Bottom-up growth of Ag/a-Si@Ag arrays on silicon as a surface-enhanced Raman scattering substrate with high sensitivity and large-area uniformity †

Bottom-up growth of Ag/a-Si@Ag arrays on silicon as a surface-enhanced Raman scattering substrate with high sensitivity and large-area uniformity †

Liwei Liu,^aMingliang Jin,^bQingwei Zhou,^aRunze Zhan,^cHuanjun Chen,^c Xingsen Gao,^aStephan Senz,^dZhang Zhang*^aand Junming Liu*^e

In this work, we demonstrated a bottom-up growth of Ag/a-Si@Ag nanosphere arrays on Si to work as a high performance SERS substrate. By a solid-state dewetting process, high density Ag nanoparticle arrays were formed on the Si substrate with large-area uniformity. With well-controlled thicknesses of a-Si and Ag outer-layer coating, the SERS properties can be optimized with the inter-particle spacing of the Ag/a- Si@Ag NSs limited to several nanometers. Based on the analysis of optical properties and SERS activities top-Tc molecules, we concluded that the optimized SERS substrate performed with both high sensitivity and large-area uniformity. A detection limit as low as 10¹⁴M and an enhancement factor (EF) up to 10⁸ were obtained. Furthermore, the wafer scale large-area uniformity and reproducibility of the SERS signal were confirmed with small standard deviation,e.g., 8% at 1076 cm¹and 9% at 1593 cm¹. The generic approach presented here can be extended to create a new class of highly sensitive SERS sensors with large-area output, and may play an important role in device design and the corresponding diverse chemical and biological detection tasks.

Introduction

Surface-enhanced Raman scattering (SERS), one of the most powerful and promising techniques, has been widely used in a variety of applications due to its high sensitivity, excellent selectivity, and rapid detection capability.¹ SERS effects are usually found in adjacent noble metallic nanostructures that are called“hot spots”, where electromagneticelds are largely enhanced. Recent studies have shown that“hot spots”have a large contribution to enormous Raman signal enhancement,^2,3 allowing rapid detection of Raman signals and highly sensitive identication of single molecules.^4,5

Numerous approaches have been widely investigated for the design of diverse noble metallic nanostructures as SERS

substrates, which can be classied into wet and dry methods or bottom-up and top-down methods. For the wet methods, substrates based on metal colloids stabilizations have been reported to achieve a SERS enhancement with a high sensi- tivity.^6,7In order to realize a stable solid SERS substrate with large-area uniformity, patterning of“hot spots”viadry methods had been developed, such as nanospheres (NSs) lithography,^8,9 reactive ion etching (RIE),¹⁰electron-beam lithography (EBL),¹¹ and nano-imprint lithography.¹² Typically, the top-down methods are time-consuming with a high fabrication cost.

Besides, to control the feature size of“hot spots”into the sub-10 nm regime to enhance SERS signal, approaches based on the bottom-up methods have been extensively exploited. Ordered arrays of Ag nanoparticles (NPs) had been grown in porous anodic aluminum oxide (AAO) templates with tunable sub-10 nm inter-particle gaps.¹³ Self-assembling of polystyrene (PS) spheres is a feasible choice instead of AAO with a pre-etching, which has a good control over the inter-particle gaps.^14–16 However, the PS spheres are usually several hundred nanome- ters in diameter, which reduces the density of“hot spots”under the Raman laser spot resulting in a limited sensitivity of SERS substrate.

Recently, substrate-supported metallic nanoparticle arrays have been fabricated over a large area by a solid-state dewetting, which attracted considerable interest for applications in solar cells, plasmonics and chemical etching.^17–19In particular, SERS substrate is one of the most important applications of substrate-supported metallic nanoparticle arrays with large-

aInstitute for Advanced Materials, Laboratory of Quantum Engineering and Quantum Materials, South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou 510006, People's Republic of China. E-mail: zzhang@scnu.

edu.cn

bElectronic Paper Display Institute, South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou 510006, People's Republic of China

cState Key Laboratory of Optoelectronic Materials and Technologies, Guangdong Province Key Laboratory of Display Material and Technology, School of Physics and Engineering, Sun Yat-sen University, Guangzhou 510275, People's Republic of China

dMax Plank Institute of Microstructure Physics, Weinberg 2, Halle 06120, Germany

eLaboratory of Solid State Microstructures, Nanjing University, Nanjing 210093, China. E-mail: liujm@nju.edu.cn

†Electronic supplementary information (ESI) available: Fig. S1–S9, calculation of enhancement factor (EF). See DOI: 10.1039/c5ra00512d

Cite this:RSC Adv., 2015,5, 19229

Received 10th January 2015 Accepted 10th February 2015 DOI: 10.1039/c5ra00512d www.rsc.org/advances

RSC Advances

PAPER

area uniformity.²⁰For example, one feasible SERS substrate of double-layer stacked Au/Al2O3@Au nanosphere arrays was realized on Si wafer.²¹ However, applying a simple annealing process, the inter-particle spacing of nanoparticles were too large to work efficiently as “hot spots”. Both theoretical and experimental data demonstrated that an inter-particle gap of less than 5 nm was essential for strong enhancement of Raman signals.^22,23Therefore, to meet the requirements of high quality SERS substrate, much effort need to be put on tailoring the inter-particle spacing of substrate-supported metallic nano- particle arrays.

In this paper, we present a bottom-up method to realize Ag/a- Si@Ag nanosphere arrays on Si as a SERS substrate with high sensitivity and large-area uniformity.Viaa solid-state dewetting, high density Ag nanoparticle arrays were formed on Si substrate with large-area uniformity. By a chemical vapor deposition (CVD), a-Si grew homogeneously on the surfaces of Ag NPs and Si substrate in between. With a controlled thickness of a-Si and Ag out-layer coating, we demonstrated that the inter-particle spacing of the Ag/a-Si@Ag nanosphere arrays could be reduced to several nanometers to work as a SERS substrate.

