Fabricating Silver Nanoparticles on Thin Silicon Nanowallsfor Highly Sensitive Surface-Enhanced Raman Scattering

Shu Ning Wen and Jiann Shieh+

Department of Materials Science and Engineering, National United University,1, Lienda, Miaoli 36003, Taiwan

Metal nanoparticles with nanoscale spacing are promising materials for the detection of single molecules through surface-enhanced Ramanscattering. To increase the sensitivity of nanoparticles through the use of a nanoscale substrate, we fabricated various Ag NP­decorated siliconnanowalls for the Raman spectroscopic detection of rhodamine 6G (R6G). The sensitivity of detection was affected by the nanowall depth andwas influenced by several parameters: the AgNO3 concentration for metal-assisted etching, the HF/H2O2 etching time for nanowall formation,and the Ag evaporation time for nanoparticle growth. For an approximately 400-nm-deep nanowall substrate having the optimal surface fillingratio and etching depth, we obtained an ultrahigh enhancement factor of 1.1 © 109 for the detection of R6G at a concentration of 10¹11M.[doi:10.2320/matertrans.M2014259]

(Received July 14, 2014; Accepted September 11, 2014; Published October 18, 2014)

Keywords: silver, surface-enhanced Raman scattering, nanowall, metal-assisted chemical etching

1. Introduction

Surface-enhanced Raman scattering (SERS) is an excellentmethod for detecting molecules at low concentration.1) As aresult of surface chemical enhancement and electromagneticeffects associated with surface plasmon resonance, especiallyfrom “hot spots” between nanoparticles (NPs), nanostruc-tured metal­assisted SERS can increase the Raman scatteringcross-sections and, thereby, improve sensitivity.2) Manymethods, including chemical reduction, electrochemicaldeposition, and vapor deposition,3) are available for thepreparation of metal NPs. On two-dimensional (2D)substrates, the Raman sensitivity can be enhanced by greaterthan 105 through the assistance of deposited NPs.4) Toimprove the sensitivity further, several three-dimensional(3D) nanostructures, including nanotips,5,6) nanotubes,7)

nanowires,8­10) nanowalls,11) nanoforests,12) and nanoca-nals,13) have been investigated in conjunction with variousnanofabrication processes. Among them, metal-assistedchemical etching (MACE) has attracted much attentionbecause it is a simple and inexpensive process that canproduce structures over large areas.14) Unfortunately, theetching rate is so high that thick nanostructures (several tensof micrometers) are usually formed, thereby limiting theSERS applications to microdevices. In addition, only limitedinvestigations have been made into the effects of micro-structures on the SERS sensitivity. In this present study, wesystematically investigated the preparation of thin-layernanostructures®varying the AgNO3 concentration, the HF/H2O2 etching time, and the Ag deposition time®to optimizethe sensitivity of SERS toward rhodamine 6G (R6G). Usingthermal evaporation to decorate Ag NPs on Si nanowalls,we found that there exists an optimal profile to obtain thehighest SERS sensitivity, confirmed through microscopyimages of the silicon nanowall structures. This optimalprofile, characterized by a silicon nanowall depth of only400 nm, can be used to detect low concentrations of R6Gwith ultrahigh sensitivity [enhancement factor (EF): >1011].

2. Experimental Procedure

(100) Silicon wafers having a resistance of 1.5­100ohm-cm (test grade) were washed sequentially withacetone, ethanol, and DI water and then immersed in asolution containing H2SO4 (98mass%) and H2O2 (30mass%), at a volume ratio of 4 : 1, at 90°C for 15min toremove any organic impurities and also to form a thinoxide layer on the surface. A buffer containing HF(5mass%) was then used to remove the oxide and tocreate a H-terminated surface. The samples were thenimmersed in a solution containing HF (5M) and AgNO3

(0.001­0.06M) to form Ag NPs on the wafer surface. Theunadsorbed Ag NPs that formed in solution were washedaway with DI water to slow down the etching rate and toimprove the uniformity in height.15) An etching solutioncomprising HF (5M) and H2O2 (0.1M) was applied for10­1800 s to etch down the silicon to form the siliconnanowalls with assistance from the Ag NPs. The residualAg NPs were removed using a HNO3 solution and then thesamples were washed with DI water and dried under aflow of N2.

