You are on page 1of 13

788

Langmuir 1996, 12, 788-800

Surface Plasmon Spectroscopy of Nanosized Metal Particles


Paul Mulvaney
Advanced Mineral Products Research Centre, School of Chemistry, University of Melbourne, Parkville, Victoria, 3052, Australia Received April 4, 1995. In Final Form: July 7, 1995X
The use of optical measurements to monitor electrochemical changes on the surface of nanosized metal particles is discussed within the Drude model. The absorption spectrum of a metal sol in water is shown to be strongly affected by cathodic or anodic polarization, chemisorption, metal adatom deposition, and alloying. Anion adsorption leads to strong damping of the free electron absorption. Cathodic polarization leads to anion desorption. Underpotential deposition (upd) of electropositive metal layers results in dramatic blue-shifts of the surface plasmon band of the substrate. The deposition of just 0.1 monolayer can be readily detected by eye. In some cases alloying occurs spontaneously during upd. Alloy formation can be ascertained from the optical absorption spectrum in the case of gold deposition onto silver sols. The underpotential deposition of silver adatoms onto palladium leads to the formation of a homogeneous silver shell, but the mean free path is less than predicted, due to lattice strain in the shell.

Introduction Interest in the optical properties of colloidal metals dates back to Roman times. Nanosized gold particles were often used as colorants in glasses, and quite complex optical effects were created using metal particles.1 In the seventeenth century, Purple of Cassius, a colloid of heterocoagulated tin dioxide and gold particles, became a popular colorant in glasses.2 These early manifestations of the unusual colors displayed by metal particles prompted Faradays investigations into the colors of colloidal gold in the middle of the last century. Today his studies are generally considered to mark the foundations of modern colloid science.3 The formation of color centers and small colloidal metal particles in ionic matrices and glasses has remained an area of very active research,4-6 driven, in part, by the technical importance of the photographic process.7 However, colloid chemists have tended to neglect the study of metal particles in aqueous solution because of their complicated double layer structure, which is more amenable to direct electrochemical investigation. The more recent discovery that the surface plasmon absorption band can also provide information on the development of the band structure in metals8-11 has led to a plethora of studies on the size dependent optical properties of metal particles, particularly those of silver and gold,12-17 while advances in molecular beam techniques now enable
X Abstract published in Advance ACS Abstracts, December 15, 1995.

spectroscopic analysis of metal clusters to be carried out in vacuum.18,19 Although many of the optical effects associated with nanosized metal particles are now reasonably well understood, there are large discrepancies between the optical properties of metal sols prepared in water, particularly those of silver, and sols prepared in other matrices.6,20-27 In a recent review Kreibig noted that while much work has been done to isolate matrix effects and to determine the roles of defects, grain boundaries, crystallinity, and polydispersity on the optical properties of sols, little is known about the way specific surface chemical interactions may influence the absorption of light by small metal particles.28 These differences are attributed to unique double layer effects present at the metal-water interface. This review focuses on some of these surface chemical effects, and attempts to show how changes to the surface plasmon absorption band of aqueous metal colloids can be related to electrochemical processes occurring at metal particle surfaces. Simple models are proposed to explain some of these chemical changes within the Drude framework for surface plasmon absorption. 1. Light Absorption by Colloids In the presence of a dilute colloidal solution containing N particles per unit volume, the measured attenuation of light of intensity Io, over a pathlength d cm is given by
(15) von Fragstein, C.; Schoenes, F. J. Z. Phys. 1967, 198, 477. (16) Kreibig, U. Z. Phys. B: Condens. Matter Quanta 1978, 31, 39; J. Phys. (Paris) 1977, 38, C2-97. (17) Yanase, A.; Komiyama, H. Surf. Sci. 1991, 248, 11, 20. (18) Fallgren, H.; Martin T. P.; Chem. Phys. Lett. 1990, 168, 233. (19) (a) Tiggesbaumker, J.; Koller, L.; Meiwes-Broer, K.-H.; Liebsch, A. Phys. Rev. A 1993, 48, R1749. (b) Huffman, D. R. Adv. Phys. 1977, 26, 129. (20) Frens, G.; Overbeek, J. Th. G. Kolloid Z. Z. Polym. 1969, 233, 922. (21) Berry, C. R.; Skillman, D. C. J. Appl. Phys. 1971, 42, 2818. (22) Miller, W. J.; Herz, A. H. In Colloid and Interface Science; Academic Press: New York, 1976; Vol. 4. (23) Heard, S. M.; Grieser, F.; Barraclough, C. G.; Sanders, J. V. J. Colloid Interface Sci. 1983, 93, 545. (24) Henglein, A. J. Phys. Chem. 1979, 83, 2209. (25) Lee, P. C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391. (26) Creighton, J. A.; Blatchford, C. G.; Albrecht, M. G. J. Chem. Soc., Faraday Trans. 2 1979, 75, 790. (27) Linnert, T.; Mulvaney, P.; Henglein, A. J. Phys. Chem. 1993, 97, 679. (28) Kreibig, U.; Genzel, U. Surf. Sci. 1985, 156, 678.

(1) See, for example: Savage, G. Glass and Glassware; Octopus Books: London, 1975. One of the most famous examples is the Lycurgus Cup which is ruby red in transmitted light but appears green in reflected light. The color is due to colloidal gold. It was manufactured in the 4th century AD. (2) See: Thiessen, P. A. Kolloid Z. 1942, 101, 241, for micrographs of this composite. (3) Faraday, M. Philos. Trans. R. Soc. 1857, 147, 145. (4) Siedentopf, H. Z. Phys. 1905, 6, 855. (5) Mott, N. F.; Gurney, R. W. Electronic Processes in Ionic Crystals; Oxford University Press: Oxford, 1948. (6) Hughes, A. E.; Jain, S. C. Adv Phys. 1979, 28, 717. (7) The Theory of the Photographic Process, 4th ed.; James, T. H., Ed.; MacMillan Press: New York, 1977. (8) Scott, A. B.; Smith, W. A.; Thompson, M. A. J. Phys. Chem. 1953, 57, 757. (9) Doremus, R. H. J. Chem. Phys. 1965, 42, 414. (10) Doyle, W. T. Phys. Rev. 1958, 111, 1067. (11) Romer, H.; von Fragstein, C. Z. Phys. 1961, 163, 27. (12) Perenboom, J. A. A.; Wyder, P.; Meier, F. Phys. Rep. 1981, 78, 173. (13) Papavassiliou, G. C. Prog. Solid State Chem. 1980, 12, 185. (14) Kreibig, U. J. Phys. F: Met. Phys. 1974, 4, 999.