Through a study on their optical properties and SERS activities, the self-assembly bottom-up growth of Ag/a-Si@Ag nanosphere arrays as a SRES substrate exhibits high sensitivity of 10¹⁴M and large-area uniformity.

Experimental

Sample preparation

The n-doped Si(111) wafers were prepared by a standard RCA cleaning, and then dipped into 5% hydrouoric acid to have a silicon oxide removed and obtain a hydrogen-terminated Si surface. A thin Aglm with a 15 nm thickness was deposited by a thermal evaporation process at a growth rate of 0.1A s˚ ¹. Subsequently, the substrate was loaded to a LP-CVD system (rstnano ET-3000 EXT, CVD Equipment Corp.), and was heated up to 540C for 30 min at 10 Torr pressure to form Ag nano- particle arrays. The Ag/a-Si NSs were synthesized by a CVD process with diluted silane (5% SiH₄in H₂) as gas precursor and 99.999% pure H₂as carrier gas. The pressure was set to be 10 Torr with axed massow of 40 sccm H2, 20 sccm SiH₄. The growth time of Si layer was 30, 60, 90 min, respectively. Finally, an out-layer of Ag with thickness from 5 nm, 10 nm, 15 nm, 20 nm, 25 nm to 30 nm was coated onto the Ag/a-Si NSs to form a SERS substrate. The nanostructures were characterized by SEM (ZEISS Ultra 55), TEM (JEOL JEM-2100HR), and AFM (Cypher™, Asylum Research). Ultraviolet-visible (UV-vis) mirror reection spectra were obtained by a SpectraSuite (Ocean Optics HR4000).

SERS detection

For the measurement of SERS, all the vibration experiments were carried out with 633 nm excitation laser lines. The inelastic scattered radiation was collected on a microscopy Raman spectrometer (Renishaw 42K864) with a CCD detector and an optical microscope using a 50 microscope objective with a numerical aperture (NA) value of 0.50. Regarding the SERS

measurements, the analyte molecules p-Tc at different concentrations, ranging from 0.4 mM to 10¹⁴M were prepared in ethanol solutions. For the 0.4 mM the substrates were dipped into a solution of molecules dissolved in ethanol for 1 h, then washed in ethanol to remove excess molecules and dried in the air. For the measurements at lower concentrations (from 10¹⁰ M to 10¹⁴M), the substrates were dipped into solutions for 1 hour, then were immersed into an de-ionized water for 5 hours to remove the adsorbed ethanol that attached to the metal surface, followed by drying in the air. All the samples were soaked into identical bottles containingp-Tc (from 410⁴M to 110¹⁴M, 50 ml).

Results and discussion

Fabrication procedures

The fabrication procedures of Ag/a-Si@Ag nanosphere arrays are illustrated schematically in Fig. 1. First, a thin Aglm with a thickness of 15 nm was deposited on Si(111) substrateviaa high vacuum thermal evaporation (Fig. 1a). Ag nanoparticle arrays were formed by an annealing at 540C in the CVD chamber (Fig. 1b). In general, with a lower annealing temperature and a shorter annealing time, a higher density of metal nanoparticle arrays was achieved.²⁴

Aer annealing, a reactive gas precursor of SiH4(5%) mixed with a carrier gas of H₂wereuxed into the CVD chamber. With SiH₄decomposed on the hot surfaces, Si atoms homogeneously deposited on surfaces of Ag NPs and Si substrate in between.

Depictured in Fig. 1c, a-Si layer covered on the Si substrate homogeneously with embedded Ag NPs. The CVD growth of a-Si layer further reduced the inter-particle spacing of Ag NPs.

Meanwhile, the continuously grown a-Si immobilized the Ag NPs from further agglomeration. Then, an outer layer of Ag covered onto the a-Si by thermal evaporation to form Ag/a- Si@Ag nanosphere arrays (Fig. 1d), which can work as SERS substrate.p-Thiocresol (p-Tc), a kind of nonuorescent mole- cule, was trapped on the gaps of Ag/a-Si@Ag nanosphere arrays by chemisorption of molecules dissolved ethanol solution (as illustrated in Fig. 1e). In terms of SERS signals, the excitation laser was used to interact with the trappedp-Tc molecules to characterize the Ag/a-Si@Ag nanosphere arrays as a SERS substrate (Fig. 1f).

Characterization of Ag/a-Si@Ag nanosphere arrays

In order to maximize the density of“hot spots”, morphologies of tunable growth of Ag/a-Si@Ag nanosphere arrays on Si substrate are shown in scanning electron microscopy (SEM) images in Fig. 2. In top view (Fig. 2a) and 75-tilt view (Fig. 2b) SEM images, we conrmed that, the 15 nm solid-state Aglm on Si had dewetted into Ag nanoparticle arrays of high density and large-area uniformity. Inter-particle spacing between the Ag NPs was tunable with different annealing time under a given temperature (Fig. S1†). The inset of Fig. 2b illustrates the size distribution of about 550 Ag NPS, and the mean diameter is about 73 nm. However, aer a direct heating up to 540C for 30 min without annealing, most of the inter-particle spacing was

19230|RSC Adv., 2015,5, 19229–19235 This journal is © The Royal Society of Chemistry 2015