For SERS measurements, Ag NPs were deposited on thenanowalls through thermal evaporation (KD-thermal system)for 30­300 s. Nanoparticles can be immobilized on Sisubstrates through the interactions between Ag atoms andSi wafer during deposition.16) The samples were placed in0.1M KCl for 60min to activate the surface for Ramanmeasurements. They were immersed in a 100-ml R6Gsolution (from 10¹7 to 10¹11M) for 1 h and then dried undera flow of N2 prior to SERS measurements.

The morphologies of the samples were characterizedthrough scanning electron microscopy (SEM) using aJEOL-6700F microscope. SERS measurements were per-formed using a micro-Raman system (Jobin Yvon-LabramHR) with excitation of light (wavelength: 532 nm) from aNd:YAG laser. The spot size was focused to approximately1 µm; the objective lens was 100©; the power was 3mW; andthe accumulating time was 1 s.

+Corresponding author, E-mail: jshieh@nuu.edu.tw

Materials Transactions, Vol. 55, No. 12 (2014) pp. 1800 to 1805©2014 The Japan Institute of Metals and Materials

3. Results and Discussions

Figure 1(a) displays a representative SEM image of asilicon wafer surface presenting Ag NPs reduced from0.02M AgNO3 and HF solution. As the AgNO3 concen-tration increased, these Ag NPs aggregated into a dendriteblock on the surface. Embedded Ag NPs were evident withinthe holes between the nanowalls after removing the non-adsorbed Ag NPs (most existed in the form of a dendriteblock) through washing with DI water and then etching for1min in HF/H2O2 solution (Fig. 1(b)). After using HNO3

solution to remove the Ag NPs, we obtained a fresh nanowallmorphology lacking any NPs (Fig. 1(c)). Interestingly, theholes between the nanowalls did not correspond exactly tothe distribution of Ag NPs, suggesting that the nanowallmorphology was not defined by the NP shape or distributionalone. In fact, the nanowall morphology beneath the NPswas evident prior to the HF/H2O2 etching step (Fig. 1(a)).We suspect that both the NPs and atomic-scale Ag speciesacted as catalysts to promote the etching of the underlyingsilicon.17) Figure 1(d) displays the cross-section of the Sinanowall after etching for 1min. The uniform heightindicates that the etching rate was nearly constant over thesilicon surface. A tilted-view SEM image at highermagnification (Fig. 1(e)) also clearly reveals the nanowallmorphology.

Figure 2 displays the variations in morphology obtainedafter varying both the AgNO3 concentration and the H2O2/HF etching time. At a low AgNO3 concentration (0.001M),nanoholes were formed initially. Upon extending the etchingtime, the surface roughness increased, implying that theetching process was determined by the Ag content and that aninsufficient number of Ag NPs might cause a rough surfaceas a result of nonuniform distribution. On the other hand, asthe concentration of AgNO3 increased, most the surface wascovered with Ag species, thereby decreasing the surfacefilling ratio (defined as the surface ratio of Si to air). Thenanowalls may have buckled at longer etching times becauseof the low surface filling ratio.

Because the Ag NPs obtained from the reduction processwere large (Figs. 1(a) and 1(b)), we removed them anddeposited smaller Ag NPs through a thermal evaporationprocess. First, we deposited Ag on a flat silicon wafer todetermine the deposition duration required for nanogapformation. Figures 3(a)­(c) display top-view SEM imagesof the samples deposited on flat silicon wafers for 30, 60, and300 s, respectively. As the deposition time increased, thediscontinuous islands aggregated into a continuous film(Fig. 3(c)). Figure 3d presents the gap width on the 60-ssample mapped using ImageJ software; it reveals that manyof the gaps were characterized by a distance of less than10 nm. These narrow gaps have been suggested to accountfor the formation of hot spots for SERS applications.18)

Accordingly, we deposited Ag NPs onto the Si nanowallsfor 30 or 60 s. Figures 4(a) and 4(b) display tilted-view SEMimages of the samples (AgNO3 concentration, 0.06M;etching time, 5min; evaporation time, 60 s) prepared beforeand after Ag decoration, respectively; in the latter, weobserve discrete NPs on the tops and sides of the nanowalls.Moreover, Fig. 4(c) reveals that the Ag NPs were alsodeposited within the pores and on the bottom of thenanowalls, demonstrating that the nanowalls were 3Dsubstrates that could support more NPs than could appearon the flat surface.

