Atomic force microscopy and surface-enhanced Raman

Atomic force microscopy and surface-enhanced Raman spectroscopy. I. Ag island films and Ag film over polymer nanosphere surfaces supported on glass

R. P. Van Duyne,=) J. C. Hulteen, and D. A. Treichelb) Department of Chemistry and Materials Research Center, Northwestern University, Evanston, Illinois 60208

(Received 10 February 1993; accepted 8 April 1993)

The surface roughness and nanometer scale structure of Ag films used for surface-enhanced Raman scattering (SERS) are characterized using atomic force microscopy (AFM). Two important types of thin film based SERS-active surface have been examined in this study: ( 1) Ag island films (AgIF’s) on smooth, insulating substrates and (2) thick Ag films evaporated over both preroughened and smooth substrates. AFM is demonstrated to be capable of quantitatively defining the three-dimensional (3D) structure of these roughened surfaces. The effects of mass thickness, d, , and thermal annealing on the nanostructure of AgIF’s are studied in detail. Particle size histograms are calculated from the AFM images for both “as-deposited” and annealed IF’s with d,,,= 1.8 and 3.5 nm. Quantitative measurements of the SERS enhancement factor (EF) are coupled with the AFM data and interpreted within the framework of the electromagnetic theory of SERS. AFM images for thick evaporated Ag films over a monolayer of polymer nanospheres (AgFON) shows the clear presence of “random substructure roughness” reducing their utility as controlled roughness surfaces. Similar roughness structures are observed for thick evaporated Ag films on smooth, insulating substrates. Nevertheless, AgFON surfaces are demonstrated to be among the most strongly enhancing thin film based surfaces ever studied with EF’s comparable to those found for electrochemically roughened surfaces. Appli- cations of FON surfaces to ultrahigh sensitivity SERS, anti-Stokes detected SERS, and surface- enhanced hyper-Raman spectroscopy (SEHRS) are reported.

I. INTRODUCTION

Surface-enhanced Raman spectroscopy (SERS) is a widely used technique for obtaining vibrational spectra at some roughened metal surfaces. Several recent review ar- ticleslg provide a comprehensive overview of the more than 2000 papers on SERS dealing with fundamental mechanistic studies of photon/molecule/surface interactions as well as applications in electrochemistry, surface science, materials science, and biochemistry. Recently, fundamental work has been far outpaced by applications development in spite of the circumstance that, in the opinion of the authors, a full, quantitative, and predictive understanding of all features of the SERS enhancement mechanism(s) has yet to emerge.

Currently, it is believed that two simultaneously oper- ative mechanisms, a long range classical electromagnetic (EM) effect and a short range “chemical” (CHEM) effect are responsible for SERS. The partitioning of the total enhancement between these two mechanisms is still a matter of debate but at least for Ag surfaces excited with visible laser excitation wavelengths, it appears that the EM mechanism contributes about lo4 with the CHEM mechanism contributing an additional factor of about 10’. Schatz6 has reviewed the current state of the EM mechanism while Ottotp7 has reviewed the recent progress in conceptualizing

._. - . . . . .- ‘)Author to whom correspondence should be addressed. b)Present address: Department of Chemistry, University of Texas at Aus-

tin, Austin, Texas.

the CHEM mechanism as a metal electron-mediated resonance Raman effect that involves increased electron- photon coupling at atomically rough surfaces and transient charge transfer into affinity levels of the adsorbate.

As a spectroscopic probe of surface chemistry and dynamics, SERS has its strengths.‘” (i) The ability to determine the, identity, bonding, orientation, and monitor the dynamics of adsorbed molecules. (ii) Extraordinary sensitivity and selectivity to the inter-facial species. (iii) Adsor- bate generality. (iv) Operation at solid/UHV, solid/gas, solid/liquid, and even solid/solid interfaces. (v) Operation at metal surfaces in addition to Ag, Cu, and Au: Li,8*g ~~ 10-14 K 12,14,15

pt ;2-24 ,,& a.25

cs 14 R,, 14 4 11,16-18 ~~ 11,19 1n 9,11,20,21

(6) The scope of SERS has been extended to include other surface-enhanced spectroscopies ( SES ) such as the nonlinear processes of surface-enhanced second harmonic generation22 (SESHG) and surface- enhanced hyper-Raman scattering (SEHRS ) .26

SERS’also has its weaknesses.‘” (i) Surface roughness on a nanometer length scale is required thereby excluding the smooth, low index, single crystal surlaces commonly used in UHV surface science. (ii) The magnitude of the surface enhancement factor (EF) and, hence, the sensitivity of SERS is a strong function of the chemical and optical properties of the metal surface as well as its detailed surface- morphology. (iii) It is not clear whether semiconduc- tor surfaces are intrinsically SERS active or not. (iv) In- sulator surfaces are not intrinsically SERS active. (v) The stability and reproducibility of most SERS-active surfaces

J. Chem. Phys. 99 (3), 1 August 1993 0021-9606/93/99(3)/2101 /15/$6.00 @ 1993 American Institute of Physics 2101 Downloaded 22 Jul 2004 to 129.105.116.12. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

2102 Van Duyne, Hulteen, and Treichel: Atomic force microscopy. I

are far from ideal. (vi) A key weakness in the comparison between SES experiment and EM theory is the mismatch between the actual surface nanostructures used in experiments and the model surface nanostructures used for theory.6

One of our longstanding goals has been to ascertain whether these weaknesses are fundamental impediments to extending the scope of SERS or whether, through an improved understanding of the SERS enhancement mechanism( s) and/or appropriate experimental design, they can be overcome. Those problems which derive from the current lack of information concerning the detailed three- dimensional (3D) structure of SERS-active surfaces and its reproducible control will be specifically addressed in this paper.

not specifically in the context of SERS.56,57 Here- again there are inherent problems. Both TEM and HREM suffer from the environmental transfer problem cited above for SEM. In addition, these techniques require specimens that are semitransparent to electrons. Therefore, specimen thicknesses must be < 100 rim or the deposition of thin islands of SERS-active material must be on low atomic number substrates. It would be more desirable if such high resolution information could be obtained without the ne- cessity of sample thinning or special substrates.

Because SERS-active surfaces possess roughness on several characteristic length scales, x-ray diffraction and low-energy electron diffraction (LEED) methods are not generally applicable to surface structure determination. As a consequence, it has become customary in SERS studies to characterize surface structure (i.e., surface roughness) using scanning electron microscopy (SEM) . Electrochemi- tally roughened surfaces,27”4 island films,35,36 thick metal films over CaF, roughened surfaces,37 thick metal films over nanosphere (FON) surfaces,38-45 and lithographically fabricated 2D metal particle arrays’6P46947 have all been examined using SEM. While much useful information can be acquired from qualitative assessments of SEM micrographs and the measurement of particle size distributions as a function of surface preparation or SERS operating conditions, there are certain inherent limitations to SEM when it is applied to the SERS problem. For example, all of the studies cited above involve removal of the surface from the SERS operating environment and the placement of it in the vacuum chamber of the SEM instrument. Thus one must assume that there are no or minimal structural rearrange- ments that occur during this environment transfer. It would be far more desirable to carry out the microscopy under identical conditions to the spectroscopy. SEM requires electrically conducting samples. Thus SEM studies of metal island films, which are usually deposited on insulating substrates, requires overcoating the sample with thin continuous conducting films (e.g., Al). A more desirable situation would be to perform the microscopy on the “as- deposited” SERS-active surfaces. SEM has a typical resolution of 5-10 nm which is insufficient to resolve the com- plete distribution of roughness features in SERS which range from atomic scale roughness (ASR) to large scale roughness (LSR, particles with dimensions of 10 nm to 3 pm>.

Since the invention of the scanning tunneling micro- scope5’ (STM) , it has been clear that its atomic resolution capabilities and environmental compatibility with solid/ vacuum, solid/liquid, and solid/gas interfaces5g-63 offered attractive possibilities for the study of surface enhancement phenomenon. Recently several research groups have used the STM to examine the morphology of silver surfaces: ( 1) 50 nm thick vapor deposited Ag films in air,64 (2) cold deposited and warm deposited Ag films in UHV,65-67 (3) Ag island films deposited on electrically conducting indium-tin-oxide (ITO) glass in UHV,68 and (4) electrochemically roughened surfaces.6g-73 Although the rele- vance of the STM results to surface-enhanced optical processes was discussed in these papers, no SERS data were included. Several of these papers, however, have presented surface-enhanced second harmonic generation (SESHG) data.67P68P72,73 Byahut and Furtak74 have demonstrated strong SERS from electrochemically roughened and STM- analyzed Ag films; however, the SERS spectra were pub- lished without the concurrent STM images. To date, Fer- rell and co-workers75’76 are the only researchers to publish both SERS spectra and STM data from the same surfaces. Although no STM images were reported, STM-derived surface roughness spectral power density functions are given. In their work, 45 nm thick Ag films were deposited over glass and over glass preroughened with CaF, films. However, they found no measurable enhancements from the Ag films on glass, and only weak enhancements could be obtained from the CaF, roughened surfaces when using benzoic acid as the adsorbate. The main limitation to STM for the study of SERS-active surfaces is that like the SEM, it requires that the specimen be electrically conductive. Thus, the study of several popular types of SERS-active surface such as island films on insulator substrates and isolated particle arrays on insulating substrates are pre- cluded from examination.