0743-7463/96/2412-0788$12.00/0

1996 American Chemical Society

Optical Properties of Metal Particles

Langmuir, Vol. 12, No. 3, 1996 789

A ) log10 Io/Id ) NCextd/2.303

(1)

- p2/(2 + d2)

(7) (8)

where Cext is the extinction cross section of a single particle. For spherical particles with a frequency dependent dielectric function ) + i , embedded in a medium of dielectric function m, Cext is given by29-33

) p2d/(2 + d2)

Cext ) 2/k2

(2n + 1) Re (an + bn)

where is the high frequency dielectric constant due to interband and core transitions and p is the bulk plasma frequency

(2)

p2 ) Ne2/m

(9)

where k ) 2 m/ and an and bn are the scattering coefficients, which are functions of the radius R and the wavelength in terms of Ricatti-Bessel functions. The extinction cross section of a particle is often normalized to give the extinction cross section per unit area:

in terms of N, the concentration of free electrons in the metal, and m, the effective mass of the electron. d is the relaxation or damping frequency, which is related to the mean free path of the conduction electrons, Rbulk, and the velocity of electrons at the Fermi energy, vf, by

Qext ) Cext/R2

(3)

d ) vf /Rbulk

(10)

Conventionally, chemists measure the extinction coefficient of a solution in units of M-1 cm-1, where the colloid concentration is the molar metal atom concentration. This quantity is related to Qext by

When the particle radius, R, is smaller than the mean free path in the bulk metal, conduction electrons are additionally scattered by the surface, and the mean free path, Reff, becomes size dependent with

(M-1 cm-1) ) (3 10-3)VmQext/4(2.303)R

(4)

1/Reff ) 1/R + 1/Rbulk

(11)

where Vm (cm3 mol-1) is the molar volume of the metal. For very small particles where kR , 1, only the first, electric dipole term in eq 2 is significant, and

Cext )

242R3

3/2 m

( + 2 m)2 + 2

(5)

This equation can be also obtained by purely electrostatic arguments, and a clear derivation is given by Genzel and Martin.34 In many cases to be described here, it will be necessary to consider the perturbation introduced by a thin surface layer. The extinction cross section of a small, concentric sphere is given by32
Cext ) 4R2k* Im

Equation 11 has been experimentally verified by the extensive work of Kreibig for both silver and gold particles right down to a size of 2 nm.14,16,28 The advantage of the Drude model is that it allows changes in the absorption spectrum to be interpreted directly in terms of the material properties of the metal. The origin of the strong color changes displayed by small particles lies in the denominator of eq 5, which predicts the existence of an absorption peak when

) -2

(if small)

(12)

shell

m)( core m)( core

-2 +2

shell) shell)

+ (1 - g)(

core

shell)( m

+2 -

shell)

shell

+2

+ (1 - g)(2

shell

-2

m)( core

shell)

(6)

where core is the complex dielectric function of the core material, shell is that of the shell, m is the real dielectric function of the surrounding medium, g is the volume fraction of the shell layer, and R is the radius of the coated particle. When g ) 0, eq 6 reduces to eq 5 for an uncoated sphere, and for g ) 1, eq 6 yields the extinction cross section for a sphere of the shell material. In the case of many metals, the region of absorption up to the bulk plasma frequency (in the UV) is dominated by the free electron behavior, and the dielectric response is well described by the simple Drude model. According to this theory,35 the real and imaginary parts of the dielectric function may be written
(29) Toon, O. B.; Ackerman, T. P. Appl. Opt. 1981, 20, 3657. (30) van der Hulst, H. C. Light Scattering by Small Particles; John Wiley and Sons: New York, 1957. (31) Kurtz, V.; Salib, S. J. Imaging Sci. Technol. 1993, 37, 43. (32) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light by Small Particles; Wiley: New York, 1983. (33) Kerker, M. The Scattering of Light and Other Electromagnetic Radiation; Academic Press: New York, 1969. (34) Genzel, L.; Martin, T. P. Phys. Status Solidi B 1972, 51, 91. (35) Kittel, C. Introduction to Solid State Physics, 2nd ed.; Wiley: New York, 1956.

From eq 7 it can be seen that over the whole frequency regime below the bulk plasma frequency of a metal, is negative which is due to the fact that the electrons oscillate out of phase with the electric field vector of the light wave. This is why metal particles display absorption spectra which are strong functions of the size parameter, kR. In a small metal particle the dipole created by the electric field of the light wave sets up a surface polarization charge, which effectively acts as a restoring force for the free electrons. The net result is that, when condition 12 is fulfilled, the long wavelength absorption by the bulk metal is condensed into a single, surface plasmon band. In the case of semiconductor crystallites, the free electron concentration is orders of magnitude smaller, even in degenerately doped materials (i.e., p is smaller), and as a result surface plasmon absorption occurs in the IR, rather than in the visible part of the spectrum. Semiconductor crystallites therefore do not change color significantly when the particle size is decreased below the wavelength of visible light, although the IR spectrum may be affected. It should be noted that the strong color changes observed when semiconductor crystallites are in the quantum size regime (R < 50 ), are due to the changing electronic band structure of the crystal, which causes the dielectric function of the material itself to change. In Figure 1, a typical surface plasmon band is shown calculated using eq 5 with parameters typical of silver for several values of the damping parameter d. The most important parameter affecting d is the particle size. From eqs 10 and 11 it can be seen that decreases in the particle size lead to an increase in d, causing the band to broaden and the maximum intensity to decrease. The position of the peak is virtually unaffected by small changes to d

790

Langmuir, Vol. 12, No. 3, 1996

Mulvaney

Figure 1. Calculated surface plasmon absorption band of a typical free electron metal particle for several values of the damping parameter, d, within the dipole approximation, eq 6. Parameters, ) 5.0, p ) 10 eV, R ) 5.0 nm. m ) 1.77. Damping frequencies in eV: (1) 0.4, (2) 0.6, (3) 0.8, (4) 1.2, (5), 2.4, (6) 3.6.

Figure 2. Plots of the square of the observed position of the surface plasmon bands of colloidal lead and silver as a function of twice the medium dielectric function. Data from Hughes and Jain.6

2 ) p2(

+2

m)

(13)

but for large damping a slow shift to lower energies occurs. In an inert matrix, the only cause of peak shifts is a change in the dielectric properties of the metal particles themselves, due to this surface scattering or for exceedingly small particle sizes (<1-2 nm), to quantization of the energy levels within the conduction band. In the case of silver particles, quantization results in a blue-shift of the plasmon band and a break-up into discrete excitation bands.36-41 Nevertheless, in water, the experimentally measured surface plasmon absorption bands of silver sols vary enormously in position (ranging from 375 to 405 nm) and the absorption coefficients vary by factors of 3 or 420-27,42-45 These discrepancies cannot be explained on the basis of eq 2 and provided the motivation for much of the work to be described below. We will consider experimental spectroscopic data illustrating the effects of anion adsorption, electronic charging, and underpotential metal deposition, and try to interpret the spectral features in terms of eqs 5-8. Because metal particles are studied in a variety of matrices, it is worth reviewing first how the surface plasmon absorption is affected by the solvent refractive index. 2. The Effect of the Solvent Refractive Index For simple metals obeying the Drude model, the position of the plasmon absorption peak does depend on the refractive index of the surrounding medium. Using eqs 5 and 6, we find, for small , that the band position should obey46
(36) Mulvaney, P.; Henglein, A. Chem. Phys. Lett. 1990, 168, 391. (37) Mulvaney, P.; Henglein, A. J. Phys. Chem. 1990, 94, 4182. (38) Henglein, A.; Mulvaney, P.; Linnert, T. Ber. Bunsen-Ges. Phys. Chem. 1990, 94, 1449. (39) Mostafavi, M.; Marignier, J. L.; Amblard, J.; Belloni, J. Radiat. Phys. Chem. 1989, 34, 605. (40) Kreibig, U.; Genzel, L. Surf. Sci. 1985, 156, 678. (41) Ershov, B. G.; Janata, E.; Henglein, A.; Fojtik, A. J. Phys. Chem. 1993, 97, 4589. (42) Blatchford, C. G.; Campbell, J. R.; Creighton, J. A. Surf. Sci. 1982, 120, 435. (43) Blatchford, C. G.; Siiman, O.; Kerker, M. J. Phys. Chem. 1983, 87, 2503. (44) Richard, J.; Donnadieu J. Opt. Soc. Am. 1969, 59, 662. (45) Skillman, D. C.; Berry, C. R., J. Chem. Phys. 1968, 48, 3297. (46) Shklyarevskii, I. N.; Anachkova, E.; Blyashenko, G. S. Opt. Spectrosc. (USSR) 1977, 43, 427.