Next, we used R6G dye as a probe molecule to examinethe enhancement in sensitivity of Raman spectra arising fromSERS. We used the C­O stretching signal of R6G at611 cm¹1 as the characteristic peak (the strongest one in therange 550­1800 cm¹1 in this study) to calculate the EFs ofthe SERS sensitivity. We evaluated the EFs using theequation19)

EF ¼ ISERIREF

� NREF

NSER

� PREF

PSER

� tREFtSER

where I, N, P, and t are the Raman spectrum intensity, numberof molecules, laser power, and exposure time, respectively;SER and REF represent the values from the Ag/nanowallsamples and the reference sample of bare silicon, respec-

Fig. 1 SEM images of (a) Ag NPs reduced from AgNO3 onto a silicon wafer, (b) Ag NPs embedded in silicon nanowalls, (c) the siliconnanowalls obtained after removing the Ag NPs with HNO3, and (d) a cross-section of the nanowalls after etching for 1min; (e) anenlarged tilted view of the image in (d).

Fabricating Silver Nanoparticles on Thin Silicon Nanowalls for Highly Sensitive Surface-Enhanced Raman Scattering 1801

Fig. 2 Top-view SEM images of the samples obtained using various AgNO3 concentrations and HF/H2O2 etching times.

Fig. 3 (a)­(c) SEM images of Ag NPs deposited on silicon wafers for (a) 30, (b) 60, and (c) 300 s. (d) Distance map of the gaps betweenAg islands in the 60 s sample.

S. N. Wen and J. Shieh1802

tively. Because each R6G solution had the same volume(100mL), we used the ratio of the R6G concentrations,instead of the ratio of the number of molecules (NREF/NSER),to calculate the EFs. Figure 5(a) presents the 3D EFdistribution as a function of the AgNO3 concentration andthe HF/H2O2 etching time when using 10¹7M R6G solutionfor measurement on the Ag/nanowall sample; we used amuch higher concentration of R6G (10¹2M) for bare siliconas the reference because the Raman signal is very weak in the

absence of silver. The samples decorated with the Ag NPsdeposited for 30 and 60 s exhibited similar trends in terms oftheir sensitivities with respect to the AgNO3 concentrationand HF/H2O2 etching time; in general, the sensitivities of the60 s samples were higher than those of the 30 s samples,presumably because of the shorter distance, which providedstronger hot spots for SERS, between their NPs. TheEF increased upon increasing the AgNO3 concentration(Fig. 5(b)), suggesting that a greater content of AgNO3

Fig. 4 (a), (b) Tilted-view SEM images of the silicon nanowalls decorated (a) before and (b) after Ag NPs. (c) SEM image of thenanowalls with Ag NPs deposited entirely along the sidewalls.

0.000.01

0.020.03

0.040.05

0.06

1x107

2x107

3x107

4x107

5x107

0300

600900

12001500

1800

Enh

ance

men

t fac

tor

Etching

time

(s)

AgNO3 concentration (M)

(a)

0

1x107

2x107

3x107

4x107

5x107

60 s*

5min* 60 s 30 min*

5min

30 min 10 sec* 10 sec

Enh

ance

men

t fac

tor

AgNO3 concentration (M)

(b) etching time

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

0.05.0x106

1.0x107

1.5x107

2.0x107

2.5x107

3.0x107

3.5x107

4.0x107

4.5x107

5.0x107

5.5x107

6.0x107

0.06 M AgNO3; 60-s deposition of Ag 0.06 M AgNO3; 30-s deposition of Ag 0.02 M AgNO3; 60-s deposition of Ag 0.02 M AgNO3; 30-s deposition of Ag 0.005 M AgNO3; 60-s deposition of Ag 0.005 M AgNO3; 30-s deposition of Ag 0.001 M AgNO3; 60-s deposition of Ag 0.001 M AgNO3; 30-s deposition of Ag

Enh

ance

men

t fac

tor

Etching time (s)

(c)

0 100 200 300 1600 1800

0 5 10 15 20

0.0

5.0x106

1.0x107

1.5x107

2.0x107

2.5x107

0.001 M AgNO3; 60-s deposition of Ag 0.005 M AgNO3; 60-s deposition of Ag 0.02 M AgNO3; 60-s deposition of Ag 0.06 M AgNO3; 60-s deposition of Ag 0.001 M AgNO3; 30-s deposition of Ag 0.005 M AgNO3; 30-s deposition of Ag 0.02 M AgNO3; 30-s deposition of Ag 0.06 M AgNO3; 30-s deposition of Ag

Enh

ance

men

t fac

tor

Etching time (s)

(d)

Fig. 5 (a) Distribution of EFs for samples prepared at various AgNO3 concentrations and etching times. Black dots: 60 s deposition of Ag;gray dots: 30 s deposition of Ag. (b) Dependence of EF on AgNO3 concentration. An asterisk represents a sample prepared with anevaporation time of 60 s; the absence of an asterisk represents a 30 s evaporation time. (c) Dependence of EF on HF/H2O2 etching time;gray lines: 60 s deposition of Ag; black lines: 30 s deposition of Ag. (d) Enlarged view of Fig. (c) after a short etching time.