Sufficient spatial resolution not only to resolve the ASR and LSR of SERS-active particles, but to determine highly detailed structural information about metal particles, can be obtained by using transmission electron microscopy (TEM) or high resolution electron microscopy (HREM) . In the context of SERS, TEM has been applied to metal colloids,48-53 island films,54955 and photographi- cally derived Ag surfaces.2 HREM has been used to examine the internal structure of Ag and Au colloids although

In this paper we report the first application of atomic force microscopy5g~77-80 (AFM) to examine the 3D structure of a SERS-active surface. AgIF’s deposited on smooth glass substrates were chosen as the primary target surface for this combined AFM and SERS study because: their oblate structure is well suited to accurate topographic im- aging by scanning probe microscopy, they are relatively stable and easy to fabricate, and they have played an important role in the development of SERS particularly in elucidating such properties of the EM enhancement mechanism as the long-range distance dependence, coverage dependence, and correlation with optical properties. AFM structural data on AgIF’s is combined with measurements

J. Chem. Phys., Vol. 99, No. 3, 1 August 1993

Downloaded 22 Jul 2004 to 129.105.116.12. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Van Duyne, Hulteen, and Treichel: Atomic force microscopy. I 2103

of their SERS enhancement factor (EF) and optical extinction spectra. These results are compared with pub- lished conclusions from the EM theory of SERS. In addition to the island film experiments, two other nanostructured surfaces have been imaged with the AFM: a thick (viz., about d,=250 nm) Ag film deposited over a monolayer of polymer nanospheres supported on a glass substrate and, for comparison, an identical thick Ag film deposited on a smooth glass substrate. The purpose of this study was to test the hypothesis that the monolayer of polymer nanospheres would provide an easy to fabricate, reproducible, and variable sized, hemispherical template for the formation of a SERS-active metal over-layer whose EM enhancement properties could be selectively “tuned” by changing the diameter of the nanospheres.38A5

II. EXPERIMENTAL TECHNIQUES

The fabrication of the roughened Ag films that are the subject of this research was carried out in a modified Con- solidated Vacuum Corporation vapor deposition system. Pressures in the diffusion pumped chamber were < 10m6 Torr during the deposition. The deposition source con- sisted of a tungsten boat filled with the material to be deposited. Prior to deposition the substrates were cleaned by washing in 1 M NaOH followed by successive sonications in deionized water, methanol, and acetone. The temperature of the substrate, T,, was not independently controlled in this apparatus but was continuously measured during the deposition by a thermocouple attached to the substrate.

The mass thickness of Ag films was measured with a homemade quartz crystal microbalance that was calibrated by two independent methods: cyclic voltammetry and STM. The cyclic voltammetric approach followed the work of Fornarisl STM calibration of mass thickness was performed by masking half of a silver substrate and depositing silver over the other half. The STM tip was repeti- tively scanned over the boundary and the average height calculated.

Monodisperse colloidal suspensions of polystyrene spheres were purchased from three vendors: Interfacial Dynamics Corporation (Portland, OR), Bangs Laborato- ries (Carmel, IN), and Duke Scientific (Palo Alto, CA). The variation in sphere diameter was less than 4% for all polystyrene spheres used. Spheres were diluted by a factor of 2 to 5 in methanol and spin coated onto glass substrates. In all experiments, sphere surface coverage was roughly 1 monolayer.

The 683 cm-’ line of cobalt pthalocyanine (CoPc) was used as a SERS intensity standard for calculating the enhancement factors for the surfaces used in this study. The standard was fabricated by depositing 0.1 monolayer of CoPc on a thick Ag film over polymer nanosphere surface (see Sec. III B and the upper panel inset in Fig. IO). CoPc was chosen as the SERS intensity standard because of previous experience,82*83 ease of fabrication, specimen longevity, and its enormous intensity. Depositions of CoPc were carried out similarly to the method used by Car- ron.*2183 CoPc was sublimed simultaneously onto the SERS active substrate and six glass slides of known area. The

CoPc was subsequently dissolved off each slide with pyridine and diluted to 5 ml and its concentration was measured by UV-visible absorption spectroscopy using a Beck- man DU-7 spectrophotometer. Coverages were calculated using E= 1.17~ lo5 M-’ cm-’ and assuming 4x 1013 molecules per monolayer coverage.82,83 Preresonance Raman spectra of CoPc were acquired from a thick film sublimed onto a sapphire window. tram- 1 ,Zbis (4-pyridyl) -ethylene (BPE) was chosen as an adsorbate because it adsorbs strongly to Ag in air and it is not resonantly enhanced in the visible. The BPE was spin coated onto the prepared substrates from a 2~ 10v4 M solution in methanol:

A Spectra Physics model 171 Arf laser was used both for Raman excitation, and to pump the tunable lasers used in determining excitation profiles. Two coherent dye lasers; a model 490 and model 590, were combined with three dyes; rhodamine 560, rhodamine 590, and DCM, to provide wavelengths between 535 and 700 nm. A Spectra Physics model 3900 Ti:A1203 laser provided excitation wavelengths greater than 700 nm. A Spectra Physics series 3000 cw mode-locked Nd:YAG laser operating at 1064 nm and its second harmonic at 532 nm were used, respectively, for the experiment comparing SEHRS and SERS on thick film over polymer nanosphere surfaces. SER and SEHR spectra were obtained on two detection systems: a SPEX- 1400-11 double grating monochromator equipped with a cooled RCA C31034A-02 photomultiplier tube, and an ACTON VM-505 single grating monochromator equipped with a Photometrics PM-5 12 charge couple device (CCD) camera.

AFM images were all taken in air using a Digital In- struments Nanoscope II. Commercial Si3N4 cantilever tips (Digital Instruments) with spring constants of approximately 0.12 nm-’ were used. These tips are pyramidal in shape with a cone angle of 70” and an effective radius of curvature at the tip R,=20-50 nm. The images reported here are raw, unfiltered data collected in the constant force mode (i.e., the tip-sample separation is adjusted to main- tain a constant force equal to 15-25 nN) with a scan speed of 8 lines s-l. All surfaces were imaged with at least two different tips to minimize the possibility of tip induced false images. Two scan heads, with ranges of 12 ,um by 12 pm for large area scans and 640 nm by 640 nm for smaller areas, were routinely used.

III. RESULTS AND DISCUSSION

A. Island films on smooth glass substrates

“As-deposited” Ag island films were formed on glass substrates with mass thicknesses, d,,,, of 1.8, 3.5, 8, and 16 nm using a fixed deposition rate, rd, of 0.3 nm s-l. Under these conditions, radiative heating of the substrate was minimal so that T, was approximately constant at about 300 K. “Annealed” Ag island films were prepared by a post-deposition heat treatment of the same specimens at < 10B6 torr vacuum at T,=600 K for 60 min.

Although all the island film surfaces reported in this paper were prepared with a fixed Ag rd=O.3 nm s-l, a systematic examination of the effect of deposition rate on




a.

100 nm b.

100 nm

8 nm

0 nm

10 nm

0 nm

e. 15 nm

0 nm

100 nm

d. 12 nm

0 nm

FIG. 1. AFM images of “as-deposited” Ag island films on glass slides. Island films were prepared at room temperature. rd=0.3 nm s-‘; d,,,= (a) 1.8 nm, (b) 3.5 nm, (c) 8.0 nm, and (d) 16.0 nm.

island film structure was carried out varying the rate from 0.05 <rd<2.0 nm s-l. This study was carried out because, in previous work by Cotton,54 a strong effect of deposition rate on film structure as determined by TEM, optical properties of the film determined by electronic extinction spectra, and SERS activity for both electronically resonant and nonresonant adsorbates was found. They prepared Ag island films of constant d,=5.0 nm on glass substrates at deposition rates from 0.003 < rd < 0.5 nm s- ‘. Films deposited with slow rates near 0.003 nm s-l were found to have larger, more well-separated particles and result in 5 or 6 times greater SERS activity than films deposited at high rates near 0.5 nm s-l. The explanation offered was that at low deposition rates metal atoms migrate independently to their final surface sites. At higher deposition rates, additional metal atoms arrive within a capture area on the surface and influence the aggregate rate of migration to final surface sites. Thus the structure of the film would be a function of arrival rate at the substrate surface.

