Silver Nanostructures on Silicon Based on Galvanic Displacement Process

Albert Gutes, Ian Laboriante, Carlo Carraro, and Roya Maboudian*Department of Chemical Engineering, UniVersity of California, Berkeley, California 94720

ReceiVed: June 12, 2009; ReVised Manuscript ReceiVed: July 24, 2009

A simple and versatile process, based on a galvanic displacement reaction, is presented that provides controlledgrowth of different silver structures on silicon including thin films, nanocrystals, nanoparticles, and morecomplex structures termed nanodesert rose. Structure selection is achieved by choosing a suitable ratio ofAgF:KF in the plating solution, in the absence of any other additive, and by changing immersion times andprecursor concentrations. We demonstrate the usefulness of some of these nanostructures as reproduciblesurface-enhanced Raman spectroscopy substrates.

1. Introduction

The integration of metals with semiconductors plays a crucialrole in a number of technologies, ranging from integrated circuitsto micro-nanosystems technology, with implications in elec-tronic, optoelectronic, electromechanical, and sensing devices.In recent years, the deposition of metals on semiconductors bygalvanic displacement (GD) has received renewed interest, asit is a versatile process, well suited to yield films with highpurity and substrate adhesion and with substrate selectivity.1-4

In GD reactions, metal ions in the plating bath are reduced bythe substrate itself upon immersion, without external currentsources or reducing agents in the bath.

Controlled galvanic displacement processes have been em-ployed on silicon to deposit a variety of noble metals, such asgold, platinum, and copper, in thin film or nanoparticle forms.1

Silver deposition using galvanic displacement has been reportedon other substrates such as Ge, Al, or GaAs.5-7 Electrochemicalapproaches involving external current8 or complex depositionprocesses involving patterning have been carried out on silicon,9

but controlled and stable displacement on nonpatterned siliconsubstrates has not been reported. Galvanic displacement of silveron silicon, carried out most commonly in solutions containingsilver nitrate and hydrofluoric acid, tends to result in a fast anduncontrolled etching of the substrate, with formation of silverdendrites on the surface, as shown in Figure 1a. Dendriteremoval with HNO3 shows deep etching in the silicon substratewith the formation of Si nanowires arrays as shown in Figure1b,c, a phenomenon reported previously.10-12 We show in thispaper that this uncontrolled process can be changed and metaldeposition can be controlled in such a way that silver thin film,nanoparticles, nanocrystals, and other nanostructures can beobtained as desired. These silver structures can then be usedfor their respective technological applications, for example, infilms or nanoparticle forms as catalysts for silicon nanowiregrowth, while as crystalline nanostructures for surface enhance-ment Raman spectroscopy or metal-enhanced fluorescence.

2. Experimental Section

Silver Nanostructures Preparation. ⟨111⟩ silicon 5 × 5 mmchips were degreased by sonication in acetone and isopropanolfor 10 min and then rinsed with deionized water (18 MΩ) before

* Corresponding author: Tel +1 (510) 643-3489, Fax +1 (510) 642-4778, e-mail maboudia@berkeley.edu.

Figure 2. SEM images obtained for different plating concentrations:(a-c) [Ag+] ) 1 mM, [F-] ) 20 mM; (d-f) [Ag+] ) 20 mM, [F-] )20 mM. Deposition times: (a, d) 5 s, (b, e) 30 s, (c, f) 60 s. All scalebars refer to 1 µm.

Figure 1. (a) Silver dendrites formed via GD from 1 mM AgNO3 in9 M HF solution; (b) silicon nanowires arrays formed by silicon etching;(c) top view of the SiNW arrays.

J. Phys. Chem. C 2009, 113, 16939–16944 16939

10.1021/jp9055297 CCC: $40.75 2009 American Chemical SocietyPublished on Web 09/10/2009

drying in gentle N2. Native oxide layer was removed byimmersing the chips in concentrated HF for 1 min, rinsed inDI water, and dried in N2. Silver plating solution was preparedby dissolving the appropriate amount of AgF and KF in DI waterto the desired final Ag+ and F- concentrations. Silicon chipswere immersed right after native oxide etching and left insolution for the desired time. DI rinsing and N2 drying wereperformed after incubation.

Characterization of Ag Nanostructures. X-ray photoelec-tron spectroscopy (XPS) analysis was performed using anOmicrometer analyzer (EA 125) to confirm metallic Ag nature

of the formed structure. Scanning electron microscopy (SEM)images were taken using a field-effect mySEM microscope(Novel X) operated at 1 kV. Atomic force microscopy (AFM)operating in tapping mode (Digital Instruments Nanoscope IIIa)was used for roughness and film thickness measurements.