Fabricating Silver Nanoparticles on Thin Silicon Nanowalls for Highly Sensitive Surface-Enhanced Raman Scattering 1803

broadened the pores between the nanowalls and allowed thesubsequently evaporated Ag atoms to deposit readily withinthem. On the other hand, the EF reached its maximum value(4.57 © 107) upon increasing the HF/H2O2 etching time to60 s (Fig. 5(c)). This value is approximately three timeshigher than that of the system in which the Ag atoms weredeposited onto a flat surface (EF = 1.37 © 107); furtherincreases in the etching time did not lead to higher EFs.Figure 5(d) reveals that after a very short etching time (10 s)the EF was even less than that obtained from the flat wafer,indicating that at the onset of etching the number of depositedAg NPs on the rough surface was not higher than that on theflat surface. The existence of an optimal depth for SERS canbe explained by considering the trade-off between the depthand the surface filling ratio. Figure 6 displays the typicalvariation in depth and surface filling ratio plotted with respectto the etching time. We measured the depth from the cross-sectional SEM images, and obtained the surface filling ratioby analyzing the top-view SEM images with ImageJsoftware. We could control the depth with uniform heights,from several tens of nanometers to several micrometers, byvarying the duration of etching with HF/H2O2. The trade-offbetween depth and filling ratio determines the SERS EF.Typically, the EF arises from hot spots between the NPs.Therefore, we ascribe the variations in SERS sensitivity tothe different number of hot spots. For samples prepared withthe shortest etching time (10 s), the rough surface decreasedthe particle packing density; accordingly, the EF decreased.On the other hand, at longer etching times (between 5 and30min), although the sides of the nanowalls provided greaterareas for deposition of Ag NPs, the surface filling ratiosdecreased as a result of buckling between the highernanowalls (Fig. 2). Therefore, the overall number of NPswas not as high as that obtained at the optimal etching time(60 s); accordingly, we observed a lower SERS EF at longeretching times. Notably, in this study the optimal depth wasonly approximately 400 nm®for the sample prepared from0.06M AgNO3 and etched under the HF/H2O2 solution for1min; this depth, which is less than that obtained typicallythrough the MACE process, should benefit the developmentof flexible or thin-film SERS devices. SERS with highersensitivity would be expected if the high surface filling ratiocould be maintained for deeper nanowalls.

The high sensitivity of 400-nm-deep nanowalls could beused to detect lower concentrations of the dye R6G. Figure 7presents Raman spectra recorded with R6G concentrationsranging from 10¹7 to 10¹11M. The characteristic signals ofR6G were still clearly identifiable at 10¹11M, but with lowerRaman intensity. Interestingly, the Raman intensity did notdecrease proportionally upon decreasing the R6G concen-tration, such that we obtained a higher EF at lowerconcentration (EF = 1.1 © 109 at 10¹11M). We suspect thatthe opportunities for contact between the hot spots and R6Gmolecules were higher in the more-dilute R6G solutions, butthe greater number of molecules in the denser solutions werepresumably blocked by other molecules adhered to thenanowall surfaces, thereby resulting in lower EFs.

4. Conclusion

We have used metal-assisted chemical etching and thermalevaporation processes to prepare thin silicon nanowallsdecorated with Ag NPs to increase the EF for the detectionof R6G molecules on silicon wafers. SEM images indicatedthat we could obtain thin silicon nanowalls of highuniformity through MACE in this study by removing non-adsorbed Ag NPs prior to HF/H2O2 etching, allowing tinierAg NPs to be evaporated onto the tops and sides of thenanowalls. By controlling several experimental parameters®the AgNO3 concentration, the HF/H2O2 etching time, and theduration of thermal evaporation®we found that both thesurface filling ratio and the etching depth of the nanowallscould be adjusted for optimal SERS performance. Increasingthe AgNO3 concentration could improve the etchinguniformity over the surface, but higher nanowalls tended tobuckle together after longer etching times, thereby decreasing

10 100 10000

10

20

30S

urfa

ce fi

lling

rat

io (

%)

Etching time (sec)

0

1000

2000

3000

4000

Dep

th (

nm)

Fig. 6 Depth and surface filling ratio of nanowalls plotted with respect tothe HF/H2O2 etching time (logarithmic scale). These samples wereprepared using 0.06M AgNO3.