RSC Advances Paper

larger than 100 nm. In order to control the inter-particle spacing into several nanometer regime as“hot spots”for an effective SERS substrate, an a-Si layer was grown on Ag NPsviaa CVD process. Fig. 2c displays a SEM image of Ag/a-Si nanosphere arrays aer a 90 min CVD growth of Si. The Ag/a-Si nanosphere arrays on Si substrate are closely packed, and most of the gaps between adjacent Ag/a-Si nanospheres shrink to sub-10 nm. The cores of Ag/a-Si nanospheres with a bright contrast (under a high electron acceleration voltage of 15 kV) in Fig. 2c represent the Ag NPs embedded in the a-Si layer of about 50 nm thick with a dark contrast. The inset shows Raman spectra of Si substrate before (red line) and aer (blue line) the CVD growth of Si. In the Raman spectra, the sharp peak at 521 cm¹indicates the crystalline Si(111) substrate. Apart from the strong band at 521

cm¹, the additional board band at 480 cm¹ represents the growth of amorphous a-Si.²⁵Fig. 2d shows a 75-tilt view SEM image of Ag/a-Si nanosphere arrays. The inset of Fig. 2d illus- trates the size distribution of about 350 Ag/a-Si NSs with a mean diameter of about 170 nm, from which the mean size of Ag nanoparticle core was estimated to be 70 nm in diameter without further agglomeration during Si growth. Top view (Fig. 2e) and 75-tilt view (Fig. 2f) SEM images illustrate Ag/a- Si@Ag nanosphere arrays as a SERS substrate aer a 15 nm thick Ag out-layer coating. The Ag coating thickness was opti- mized to achieve the strongest SERS signals, which resulted to be 15 nm at a laser excitation wavelength of 633 nm. The inset of Fig. 2f illustrates the size distribution histograms of the Ag/a- Si@Ag nanosphere arrays. Aer the Ag out-layer coating, there Fig. 1 Schematic diagrams to illustrate the fabrication procedures of Ag/a-Si@Ag nanosphere arrays on Si substrate as a SERS substrate to detect probed molecules, such asp-thiocresol (p-Tc).

Fig. 2 Top-view SEM image (a) and 75tilted-view SEM image (b) of Ag nanoparticles (NPs), inset shows the size distribution histogram of diameters of Ag NPs. (c) Top-view SEM image of Ag/a-Si nanosphere arrays, inset shows Raman shifts of Si substrate before (red line) and after (green line) CVD process. (d) 75tilted-view SEM image of Ag/a-Si nanosphere arrays with a 90 min Si CVD growth and inset of (d) is the corresponding size distribution histogram. (e) Top-view SEM image Ag/a-Si@Ag nanosphere arrays as SERS substrate. (f) 75tilted-view SEM image Ag/a-Si@Ag nanosphere arrays with a 15 nm thick Ag out-layer, inset shows the corresponding size distribution histogram of diameters of Ag/a-Si@Ag NSs.

exists a systematic increase in the size of Ag/a-Si@Ag nano- spheres with the mean diameter of 190 nm, which is in accord with the thickness of Ag coating shown in Fig. 2e.

Crystallographic characterizations are shown in trans- mission electron microscopy (TEM) images of Fig. 3. Cross- sectional nanostructure of Ag/a-Si nanosphere arrays grown on Si(111) substrate is shown in Fig. 3a. Aer the CVD growth, an a-Si layer with a uniform thickness of 50 nm covered the surfaces of both Ag NPs and Si substrate in between. The gaps between Ag/a-Si nanospheres (NSs) on Si substrate are decided by the thicknesses of the a-Si layer, which could be adjusted by different growth time at a certain growth temperature (see Fig. S3†). Ane-grained Aglm was deposited onto the surfaces of Ag/a-Si nanosphere arrays, as shown in the cross-sectional image of Fig. 3b. From the magnied cross-sectional TEM image of an individual Ag/a-Si@Ag nanosphere (shown in Fig. 3c), we measured that the thickness of Aglm was about 15 nm. Moreover, high resolution TEM images (Fig. 3d and e) were acquired from different regions marked by the red dotted line boxes plotted in Fig. 3c. The single crystalline nature of Si(111) substrate was conrmed by the FFT pattern of Si substrate (inset of Fig. 3d). And the a-Si was grown on the Si substrate covered by a thin Si oxides layer with a thickness around 1 nm, which is consistent with the Raman peak at 480 cm¹observed in Fig. 2c.

Upon a closer inspection of Fig. 3e at the interface of Ag and a- Si, the lined fringes indicated the crystalline nature of Ag NPs aer the CVD growth, which is further conrmed by the selected area electron diffraction (SAED) pattern taken from Ag/a-Si NSs being a face-centered cubic (fcc) crystal structure (see Fig. S4†).

Optical response

Optical properties of nanostructures at different stages: (i) Ag

lm deposition; (ii) dewetted Ag NPs; (iii) Ag/a-Si nanosphere arrays were characterized by a UV-vis reection spectroscopy in Fig. 4a. The Si substrate with the continuous Ag lm has a higher reectance than the one with Ag nanoparticle arrays aer dewetting. Moreover, the Ag nanoparticle arrays on Si exhibit a distinct plasmonic reection trough at 450 nm, which is consistent with literature.²⁶Aer the Si growth, the reection trough red-shis to 617 nm related to the nanostructure of a-Si layer embedding Ag NPs. Furthermore, UV-vis reection spectra of Ag/a-Si@Ag nanosphere arrays with different Ag out-layer coating thicknesses are displayed in offset y values form (shown in Fig. 4b). With the Ag coating thickness increased from 5 nm to 10 nm, the corresponding UV-vis spectra exhibit an obvious red shifrom 622 nm to 643 nm. The red shiis mainly related to the increase of the size of discrete small Ag NPs composing the Ag out-layer,¹⁸which was observed in top- view SEM images (see Fig. S5a and b†). The reection spec- trum of Ag/a-Si@Ag nanosphere arrays with a 15 nm thick Ag coating has a blue shito 627 nm, which may be attributed to transformation of the discrete Ag NPs into connected nanois- lands. Typically, as the wavelength of the excitation laser in Raman being in accordance with the position of reection trough, Raman signals were found to be more obvious.²⁷Since the wavelength of the excitation laser in our Raman character- izations is 633 nm, the blue shito 627 nm with a 15 nm thick Ag coating has a better coupling between the incident light and surface plasmonic polarizations, resulting in stronger Raman signals. However, with increasing the Ag coating thickness, the reection trough tends to disappear and would give weak Raman signals. Therefore, for Ag/a-Si@Ag nanosphere arrays on Si as a SERS substrate, the Ag coating with a thickness of 15 nm was conrmed to be optimized with the 633 nm excitation wavelength.