where p2 ) (2c)2/p2 is the metals bulk plasma wavelength. From a plot of the observed band position vs 2 m, both the high-frequency dielectric constant and the bulk plasma frequency (or wavelength) can be extracted. Plots of 2 vs 2 m are shown for silver and lead colloids in Figure 2. From the data for lead we find that the bulk plasma energy is 11.3 eV and ) 1.1, in good agreement with electron loss spectroscopy data of Ashton and Green and that of Girault et al.47,48 In the case of silver, the band position in water is variable, but interpolating the values from the salt matrices, we predict that the true position of the silver surface plasmon band in water is 382 ( 1 nm. The high frequency dielectric constant is estimated to be 5.9, not far from other estimates of 4.7-5.3. Adherence to eq 13 demonstrates that the absorption of light by the particles over the spectral region is due to the absorption by conduction electrons, rather than interband transitions. It also suggests that the position of the surface plasmon band may be used to measure the local dielectric constant in microheterogeneous systems. Papavassiliou has used a similar procedure to analyze the optical properties of particulate metal films.49 He found that the color of the film was altered when it was immersed into solvents of differing refractive index. More recently, Wilcoxon et al.50 and Esumi et al.51,52 have prepared metal sols in nonaqueous media, and observed spectral shifts in qualitative accord with eq 13. This optical approach to the measurement of the dielectric function of the environment is analogous to the use of solvatochromic molecules such as ET30.53 Clearly, since the particles in these experiments are prepared in different media, and may have variable shapes, sizes, and defect structures, good adherence to eq 13 cannot always be assumed. Ideally, the same particles
(47) Ashton, A. M.; Green, G. W. J. Phys. F: Met. Phys. 1973, 3, 179. (48) Girault, P.; Seignac, A.; Priol, M.; Robin, S. C.R. Acad. Sci. 1968, 266B, 688. (49) Papavassiliou, G. C. Z. Phys. Chem. (Leipzig) 1976, 257, 241. (50) Wilcoxon, J. P.; Williamson, R. L.; Baughman, R. J. Chem. Phys. 1993, 98, 9933. (51) Torigoe, K.; Nakajima, Y.; Esumi, K. J. Phys. Chem. 1993, 97, 8304. (52) Esumi, K.; Shiratori, M.; Ishizuka, H.; Tano, T.; Torigoe, K.; Meguro, K. Langmuir 1991, 7, 457. (53) Drummond, C. J.; Grieser, F.; Healy, T. W. J. Chem. Soc., Faraday Discuss. 1986, 81, 95.

Optical Properties of Metal Particles

Langmuir, Vol. 12, No. 3, 1996 791

shown together with the calculated spectra using eq 2. As can be seen, the agreement is generally very good, with the intensity and position of the surface plasmon bands of colloidal cadmium and tin being well reproduced by the available dielectric data. The observed position of the band for colloidal lead at 215 nm is clearly red-shifted from the position predicted using the data of Lemmonnier63 but the observed position is in excellent agreement with the values in salt matrices using eq 13 (Figure 2), and their values of for lead must be a little too negative. Their data also predict the existence of a small peak at 280 nm, which is not observed experimentally. The fact that the intensity and peak widths can be reproduced with bulk dielectric data implies immediately that there is little damping due to surface scattering in these aqueous sols, i.e. the mean free path for conduction electrons in these metal particles is substantially smaller than 100 . 4. Electronic Equilibrium
Figure 3. Position of the plasmon band of colloidal gold in numerous solvents of different refractive index.54 Fits to eq 2 using dielectric data from Weaver et al.55 and Johnston and Christy.56 The solvents used are indicated. The full circles refer to binary mixtures of butyl acetate and CS2.

should be transferred from one medium to the next to obviate all the matrix effects. This has recently been carried out with gold sols using a comb polymer as stabilizer.54 The sols ranged in color from red to purple, depending on the solution refractive index, and were stable for many months, particularly in butyl acetate. The results are shown in Figure 3. Rather than assume free electron behavior, the predicted position of the surface plasmon band was calculated using the full Mie equations with dielectric data for gold from the literature. The dielectric data of Johnston and Christy were found to yield excellent agreement. This calibration curve allows gold sols to be used as dielectric indicators in a variety of media. 3. The Extinction Spectra of Base Metal Sols The unique sensitivity of the absorption spectrum of silver sols to double layer perturbations makes interpretation of spectral changes difficult, and we have expended considerable effort synthesizing sols of other free-electron metals in order to extend the number of systems where surface plasmon effects can be investigated. Mie calculations by Creighton and Eadon have indicated that for the majority of metals, surface plasmon absorption contributes significantly to the UV-visible spectrum.57 Very little experimental work has been done on the more electropositive metal sols, because they are so readily oxidized by air, and the work is almost entirely confined to salt matrices.6 However, more recently it has become possible to prepare well-characterized samples using radiolytically generated reductants to ensure rapid nucleation. A primary experimental difficulty is the determination of the particle size, since the sols dissolve instantaneously in air and must be transferred into electron microscopes with the rigorous exclusion of oxygen. In Figure 4, the extinction spectra of several metal sols are
(54) Underwood, S.; Mulvaney, P. Langmuir 1994, 10, 3427. (55) Optical Properties of Metals vol. 2; Weaver, J. H., Krafka, C., Lynch, D. W., Koch, E. E., Eds.; Physics Data Series, No. 18-2; Fachinformationszentrum: Karlsruhe, 1981. (56) Johnston, P. B.; Christy, R. W. Phys. Rev. B 1972, 8, 4370. (57) Creighton, J. A.; Eadon, D. G. J. Chem. Soc., Faraday Trans. 1991, 87, 3881. (58) Henglein, A.; Gutierrez, M.; Janata, E.; Ershov, B. G. J. Phys. Chem. 1992, 96, 4598. (59) Henglein, A.; Mulvaney, P.; Holzwarth, A.; Sosebee, T. E.; Fojtik, A. Ber. Bunsen-Ges. Phys. Chem. 1992, 96, 754.

The fundamental reason for the variable absorption spectra of metal sols prepared in water is the existence of specific ion adsorption and electronic coupling with the solvent. The first attempts to quantify the effects of electronic polarization on the optical properties of colloidal metals were those of Blatchford et al. as part of their studies into the origin of the SERS effect.29,42,43 They pointed out that the spectrum of colloidal silver prepared with citrate could be drastically altered by addition of borohydride ion. The band was blue-shifted and increased in intensity by a factor of 50%. They attributed this to a change in the electron density in the particles. Similar results have often been obtained in other studies but were not correlated directly with redox potential. Direct evidence for a blue-shift following electron injection was subsequently found using pulse radiolysis.66 Exposure of a silver sol to a microsecond pulse of electrons from a Van de Graaff generator produces (CH3)2COH radicals (Eo ) -1.5 V NHE). The radicals react with the sol particles according to

(CH3)2COH + Agn f Agn- + H+ + (CH3)2CO

(14)

The transfer of electrons results in the release of protons into solution, and the resultant transient conductivity signal can be used to monitor the rate of electron transfer. Absorption changes to the sol take place at the same time, with bleaching at longer wavelengths and increased absorption at shorter wavelengths than the peak, as shown in Figure 5. Further bleaching occurs at 320 nm, close to the onset of d-band to Fermi level transitions. The discharge of the sols by solvent reduction or other oxidants limits the attainable negative charge on the particles.

e-coll + H+ f 1/2H2

(15)

The transfer of electrons alters the free electron concentration and therefore the bulk plasma frequency, p. From eqs 7 and 9, it is clear that this will result in a blue-shift of the surface plasmon absorption band. The simulated
(60) Henglein, A.; Giersig, M. J. Phys. Chem. 1994, 98, 6931. (61) Sosebee, T.; Giersig, M.; Holzwarth, A.; Mulvaney, P. Ber. Bunsen-Ges. Phys. Chem. 1995, 99, 40. (62) (a) Schwarz, H. Phys. Status Solidi B 1971, 43, 755. (b) Olsen, C. G.; Lynch, D. W. Phys. Rev. B 1974, 9, 3159. (63) Lemmonnier, J. C.; Priol, M.; Robin, S. Phys. Rev. B 1973, 8, 5452. (64) Vina, L.; Hochst, H.; Cardona, M. Phys. Rev. B 1985, 31, 958. (65) Johnston, P. B.; Christy, R. W. Phys. Rev. B 1974, 11, 1315. (66) Henglein, A.; Mulvaney, P.; Linnert, T. J. Chem. Soc., Faraday Discuss. 1991, 92, 31.