600 800 1000 1200 1400 1600 1800

0

4x104

9x104

1x105

0

230

460

690

0

88

176

264

Raman shift (cm-1)

10-7 M

Inte

nsity

10-9 M

10-11 M

Fig. 7 SERS spectra of R6G collected on the 400-nm-thick nanowallsample at various R6G concentrations.

S. N. Wen and J. Shieh1804

the surface filling ratio and leading to lower EFs. For a10¹7M solution of the dye R6G, the 400-nm-deep nanowallscould be applied to increase the SERS EF up to 4.57 © 107,approximately three times higher than that of the NPs on theflat surface. The Raman spectrum identified R6G clearly evenwhen we had decreased its concentration to 10¹11M. Thisreadily processed, highly sensitive, thin nanowall substrateappears to be an attractive material for further research intolow-cost, large-area SERS sensors on flexible substrates.

Acknowledgments

We thank the National Science Council, Taiwan, Republicof China, for financial support (contract NSC 101-2628-E-239-002-MY3). We also thank Professor Yi-Sheng Lai forhelp with the thermal evaporation process, and the Centerfor Micro/Nano Science and Technology at National ChengKung University for Raman measurement.

REFERENCES

1) B. Sharma, R. R. Frontiera, A. Henry, E. Ringe and R. P. Van Duyne:Mater. Today 15 (2012) 16­25.

2) H. Ko, S. Singamaneni and V. V. Tsukruk: Small 4 (2008) 1576­1599.3) G. Cao: Nanostructures and Nanomaterials: Synthesis, Properties, and

Applications, (Imperial College Press, London, 2004) pp. 51­109.4) M. J. Natan: Faraday Discuss. 132 (2006) 321­328.5) S. Chattopadhyay, H. C. Lo, C. H. Hsu, L. C. Chen and K. H. Chen:

Chem. Mater. 17 (2005) 553­559.6) K. Sun, G. Meng, Q. Huang, X. Zhao, C. Zhu, Z. Huang, Y. Qian, X.

Wang and X. Hu: J. Mater. Chem. C 1 (2013) 5015­5022.7) X. Hu, G. Meng, Q. Huang, W. Xu, F. Han, K. Sun, Q. Xu and Z.

Wang: Nanotechnology 23 (2012) 385705.8) B. Zhang, H. Wang, L. Lu, K. Ai, G. Zhang and X. Cheng: Adv. Funct.

Mater. 18 (2008) 2348­2355.9) M. Peng, J. Gao, P. Zhang, Y. Li, X. Sun and S. T. Lee: Chem. Mater.

23 (2011) 3296­3301.10) M. Peng, H. Xu and M. Shao: Appl. Phys. Lett. 104 (2014) 193103.11) S. Siddhanta, V. Thakur, C. Narayana and S. M. Shivaprasad: ACS

Appl. Mater. Interfaces 4 (2012) 5807­5812.12) M. L. Seol, S. J. Choi, D. J. Baek, T. J. Park, J. H. Ahn, S. Y. Lee and

Y. K. Choi: Nanotechnology 23 (2012) 095301.13) H. Ko and V. V. Tsukruk: Small 4 (2008) 1980­1984.14) Z. Huang, N. Geyer, P. Werner, J. de Boor and U. Gösele: Adv. Mater.

23 (2011) 285­308.15) M. L. Zhang, K. Q. Peng, X. Fan, J. S. Jie, R. Q. Zhang, S. T. Lee and

N. B. Wong: J. Phys. Chem. C 112 (2008) 4444­4450.16) M. Ohring:Materials Science of Thin Films, (Academic Press, London,

2002) pp. 95­144.17) S. C. Tseng, H. L. Chen, C. C. Yu, Y. S. Lai and H. W. Liu: Energy

Environ. Sci. 4 (2011) 5020­5027.18) 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. 18 (2006) 491­495.19) M. Zamuner, D. Talaga, F. Deiss, V. Guieu, A. Kuhn, P. Ugo and N.

Sojic: Adv. Funct. Mater. 19 (2009) 3129­3135.

Fabricating Silver Nanoparticles on Thin Silicon Nanowalls for Highly Sensitive Surface-Enhanced Raman Scattering 1805