SERS characterization

To study the direct inuence of Ag coating thickness on the SERS performance, aer the chemisorption in a 0.4 mMp-Tc solution, Raman spectra of Ag/a-Si@Ag nanosphere arrays with

Fig. 3 Cross-sectional TEM images of (a) Ag/a-Si nanosphere arrays and (b) Ag/a-Si@Ag nanosphere arrays with Agfilm coating both on Si(111) substrates, the a-Si layer is about 50 nm thick. (c) Magnified cross-sectional TEM image of a single Ag/a-Si@Ag nanosphere on the Si substrate, the thickness of Ag coating is about 15 nm. (d) HRTEM image taken from the selected area marked in (c) of Ag/a-Si@Ag nanosphere with the corresponding FFT pattern as inset, which illus- trates the amorphous a-Si growth on the single-crystalline Si(111) substrate with a thin SiO2interface. (e) HRTEM image taken from the selected area marked in (c) with the crystalline Ag core surrounded by the a-Si shell.

Fig. 4 (a) UV-vis reflection spectra with different fabrication stages of Ag/a-Si@Ag nanosphere arrays on Si substrate: (i) Agfilm deposition, (ii) solid-state dewetting of Ag NPs, (iii) Ag/a-Si nanosphere arrays formation. (b) Reflection spectra of Ag/a-Si@Ag nanosphere arrays on Si substrate with the Ag out-layer coating thicknesses varying from 5 nm to 25 nm.

19232 |RSC Adv., 2015,5, 19229–19235 This journal is © The Royal Society of Chemistry 2015

RSC Advances Paper

different Ag coating thicknesses are compared. All the spectra show clear Raman peaks ofp-Tc (see Fig. S6a†). The intensities of Raman peaks at 1076 cm¹with different Ag coating thick- nesses are shown statistically in the bar chart of Fig. 5a, in which the strongest Raman signals are observed on the SERS substrate with the 15 nm thick Ag coating. However, the rough nanostructures of the Aglm coating on a plane Si substrate with the same depositing conditions also have a SERS effect. In order to identify that the SERS effect is mainly from the Ag/a- Si@Ag nanosphere arrays, Raman spectra from the Ag/a- Si@Ag nanosphere arrays are compared with the ones obtained on plane Si substrates with the same Ag coating thicknesses (see Fig. S6b†). It is evident that the SERS effect from the Aglm coating on a-Si layer is negligible. For a prac- tical application, we compares the average Raman spectra intensity of analyte p-Tc chemisorbed on the Ag/a-Si@Ag nanosphere arrays with that obtained on a commercial Klarite SERS substrate. As shown in Fig. 5b, the strongest Raman peaks at 1076 and 1593 cm¹belonging top-Tc are observed on both SERS substrates.²⁸The peak at 1076 cm¹is due to a combi- nation of phenyl ring-breathing mode, CH in-plane bending, and CS stretching, while the peak at 1593 cm¹can be assigned to phenyl stretching motion.^29,30With the optimized thickness of Ag out-layer coating, the intensity of Raman peak at 1076 cm¹ from the Ag/a-Si@Ag nanosphere arrays is 15 times stronger than the one from the commercial SERS substrate.

In order to conrm the Ag/a-Si@Ag nanosphere arrays on Si as a SERS substrate has a high sensitivity, Raman spectra from

the SERS substrates with differentp-TC concentrations (from 10¹⁰to 10¹⁴M) are shown in Fig. 5c. With thep-Tc concen- tration decreasing, typical Raman peaks ofp-TC can be clearly identied down to ap-Tc concentration of 10¹³M. Even with a 10¹⁴M concentration, the Raman peaks at 623 cm¹and 1593 cm¹are still distinguished from background. Compared with the double-layer stacked Au/Al₂O₃@Au nanosphere structures enabling a sensitivity down to 10⁹ M,²¹ obviously, the SERS substrate based on the Ag/a-Si@Ag nanosphere arrays has a much higher sensitivity. However, on a plane Si substrate coated with Ag lm of the identical thickness, the detection limit of p-Tc is only 10⁶ M (see Fig. S7†). Therefore, the observed high sensitivity of 10¹⁴M is attributed to the Ag/a- Si@Ag nanosphere arrays. Due to a coupling of the localized surface plasmon resonances (LSPR), the electricaleld of the adjacent nanospheres could be enhanced by more than 10⁸ folds.^31,32Thenite difference time-domain (FDTD) method was used to simulate the near-led electricaleld distribution in a model of two Ag/a-Si@Ag NSs with different inter-sphere gaps (from 5 nm to 20 nm). The calculation conrmed that the coupling of LSPR was generated in the vicinity of the gap region and decayed rapidly outside the gap (see Fig. S8†).

In order to estimate the enhancement factor (EF) of the Ag/a- Si@Ag nanosphere arrays on Si as a SERS substrate, two methods were carried out.^33,34One estimation was based on the FDTD simulation, and the total SERS EF was calculated to be 3.11 10⁷ (see corresponding calculations in the ESI with Fig. S8 and S9†). Since this value varies signicantly depending on the initial assumptions, we compared with the other esti- mation, which came directly from the comparison of the Raman spectra ofp-Tc obtained on the Ag/a-Si@Ag nanosphere arrays with the one on a planar SERS substrate (as the reference).

Considering the formula of EF¼ (ISERS/Iref)/(CSERS/Cref), CSERS

andC_refcorrespond to the concentrations of thep-Tc molecules in the SERS (10¹³M) and the reference (10⁶M), respectively.

I_SERSandI_refdenote the corresponding average intensities of the 1076 cm¹ Raman peak (shown in Fig. 5d), respectively.