792

Langmuir, Vol. 12, No. 3, 1996

Mulvaney

Figure 4. Experimental and calculated absorption spectra of a number of colloids of electropositive metals in water. The dotted curves are the experimental spectra. Data sources for experimental spectra: Cd, ref 58; Pb, ref 59; Sn, ref 60; Cu, ref 61. Dielectric function data sources: Cd, ref 62; Pb, ref 63; Sn, ref 64; Cu, ref 65.

difference spectrum is also shown. The dielectric function of bulk silver was first modified to take into account surface scattering. Then the plasma frequency was increased to mimic the polarization effects. As is evident, the general shape is similar, but the asymmetry in the experimental difference spectrum is not reproduced. Note that the citrate sols in Blatchfords work also showed a very pronounced, asymmetric increase in absorbance in the presence of borohydride. The steady-state electron concentration is determined by the redox potential of the reductant used to prepare the sol. The most blue-shifted spectra are found when the preparation utilizes borohydride as a reductant (max 376 nm). Citrate is a much weaker reductant and the Fermi level in the silver particles lies at much more positive values. In this case the maximum usually lies between 385 and 390 nm. These optical changes are analogous to the electroreflectance effects observed with silver electrodes upon cathodic polarization but are far more pronounced.67-72 The
(67) Feinleib, J. Phys. Rev. Lett. 1966, 16, 200. (68) Kolb, D. M.; Kotz, R. Surf. Sci. 1977, 64, 96. (69) Hansen, W. N.; Prostak, A. Phys. Rev. 1967, 160, 600; 1968, 174, 500.

electroreflectance studies have also revealed that anion adsorption and desorption occur concomitantly.73-76 It will be shown below that the asymmetry in the difference spectrum is associated with changes in the degree of adsorption of the stabilizing anions with particle charge. Nanosized metal particles have an enormous double layer capacity and in changing the environment in the solution from oxidizing to reducing, a large double layer charge must build up on the particles. If the double layer capacity is taken to be 20 F cm-2, a crude estimate suggests that a 14% change in the electron density is needed if the redox potential in a 30 diameter particle
(70) Takamura, T.; Takamura, K.; Yeager, E. J. Electroanal. Chem. 1971, 29, 279. (71) McIntyre, J. D. E. Surf. Sci. 1973, 37, 658. (72) McIntyre, J. D. E. In Advances in Electrochemistry and Electrochemical Engineering, Vol. 9; Muller, R. H., Ed.; Wiley-Interscience: New York, 1973; p 61. (73) Anderson, W. J.; Hansen, W. N. J. Electroanal. Chem. 1973, 47, 229. (74) Cahan, B. D.; Horkans, J.; Yeager, E. Faraday Soc. Symp. 1970, 4, 36. (75) McIntyre, J. D. E.; Aspnes, D. E. Surf. Sci. 1971, 24, 417. (76) Kolb, D. M. In Trends in Interfacial Electrochemistry; Silva, A. F., Ed.; Reidel Publishing Co.: Dordrecht, 1986; p 301.

Optical Properties of Metal Particles

Langmuir, Vol. 12, No. 3, 1996 793

Figure 5. (a) Absorption spectrum of 30 radius colloidal silver in water. (b) The calculated difference spectrum due to an increase of 0.5% in the particle electron density. Mean free path effects included in calculation. (c) The observed difference spectra obtained after electron injection into the particles by hydroxyalkyl radicals generated by pulse radiolysis.66

is to be raised from the ambient level in air of about +0.4 V NHE to around -0.4 V NHE where H2 evolution begins. This corresponds to a plasmon band shift from 390 to 375 nm. This is clearly consistent with the fact that citrate stabilized silver sols typically have a band around 385390 nm whereas extensively -irradiated silver sols (strong reducing conditions) possess absorption bands at < 380 nm and steadily evolve hydrogen. (For very small particles, the shift in the Fermi level is not due just to the double layer charging. The increase in the conduction band electron concentration also causes an additional increase in the Fermi energy, but this varies only as N2/3.) Unless a reductant is used which has a redox potential identical to the potential of zero charge of polycrystalline silver (-0.7 V NHE) chemically produced sols will always contain residual electrical charge and will possess plasmon absorption bands shifted from the position expected for an electrically neutral particle. We have already identified this wavelength by reference to peak positions in salt matrices where electrostatic charging can be neglected (Figure 2). An uncharged silver colloid should have a maximum at 382 ( 1 nm; wavelengths shorter than this are due to cathodic polarization, and longer wavelengths are due to incomplete reduction of silver ions. Blatchford et al. in fact reached similar conclusions, but their calculated spectra for the cathodically polarized silver particles had two peaks, because they assumed that the charge was confined to a very thin surface layer of thickness 2 , corresponding to the Fermi screening length.43 p was only increased in this thin shell layer, resulting in the appearance of two absorption peaks, which is not observed experimentally. Smearing out the double layer charge by increasing p throughout the particles gives better agreement with experiment. These surface plasmon shifts could readily form the basis for very efficient electrochromic switching. Electrical polarization of a film of nanosized silver colloids can alter the position of the band by 20 nm quite reversibly. 5. Chemisorption When metal particles are prepared in solution, they must be stabilized against the van der Waals forces which