In our work, no such effect of rd on film structure as determined by atomic force microscopy, or optical properties, or SERS activity was found. Why there should be such disparity between our results and those of Cotton is not at all clear at this time. Two possible sources for this effect are ( 1) differences in T, caused by differences in radiative heating of the specimen in low vs high rd experiments and (2) differences in substrate surface energy caused by different pretreatments. Cotton et al. assumed that the distance between their sample and evaporation source (12 cm) was sufficient that T, was not influenced by the different radiative heat loads on the sample at different deposition rates and consequently, they did not measure T,. If, in fact, T, was a fUnCtiOn of rd, the Complex interplay

between the rates of metal atom arrival, surface diffusion, nucleation of islands, growth of islands, and coalescence of islands would have to be modeled in detail to understand this phenomenon. Additional complexity enters this problem because the surface energies of the substrates used in our experiments are likely to be different. from those in Cotton’s work as a result of the different surface pretreat- ment’s used. Consequently, the rates of surface diffusion, island growth, island coalescence, etc., would be modu- lated by differences in substrate surface energy. Clearly, further study is required to unravel the details of these two different observations concerning the effect of rd on island film structure and SERS EF values.

1. Atomic force microscopy

A set of AFM images with (400 nm) 2 field of view is shown in Fig. 1 for “as-deposited” Ag island films of various d, on glass substrates. These images provide an overview of “as-deposited” island film morphology as a function of mass thickness. For purposes of discussing the geometry of the particles in these AFM images, as well as making contact with the EM theory of SERS based on spheroids,6984-g5 these films will be considered as a random array of individual, oblate spheroidal particles. Each particle is, therefore, characterized by two size metrics: a is the major axis or particle diameter and b is the minor axis or particle height. Particle shape is referenced by the minor-to-major axis ratio, R = b/a. The random oblate spheroidal array approximation is clearly well-suited to the films shown in Figs. l(a) and l(b) with d,=1.8 nm and 3.5 nm since they are composed of regularly shaped, individual particles. The z-contrast scale provides a qualitative




a. 10 urn

0 nm

100 nm b.

$8 nm

0 nm

c. 20 nm

0 nm

100 nm

d. 3 nm

0 nm

100 nm

FIG. 2. AFM images of Ag island films annealed at 600 K for 60 min under vacuum. Depositions performed at room temperature. r,=O.3 nm s-’ and d,,,=(a) 1.8 nm, (b) 3.5 nm, (c) 8.0 nm, and (d) 16.0 nm.

indication of the particle heights for these films which range up to 8 and 10 nm, respectively. When the mass thickness is increased to 8 nm [Fig. 1 (c)l, the film is still composed of individual particles but there is now great irregularity to the particle shape and the distribution of particle sizes and shapes has increased. The z-contrast scale shows that particle height now ranges upward of 15 nm. This is presumably due to the onset of particle-particle coalescence. At d,n = 16 nm [Fig. 1 (d)], the thickest films imaged in this series, the Ag particles appear more spherical but now strongly overlap each other. Particle heights, which really correspond to film thickness now that the particles overlap, are somewhat lower than those in the 8 nm film with a range up to 12 nm. Particle-particle overlap and/or coalescence pigs. 1 (c) and 1 (d)] reduces the ease of particle geometry interpretation in terms of a random array of spheroids.

A corresponding set of (400 nm)2 AFM images that reveals the overall behavior of Ag island films subjected to post-deposition, vacuum annealing to T,=600 K is shown in Fig. 2. These images show that annealed 1.8 nm [Fig. 2(a)] and 3.5 nm [Fig. 2(b)] island films are still composed of individual spheroidal particles but that they have grown in diameter and height compared to those in the corresponding “as-deposited” film. The 8 nm [Fig. 2(c)] annealed film still shows the irregular particle shapes pro- duced by coalescence but they are larger and higher than those of the “as-deposited” film. The particles composing the 16 nm annealed [Fig. 2(d)] film, in contrast to those of the 8 nm annealed film, are reduced in diameter with particle height or film thickness now much smaller at approximately 3 nm after annealing.

In order to quantitate the effect of mass thickness and

thermal annealing on particle geometry, a series of higher resolution AFM images for these same films in both their “as-deposited” and annealed states were acquired. Figures 3 and 4 show AFM images with ( 100 nm)2 and (200 nm)2 fields of view, respectively, for 1.8 nm and 3.5 nm films along with selected AFM line scans which provide an independent measure of the particle height distribution. Since these surfaces are composed of isolated, regularly shaped particles, it was possible to obtain histograms of the particle diameter distribution which are shown in Fig. 5. The insets to Fig. 5 schematically illustrate the particle geometry. Particle diameters, a, are given as the mean for 40-60 particles in the field of view. Particle heights, b, are derived from the peaks in AFM line scan data and are reported as average values from at least five randomly se-

b. 10 nm 5 nm

ONn OllUl

c- 100 run ------+ +----- 1oonm----+

FIG. 3. AFM images and linescans of Ag island films on glass slides. Depositions were performed at room temperature. rd=0.3 nm s-’ and d,= 1.8 nm (a) unannealed and (b) annealed at 600 K for 60 min.



2106

a.

Van Duyne, Hulteen, and Treichel: Atomic force microscopy. I

b.

FIG. 4. AFM images and linescans of Ag island films on glass slides. Depositions performed at room temperature. r,=O.3 nm s-’ and d,=3.5 nm (a) unannealed and (b) annealed at 600 K for 60 min.

lected lines in the field of view. These values and their standard deviations are also given in Table I. Figures 6 and 7 present higher resolution AFM images and line scans for the 8 and 16 nm mass thickness films. Particle diameter histograms were not computed for these surfaces due to the small number, irregular shape, and/or overlapping of the particles. Estimates of a were obtained by measuring the diameter of a circle drawn to circumscribe the particle on at least ten particles. Particle height was again taken directly from the AFM line scan maxima. Table I summa- rizes all the particle diameter, height, and shape, data obtained for these films. Several trends in particle structure are clearly revealed in these studies as a function of increasing d, . In “as-deposited” films the average particle diam-

, --- I 15.0- . - - lO.O-

5.0- 0.0 r

s= 1010 nnl. . b=5,&

R 7 0.50 . _I’ : I-

10.0 20.0

Particle Diameter (nm.)

FIG. 5. Histograms of size distributions from AFM micrographs for FIG. 6. AFM images and linescans of Ag island films on glass slides. d,,,= 1.8 and 3.5 nm Ag island films. Schematic illustrations of these Depositions performed at room temperature. r,=O.3 nm s-’ and d,,,=8.0 surfaces are shown in the insets. nm (a) unannealed and (b) annealed at 600 K for 60 min.

TABLE I. Hemispheroidal particle size and shape properties for Ag island films.

Film thickness

(nm)

1.8 3.5 8.0

16.0

As-deposited Annealed a h-n) b (nm) R a (nm) b (nm) R

10.0* 1.3 5 0.50 13.3*2.0 10 0.15 14.6A2.1 8 0.55 20.8=‘=3.0 16 0.77

33*4 12 0.36 63*7 18 0.29 100* 12 10 0.10 40*5 3 0.08

eter increases; distribution of particle diameters is quite narrow and nearly Gaussian (d,= 1.8 and 3.5 nm films); particle height increases more rapidly than the diameter (the 16 nm film is an exception probably due to the effects of particle-particle overlap); and particle shape becomes increasingly oblate with the minor-to-major axis ratio sys- tematically varying by a factor of 5. Thermal annealing to 600 K for 60 min produces the following changes in particle structure: particle diameter increases by about 35%, 45%, and 90% for the 1.8, 3.5, and 8 nm films, respectively, but decreases by a factor of 2.5 for the 16 nm film; particle diameter distributions are significantly broadened; particle height doubles for the 1.8 and 3.5 nm films, increases by only 50% for the 8 nm film, and actually decreases for the 16 nm film; oblate particle shape is retained with particles in the 1.8 and 3.5 nm films becoming significantly more prolate.

This reexamination of the structure of Ag island films from the new vantage point afforded by atomic force microscopy shows that the AFM is a powerful tool for non- conducting specimens with lateral resolution superior to that for SEM and comparable to that for TEM. In addition the clear advantage of the AFM- is its capability of independently measuring both particle diameter and particle height. Thus, we are able to easily extract the particle shape parameter, R, that is so important in determining the optical properties of these surfaces and, in particular, their SERS activity.