SERS Sample Incubation and Characterization. Silvernanostructures were immersed in a 5 mM BPE solution inmethanol for 24 h, then rinsed in methanol to remove physicallyadsorbed BPE, and dried in gentle N2 flow. XPS analysis wasperformed to estimate the surface coverage using C to N ratioas target parameter.

Figure 3. AFM images for different plating concentrations: (a-c) [Ag+] ) 1 mM, [F-] ) 20 mM; (d-f) [Ag+] ) 20 mM, [F-] ) 20 mM.Deposition times: (a, d) 5 s, (b, e) 30 s, (c, f) 60 s. In all cases, image area is 10 µm × 10 µm with the z-scale of 400 nm.

16940 J. Phys. Chem. C, Vol. 113, No. 39, 2009 Gutes et al.

Raman Measurement. Raman spectroscopy (JYHoriba La-bRAM) was performed in backscattering configuration with anexcitation line provided by a HeNe laser (632.8 nm wavelength,10 mW at the sample) through an Olympus BX41 100×confocal microscope (numerical aperture ) 0.8).

3. Results and Discussion

The process described in this paper is a simple and inexpen-sive dip-and-rinse galvanic displacement process that enablesthe synthesis of silver films, nanoparticles, nanocrystals, or morecomplex structures depending predictably on bath compositionand immersion times. Taking into account that the two half-cell reactions involved are

we note that for each single silicon atom that is oxidized fourelectrons have to be captured by four silver cations; thus, theglobal kinetics of the reaction is substantially controlled by thediffusion rate of silver cations to the surface that are reducedsubsequently by electrons produced in the oxidation of silicon.In the conventional HF-based silver GD processes, silicon iseasily oxidized and the oxides are quickly removed. Electronsprovided by the oxidation process are then consumed by the

nearby silver cations in a ratio of 4 silver atoms per oxidizedsilicon atom. This phenomenon impoverishes the silver cationsof the surrounding solution. The reaction evolves via thediffusion of electrons (through the metallic silver previouslydeposited) that can reach other regions where silver cations areavailable. Silver dendrite formation enables a long-term chainreaction because of the high porosity of the dendrites that createsa pathway for HF to reach silicon and the conduction ofelectrons through the dendrites to sustain the two half-cellreactions. The novelty of our process, which departs significantlyfrom the previous work, is in the use of AgF as a source forboth silver and fluoride species and in the addition of KF whenhigher fluoride concentrations are desired. The ratio of concen-trations of these salts is varied in order to obtain different depositmorphologies. In all cases, no HF is added directly to the bath,and given the basic behavior of F- ions, the plating solution isalkaline instead of acidic, as commonly used in the past. Bycontrolling the rates of silicon etching as well as the amount ofsilver that is near the surface, it is possible, as reported in thispaper, to grow a variety of silver structures.

First, the formation of stable silver thin films is demonstrated.To ascertain the dependence of reaction rate on Ag+ to F-

concentration ratio, two sets of experiments are carried out. Thefirst set involves using a 1 mM AgF with added KF up to 20mM final concentration, while the second set involves 20 mMAgF and no added KF. Figures 2 and 3 show respective SEMand AFM images obtained on samples from baths consistingof [Ag+] ) 1 mM (left-hand side images) and 20 mM (right-hand side images), while maintaining [F-] ) 20 mM in bothcases. On freshly prepared samples, the XPS analysis indicatesthe formation of Ag(0) in all cases. As shown in Figures 2aand 3a, for the 1:20 Ag:F ratio, smooth thin silver films result,while much rougher films, with the presence of silver nano-particles, are obtained when silver concentration is equal to thatof fluoride, with a final concentration of 20 mM. As discussedlater, a 1:20 Ag:F ratio but with higher total concentrations leadsto the formation of different crystalline structures. Figure 4ashows the Ag 3d region of the X-ray photoelectron spectrumobtained on the freshly prepared Ag film shown in Figure 2a.The Ag 3d5/2 and 3d3/2 observed at 368.3 and 374.3 eV confirmthe Ag(0) state.

To determine the film thickness, Teflon tweezers are used toscratch the Ag film without affecting the substrate, and the stepheight between the silver surface and the substrate is measuredusing AFM, as done previously.13 In all cases, a silver thin filmcovers the whole surface. As summarized in Table 1, AFMresults reveal that both film thickness and rms roughness increasewith deposition time for samples prepared from 1:20 Ag:F ratio.A similar trend is observed in thickness for films formed fromequal concentration of Ag+ and F-, without systematic changein rms roughness. Apart from the increase in thickness withtime, it is observed that for a given concentration condition thenumber of nanoparticles and their mean diameters increase withtime. Comparing films deposited at different concentrations butequal immersion times, higher silver concentrations are foundto yield higher density of nanoparticles formed on the film.