cause coagulation.77 This can be achieved in a number of ways: physisorbed surfactant and polymers create steric or electrosteric barriers,78 physisorption of ions produces purely electrostatic barriers, while depletion stabilization occurs in the presence of some polymers due to osmotic pressure.78 In many cases, the distinction between chemical adsorption (involving direct covalent bonding with the surface metal atoms) and more subtle electrostatic mechanisms (e.g. charge-induced dipole mechanisms and dispersion force mechanisms79 ) is largely a matter of degree. The problem distinguishing chemisorption and physisorption is further exacerbated in the case of metal clusters containing only a few tens of metal atoms, where it is not clear whether the stabilizing molecules should be denoted adsorbates or ligands. Is the electrostatic charge on such a cluster containing, say, ten metal atoms to be considered as a double layer charge or as an intrinsic molecular charge? What is clear is that for the coinage metals, silver and gold, sols prepared with different stabilizers often have quite different absorption spectra even though the particle size distributions appear similar. It is therefore pertinent to ask whether chemisorption of ligands can alter the optical properties of the sol. The gelatin used to stabilize photographic emulsions certainly affects the absorption of silver sols, as was shown by Berry et al.21,45 Changes to the absorption spectra in the presence of specifically adsorbed sulfides or thiols were reported by Herz.22 They showed that chemical complexation and adsorption could be used to drive the aerial oxidation of the silver. In Figure 6a, the adsorption of iodide ion onto a silver sol is shown under nitrogen purging (<0.1 M O2). A drastic damping of the plasmon band is evident. The band initially remains at 382 nm but at higher coverages slowly red-shifts. Solvated iodide ion possesses a CTTS band at 229 nm, but there is initially no increase at 229 nm when iodide ion is added. It is therefore desolvated as it adsorbs. Close to a monolayer of iodide ion can adsorb before the further addition of iodide leads to a clear increase in the 229 nm CTTS band. Even stronger damping is observed with mercaptans and sulfide ions. In all cases there is initially almost no shift in the position of the surface plasmon band with adsorption, but close to monolayer coverage, a shift to longer wavelengths is observed. These changes can be reversed by exposing the coated sol to -radiation in the presence of 2-propanol. Organic radicals transfer electrons to the sol (eq 14), cathodically polarizing them (Figure 6b). In the case of iodide adsorption, the peak immediately recovers to its initial intensity and position, and simultaneously the CTTS absorption band of the solvated ion increases in intensity until it too attains its initial intensity. In the case of sulfide ion, complete reversibility is not achieved. Sulfide ion remains partly adsorbed even during vigorous hydrogen evolution from silver sols.81 In order to model this chemical damping, a number of simple models can be suggested, based on a core-shell structure. We treat the adsorbed iodide ion as a nonabsorbing layer with an ionic diameter of 2.5 and assume a real refractive index in this layer of n ) 1.66. (This is
(77) (a) Biggs, S.; Mulvaney, P. J. Chem. Phys. 1995, 100, 8501. (b) Mahanty, J.; Ninham, B. N. Dispersion Forces; Academic Press: London, 1976. (78) (a) Heller, W.; Pugh, T. L. J. Polym. Sci. 1960, 47, 203. (b) Pugh, T. L.; Heller, W. J. Polym. Sci. 1960, 47, 219. (79) The polarization of neutral molecules at a metal surface has been considered in Antoniewicz, P. R. Surf. Sci. 1975, 52, 709. (80) Mulvaney, P. In Electrochemistry of Colloids and Dispersions; Mackay, R. A., Texter, J. Eds.; VCH: New York, 1991. (81) Linnert, T.; Mulvaney, P.; Henglein, A. J. Phys. Chem. 1993, 97, 679.

794

Langmuir, Vol. 12, No. 3, 1996

Mulvaney

Figure 7. Effect of a dielectric layer of KI on the absorption spectrum of colloidal silver in water. Ag radius ) 3 nm, m ) 1.77, shell ) 2.76. Numbers refer to the shell layer thickness. Inset: Position of the surface plasmon band as a function of the coating thickness.

Figure 6. (a) Effect of added iodide ion on the surface plasmon band of colloidal silver. Experiment carried out under nitrogen, pH 5. Figures refer to added [KI]/M (1) 0; (2) 2; (3) 6; (4) 20; (5) 30; (6) 40. (b) Effect of cathodic polarization on the spectrum of iodide damped silver colloids.80 Spectrum a is that of iodidecoated silver particles. Spectrum b is obtained after electron transfer to the colloid using radiolytically generated reductants. The iodide ion desorbs which causes a strong decrease in the surface plasmon damping, and the blue shift is due to the increase in the conduction band electron density.

Figure 8. Effect of an AgI shell on the spectrum of colloidal silver in water. Thickness of shell layers are indicated. Ag radius ) 3 nm, m ) 1.77. Dielectric data for AgI taken from ref 83.

the value for KI, but the polarizability due to the iodide ion dominates.) The results of such a calculation are shown in Figure 7 where the spectra for several coating thicknesses are shown. They do not in any way explain the observed damping. The nonabsorbing coat cannot introduce any damping, and the model predicts an almost linear shift in the resonance frequency with coating thickness. However the increased absorption and shift to longer wavelengths do account for the effects of some stabilizers such as gelatin (n ) 1.5), PVP,21,45 and hydrophobically bound surfactant stabilizers.82 The damping can also be modeled by treating the shell layer as bulk silver iodide. Typical spectra are shown in Figure 8. It is clear that the presence of silver iodide will induce some damping but also introduces absorption peaks due to exciton formation in the shell layer, as well as predicting pro(82) Huy, T.; Mulvaney, P. To be submitted for publication. (83) Bedikyan, L. D.; Miloslavskii, V. K.; Ageev, L. A. Opt. Spectrosc. (USSR) 1979, 47, 225.

nounced red-shifts in the band position. Thus the formation of a bulk silver iodide surface phase appears to be ruled out. Chemisorption of anions must alter the metallic properties of the silver core. A number of authors have, in the past, invoked demetalization to explain phenomena associated with the surface of metals.84-86 The adsorption of gases onto thin metal films has long been known to increase the resistivity of the film, and this has been variously attributed to changes in the free electron concentration, to changes in specularity, or to the formation of unusual surface phases. In fact, comparing the spectra in Figure 6a with those in Figure 1, it is evident that these model spectra mimick quite well the effects of chemisorption. Chemisorption of anions appears to result in an increase in the damping frequency of the core conduction electrons. The major difficulty is that if we interpret the data purely in terms of a reduction in electron mobility or electron mean free path via eqs 10 and 11, then for SH- adsorption onto Ag sols at pH 10.5, where the surface plasmon intensity
(84) Gordon, J. G., II; Ernst, S. Surf. Sci. 1980, 101, 499. (85) Sonderheimer, E. H. Adv. Phys. 1952, 1, 1. Watanabe, M. Surf. Sci. 1973, 34, 759. (86) Ehrlich, G. J. Chem. Phys. 1961, 35, 2165.

Optical Properties of Metal Particles

Langmuir, Vol. 12, No. 3, 1996 795

decreases by about a factor of 5, the mean free path would have to be decreased to just 6 .81 This implies almost complete demetalization of the entire silver metal particle! One might argue that the sulfide ion can diffuse into the particle, but we consistently find that damping ceases at around one monolayer coverage (in the absence of air). There is no doubt that the surface atoms themselves are demetalized because it has been well established from studies of the underpotential deposition of anions onto silver that the anion is oxidized as it adsorbs.87,88 Under open circuit conditions, the oxidation of the anion is coupled to oxygen or proton reduction.89 Underpotential deposition of iodide occurs at -0.6 V NHE,87 which is sufficient to drive hydrogen evolution in acid solution, until monolayer coverage is attained. The net effect is the formation of a monolayer of AgI via reaction 16.

Agn + mH+ + mI- f Agn-mAgIm + (m/2)H2

(16)

due to the increased refractive index of the dye layer but the position of the dye absorption band is also red-shifted.95 In summary, the adsorption of complexing anions onto colloidal silver results in strong damping of the surface plasmon absorption band. Demetalization of several atomic layers would be necessary to account for this process. The damping is stronger for anions that complex more strongly with silver ions, suggesting that plasmon energy is channeled into the electronic absorption bands of the adsorbate complex. The degree of damping is controlled by the redox potential of the sol particles since this controls the degree of adsorption from solution. Electrostatic charging of the particles by reductants causes the specifically adsorbed anions to desorb, which drastically reduces the amount of surface plasmon damping. The intensity, position, and width of the surface plasmon band are then similar to the values found for sols prepared in salt matrices. 6. Underpotential Deposition (upd) of Metals