+200nm + - 200 nm +

FIG. 7. AFM images and linescans of Ag island films on glass slides. Depositions were performed at room temperature. i-,=0.3 nm SC’ and dm= 16 nm (a) unannealed and (b) annealed at 600 K for 60 min.

2. Extinction spectra

The optical properties of submicron structures composed of metal particles have been extensively studied.35*36*87-g5 It is well known, for example, that the optical absorbance of Ag island films composed of oblate spheroidal particles change dramatically as a function of 1 nm < dttl < 16 nm.35p36*96 Normal incidence, unpolarized extinction spectra of the “as-deposited” Ag island films studied above by AFM are shown in Fig. 8 (a). The 1.8 nm thick film shows a maximum extinction at 510 nm. Since the particles in this film are oblate with R=0.50, this extinction maximum can be assigned to surface-plasmon excitation along the major axis. As d, increases to 3.5, 8.0,

1.0 I. ,

a. 0.8 -

5 nm. 0.6 8.0 - ‘42 “e *- 04 R ’ w

0.2 -

:

b. a&) R 33 0.36 ,. d,=

40.0- -is- 8Onm “2 + 160”rn 100

--A==

0.10 x t 3.5l-m -o- ii 1e.m E-4 20.0-

o.oj Lc*:: ,::/ 400

I I I 500 600 700 800

Wavelength (nm.)

FIG. 8. (a) Unpolarized, normal incidence extinction spectra of “as- deposited” Ag island films with d,,,= 1.8, 3.5, 8.0, and 16.0 nm; rd=0.3 rims -‘. (b) SERS EF’s for the same films at laser excitation wavelengths 488.0, 5 14.5, 641.3, and 722.0 nm. The values of d,,, , CI, and R are listed from top to bottom in order of decreasing EF at /1,x=722 nm. c

1.0 . . . . . I a.

.g 0.6 - z .g $J 0.4 -

b. .,.

-I I +I- e.onm. 15.0 + 3snm. I

“z 10.0

1

I= :,“,“I.1

z w 5.0

~%~s

0.0 i

20.8 0.77 =* ;;s x:7;

400 I I I

500 600 700 800 Wavelength (nm.)

FIG. 9. (a) Unpolarized, normal incidence extinction spectra of Ag island films annealed at 600 K for 60 min with d,,,= 1.8, 3.5, 8.0, and 16.0 nm; rd=0.3 nm s-‘. (b) SERS EF’s for the same films at laser excitation wavelengths 488.0, 514.5, 641.3, and 722.0 nm. The values of d,,, , a, and R are listed from top to bottom in order of decreasing EF at A,=722 nm.

and 16 nm, the surface-plasmon resonance energy red shifts and its width broadens significantly. These results, when combined with our AFM structural data (Table I) which showed a systematic decrease in R with increasing d,, are consistant $th EM theory (see, for example,.Fig. i3 in Ref. 67). ~~

The corresponding set of normal incidence, unpolarized extinction spectra for Ag island films after post- deposition annealing at T,=600 K for 60 min is shown in Fig. 9(a). Three features are observed: ( 1) the extinction maxima for the annealed samples are blue-shifted and reduced in intensity With respect to those for the “as- deposited” samples, (2) the surface-plasmon resonance widths are decreased, a@ (3). there is a general loss of -_.. light absorption in the red for all samples. Blue-shifted extinction or absorption maxima have been previously reported for thermally annealed Ag island films on rock salt9’ and on g1ass98 substrates. McCarthy measured Ag particle shape and height using the Pt shadow casting technique with TEM observation. He interpreted the blue shift to be a consequence of thermally induced Ag particle breakup and regrowth at new nucleation sites resulting in Ag particles with increased height. Aussenegg did not characterize annealed AgIF’s using microscopic methods and did not offer a structural interpretation of the blue shift. Red-shifted extinction or absorption maxima have also been observed for thermally annealed Ag island films.36J87 The red shifts were interpreted as being consistent with EM theory based on SEM and TEM micrographs showing the particles in the annealed film to be larger in

J. Chem. Phys., Vol. 99, No. 3, I August 1993



diameter than those in “as-deposited” films. In order for this to be the case for oblate spheroidal particles, the particle shape factor, R, would have had to be approximately constant before and after annealing. No independent measurements of the particle height were made to vary the constancy of R.

Our results for low d,, thermally annealed Ag island films on glass are similar to those of McCarthy on rock salt. The AFM shows that annealed AgIF’s are composed of larger but significantly less oblate particles than “as- deposited” films. This change in the minor-to-major axis ratio of the Ag particles accounts for the blue shift in the extinction spectra according to EM theory.

Why the annealing of Ag island films in some experiments should be dominated by increases in particle height, whereas, the annealing of similar films should be dominated by increased particle diameter is not clear. One possible explanation centers on substrate energies. Different substrates and/or pretreatments for the same substrate are likely to result in different substrate surface energies and consequently contact angles between the metal overlayer and the substrate. If, for example, the Ag islands in the Royer experiments were to “wet” the substrate more effectively at higher temperatures, then they will grow in diameter and decrease in height producing the red shift as they are annealed. If, on the other hand, the Ag islands in our experiments or those of McCarthy and Aussenegg were to “wet” the substrate less effectively at the annealing temperature, the opposite would be expected to occur.

3. SERS enhancement factor

In this section we examine the dependence of the surface enhancement factor for AgIF’s on film thickness and laser excitation wavelength in light of the nanostructural information provided by the AFM. Correlations with optical properties characterized by extinction spectra and with the results of the EM theory of SERS are also made. trans- 1,2-bis (4-pyridyl) -ethylene (BPE) , rather than the more familiar pyridine system, was chosen as the adsorbate molecule for these island film EF measurements because it adsorbs to Ag more strongly than pyridine; adsorbs irre- versibly on Ag as compared with the reversible adsorption of pyridine; has a larger spontaneous Raman cross section than pyridine; and is a nonresonant adsorbatelike pyridine when excited at excitation wavelengths in the visible and near infrared. A representative SER spectrum of BPE adsorbed onto a 8.0 nm thick Ag on glass island film is shown in Fig. 10.

The EF is usually defined and directly measured as the ratio of the adsorbate’s SERS intensity to its bulk, normal Raman intensity, where the intensities are in units of counts s -’ W-i mol-’ to normalize for the effects of data aquisition time, laser power, and surface vs volume number density, as shown for 1640 cm-’ symmetric C==C! stretch- ing mode of BPE in Eq. ( 1) . An alternative approach was adopted here because of the thin film nature of our samples:

1500 1000 500

Wavenumbers (cm-‘)

FIG. 10. SER spectrum of 1.0 monolayer of BPE spin coated on a Ag island film, d,,,=8 nm. 10 mW of A,,=722 nm.

EFBPE, 1640 cm-l(~>= G$m g#-) *

(1)

The EF of BPE adsorbed on AgIF’s was obtained by first measuring the ratio of its SERS intensity to that of a thin film SERS intensity standard based on the 683 cm-’ line of cobalt phthalocyanine (CoPc) .82p83 Combining this ratio with the enhancement factor for the CoPc intensity standard excited at 753 nm and the ratio of the bulk, normal Raman intensities for BPE and CoPc we get the desired EF from the following identity:

EFBPE(A)= Gi&(N

)(

1&9A.=753 nm) Is~$,(A=753 nm) 1 &$#>

x ( I”z$(,,(n=753 nm) Igic(,(n=753 nm) ’ 1

(2)

The EF for the 683 cm-’ CoPc standard is calculated to be 1 X 10’ from the spectra shown in Fig. 11. The inset in Fig. 11 (b) shows the electronic absorption spectrum for an 80 monolayer thick film of CoPc on glass and reveals that the unenhanced Raman reference spectrum is actually an electronically preresonance Raman spectrum.

