, “Comparison of the agglomeration behavior of thin metallic films on SiO2,” J. Vac. Sci. Technol., A 23(4), 1152–1161 (2005). [CrossRef]
18. D. Kim, A. L. Giermann, and C. V. Thompson, “Solid-state dewetting of patterned thin films,” Appl. Phys. Lett. 95(25), 251903 (2009). [CrossRef]
19. A. L. Giermann and C. V. Thompson, “Solid-state dewetting for ordered arrays of crystallographically oriented metal particles,” Appl. Phys. Lett. 86(12), 121903 (2005). [CrossRef]
20. A. L. Giermann and C. V. Thompson, “Requirements for graphoepitaxial alignment through solid-state dewetting of Au films,” J. Appl. Phys. 109(8), 083520 (2011). [CrossRef]
21. J. H. Chen, T. F. Lei, D. Landheer, X. Wu, and T. S. Chao, “Si nanocrystal memory devices self-assembled by in situ rapid thermal annealing of ultrathin a-Si on SiO2,” Electrochem. Solid-State Lett. 10(10), H302–304 (2007). [CrossRef]
22. J. Basu, C. B. Carter, R. Divakar, B. Mukherjee, and N. Ravishankar, “Nanopatterning by solid-state dewetting on reconstructed ceramic surfaces,” Appl. Phys. Lett. 94(17), 171114 (2009). [CrossRef]
23. Q. Jiang and H. M. Lu, “Size dependent interface energy and its applications,” Surf. Sci. Rep. 63(10), 427–464 (2008). [CrossRef]
24. X. J. Liu, L. Y. Cao, W. Song, K. Ai, and L. H. Lu, “Functionalizing metal nanostructured film with graphene oxide for ultrasensitive detection of aromatic molecules by surface-enhanced Raman spectroscopy,” ACS Appl. Mater. Interfaces 3(8), 2944–2952 (2011). [CrossRef]