A similar reaction takes place during the formation of self-assembled alkanethiol monolayers on gold.90 There is no simple way to account for the intensity of this chemical damping at present. However, the fact that the damping strength is correlated with the strength of the Ag-anion bond suggests that the effect is not simply due to a reduction in the electron mean free path but rather to the channeling of the surface plasmon energy into excitation modes of the surface metal-adsorbate complex. Persson has recently presented a model along these lines.91 These damping effects are not confined to silver particles, although they are most pronounced for silver sols. Similar damping experiments have now been carried out on colloidal Cd and Pb.92 It is worth noting that the UVvisible spectrum of the water soluble metal cluster [Au55{Ph2P(m-C6H4SO3Na)}12Cl6] shows no surface plasmon resonance. This has been attributed to size dependent single electron 5d-6s interband transitions,93 but it is equally likely that the covalently bonded ligands are responsible for the damping, which is about twice that predicted by surface scattering (eq 11). Comparing the silver colloid spectra in Figure 6b with those of Blatchford et al.,43 we see that in both cases cathodic polarization produces an asymmetric increase in absorbance, and we conclude that when borohydride is added to the citrate stabilized sol, the particles become negatively charged, but in addition, citrate ions desorb, reducing the plasmon damping. Conversely, the addition of silver ions to a silver sol will drive the redox potential of the particles to very positive potentials, and a red-shift is expected due to a decrease in electron density. However this anodic shift also induces adsorption of anionic stabilizers from solution causing plasmon damping, so the band will broaden considerably, as indeed is observed.24,94 Finally, it is worth noting that in the case of dye adsorption onto silver sols, strong coupling of the electronic states of the dye with the metal is observed.95 Not only is the colloid surface plasmon band red-shifted
(87) Wierse, D. G.; Lohrengel, M. M.; Schultze, J. W. J. Electroanal. Chem. 1978, 92, 121. (88) Koppitz, F. D.; Schultze, J. W. Electrochim. Acta 1976, 21, 337. (89) Nazmutdinov, R. R.; Spohr, E. J. Phys. Chem. 1994, 98, 5956. (90) Widrig, C. A.; Chung, C.; Porter, M. D. J. Electroanal. Chem. 1991, 310, 335. (91) Persson, B. N. J. Surf. Sci. 1993, 281, 153. (92) Henglein, A.; Holzwarth, A. Submitted to J. Phys. Chem. (93) Schmid, G. In Clusters and Colloids; Schmid, G., Ed.; VCH Publishers: Weinheim, 1994. (94) Henglein, A.; Mulvaney, P.; Linnert, T. Discuss. Faraday Soc. 1991, 92, 31.

In the previous sections we have examined the effects of electron transfer and anion adsorption on the optical spectrum of nanosized particles. Dramatic optical shifts are also associated with the electrodeposition of metal atoms onto metal particles to form bilayer colloids. When electrical charge is accumulated on metal particles, they are cathodically polarized, and the surface plasmon band is blue-shifted. In the presence of metal ions, electrodeposition of metal atoms can compete with proton reduction for the stored electrons.96 A number of papers have demonstrated that the formation of well-defined bilayer or multilayer metal particles is possible by this technique.97-101 The first case was Zsigmondys classic work on the use of gold nuclei to form monodisperse silver particles.102 Bradley has recently presented a thorough review of much of the work on the preparation of mixed metal particles in water and organic solvents.103 Most of the catalytically important metals do not show pronounced surface plasmon absorption because of damping by the d-d interband transitions. This leads to large values of in the visible part of the spectrum which, through eq 5, causes damping of the plasmon oscillations. Effectively, the plasmon energy is lost by excitation of the single electron interband transitions. In Figure 9, the calculated spectra of lead-coated silver sols is shown using dielectric data from Lemonnier et al.63 The spectra reproduce the experimental results exceedingly well.59 Initial reduction leads to a blue-shift and increase in the band intensity as the particles are cathodically polarized. When the redox potential is sufficiently negative, reduction of the metal ions begins, possibly in competition with H2 formation. The metal deposition causes strong blue-shifting and damping, which distinguishes it from simple double-layer charging. The shifts are extremely dramaticsa monolayer (ML) of lead atoms is capable of shifting the surface plasmon band by
(95) Kerker, M. J. Colloid Interface Sci. 1985, 105, 297. (96) Gerischer, H.; Kolb, D. M.; Sass, J. K. Adv. Phys. 1978, 27, 437. (97) Morris, R. H.; Collins, L. F. J. Chem. Phys. 1964, 41, 3357. (98) Mulvaney, P.; Giersig, M.; Henglein, A. J. Phys. Chem. 1992, 96, 10419. (99) Henglein, A.; Mulvaney, P.; Linnert, A.; Holzwarth, A. J. Phys. Chem. 1992, 96, 2411. (100) Henglein, F.; Mulvaney, P.; Henglein, A. Ber. Bunsen-Ges. Phys. Chem. 1994, 98, 180. (101) Mulvaney, P.; Giersig, M.; Henglein, A. J. Phys. Chem. 1993, 97, 7061. (102) Zsigmondy, R.; Thiessen, P. A. Das Kolloidale Gold; Akadem. Verlag: Leipzig, 1925. (103) Bradley, J. S. In Clusters and Colloids; Schmid, G., Ed.; VCH Publishers: Weinheim, 1994.

796

Langmuir, Vol. 12, No. 3, 1996

Mulvaney

Figure 9. Calculated spectra of Pb-coated Ag particles in water for the first four monolayers of deposition assuming bulk dielectric behavior in the lead shell. Letters refer to spectra of 3 nm radius silver with (a) no Pb, (b) 0.1 nm Pb, (c) 0.2 nm Pb, (d) 0.3 nm Pb, (e) 0.7 nm Pb, and (f) 1.5 nm Pb shell. Note these dielectric data reproduce the spectra of the pure metal sols well (Figure 4), but the small peak at 280 nm is an artifact of the dielectric data.63

Figure 10. Plot of 2 vs the mole fraction of lead in the coated particles for first four monolayers: (a) experimentally observed position of the band;59 (b) using eq 19; (c) using bulk dielectric data. The simple adatom model (eq 20) predicts a slope of around -1. Note no account is taken of the double layer charge which will cause a slight, additional blue-shift.

40 nm! This changes the color of the sol from yellow to brown, or in the case of deposition onto gold sols from ruby red to brown. The deposition of just 0.1 ML is readily discernible to the naked eye. We can gain a more concrete understanding of the origin of these dramatic shifts by appealing to eq 6, the extinction cross section of a small coated particle. The condition for excitation of the surface mode is that the denominator equal zero, or
core

The use of the bulk dielectric constants to model the shift due to deposition of less than a monolayer of metal atoms is clearly questionable, although it yields almost the correct behavior. A simpler chemical explanation of the surface plasmon shifts during the deposition of the first monolayer is as follows. Both lead and cadmium are divalent, and each adatom occupies a single lattice site, but contributes two electrons to the particle conduction band. The overall electron density increases (though the electrical double layer is not being charged significantly because the extra charge is neutralized by the divalent lattice cation). The rate of plasmon band shift is then predicted to be just

) -2

shell

shellg

m(3

- g)]/[

shell(3

- 2g) + 2 mg] (17)

o2/2 ) 1 + zVm(Pb)[Pb]/(Vm(Pb)[Pb] + Vm(Ag)[Ag]) (20)


where z is the difference in valency of the Pb and Ag lattice ions and Vm values are the molar volumes. This model predicts that all divalent metals will cause a blue shift of the surface plasmon band, but that monovalent adatoms will cause a much weaker shift. This simple model is only valid for submonolayer coverage, because it treats the adatoms as dopants, not as a discrete layer. 7. Catalyst Poisoning The deposition of electropositive metals increases the overpotential for hydrogen formation by blocking the active sites for H atom formation. It is therefore of interest to ask whether deposition of metal atoms can compete with hydrogen evolution on colloidal Ptn, which has a very small overpotential for hydrogen formation. It has been shown that radiolysis of Ptn sols in the presence of 2-propanol leads to the following reactions:104,105