The EF’s for BPE on the “as-deposited” AgIF’s as a function of /2,, and d, are listed in Table II as the mean and standard deviation for 6 replicate surfaces at each film thickness. The wavelength dependencies of the EF, which are also plotted in Fig. 8(b) as excitation profiles to facil- itate the visualization of the trends. The following characteristics are observed in the 488 < il,, < 722 nm range: ( 1) a systematic increase in EF as &., is swept to the red, (2) no well-defined EF maxima are observed, and (3) the largest EF=5.3 X lo5 for d,=8.0 nm and a,,=722 nm. The film thickness dependence of EF at constant ;1, shows a systematic increase with d, from 1.8 to 8.0 nm followed by a falloff as d, is further increased to 16.0 nm. In order to interpret this behavior within the framework of the EM theory of SERS for isolated oblate spheroidal particles, it is




a. --~ in -- 52

rmrrj j 3.0 x 10’

I counts s“

b

.--

L

G

‘_

I

1500 1000 500

Wavenumbers (cm-‘)

FIG. 11. Surface enhanced and preresonance Raman spectra of CoPc. (a) SER spectrum of 0.10 monolayer of CoPc. Inset in (a) shows a schematic illustration of the surface composed of 200 nm of Ag evaporated over a monolayer of 542 nm polymer nanospheres. Substrate composed of 200 nm of Ag evaporated over 542 mn polystyrene microspheres. 1.0 mW of /1,,=753.2 nm. (b) Preresonance Raman spectrum of an 80 monolayer thick CoPc film. 300 mW of &,=753.2 nm. Inset in (b) is UV-vis absorption spectrum of 80 monolayer thick CoPc film on glass.

necessary to keep in mind that the EF for each AgIF is actually characterized by the two independent particle structure parameters, a and R, rather than a single d, value. For convenience the a and R structural parameters from Table I are reproduced in Fig. 8 (b) . The EM theory of SERS predicts that,6*86J89 for a fixed value of the particle’s major axis, a, the plasmon frequency will shift to the red as the spheroid becomes more oblate (decreasing R) increasing the Raman EF with red laser excitation compared to blue, and, similarly, for a,fixed particle shape, R, the plasmon frequency will also shift to the red as the spheroid’s major axis, a, becomes larger again increasing the Raman EF with red laser excitation compared to blue. Thus the systematic increase in EF as ;1,, is swept to the red observed in Fig. 8(b) for AgIF’s with d,=1.8, 3.5, and 8.0 nm can be understood as the consequence of in-

creasing the major ‘axis length’and decreasing the R bG-- ~- rameter for the film’s constituent particles. The EF for the d, = 16.0 nm film does not follow this trend. This is probably a consequence of the high packing density of the particles in these films. Densely packed particle arrays, either periodic or random, exhibit surface plasmon resonance features that are broadened considerably by dipole-dipole interactions and may have either red- or blue-shifted excitation spectra compared with those of isolated particles.6’99-‘03 Thus, for /2,, that produces a large EF in a system of well isolated particles, a diminished EF is likely to result from increased particle-particle interactions in an otherwise. similar system.

Two previous studies of the SERS enhancement proper$es of AgIF’s provide relevant data for comparison. Otto’ (see Fig. 8) reported excitation spectra for d,= 2.2, .6.0]-and 15.6’nm.“AgIF’s prepared in UHV that exhibit EF -maxima:= 1,.4X $.at 475’ nm, 1.4X 104 at 540 nm, and 0.6~10~. at 575 nm, respectively. Optical transmission spectra were reported for each of these UHV-prepared AgIF’s but no nanostructural characterization was carried out. As far as we are aware, our observation of EF=5.3 X 105, is the largest yet reported for a AgIF. Although EF’s of this magnitude are commonly found for electrochemically roughened surfaces most AgIF EF’s are about 1 x lo4 as found by Otto. Otto’s results in comparison with ours clearly attest to the high degree of sensitivity that the enhancement properties of a SERS-active surface have with respect to its method of preparation and therefore its detailed nanostructure. In addition, they suggest that it would be profitable to examine the 700 <;lex < 1000 nm range more extensively to search for an optimum excitation wavelength that would maximize the SERS EF. Cotton54 found a linear relationship between SERS intensity and optical density for AgIF’s with constant d,,,=5.0 nm but deposited with varying rd. However, no quantitative characterization of particle structure vs rd was reported. Our results show that although SERS intensities are generally stronger for more strongly absorbing films there is no quantitative linear relationship manifested. Our conclusion is that one must supplement the measurement of an island film’s optical absorption with the direct measurement of nanostructural parameters to provide a firm basis for the interpretation of its surface enhancement properties.

The EF’s for BPE on the annealed AgIF’s as a function of il, and d, are listed in Table III as the mean and standard deviation for six replicate surfaces at each film

TABLE II. SERS enhancement factors (EF) for BPE from “as-deposited” Ag island films at various laser excitation wavelengths (nm) .

Film thickness (nm)

1.8 3.5 8.0 16.0

“EF < ld. bNo data available.

488.0 (nm) 514.5 (nm) 641.3 (nm) 722.0 (nm)

1.0*0.2x 104 1.9*0.8x IO4 Oa 0” 1.0*0.8x lo4 4.8&0.6x lo4 7.1*1.9x 104 9.7+3.8x104 3.8*l.5x104 9.5=‘=2.6x104 23.7*5.6x lo4 52.6*2.3x lo4

b 1.8=!=0.5x1Q4 8.2* 1.7~ lo4 34.8*6.7x lo4




TABLE III. SERS enhancement factors (EF) for BPE from annealed Ag island films at various laser excitation wavelengths (nm) .

Film thickness (nm) 488.0 (nm) 514.5 (nm) 641.3 (nm) 722.0 (nm)

1.8 0.70*0.18x lo4 3.5 0.70*0.20x lo4 8.0 b 16.0 Oa

1.2*0.3x104 0” 0” 1.4&0.4X 104 o.66+o.16x104 0.25 ho.07 x lo4 7.8+2.5x104 4.0*2.0x lo4 7.8*2.3x104

Oa 0= 0”

aEF < 102. bNo data available.

thickness. Excitation profiles are plotted in Fig. 9 (b). The EF’s for annealed films are essentially unchanged at the low d, and blue it,, end of the data set but attenuated by factors as large as 40 at the high d, and red il,, end of the data set compared to “as-deposited” films, show very little wavelength dependence, have no maxima, and have the largest value equal to 7.8 X lo4 for d,= 8.0 nm. The film thickness dependence of EF at constant A,, shows a systematic increase with d, from 1.8 to 8.0 nm followed by a falloff to EF < lo* as d, is further increased to 16.0 nm. For the d,= 1.8 and 3.5 nm films, the attenuation of EF in Fig. 9(b) can be understood in terms of an annealing induced change to significantly less oblate particle shapes with only slightly increased major axis size. This is consistent with the annealing induced blue shift observed in the extinction spectra. For d,=8.0 nm film, the annealing doubles the major axis size while leaving R little changed thus offsetting the blue-shift trend and maintaining a relatively large EF. The lack of SERS activity for the d, = 16.0 nm film is likely to be due to a shift, probably to the red, of the plasmon resonance caused by a combination of R=0.08 and the effects of dipole-dipole interactions in densely packed random particle arrays. So far as we are aware there have been no previous studies of SERS EF’s for annealed Ag island films for comparison; however, Aussenegg9* has reported the effects of island film annealing on the fluorescence intensity and lifetime for adsorbed dye molecules. One practical consequence of this annealing study, which should be noted in addition to the important results connecting AFM structural data, SERS EF’s, and EM theory, is that AgIF’s are viable candidates for SERS- active surfaces in high temperature applications such as in situ heterogeneous catalytic studies and combined thermal desorption/SERS investigations in UHV experiments.

B. Thick films over polymer nanospheres

It is clear from the experiments presented earlier that even with the improved structural information afforded by AFM, island films are still not quite ideal surfaces either for studying more detailed aspects of the EM theory of SERS or for preparing optimized, reproducible surfaces for purposes of chemical analysis. Comparison of theory and experiment for island films is complicated by the distribution of particle sizes and shapes in thin films with isolated particles as well as the overlapping particle problem in thicker films. Analytical studies are complicated by the need to control the many deposition parameters that control the details of the particle size, shape, and spacing dis-

tribution. Several attempts have been made in the literature to overcome these problems by using nanofabrication tech- ~quesWW to predefine the roughness of the substrate so that an overlayer film of SERS-active material will adopt that ~roughness. Significant success in EM theory/ experiment comparison has to be achieved with this approach but it is far too difficult to prepare these samples for routine use. An approach to nanofabricated surfaces that would seem to combine the elegance of nanofabrication as a means of defining surface roughness with the simplicity of surface preparation associated with island film or CaF2 roughened surfaces, involves depositing a “thick, smooth” SERS-active metal film over a monolayer of polymer nanospheres supported on glass, quartz, or silicon substrates.38’39’41 These studies which relied on SEM to define the nanostructure of the surface have left the impression that the electromagnetic enhancement properties of these surfaces can be selectively “tuned” by changing the size of the polymer spheres. A schematic representation of such a surface is given in the inset to Fig. 11 (a).