Therefore, the total SERS EF was estimated to be 2.2210⁸. Fig. 6a illustrates 2.5 cm5 cm cut Si wafers with the growth of Ag/a-Si nanosphere arrays and with the growth of Ag/a-Si@Ag nanosphere arrays, respectively. Aer Aglm coating, the wafer changed colour from blue (up) to grey-blue (down). The large- area uniformity of Ag/a-Si@Ag nanosphere arrays on Si was further supported by the atomic force micrograph (AFM) char- acterization. Fig. 6b displays AFM topography of a randomly selected area on the SERS substrate. A uniform distribution of Ag/a-Si@Ag NSs is seen over a 10mm10mm areas on the Si substrate. Besides the large-area uniformity in nanostructure, the good reproducibility of SERS signals across the wafer was also tested. The point by point scanning mode was carried out using a laser spot diameter of 1.54mm (50 objective lens with NA¼ 0.5) and a step-size of 100mm (inset in Fig. 6a). Fig. 6c demonstrates the 3D waterfall plot of enhanced Raman spectra obtained from 10 locations (along arrow marked in the inset of Fig. 6a), displaying almost the same intensity for each charac- teristicp-Tc peak. Raman intensities at 1076 cm¹peak from 10 10 locations (inset in Fig. 6a) were then plotted to a planar Fig. 5 (a) Raman intensities ofp-Tc at 1076 cm¹peak with different

Ag coating thicknesses (5 nm to 30 nm). (b) Comparison of Raman spectra ofp-Tc chemisorbed on the Ag/a-Si@Ag SERS substrate with the commercial SERS substrate and bare Ag/a-Si NSs. (c) Raman spectra of the Ag/a-Si@Ag SERS substrate withp-Tc concentrations ranging from 10¹⁰M to 10¹⁴M, and a plane SERS substrate withp-Tc concentration of 10⁶ M. (d) An enlarged view of Raman spectra obtained from Ag/a-Si@Ag SERS substrate (10¹³Mp-Tc) and from the plane SERS substrate (10⁶Mp-Tc) at 1076 cm¹peak.

Raman map of enhancement. The Raman map in Fig. 6d sug- gested that most of the Raman intensities of the 1076 cm¹ peak are in 6800–7800 a.u. range (blue color). According to the statistical analysis of Raman peak at 1076 cm¹and 1593 cm¹ (Fig. S10†), the calculated relative standard deviation (RSD) of the Raman intensity is 8% and 9%, which prove strongly the reproducibility of the SERS substrate.³⁵The RSD less than 10%

demonstrated that the Ag/a-Si@Ag nanosphere arrays on Si was suitable as a highly uniform SERS substrate.³⁶

Conclusions

For the rst time, we demonstrated growth of Ag/a-Si@Ag nanosphere arrays on a Si substrate by a bottom-up method to work as a high-performance SERS substrate. With the help of the solid-state dewetting, high density Ag nanoparticle arrays were formed on Si substrate with large-area uniformity. By the adjustments of thicknesses of a-Si and Ag out-layer coating, the SERS properties can be optimized with the inter-particle spacing of the Ag/a-Si@Ag NSs limited to several nanometers.

Based on the analysis of optical properties and SERS activities to p-Tc, we concluded that, the optimized SERS substrate per- formed with a high sensitivity with detection limit as low as 10¹⁴M and an enhancement factor up to 10⁸. Furthermore, a wafer scale large-area uniformity and reproducibility of SERS signal were conrmed with small standard deviation,e.g., 8% at 1076 cm¹and 9% at 1593 cm¹. The method present here can be extended to create a new class of highly sensitive SERS sensor

with large-area output, and may play an important role in device design and the corresponding diverse chemical and biological detection.

Acknowledgements

We acknowledge partial support by thenancial support by the Natural Science Foundation of China (Grant no. 51202072), the State Key Program for Basic Researches of China (Grant no.

2015CB921202), the International Science & Technology Coop- eration Platform Program of Guangzhou (Grant no.

2014J4500016), the Project for Guangdong Province Universities and Colleges Pearl River Scholar Funded Scheme (2014), the Program for Changjiang Scholars and Innovative Research Team in University (Grant no. IRTI1243), the European NODE project 015783 and by the DFG project GO 704/5-1. The authors would also like to acknowledge Mr Guohui Lin for preparation of SEM samples.

Notes and references

1 J. E. M. Matthew, J. Banholzer, L. Qin and C. A. Mirkin,Chem.

Soc. Rev., 2008,37, 885.

2 Y. Fang, N.-H. Seong and D. D. Dlott,Science, 2008,321, 388.

3 F. J. Garc´ıa-Vidal and J. B. Pendry,Phys. Rev. Lett., 1996,77, 1163.

4 T. Haibin, M. Guowen, H. Qing, Z. Zhuo, H. Zhulin and Z. Chuhong,Adv. Funct. Mater., 2012,22, 1.

5 J. A. Dieringer, R. B. Lettan, K. A. Scheidt and R. P. Van Duyne,J. Am. Chem. Soc., 2007,129, 16249.

6 S. Nie and S. R. Emory,Science, 1997,275, 1102.

7 Z. Liu, Z. Yang, B. Peng, C. Cao, C. Zhang, H. You, Q. Xiong, Z. Li and J. Fang,Adv. Mater., 2014,26, 2431.

8 Y. Shikuan, C. Weiping, K. Lingce and L. Yong,Adv. Funct.

Mater., 2010,20, 2527.

9 J. A. Huang, Y. Q. Zhao, X. J. Zhang, L. F. He, T. L. Wong, Y. S. Chui, W. J. Zhang and S. T. Lee,Nano Lett., 2013,13, 5039.