Around monolayer coverage, g , 1 and we can simplify this to


core

-2

- 2g/3 (

shell

m)

(18)

Assuming the real part of the dielectric functions for both metals is given by eq 7, we find on substitution into eq 18 that

o / ) 1 + 2g/3 psc /pc

(19)

where o is the position of the surface plasmon band prior to deposition, pc is the core metal bulk plasma frequency, and psc is the difference in the bulk plasma frequencies of the shell and core metals. (We have neglected the small shift due to the imaginary part of the metal dielectric functions.) For Cd and Pb, which have similar bulk plasma energies of about 12 eV48,62 the slope of a plot of 2 vs g is predicted to be around 0.8, which is fairly close to the experimental value. In Figure 10, the experimentally observed position of the plasmon band is shown for Pb deposition onto Ag together with the calculated position using bulk dielectric data and the simpler analytic expression, eq 19. It is clear from eq 18 that strong surface plasmon shifts are associated with large values of ( shell - m). For dielectric coatings such as gelatin, surfactant molecules, or citrate ions, ( shell - m) is at most about 0.5, whereas for Pb, ( shell - m) -16 at 380 nm.

2(CH3)2COH + Ptn f (CH3)2CO + Ptn(H)2 (21) (H)2-Ptn f Ptn + H2(aq) (22)

The transfer of electrons to the sol is accompanied by proton uptake, converting the electrons to adsorbed
(104) Westerhausen, J.; Lindig, B.; Henglein, A. J. Phys. Chem. 1981, 85, 1627. (105) Matheson, M. S.; Lee, P. C.; Meisel, D.; Pelizzetti, E. J. Phys. Chem. 1983, 87, 394.

Optical Properties of Metal Particles

Langmuir, Vol. 12, No. 3, 1996 797

Figure 11. Conductivity of ion-exchanged and evacuated solutions of 0.3 mM colloidal Pt at pH 4.0 as a function of the irradiation time at 8.5 104 rad h-1 in the presence of 0.1 M 2-propanol and 0.01 M acetone in the absence of lead and in the presence of 0.1 mM Pb2+ and 0.2 mM Pb2+. The plateau values correspond to complete reduction of lead ions and the slopes yield G(Pb0) ) 2.04. Sodium poly(vinylsulfonate) (0.1 mM) was used to stabilize the sols.

hydrogen atoms. No conductivity changes are observed due to particle charging, and hydrogen is evolved stoichiometrically, as seen from Figure 11.104,105 When the same experiment is carried out in the presence of 0.1 mM Pb2+, conductivity changes are seen immediately, due to the deposition of lead adatoms:

2(CH3)2COH + Pb2+ + Ptn f PtnPb + 2H+ + 2(CH3)2CO (23)


The deposition of Pb poisons the hydrogen evolution reaction, and the particles become coated with lead. In Figure 12, electron micrographs of the Pt sol particles after Pb deposition are shown.106 The lead mantle is clearly visible. Thus, underpotential deposition is also capable of interfering with very facile catalytic processes on nanosized metal particles. 8. Optical Detection of Alloy Formation Although core-shell structures have generally been clearly identified by HRTEM following metal deposition on nanosized metal particles, alloy formation has sometimes been postulated to account for differences between experimental and observed spectra or to explain sluggish reoxidation of shell layers. The melting point of metal particles decreases rapidly once the diameter is less than 100 ,107 which means that alloying and surface diffusion of adatoms will be more facile on nanosized particles than on bulk electrodes. Belloni et al. reported that simultaneous reduction of Cu and Pd resulted in colloidal alloy formation,108 and recent work on coreduction of Ag, Au, and Pt salts by Liz-Marzan et al.109 suggested that alloy formation also occurred in these sytems. Alloy formation was also reported recently for Sn reduction on colloidal Au.60 Berry and Skillman suggested that reaction of lead
(106) The author is indebted to M. Giersig for providing the electron micrographs. (107) (a) Buffat, P. A.; Borel, J.-P. Phys. Rev. A 1976, 13, 2287. (b) Sambles J. R. Proc. R. Soc. London, A 1971, 324, 339. (108) Marignier, J.; Belloni, J.; Delcourt, M.; Chevalier, J. Nature 1985, 317, 344. (109) Liz-Marzan, Luis; Philipse, A. P. J. Phys. Chem. 1995, 41, 15120.

and silver sols resulted in mixed metal particles,110 although it was subsequently found that the equilibration of lead and silver sols takes place via underpotential deposition of lead onto the more noble metal.111 Duff et al. carried out extensive work on the nucleation of mixed noble metal sols, and presented TEM results and optical spectra on a number of systems of catalytic interest.112 Bradley and co-workers demonstrated in a series of elegant papers on CO adsorption onto metal sols that the bonding to the different surface metals could be readily distinguished.113 This may prove a useful chemical technique for distinguishing alloys and shell structures. The question arises whether the optical spectra alone can reveal whether alloying has taken place during deposition of a second metal onto a seed metal particle. In the case of mixed gold-silver colloid particles, this can be done by comparing the experimental absorption spectra of the coated particles with calculated spectra for the gold-coated silver sol and the calculated gold-silver alloy sol based on the alloy dielectric data.114,115 The calculated spectra are shown in Figure 13, and it is obvious that whereas gold deposition should result only in damping of the underlying silver surface plasmon band, alloy formation is accompanied by a continuous red-shift in the band with increasing gold content. In Figure 14 the position of the surface plasmon bands and the experimentally observed positions101 are plotted as a function of the mole fraction of Au in the bimetallic particles. It is apparent that the strong red-shifting observed experimentally is best explained by spontaneous alloy formation during the electrodeposition process. Papavassiliou has prepared alloy colloids of gold and silver by evaporation and condensation of the alloys,116 and the observed position of the plasmon bands are indeed consistent with the positions predicted from the dielectric data of Ripken.115 The origin of the red-shift in the case of the AuAg alloy spectra is quite interesting. Silver and gold have identical bulk plasma frequencies, so a peak shift due to a changing electron density is not expected. However, the highfrequency dielectric constants are quite different, primarily because the interband transitions in gold extend across most of the visible spectrum. The absorption band shift in this case is due to the perturbation of the d-band energy levels and not to changes in the free electron concentration. This results in a steady increase in the effective value of for the alloy and, consequently, a red-shift in the position of the absorption band. The fact that a linear shift is found, as observed experimentally by Papavassiliou and as predicted using Ripkens data, can be explained if the alloy dielectric function takes the form, (R) ) (1 - R) Ag + R Au, where R is the mole fraction of Au in the particle. 9. The Mean Free Path in the Shell Layer While electron microscopy reveals quite homogeneous deposition once the particle size has significantly increased, it is difficult to assess how homogeneously the first few monolayers are deposited because the small size changes are dwarfed by the core particle size distribution. We can in principle characterize the quality of the shell layer in terms of the electron mean free path. For a sphere,
(110) Berry, C. R.; Skillman, D. C. J. Photogr. Sci. 1969, 17, 145. (111) Henglein, A.; Holzwarth, A.; Mulvaney, P. J. Phys. Chem. 1992, 96, 8700. (112) Duff, D. Ph.D. Thesis, Cambridge University, 1989. (113) (a) Bradley, J. S.; Hill, E. W.; Behal, S.; Klein, C.; Chaudret, B.; Duteil, A. Chem. Mater. 1992, 4, 1234. (b) Bradley, J. S.; Millar, J. M.; Hill, E. W.; Melchior, M.; J. Chem. Soc., Chem. Commun. 1990, 705. (c) Mucalo, M. R.; Cooney, R. P. J. Colloid Interface Sci. 1992, 150, 486. (114) Schluter, M. Z. Phys. 1972, 250, 87. (115) Ripken, K. Z. Phys. 1972, 250, 228. (116) Papavassiliou, G. C. J. Phys. F: Met. Phys. 1976, 6, L103.