Ag film over polymer nanosphere surfaces were prepared with a variety of sphere diameters, D, and film thicknesses, d, . Figure 12(a) shows a low-resolution AFM image for a d,= 250 nm. Ag film deposited at a rdZO.3 nms -’ over 542 nm diameter polymer nanospheres supported on glass. The radiative heating of the substrate resulted in T, rising by less than 30 K so that no melting or deformation of the nanosphere was observed to occur under these deposition conditions. This image with its 2000 nm by 2000 nm field of view clearly shows the well-ordered ( 111 )-type structure of the nanosphere monolayer that de- fines a roughness scale dictated by the size of the underly- ing spheres. Figure 12(b) shows a higher-resolution AFM image of the top of one of the nanospheres that reveals random substructure roughness (RSR) on the 40-50 nm scale, superimposed on the roughness scale defined by the nanospheres. Clearly, the schematic representation of these surfaces in Fig. 11 (a) is not correct. For comparison purposes, similar AFM images were recorded for d,=250 nm Ag films deposited at the same rate directly on the glass substrate. Figures 13(a) and 13 (b) show the surface of the film to be composed of roughly spherical Ag particles about 40 nm in diameter, not unlike those in the 16 nm thick island films and nearly identical to the RSR on the nanosphere surfaces. Large area images, not shown, reveal the structure to be consistent over macroscopic regions. Figure 13 (c) is an AFM line scan image with (50 nm)2

J. Chem. Phys., Vol. 99, NO. 3, 1 August 1993



a.

RIO b.

50 rim

033Ill

FIG. 12. AFM images of d,,,=250 nm thick Ag film vacuum evaporated over close-packed 542 nm diameter polystyrene nanospheres. rd=O.3 nm s-‘. (a) 2 pm by 2 pm image showing patterning caused by spheres and (b) 200 nm by 200 nm image of “random substructure roughness” on a single sphere.

field of view showing that the individual 40 nm particles are free of substructure roughness down to about the 1 nm level.

SERS intensities for CoPc adsorbed on sphere coated substrates were measured using il,=753.2 nm from a cw Ti:A1203 laser as a function of D, d,, and nanosphere coverage. The values used were D= 121,250, 542, and 870 nm and 0 <dm <4CQ nm. SERS EF’S were then calculated from the ratio of the intensities of the 683 cm-’ line in the surface and bulk samples. Figure 14(a) shows the dependence of SERS intensity on Ag film thickness for various nanosphere sizes. The maximum intensity was obtained for d,=200 mn and D=542 nm. This intensity maximum corresponds to an EF= 1 X 10’. Given that the AFM images for Ag film over polymer nanosphere surfaces clearly show two distinct roughness scales, the question naturally arises as to their relative contributions. This question was addressed in two ways. First, EF’s were compared for the d,,, =200 nm and D= 542 nm film over polymer nanosphere system with a d,=200 nm film directly deposited on glass. These numbers were EFtotal= 1 X lo7 and EFfi~rn only - - 1 X 105. Therefore the incremental enhancement is EFnanospheres = 1 X 102. Next the dependence of E%al on nanosphere coverage and size were determined. Figure 14(b) shows the coverage data for two sphere sizes. This incremental enhancement due to the nanospheres is approximately linear in sphere coverage and clearly size dependent. From these studies we conclude that EFtotal= 1

8 nm.

FIG. 13. AFM images of d,,,=250 nm thick Ag film vacuum evaporated over cleaved mica. rd=0.3 nm s-‘. (a) 400 run by 400 nm image; (b) 200 nm by 200 nm image; (c) 50 nm by 50 nm linescan image.

X lo7 may be partitioned into a lo5 contribution from the RSR of the thick film and a size dependent contribution from the roughness defined by the nanospheres. It is likely that this incremental enhancement due to the large scale but well-defined roughness of the nanospheres is the consequence of extended or delocalized, propagating surface plasmon polaritons (SPP). Studies of the wavelength dependence of EFnanospheres along with theoretical and experimental examination of the gratinglike properties of these surfaces will be required to further elucidate the role of SPP’s in the incremental enhancement mechanism.

Before moving on to the section dealing with applications of film over polymer nanosphere surfaces, a comment is directed to the observation that d,=200 nm Ag films directly deposited on glass at room temperature support SERS EF’s on the order of 105. While this may not be the first observation of this effect, we were unable to readily locate literature references to previous studies. The closest was the case of cold-deposited thick Ag films which are well-known to be SERS active.73 In fact, the usual impression is that thick Ag and Au films are too smooth to support SERS. Clearly this is also highly dependent on depo-



2112

I q

Van Duyne, Hulteen, and Treichel: Atomic force microscopy. I

- WC0 a. D- -m -o- mnm

JO

B + 2W”rn 8

-+- w2nm -la- 8mnm

= o.mv 100 x0 300 -400

dm(m.)

aol 040 0.80 Nanosphere Coverage

FIG. 14. SERS intensity for 0.1 monolayer of CoPc adsorbed on Ag film over polymer nanosphere surfaces, (a) d, dependence at 1.0 monolayer nanosphere coverage for various nanosphere diameters: D=121, 264, 542, and 870 nm; 1.0 mW of &,=753.2 nm. (b) Nanosphere coverage dependence at optimal d ,,,E0.4 D for nanosphere diameters: D=264 and 542 nm; 14 mW of L-=753.2 nm.

sition and substrate conditions. That thick films of Ag and Au support SERS directly may be extremely relevant to studies of the molecular structure of alkanethiol-based self- assembled monolayers (SAM).

C. Applications for film over polymer nanosphere surfaces

Studying the EM enhancement mechanism is only one use for nanosphere roughened substrates. Because they support large enhancements, these surfaces can be used for a variety of other studies. The fact that the enhancements are greatest in the red are a further advantage. In many cases it is desirable to study adsorbates at longer wavelengths to avoid electronic resonances and reduce laser degradation of adsorbates-particularly with biological molecules.

In addition, the introduction of new detectors, most importantly, the charge coupled device (CCD), have en- abled efficient detection of photons into the near IR. Cou- pled with CCD detections systems, nanosphere surfaces can be valuable for improving detection limits of SERS, and studying scattering processes weaker than Stokes Ra- man; i.e., anti-Stokes Raman, and hyper-Raman processes.

Figure 15 shows a SERS spectra of BPE taken with only 5 PW of input 753 nm radiation. A data acquisition time of 1000 s resulted in a S/N> 100. Previously, the lowest reported laser power for a SERS experiment was 0.7 mW at 632.8 nm for a monolayer of benzoic acid adsorbed on a stochastic Ag particle array surface yielding a spec-

a. g

0.30 ADU d mW’

I I I I I 16.96 ~1600 1.550 1500

Wavemnmbers (cm-‘)

FIG. 15. SER spectra of BPE adsorbed on a 200 nm Ag film over 542 nm diameter polymer nanosphere surface using a CCD detector. (a) 2.5 nW of a,=753 nm, 5 s integration. (b) 5 PW of /2,,=753 nm, 1000 s integration.

trum with S/N=4-5 (data acquisition time not given).** A loo-fold reduction in the minimum laser power require- ments for Raman spectroscopy has been achieved. This suggests that PW power level SERS may permit experiments on adsorbates that readily decompose by photother- ma1 or photochemical pathways. In addition, low power diode lasers or even nonlaser light sources should now be usable for SERS.

The lo7 enhancements for SERS also increases the intensities of an adsorbate’s anti-Stokes spectrum. Classical Raman theory predicts that both the Stokes and anti- Stokes branches contain the same vibrational information, but the relative intensities are related as

Is(%) (Q--yfJ4 -ZZ IAs (YL-t2rK)

4 ew&w ( (YL--Ye+YK)2+r2

1 (YL--Ye-vfJ2+P ’ (3)

where I = intensity, yL = laser frequency, vK= vibration frequency, h =Planck’s constant, k=Boltzmann’s constant, and T = temperature. The term in large parentheses is a correction factor for heating induced by the resonance Raman effect, lo4 where Y, is the frequency of the absorption-for the resonance and I is the full width at half maximum (FWHM) for the resonance. With the exception of colloidal systems, the importance of this correction is difficult to quantify. Furthermore, although this correction factor has been used by Pemberton and co-workers27 to measure adsorbate temperatures, the correction was small. SERS spectra of the anti-Stokes bands of BPE at 1609 and 1640 cm-’ demonstrate the large intensities obtainable




I I I I I I I

1650 l&O0 1550 1500 I I I I , 8 I

-1500 -1550 -1600 -1650

Wavenumber (cm-‘)

FIG. 16. SERS of 1.0 monolayer of BPE adsorbed on a 200 nm Ag film over 542 nm diameter polymer nanosphere surface in the 1600 cm-’ range using a CCD detector. (a) Stokes region, 30 mW of AC,=753 nm, 1 s integration. (b) Anti-stokes region, 30 mW of &=753 nm, 30 s integration.

(Fig. 16). At room temperature the anti-Stokes bands are predicted to be approximately lo3 weaker than the Stokes intensities for 753 nm excitation.