10 C. Qian, C. Ni, W. Yu, W. Wu, H. Mao, Y. Wang and J. Xu, Small, 2011,7, 1800.

11 M. Khorasaninejad, J. Walia and S. S. Saini,Nanotechnology, 2012,23, 275706.

12 R. Alvarez-Puebla, B. Cui, J. P. Bravo-Vasquez, T. Veres and H. Fenniri,J. Phys. Chem. C, 2007,111, 6720.

13 H. H. Wang, C. Y. Liu, S. B. Wu, N. W. Liu, C. Y. Peng, T. H. Chan, C. F. Hsu, J. K. Wang and Y. L. Wang, Adv.

Mater., 2006,18, 491.

14 S. Kim, L. Piao, D. Han, B. J. Kim and T. D. Chung, Adv.

Mater., 2013,25, 2056.

15 F. Cosmin and A. Simion,J. Phys. Chem. C, 2010,114, 11717.

16 W. Chunxu, R. Weidong, J. Nan, J. Wei, L. Sa, Z. Chun and Z. Bing,J. Phys. Chem. C, 2010,114, 2886.

17 M. C. G¨unendi,˙I. Tanyeli, G. B. Akg¨uç, A. Bek, R. Turan and O. G¨ulseren,Opt. Express, 2013,21, 18344.

18 B. T. Alexander, C. Lev, K. Tanya, A. B. Tatyana, H. Gilad, V. Alexander and R. Israel,J. Phys. Chem. C, 2011,115, 24642.

Fig. 6 (a) Color change of the wafer-scale (2.55 cm cut) samples with the growth of Ag/a-Si nanosphere arrays and Ag/a-Si@Ag nanosphere arrays on Si substrates, respectively, inset shows an optical magnified view (50 objective lens with NA ¼0.5) of the red dot region with the point by point scanning mode (step-size of 100mm). (b) AFM topography of a randomly selected area on the Ag/a-Si@Ag SERS substrate with a uniform distribution of NSs over a 10mm10mm area on the Si substrate. (c) Raman spectra from 10 different spots along the arrow as shown in inset of (a) with steps of 100mm. (d) Planar Raman map of the intensities of Raman signals at 1076 cm¹on the Ag/a- Si@Ag SERS substrate of the inset area marked in (a).

19234|RSC Adv., 2015,5, 19229–19235 This journal is © The Royal Society of Chemistry 2015

RSC Advances Paper

19 Z. Huang, N. Geyer, P. Werner, J. Boor and U. Gosele,Adv.

Mater., 2011,23, 285.

20 F. Ruffino and M. Grimaldi,Appl. Surf. Sci., 2013,270, 697.

21 Z. S. Hu, Z. Liu, L. Li, B. G. Quan, Y. L. Li, J. J. Li and C. Z. Gu, Small, 2014,10, 3932.

22 M. M. Jeffrey, L. Shuzhou, K. A. Logan and C. S. George,J.

Phys. Chem. C, 2012,116, 1627.

23 M. Chirumamilla, A. Toma, A. Gopalakrishnan, G. Das, R. P. Zaccaria, R. Krahne, E. Rondanina, M. Leoncini, C. Liberale, F. De Angelis and E. Di Fabrizio,Adv. Mater., 2014,26, 2353.

24 C. M. M¨uller, F. C. Mornaghini and R. Spolenak, Nanotechnology, 2008,19, 485306.

25 M. Schade, O. Varlamova, J. Reif, H. Blumtritt, W. Erfurth and H. S. Leipner,Anal. Bioanal. Chem., 2010,36, 1905.

26 H. Dai, M. Li, Y. Li, H. Yu, F. Bai and X. Ren,Opt. Express, 2012,20, 9.

27 J. Ye, F. Wen, H. Sobhani, J. B. Lassiter, P. V. Dorpe, P. Nordlander, N. J. Halas, P. Nordlander and N. Halas, Nano Lett., 2012,12, 1660.

28 W. Li, P. H. Camargo, X. Lu and Y. Xia,Nano Lett., 2009,9, 485.

29 G. Sauer, G. Brehm and S. Schneider,J. Raman Spectrosc., 2004,35, 568.

30 O. Masatoshi, M. Naoki, Y. Katsumasa and U. Isamu,J. Phys.

Chem., 1994,98, 12702.

31 M. Hentschel, D. Dregely, R. Vogelgesang, H. Giessen and N. Liu,ACS Nano, 2011,5, 2042.

32 S. Kleinman, R. Frontiera, A. Henry, J. Dieringer and P. V. Duyne,Phys. Chem. Chem. Phys., 2013,15, 21.

33 E. Ru, M. Blackie, M. Meyer and E. Etchegoin,J. Phys. Chem.

C, 2007,111, 13794.

34 Z. Ming-Liang, F. Xia, Z. Hong-Wei, S. Ming-Wang, J. A. Zapien, W. Ning-Bew and L. Shuit-Tong,J. Phys. Chem.

C, 2010,114, 1969.

35 L. Hongyan, L. Zhipeng, W. Wenzhong, W. Youshi and X. Hongxing,Adv. Mater., 2009,21, 4614.

36 M. J. Natan,Faraday Discuss., 2006,132, 321.

Supporting information

Bottom-up growth of Ag/a-Si@Ag Arrays on Silicon as a Surface-Enhanced Raman Scatting Substrate with High Sensitivity and Large-area Uniformity

Liwei Liu,^a Mingliang Jin,^b Qingwei Zhou,^a Runze Zhan,^c Huanjun Chen,^c Xubing Lu,^a Xingsen Gao,^a Stephan Senz,^d Zhang Zhang,^a*

and Junming Liu,^e*

a Institute for Advanced Materials and Laboratory of Quantum Engineering and Quantum Materials, South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou 510006, People’s Republic of China.

b Electronic Paper Display Institute, South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou 510006, People’s Republic of China.

c State Key Laboratory of Optoelectronic Materials and Technologies, Guangdong Province Key Laboratory of Display Material and Technology, School of Physics and Engineering, Sun Yat-sen University, Guangzhou 510275, People’s Republic of China.

d Max Plank Institute of Microstructure Physics, Weinberg2, Halle06120, Germany.

e Laboratory of Solid State Microstructures, Nanjing University, Nanjing 210093, China

International Center for Materials Physics, Chinese Academy of Sciences, Shenyang 110016, China

*email: zzhang@scnu,edu.cn

Electronic Supplementary Material (ESI) for RSC Advances.