Silver crystals in the micrometer-size range can also beobtained by Ag GD process with appropriate bath conditions.To this end, a 1:20 Ag+ to F- concentration ratio is again usedbut at higher absolute concentration values. Figure 5 shows theresults obtained from a bath consisting of a Ag+ concentrationof 10 mM and a final F- concentration of 200 mM obtained bythe addition of the appropriate amount of KF. The figure showsthe silver crystals obtained under these conditions when a silicon

Figure 4. Silver 3d region of X-ray photoelctron spectra obtained onAg thin film shown in Figure 2a: (a) fresh sample after the galvanicdisplacement process, (b) after 24 h incubation on a 5 mM BPE solution.

Si(s) + 6F-(aq) f SiF6

2-(aq) + 4e-

Ag+(aq) + e- f Ag(s)

Silver Nanostructures on Silicon J. Phys. Chem. C, Vol. 113, No. 39, 2009 16941

substrate was incubated for 24 h. SEM and XPS characteriza-tions reveal the formation of metallic silver crystallites. XPSalso shows the absence of Si signal, indicating the formationof a continuous Ag film in between the Ag crystals.

As shown above, a 1:20 Ag+ to F- concentration ratio canlead to Ag thin films or Ag microcrystals with the appropriatenominal concentration values. A higher AgF concentration isthen used to a final value of 30 mM, with an increased F-

concentration of 600 mM obtained again by adding theappropriate amount of KF. Figure 6a,b shows the structuresachieved by a 24 h immersion of silicon substrates in this platingsolution. The formed structures are named nanodesert rose, asthey resemble the crystalline gypsum formations found in somedeserts, shown as an inset to Figure 6. A ratio of 1:20 for Ag+

and F- concentrations provides again enough fluoride to oxidizethe available silicon surface but not fast enough to drive adendrite-based reaction. This stems from the different pH usedin this plating solution (pH ) 7.32) which slows down theoverall silicon oxidation rate. XPS and SEM characterizationsconfirm the formation of crystalline silver flakes. Figure 6cshows the resulting silver plating when no additional fluoride

is added, and thus the final silver and fluoride ions concentrationsare 30 mM and the 1:20 ratio is not maintained. In this case,no flakes are formed due to the fast coverage of the surfacethat takes place in the early stages of reaction.

High-quality nanotextured Ag substrates confirm usefulapplications for example, as surface enhancement Ramanspectroscopy substrates. In this regard, since silver is known tobe the most effective noble metal for SERS, great efforts areunderway for improving the yield and uniformity of SERS hotspots using silver nanostructures as substrates.14 Thus, we haveevaluated the SERS activity of the nanodesert rose sample. Ahigh activity might be anticipated on this sample because ofthe large density of hot spots that the intersecting flakes couldprovide. Figure 7a shows the Raman spectrum of bulk trans-1,2-bis(4-pyridyl)ethylene powder (BPE) (Aldrich, 97% purity),and Figure 7b shows the Raman spectrum obtained on the desertrose substrate after 24 h of incubation and thorough rinsing ofthe BPE solution. XPS analyses show that this procedure yieldsabout one BPE monolayer on silver, and thus the Raman signalis due to a very small number of molecules compared to theRaman signal from the bulk substance. For the bulk sample,using a 200 µm laser depth and the BPE bulk density of 5 ×1021 molecules cm-3, the number of molecules per unit areaprobed is ∼1020 molecules cm-2. Assuming the areal densityof BPE monolayer adsorbed on the Ag surfaces to be ∼1014

molecules cm-2, the number of molecules present in the laserspot is estimated to be approximately 1/106 of the bulksubstance. Despite the large reduction in the number ofmolecules probed, a much higher Raman signal is achieved fromthe BPE monolayer adsorbed on Ag nano-desert rose surfacethan from the bulk BPE sample. In addition, no Raman signalis recorded when using the Ag thin film shown in Figure 2a asthe substrate, although XPS indicates a BPE coverage similarto that present on the desert rose structure. In order to check

Figure 5. Silver crystals produced on silicon after 24 h of GD with plating concentrations [Ag+] ) 10 mM and [F-] ) 200 mM.