Large enhancements from nanosphere roughened substrates are not limited to experiments in air or vacuum. Figure 17 reveals that not only SERS spectra, but also surface-enhanced hyperRaman spectra (SEHRS), can be readily acquired in electrochemical environments. It is sig-

3 a. SEERS

-82 L

&a 4 gi I 0.01 ADU 8” mW’

1600 1300 1000 700

Wavenumbers (cm-‘) FIG. 17. Electrochemical SERS and SEHRS spectra of 1.0 monolayer of BPE on a 200 nm. Ag film over 542 nm diameter polymer nanosphere electrode surface using a CCD detector. (a) 1.0 W of A,,= 1064 nm from a cw mode-locked Nd:YAG laser, E= -0.6 V vs SCE and (b) 10 mW of da=532 nm from second harmonic of a cw mode-locked Nd:YAG laser, E= -0.6 V vs SCE.

nificant to note that although comparable SERS intensities are seen for SERS spectra taken in air and electrochemical environments, SEHRS was only obtainable when the surface was under electrochemical control. It is possible that electromagnetic enhancements are greatly increased at 1064 nm excitation for electrochemical systems and not ones in air. However, SERS spectra taken by FT-Raman indicate that enhancements actually decrease between 750 and 1064 nm.ro5 Another possibility is that chemical enhancement mechanisms are major contributors to hyper- Raman enhancements. Further studies are underway to determine the relative role of enhancements in SEHRS.

IV. CONCLUSIONS

This is the first study in which atomic force microscopy has been used to directly probe the three-dimensional nanostructure of SERS-active surfaces. AFM is clearly demonstrated to be a powerful new tool for determining the shape factor, R, for isolated particles. The direct measurement of R afforded by AFM greatly simplifies the interpretation of the SERS EF’s for Ag island films in terms of the electromagnetic theory of SERS for isolated oblate spheroids.

This is one of the first AFM studies of island film formation and nanostructural modification upon thermal annealing. Since many island films are grown on insulating substrates, these systems would normally require shadow coating with conductive metal for conventional scanning electron microscopy or scanning tunneling microscopy work.

The SERS activity of Ag island films is not lost when they are subjected to quite vigorous thermal annealing conditions (viz., 600 K for 60 min.). EF’s with blue excitation are essentially the same as “as-deposited” surfaces while EF’s with red excitation are attenuated from the 5 X lo5 level down to the lo4 level.

AFM studies show that Ag film over polymer nanosphere systems do not possess the well-controlled roughness characteristics desired in nanofabricated surfaces. However, SERS EF measurements show that they are extremely active. Under “optimized” conditions EF’s in the lo7 range can be obtained. The EFtotal= lo7 is partitioned into a 10’ contribution from random substructure roughness supporting localized surface plasmons and a lo2 contribution which is nanosphere size dependent and may by the consequence of delocalized surface plasmon polaritons. This is the first thin film based SERS-active surface with an EF comparable to that found for electrochemically roughened surfaces.

Ag film over polymer nanosphere surfaces are shown to have applications in ultrahigh sensitivity SERS experiments involving only 5 PW of laser power, detecting anti- Stokes SERS signals for W,ib < 1650 cm-‘, and surface- enhanced hyper-Raman scattering (SEHRS).

ACKNOWLEDGMENTS

We thank Professor George C. Schatz for helpful com- ments. This work was supported by the National Science




Foundation through the Solid State Chemistry Program, Grant No. DMR-8802706 and by the Northwestern Uni- versity Materials Research Center, Grant No. DMR- 9120521.

‘A. Otto, I. Mrozek, H. Grabhom, and W. Akermann, J. Phys. Con- dens. Matt. 4, 1143 (1992).

‘T. M. Cotton and E. S. Brandt, in Physical Methods of Chemistry (WiIey, New York, 1992).

3 J. E. Pemberton, in Electrochemical Interfaces: Modern Techniques for In-Situ Interface Characterization, edited by H. D. Abrulia (VCH Ver- lag Chemie, Berlin, 1991), p. 193.

4T. M. Cotton, J.-H. Kim, and G. D. Chumanov, J. Raman Spectrosc. 22, 729 (1991).

‘R. L. Birke, T. Lu, and J. R. Lombardi, in Techniques For Character- ization of Electrodes and Electrochemical Processes, edited by R. Varma and J. R. Selman (Wiley, New York, 1991), p. 211.

6G. C. Schatz, in Fundamentals and Applications of Surface Raman Spectroscopy, edited by R. L. Garrell, J. E. Pemberton, and T. M. Cotton (VCH Publishers, Inc., Deerfield Beach, FL, 1993).

‘A. Otto, J. Raman Spectrosc. 22, 743 ( 1991). *M. Moskovits and D. P. DiLella, in Surface Enhanced Raman Scatter-

ing, edited by R. K. Chang and T. E. Furtak (Plenum, New York, 1982), p. 243.

9M. Moskovits, Rev. Mod. Phys. 57, 783 (1985). ‘OP. A. Lund, R. R. Smardzewski, and D. E. Tevault, Chem. Phys. Lett.

89, 508 (1982). “G T. Boyd, T. Rasing, J. R. R. Leite, and Y. R. Shen, Phys. Rev. B 30,

5;9 (1984). ‘*W. Schulze, B. Breithaupt, K.-P. Charlb, and U. Kloss, Ber. Bunsenges.

Phys. Chem. 88, 308 (1984). 13W. Schulze, B. Breithaupt, and F. Frank, Surf. Sci. 156, 963 (1985). 14B. Bozlee, B. Lian, J. Kahn, R. L. Garrell, T. Heme, A. Leiden, P.

Palko, N. Hess, and G. Exarhos, Chem. Phys. Lett. 196, 437 (1992). “W. Schulze, K.-P. Charlt, and U. Kloss, Surf. Sci. 156, 822 (1985). 16P. F. Liao and M. B. Stem, Opt. Lett. 7, 483 (1982). “T Lopez-Rfos, Y. Gao, and G. Vuye, Chem. Phys. Lett. 111, 249

(i984). ‘*Y. Gao and T. Lopez-Rios, Surf. Sci. 198, 509 (1988). I’D. P. Dilella and P. Zhou, Chem. Phys. Lett. 166, 240 (1990). zoC. Jennings, R. Aroca, A. M. Nor, and R. 0. Loutfy, Anal. Chem. 56,

2033 (1984). “R. Aroca, C. Jennings, G. J. Kovacs, R. 0. Loutfy, and P. S. Vincent,

J. Phys. Chem. 89, 4051 (1985). **K. L. Haller, L. A. Bumm, R. A. Altkom, E. J. Zeman, G. C. Schatz,

and R. P. Van Duyne, J. Chem. Phys. &l, 1237 (1989). 23S. A. Bilmes, J. C. Rubim, A. Otto, and A. J. Arvia, Chem. Phys. Lett.

159,89 (1989). 24M. Fleischmann, D. Sockalingum, and M. M. Musiani, Spectrochim.

Acta 46A, 285 (1990). 25S. A. Bilmes, Chem. Phys. Lett. 171, 141 (1990). 26J. T. Golab, J. R. Sprague, K. T. Carron, G. C. Schatz, and R. P. Van

Duyne, J. Chem. Phys. 88, 7942 (1988). *‘J E Pemberton, A. L. Guy, R. L. Sobocinski, D. D. Tuschel, and N. A. . .

Cross, Appl. Surf. Sci. 32, 33 (1988). *sD. D. Tuschel and J. E. Pemberton, Langmuir 4, 58 (1988). *‘J E Pemberton and M. M. Girand, J. Electroanal. Chem. 217, 79

(i98-7). “D. D. Tuschel, J. E. Pemberton, and J. E. Cook, Langmuir 2, 380

(1986). 31T M. Devine, T. E. Furtak, and S. H. Macomber, J. Electroanal. Chem.

164, 299 (1984). 32D V. Murphy, K. U. von Raben, T. T. Chen, J. F. Owen, R. K. Chang,

and B. L. Laube, Surf. Sci. 124, 529 (1983). 33C C Busby and J. A. Creighton, J. Electroanal. Chem. 140, 379

(i98i). 34S. G. Schultz, M. Janik-Czachor, and R. P. Van Duyne, Surf. Sci. 104,

419 (1981). “R. S. Sennett and G. D. Scott, J. Opt. Sot. Am. 40, 203 (1950). 36P. Royer, J. P. Goudonnet, R. J. Warmack, and T. L. Ferrell, Phys.

Rev. B 35, 3753 (1987). 37C. A. Murray, D. L. Allara, A. F. Hebard, and F. J. Padden, Surf. Sci.

119,449 (1982).

38J. P. Goudonnet, G. M. Begun, and E. T. Arakawa, Chem. Phys. Lett. 92, 197 (1982).