This journal is © The Royal Society of Chemistry 2015

2 um 2 um

Fig. S1 Low magnification SEM images of Ag NPs arrays on Si substrate to illustrate the solid-state dewetting behavior. The samples were annealed at 540 ⁰C with different annealing time: (a) 20 min, (b) 60 min. The Ag NPs were obtained through the dewetting of continuous Ag films with a 15 nm thickness. The substrates were heated up to 540 ⁰C in 30 min in the CVD chamber and annealed with a pressure of 10 Torr and an H2 mass flow of 40 sccm.

200 nm 200 nm

200 nm 200 nm

(a) (b)

(c) (d)

Fig. S2 Top-view and 75⁰ tilted-view SEM images of Ag/a-Si nanosphere arrays grown on Si substrate by a CVD with growth times:

(a), (b) 30 min, (c), (d) 60 min.

Fig. S3 TEM images of individual Ag/a-Si nanosphereS with different growth times of Si: (a), (b) 30 min, (c), (d) 60 min. The thickness of a-Si shell increased with the growth time.

Fig. S4 (a) TEM image of Ag/a-Si NSs after CVD process with a growth time of 30 min, which were separated from the substrate.

The average diameter of Ag NPs is 73 nm, and the a-Si layer was uniformly grown onto the Ag cores. (b) Selected-area electron diffraction (SAED) pattern from Ag/a-Si NSs showed a polycrystalline structure of Ag cores, with,{2 0 0},{2 2 0},{3 1 1} planes according to the JCPDS file number 04-0783. ^{1, 2}

200 nm 200 nm

200 nm 200 nm

Fig. S5 Top-view SEM images of the Ag/a-Si@Ag nanosphere arrays with different Ag out-layer coating thickness: (a) 5nm, (b) 10nm, (c) 15nm, (d) 20nm, (e) 25 nm, (f) 30 nm. With the increase of coating thickness, the Ag out-layer was from a serious of discrete NPs to a continuous film.

5 10 15 20 25 30

0 1k 2k 3k 4k 5k 6k 7k 8k 9k 10k

Intensity(a.u.)

Ag coating thickness (nm) Ag films on plane Si Ag films on Ag/Si NSs

400 600 800 1000 1200 1400 1600 1800 2000 0.0

5.0k 10.0k 15.0k 20.0k 25.0k 30.0k

30 nm 25 nm 20 nm 15 nm 10 nm

Intensity (a.u.)

Raman shift (cm-1)

5 nm

(a) (b)

Fig. S6 (a) SERS spectra of p-Tc molecules chemisorbed on the Ag/a-Si@ Ag nanosphere arrays with different Ag coating thickness (5nm, 10nm, 15nm, 20nm, 25nm, 30nm). (b) Plots of the Raman peak intensities of 1076 cm-1 with different Ag coating thicknesses, on SERS substrate (red line) and plain Si substrate (black line). The Raman intensity obtained on plane Si substrate is small compared with the ones on SERS substrate.

600 800 1000 1200 1400 1600 1800 2000 0

200 400 600 800 1000 1200

10-9 M 10-8 M 10-7 M 10-6 M

Raman intensity

Raman shift (cm-1)

10-5 M

1076 1593

400 800 1200 1600 2000

0.0 2.0k 4.0k 6.0k 8.0k 10.0k 12.0k 14.0k

Intensity (a.u.)

Raman shift (cm-1)

SERS : 10-13 M

1076 1593

(a) (b)

Fig. S7 (a)SERS spectra of p-Tc (chemisorbed at a concentration of 10^-13 M) on different positions of the sameAg/a-Si@Ag nanosphere arrays as SERS substrate. (b) Raman spectra of p-Tc (chemisorbed at concentrations from 10^-5 M to 10^-9 M) on a plane Si substrate with a 15 nm thick Ag film coating. The detection limit of the plane Si substrate is 10^-6 M. The laser wavelength is 633 nm and the acquisition time is 10 s.

2 4 6 8 10 12

0

-0.1 0.0 0.1 0.2 0.3

-0.04 0.00 0.04 0.08

Z(microns)

X(microns)

0 1 2 3 4

-0.1 0.0 0.1 0.2 0.3

-0.04 0.00 0.04 0.08

Z(microns)

X(microns)

0.40 1.2 2.42.8

0.8 1.61.8

-0.1 0.0 0.1 0.2 0.3

-0.04 0.00 0.04 0.08

Z(microns)

X(microns)

-0.1 0.0 0.1 0.2 0.3

0 2 4 6 8 10 12

E/E0

X(microns)

(a) (b)

(c) (d)

Fig. S8 Schemes of 2-D FDTD simulations illustrate the electric field distribution in a model of two adjacent Ag/a-Si@Ag NSs with different inter-sphere gaps: (a) 5 nm, (c) 10 nm, (d) 20 nm. (b) Electric field intensity distribution curve along the X axis corresponding to (a), calculated at an excitation wavelength of 633 nm.

r a

Ntrap

=

Molecule footprint of p-Tc Area of two caps

=

^4_f^rh

r= radius of Si nanosphere h= heights of the cap (red color)

(b)

Fig. S9 (a) A schematic drawing of our approach to estimate the number of analyte molecules trapped in the hot spots between two NSs. The hot spot region is assumed to comprised a cap (red area) on the surface of each sphere.³ The number of trapped molecules (Ntrap) was obtained by calculating the total surface area of the two caps. In our FDTD simulation, bottom radius of the cap is about 25 nm, thus h = 4 nm (where h is the height of the cap and r is the mean radius of the NSs). (b) Raman spectra obtained from the SERS substrate (red curve) and pure powder of p-Tc on a plane glass substrate (black curve). For clarity, the signal of pure powder of p-Tc has been multiplied by 10 marked in the curve.