TABLE 1: Average Thicknesses and RMS RoughnessValues for the Films Shown in Figure 3a

[Ag+] ) 1 mM;[F-] ) 20 mM

[Ag+] ) 20 mM;[F-] ) 20 mM

immersiontime (s)

thickness(nm)

rms(nm)

thickness(nm)

rms(nm)

5 18.8 5.0 52.8 41.630 57.2 15.4 84.9 36.160 81.4 45.4 98.4 47.4

a The rms values calculated based on a 10 × 10 µm2 image size.Each value is calculated as the average from three different spots onthe same sample.

16942 J. Phys. Chem. C, Vol. 113, No. 39, 2009 Gutes et al.

the purity of Ag after the monolayer formation process, the flatAg film after BPE incubation, but prior to Raman, is analyzedusing XPS. Figure 4b shows the Ag 3d region of the X-rayphotoelectron spectrum. The positions of the Ag 3d5/2 and 3d3/2

observed at 368.3 and 374.3 eV confirm that Ag preserves theAg(0) state after 24 h incubation. The same Ag(0) state isobserved on the nano-desert rose structure. The much higher

Raman signal obtained on the nano-desert rose sample highlightsthat this structure is a highly effective SERS substrate.

4. Conclusions

To conclude, different crystalline silver structures have beenproduced via silver galvanic displacement using aqueous AgFand additional KF in the absence of HF to control the depositionprocess. Different nominal concentrations of silver and fluorideions have been tested maintaining a 1:20 ratio, leading to verydifferent silver structures. Promising applications in SERS havebeen presented, and other implications of the present studies,for example as catalyst for Si nanowire growth, are presentlyunder investigation.

Acknowledgment. This work was funded by DARPA SERSS&T Fundamental Program under LLNL Subcontract # B573237.Albert Gutes thanks Comissionat per a Universitats i Recerca(CUR) del Departament d’Innovacio, Universitat i Empresa dela Generalitat de Catalunya, for funding through the Beatriu dePinos postdoctoral program.

References and Notes

(1) Carraro, C.; Maboudian, R.; Magagnin, L. Surf. Sci. Rep. 2007,62, 499.

(2) Porter, L. A.; Choi, C. H.; Schmeltzer, J. M.; Ribbe, A. E.; Elliott,L. C. C.; Buriak, J. M. Nano Lett. 2002, 2, 1369.

(3) Porter, L. A.; Choi, H. C.; Ribbe, A. E.; Buriak, J. M. Nano Lett.2002, 2, 1067.

(4) Carraro, C.; Magagnin, L.; Maboudian, R. Electrochim. Acta 2002,47, 2583.

Figure 6. (a) Silver desert rose produced by a 24 h GD with platingconcentrations [Ag+] ) 30 mM and [F-] ) 600 mM. Inset: naturaldesert rose [wikipedia.org]. (b) Details of the desert rose flakes. (c)Absence of flakes when [Ag+] ) 30 mM and [F-] ) 30 mM.

Figure 7. (a) Raman shift for the pure BPE powder. (b) SERS BPEspectrum after incubation on the nano-desert rose substrate and rinsing.

Silver Nanostructures on Silicon J. Phys. Chem. C, Vol. 113, No. 39, 2009 16943

(5) Aizawa, M.; Cooper, A. M.; Malac, M.; Buriak, J. M. Nano Lett.2005, 5, 815.

(6) Brevnov, D. A.; Olson, T. S.; Lpez, G. P.; Atanassov, P. J. Phys.Chem. B 2004, 108, 17531.

(7) Sayed, Y. S.; Daly, B.; Buriak, J. M. J. Phys. Chem. C 2008, 112,12291.

(8) Wang, L.; Shaojun, G.; Xiaoge, H.; Shaojun, D. Electrochem.Commun. 2008, 10, 95.

(9) Liu, F.-M.; Green, M. J. Mater. Chem. 2004, 14, 1526.

(10) Peng, K. Q.; Yan, Y. J.; Gao, S. P.; Zhu, J. AdV. Funct. Mater.2003, 13, 127.

(11) Fang, H.; Wu, Y.; Zhao, J.; Zhu1, J. Nanotechnology. 2006, 17, 3768.(12) Peng, K.; Zhu, J. Electrochim. Acta 2004, 49, 2563.(13) da Rosa, C. P.; Maboudian, R.; Iglesia, E. J. Electrochem. Soc.

2008, 155, E70.(14) Sun, Y.; Wiederrecht, G. P. Small 2007, 3, 1964.

JP9055297

16944 J. Phys. Chem. C, Vol. 113, No. 39, 2009 Gutes et al.