3gR. L. Moody and T. Vo-Dinh, Appl. Spectrosc. 41, 966 (1987). 40R. Aroca and G. J. Kovacs, J. Mol. Struct. 174, 53 (1988). 4’T. Vo-Dinh, G. H. Miller, J. M. Belle, R. Johnson, R. L. Moody, and

A. Alak, Talanta 36, 227 (1989). 42U Guhathakurta-Ghosh and R. Aroca, J. Phys. Chem. 93, 6125

(l.989). 43R Aroca and U. Guhathakurta-Ghosh, J. Am. Chem. Sot. 111, 7681

(i989). 44P. Bamickel and A. Wokaun, Mol. Phys. 67, 1355 (1989). 45M. Janik-Czachor and J. Jasny, Electrochim. Acta 37, 2347 (1992). 46P. F. Liao, J. G. Bergman, D. S. Chemla, A. Wokaun, J. Melngailis, A.

M. Hawryluk, and N. P. Economou, Chem. Phys. Lett. 81,355 (1981). “P. F. Liao, in Surface Enhanced Raman Scattering, edited by R. K.

Chang and T. E. Furtak (Plenum, New York, 1982), p. 379. 48P. X. Zhang, Y. Fang, W. N. Wang, D..H.. Ni, and S. Y. Fu, J. Raman

Spectiosc. 21, 127 (1990). 49J. Wiesner and A. Wokaun, Chem. Phys. Lett. 157, 569 (1989). 5o0. Siiman and I-L Feilchenfeld, J. Phys. Chem. 92, 453 (1988). ‘I C!. G. Blatchford, J. R. Campbell, and J. A. Creighton, Surf. Sci. 120,

435 (1982). ‘*K U Von Raben, R. K. Chang, and B. L. Laube, Chem. Phys. Lett. 79,

465 (1981). 53D. A. Weitz, M. Y. Lin, and C. J. Sandroff, Surf. Sci. 158, 147 (1985). 54V. L. Schlegel and T. M. Cotton, Anal. Chem. 63, 241 (1991). 55C. A. Murray, J. Electron. Spectrosc. Relat. Phenom. 29, 371 (1983). 56D. J. Wales, A. I. Kirkland, and D. A. Jefferson, J. Chem. Phys. 91,603

(1989). 57D. G. Duff, A. C. Curtis, P. P. Edwards, D. A. Jefferson, B. F. G.

Johnson, A. I. Kirkland, and D. E. Logan, Angew. Chem. Int. Ed. Engl. 26, 676 (1987).

‘*G Binnig, H. Rohrer, C. H. Gerber, and E. Weibel, Phys. Rev. Lett. 49, 5; (1982).

59P K. Hansma, V. B. Elings, 0. Marti, and C. E. Bracker, Science 242, 209 (1988).

“P. K. Hansma, J. Appl. Phys. 61, Rl (1987). “J. A. Golovchenko, Science 232, 48 (1986). 62G. Binnig and H. Rohrer, Rev. Mod. Phys. 59, 615 (1987). 63 C. F,Quate, Physics Today August, 26 ( 1986). 64H. Raether, Surf. Sci. 140, 31 (1984). 65J. K. Gimzewski, A. Humbert, J. G. Bednorz, and B. Reihl, Surf. Sci.

162,961 (1985). 66J. K. Gimzewski, A. Humbert, J. G. Bednorz, and B. Reihl, Phys. Rev.

L.&t. 55, 951 (1985). 670. A. Aktsipetrov, A. A. Nikulin, V. I. Panov, and S. I. Vasil’ev, Solid

State Commun. 73, 411 (1990). 68R. Moller, C. Baur, U. Graf, P. Kiirz, A. Leitner, J. D. Pedamig, and

F. R. Aussenegg, J. Phys. D 23, 1267 (1990). 6g I. Otsuka and T. Iwasaki, J. Vat. Sci. Technol. A 8, 530 (1990). “‘OK. Sakamaki, K. Itoh, A.. Fujishima, and Y. Gohshi, J. Vat. Sci. Tech-

nol. A 8, 525 (1990). ‘II. Otsuka and T. Iwasaki, J. Microscopy 152, 289 (1988). ‘*O A. Aktsipetrov, S. I. Vasil’ev, and V. I. Panov, JETP Lett. 47, 226

(i988). 730. A. Aktsipetrov, A. A. Niiulin, V. I. Panov, S. I. Vasil’ev, and A. V.

Petukhov, Solid State Commun. 76, 55 (1990). 74S Byahut and T. E. Furtak, Langmuir 7, 508 (1991). “P Dawson, K. B. Alexander, J. R. Thompson, J. W. Haas, and T. L.

Ferrell, Phys. Rev. B 44, 6372 (1991). 76P. Dawson, J. W. Haas, K. B. Alexander, J. Thompson, and T. L.

Ferrell, Surf. Sci. Lett. 250, L383 (1991). “G. Binnig, C. F. Qua@ and C. H. Gerber, Phys. Rev. Lett. 56, 930

(1986). “D. Rugar and P. Hansma, Physics Today October, 23 (1990). “Scanning Force Microscopy With Applications to #Electric, Magnetic, and

Atomic Forces, edited by D. Sarid (Oxford University, New York, Ox- ford, 1991).

*‘E. Meyer and H. Heinzelmann, in Scanning Tunneling Microscopy II, edited by R. Wiesendanger and H.-J. Giintherodt (Springer-Verlag, New York, 1992), p. 99.

*‘B. Fomari, G. Mattei, and M. Pagannone, J. Vat. Sci. Technol. A 6, 167 (1988).




*‘K. T. Carron, Ph.D. thesis, Northwestern University, Evanston, Illi- nois, 1985.

*jE. J. Zernan, K. T. Carron, G. C. Schatz, and R. P. Van Duyne, J. Chem. Phys. 87, 4189 (1987).

84G. C. Schatz, Act. Chem. Res. 17, 370 (1984). ssE. J. Zeman and G. C. Schatz, in Dynamics on Surfaces: The Jerusalem

Symposia on Quantum Chemistry and Biochemistry, edited by B. Pull- man, J. Jortner, B. Gerber, and A. Nitzan (Reidel, Dordrecht, 1984), Vol. 17, p. 413.

86E. J. Zeman and G. C. Schatz, J. Phys. Chem. 91, 634 (1987). *‘P. Royer, J. L. Bijeon, J. P. Goudonnet, R. Inagaki, and E. T. Arakawa,

Surf. Sci. 217, 384 (1989). **J. P. Goudonnet, T. Inagaki, T. L. Ferrell, R. J. Warmack, M. C.

Buncick, and E. T. Arakawa, Chem. Phys. 106, 225 (1986). s9J. P. Goudonnet, J. L. Bijeon, R. J. Warmack, and T. L. Ferrell, Phys.

Rev. B 43, 4605 (1991). 9oJ. P. Goudonnet, J. L. Bijeon, and M. Pauty, Thin Solid Films 177, 49

(1989). 9’M. J. Bloemer, T. L. Ferrell, M. C. Buncick, and R. J. Warmack, Phys.

Rev. B 37, 8015 (1988).

92B K. Russell, J. G. Mantovani, V. E. Anderson, R. J. Warmack, and T. L: Ferrell, Phys. Rev. B 35, 2151 (1987).

q3J. L. Bijeon, P. Royer, J. P. Goudonnet, R. J. Warmack, and T. L. Ferrell, Thin Solid Films 155, Ll (1987).

94R. J. Warmack and S. L. Humphrey, Phys. Rev. B 34, 2246 (1986). 9sS. W. Kennerly, J. W. Little, R. J. Warmack, and T. L. Ferrell, Phys.

Rev. B 29, 2926 (1984). 96A. M. Glass, P. F. Liao, J. G. Bergman, and D. H. Olson, Opt. Lett. 5,

368 (1980). “S. L. McCarthy, J. Vat. Sci. Technol. 13, 135 (1976). 98F. R. Aussenegg, A. Leitner, M. E. Lippitsch, H. Reinisch, and M.

Riegler, Surf. Sci. 189/190, 935 (1987). “U. Laor and G. C. Schatz, Chem. Phys. Lett. 82, 566 (1981). “%. Laor and G. C. Schatz, J. Chem. Phys. 76, 2888 (1982). *‘iM. Inoue and K. Ohtaka, J. Phys. Sot. Jpn. 52, 1457 (1983). “*M. Inoue and K. Ohtaka, J. Phys. Sot. Jpn. 52, 3583 (1983). “‘M. Meier, A. Wokaun, and P. F. Liao, J. Opt. Sot. Am. B 2, 931

(1985). lo4L. A. Nafie, P. Stein, and W. L. Peticolas, Chem. Phys. Lett. 12, 131

(1971). “‘D. B. Chase and B. A. Parkinson, J. Phys. Chem. 95, 7810 (1991).



Documents

Atomic force microscopy and surface-enhanced Raman