(a) (b)

1076 cm^-1 1593 cm^-1

Fig. S10 Intensities of Raman peaks p-Tc molecules at 1076 cm^-1 (a) and 1593 cm^-1 (b) of the at 100 locations. The calculated relative standard deviation (RSD) of the Raman intensity is 8 % and 9 %, respectively.

Calculation of SERS EF

To estimate the Enhancement Factor (EF) of the SERS substrate, p-Tc solid films of known volume and weight was placed on a clean glass slide for the measurement of normal Raman spectrum. The SERS peaks of 1076 cm^-1 and 1593 cm^-1 correspond to the p- Tc Raman peaks of 1093 cm^-1 and 1595 cm^-1, respectively. EF of p-Tc on Ag/a-Si@Ag nanosphere arrays was calculated for the SERS peaks at 1076 cm^-1 and 1593 cm^-1.

The experimental Raman EF of the SERS substrate was estimated using the equation:

(Eq. S1)

SERS bulk bulk SERS

I N

EF  I N

Where ISERS denotes the Raman intensity of the p-Tc molecules under the condition of Ag/a-Si@Ag nanosphere arrays and Ibulk

represents the Raman intensity of the p-Tc molecules under non-SERS conditions. NSERS is the number of the p-Tc molecules contributing to the SERS signals, Nbulk is the number of the p-Tc molecules used for the non-SERS measurement. For the calculation of Nbulk, different methodologies have been reported to estimate it.^{4, 5} Here, we used pure p-Tc powder to get values of Nbulk. The number of molecules excited in the non-SERS substrate was determined by assuming the excitation volume, which had the shape of cylinder with circle diameter equal to the focused laser beam spot diameter and height equal to the effective probe depth (Hobj).The objective lens (50×, 0.6) was used, and the focus laser beam diameter is 1.54 μm.⁶ The Hobj was determined by changing the substrate sample stage out of the laser focus plane at 1μm increments and recording the Raman peak value at 1076 cm^-1 at each position.⁷ While integrating the maximum Raman intensities over all positions, the 1/2 power points were recorded, yielding Hobj =16.5 μm. The number of molecules Nbulk in the bulk sample inside the interaction volume (30.88 μm³) was calculated to be 1.50×10¹¹ molecules.

While focusing on the centre of the non SERS substrate, Ibulk was measured at the excitation wavelength of 633 nm. The average background intensity was subtracted from the Raman spectrum to obtain the real peak value. The Raman intensity (Ibulk) at 1076 cm^-1 and 1595 cm^-1 was calculated to 224 counts and 70 counts, respectively.

To measure the number of NSERS, p-Tc molecules chemisorbed on the surface of Ag/a-Si@Ag nanosphere arrays (after 1 hour incubation), the density of p-Tc molecules can be estimated to be sub-monolayer. Here the coverage of p-Tc molecule for a dense monolayer was estimated to be about 4 molecules/nm².⁸ Taking the topography of the Ag/a-Si@Ag NSs surface into consideration, as shown in SEM image, approximately 30 spheres were seen in the laser focus area. It is assumed that all the spheres in the laser spot have contributed to the SERS signals and the probe molecules are distributed evenly across the surface of spheres. The effective surface area underneath the laser was 1.36 μm². This gives an estimate of NSERS calculated to be 5.44×10⁶. Using identical conditions as for ISERS, a Raman intensity of 6834 counts at 1076 cm^-1 and 4641 counts at 1593 cm^-1 were measured. Hence, the mean EF came to 8.4×10⁵ and 1.83×10⁶, respectively.

If we assumed that it was only molecules located on the gaps between the Ag/a-Si@Ag NSs, i.e. the “hot spots” were the main factors that contributed to the Raman enhancement signals a higher enhancement factor is estimated. Finite difference time domain (FDTD) calculations had been used for further insight into the possible geometries of highly active sites. 2-D FDTD simulations of the electromagnetic fields distribution showed that the hot spot region was a cap on the surface of each nanosphere in the inter-particle region (red color), as illustrate in Figure S9. That is, just the probe molecules trapped in the hot spot regions contributed to the detected SERS signals. In our estimation, Ntrap is obtained by calculating the total surface area of the cap among average 30 NSs. Here we assume h = 5 nm (where h is the height of the cap and r = 85 nm is the radius of the spheres), and yield Ntrap = 3.20×10⁵. This number represents a theoretical maximum number of molecules. For the gap distance between two spheres exceeding 15 nm, no distinct Raman enhancement signals were observed. However, we neglected this. According to such an approach, the EFs of the hot spot were calculated to be: EF (1076 cm^-1) = 1.43×10⁷; EF (1593 cm^-1) = 3.11×10⁷.

2 N. Ahmad,S. Sharma,M. K. Alam,V. N. Singh,S. F. Shamsi,B. R. Mehta,A. Fatma, Colloids Surf.B: Biointerfaces 2010, 81, 81.

3 W. Li,P. H. Camargo,X. Lu,Y. Xia, Nano lett. 2009, 9, 485.

4 M. S. Schmidt,J. Hübner,A. Boisen, Adv. Mater. 2012, 24, 8.

5 H. Mao,W. Wu,D. She,G. Sun,P. Lv,J. Xu, Small 2014, 10, 127.

6 A.-P. Ramón, Phys. Chemi. Lett. 2012, 3, 857.

7 M. A. Khan,T. P. Hogan,B. Shanker, JNST 2009, 1,1.

8 C. Jianing,Y. Weisheng,D. Kimberly,D. Knut,H. Q. Xu,S. Lars,X. Hongxing, Appl. Phys. Lett. 2008, 92, 93.