798

Langmuir, Vol. 12, No. 3, 1996

Mulvaney

Figure 12. High-resolution electron micrographs of colloidal Pt particles after radiolysis in the presence of Pb2+. Well-defined core shell particles of PtPb are formed. The spectrum on the left shows the coating to be fairly homogeneous, the higher resolution picture on the right clearly shows the lattice planes of bulk lead. The PtPb particles were prepared as outlined in Figure 11.

Figure 13. (left) Calculated spectra of 6 nm sized particles of AuAg alloys of various composition in water using full Mie equations, and the dielectric data for the alloys from refs 114 and 115. No mean free path effects were used, since the damping frequency in the alloys is unknown. The numbers refer to the mole fraction of gold. (right) Calculated spectra of Au-coated Ag colloids in water. Core radius ) 3.0 nm. Note the silver plasmon band is strongly damped but does not shift. Thick coatings show a surface plasmon resonance in the gold layer close to 500 nm. Surface scattering in the core is included. Gold layer thicknesses used were 0, 0.32, 0.6, 1.0, 1.5, 2.0, and 3.0 nm.

the mean free path was shown by Euler to be equal to the radius of the sphere, R.117 Granqvist et al. proposed that the mean free path in the shell layer of a layered particle should be given by

R ) {(dcoat - dcore)(dcoat2 - dcore2)}1/3/2

(24)

where dcoat is the diameter of the coated particle and dcore the diameter of the core.118 In Figure 15, the absorption spectra obtained from the chemical reduction of silver
(117) Euler, J. Z. Phys. 1954, 137, 318. (118) Granqvist, C. G.; Hunderi, O. Z. Phys. 1978, B30, 47.

onto Pd colloids are shown, and in Figure 16, the calculated spectra are shown. Pd was chosen as a core because it is noble and has a fairly featureless absorption spectrum. Silver was chosen as the shell material because it displays strongly size-dependent, surface plasmon broadening. For each coating thickness of Ag (found by HRTEM) the mean free path was calculated according to eq 24, and the dielectric function of the silver layer was corrected using eqs 7, 8, and 11. The chemical deposition of silver onto
(119) Michaelis, M.; Henglein, A.; Mulvaney, P. J. Phys. Chem. 1994, 98, 6212.

Optical Properties of Metal Particles

Langmuir, Vol. 12, No. 3, 1996 799

Figure 14. (a) Experimentally observed position for gold electrodeposited onto silver sols.101 (b) Position of the surface plasmon band in nanosized colloids of AuAg alloys for various mole fractions of Au in the particles. (c) Positions predicted for Au-coated Ag sols. The coated particle shows two resonances due to the silver core and for higher mole fractions of Au a band due to the shell.

Figure 16. Calculated extinction spectra of Ag-coated Pd particles assuming a core radius of 4.6 nm. 1 monolayer ) 0.257 nm. Dielectric data for Pd from ref 55. Mean free path in shell layer calculated from eq 24. Figures refer to monolayers of Ag.

Figure 17b, the mean free path, as calculated from the peak intensity, is shown together with the values predicted by eq 24. It is clear that in all cases the MFP in the shell layer is at least a factor of two smaller than the Granqvist model predicts. Underpotential deposition leads to a large number of nucleation sites and, as these patches merge to form the apparently homogeneous shell layer, considerable lattice strain must exist. The MFP is governed by the patch size. Interestingly, when the coated sol particles were subsequently aged at room temperature after complete reduction of the silver in solution, the plasmon band increased in intensity. Aging must allow some annealing of the patches, and this leads to a slow increase in the crystallinity of the shell layer and a corresponding increase in the electron mobility, which results in an increase in the intensity of the surface plasmon absorption band. Conclusions In this review, a number of chemical effects that alter the optical behavior of small metal particles have been identified and discussed. It is obvious that optical measurements can provide important insights into the processes of redox catalysis on nanosized metal particles. Turkevich long ago mused that it may be simpler and perhaps also more accurate to determine the dielectric function of a metal from the colloid extinction spectrum, rather than from reflectivity measurements made in vacuo. It is clear from Figure 4 that whereas this is quite reasonable in some cases, for metals where double layer effects are important, this procedure can only be adopted once the chemical perturbations to the extinction spectra discussed in this review have been accounted for. For this reason, it is still worthwhile comparing the measured spectra with ones calculated using dielectric data obtained independently, and this is the procedure we have adopted. Particularly interesting is the fact that in the cluster regime where covalent bonding is so important to prevent coalescence, plasmon damping is very pronounced.93 The results of the experiments on anion adsorption onto metallic silver particles substantiate the important hypothesis that the ligands may largely determine the conduction electron mobility in metal clusters. It is also worthwhile reflecting on the implications that any dielectric modulation has on double layer interactions. Metals have large Hamaker constants precisely because the free electrons contribute enormously to the polariz-

Figure 15. Experimental absorption spectra of Ag-coated Pd colloids in water.119 Numbers refer to monolayers of Ag calculated from particle sizes found by HRTEM. The Pd colloids had a radius of 4.6 nm. [Pd] ) 2.9 10-5 M.

colloidal palladium results in the appearance of a surface plasmon resonance at about 340 nm which red-shifts toward the bulk value as the coating thickness increases. The Mie calculations reproduce the shift in the peak position very well as seen in Figure 17a, which suggests the shell thicknesses determined by HRTEM were quite accurate. However, the observed peaks are very much broader than those predicted. The plasmon oscillations should have been evident at 340 nm and just 2 ML coverage. The broader peaks imply larger damping and therefore that the electron mean free path (MFP) in the shell layer is smaller than that predicted by eq 24. (Note the decrease in the MFP causes only a very small shift in the peak position of (2 nm, as shown in Figure 17a.) In

800

Langmuir, Vol. 12, No. 3, 1996

Mulvaney

Figure 17. (a) Comparison of experimental and calculated positions of silver shell surface plasmon absorption band. Curve a assumes damping is given by eq 24. Curve b is with increased damping to give correct intensity of the plasmon band. (b) Predicted and observed values of the electron mean free path in the silver shell layer, based on eq 24. The experimental values indicate the electron mobility in the shell is a factor of 2 lower than that predicted by eq 24.

ability of the particles. Shifts in p induced by electron transfer will alter the value of the Hamaker constant of the colloidal metal particle and, thus, the strength of the van der Waals interaction becomes a function of the electrostatic charge on the particle.120 The fundamental tenet of DLVO theorysthe additivity of van der Waals and electrostatic contributions to the double layer
(120) Within the dipole approximation, it can be shown that for two small metal particles, the van der Waals interaction energy takes the form U ) 3R6pm/4d6, where R is the particle radius, d the separation, and m2 ) p2/3. An increase in the stored electron concentration will help stabilize the particles electrostatically but is partially offset by an increase in the van der Waals interaction energy between the particles. Conversely however, sols stabilized by excess silver ions should experience smaller dispersion forces for similar (positive) surface charge densities.

forcessdoes not strictly apply to metal particles. The effect is, of course, most pronounced in the case of very small particles, where the ratio of the double layer to volume charge is most significant.

Acknowledgment. The author thanks his colleagues at the Hahn-Meitner Institute in Berlin for many stimulating discussions. The receipt of an ARC QEII Research Fellowship and the financial support of the Advanced Mineral Products Research Centre are gratefully acknowledged.
LA9502711

You might also like