J. Phys. Chem. 1993,97, 8495-8503
8495
Optical Characterization of Anodic Aluminum Oxide Films Containing Electrochemically Deposited Metal Particles. 1. Gold in Phosphoric Acid Anodic Aluminum Oxide Films Carolyn K. Preston Western Research Centre, Canada Centre for Mineral and Energy Technology (CANMET), Energy, Mines and Resources Canada, Devon, Alberta, Canada TOC 1 EO
Martin Moskovits' Department of Chemistry and The Ontario Laser and Lightwave Research Centre, University of Toronto, Toronto, Ontario, Canada MSS 1Al Received: February 2, 1993; In Final Form: April 29, 1993
In a previous paper (Preston, C. K.; Moskovits, M. J. Phys. Chem. 1988,92,2957) we presented a technique, based on the width of the surface plasmon absorption measured by reflectance spectroscopy, for determining the average size of noble metal particles electrochemically deposited in anodic aluminum oxide films. This method is expanded, here, to take into account the optical anisotropy and nonuniformity of the film which arises from nonspherical nature of the particles themselves as well as from their distribution. Equations describing the reflectance spectrum expected for such films, assumed to consist of an alumina overlayer, a metal-filled oxide layer, and an aluminum substrate, are developed. Experimentally measured s- and p-polarized visible reflectance spectra of gold-filled anodic alumina films prepared in phosphoric acid are found to fit these expressions very well, yielding, as adjustable parameters, the optical thickness of the film, the average size and shape of the gold metal particles, and the equivalent-mass thickness of deposited gold. Particle sizes are found to be more reliably determined from the p-polarized spectra, while the thickness of the alumina overlayer was more accurately derived from the s-polarized spectra. ac electrodeposition was used to introduce gold into the uniform pores of the anodic oxide films. The quantity of metal deposited and the average size of the particles and their eccentricity were found to increase with increasing deposition time. However, these parameters did not vary significantly with the frequency of the ac depositionvoltage over the range 50-500 Hz. Several of the parameters determined by the optical method were corroborated by transmission electron microscopy of sections of the films and of isolated gold metal particles and by atomic absorption measurements of the total gold metal content in the films.
Introduction When anodized in an acidic electrolyte, aluminum forms a porous oxide with very uniform and parallel pores open at one end and sealed at the other. Anodic aluminum oxide has been the subject of numerous studies over more than half a century.'** Recently there has been renewed interest in this material as a medium for creating uniform nanostructures since its pores can function as "nanotemplates" in which small meta13s4and semiconductors particles can be electrochemically deposited. These deposits have been studied in the context of a wide spectrum of scientific goals ranging from catalysis6 to magnetic properties and magnetic data storage.' In order to understand the nature of the chemistry and physics of the metal particles deposited by ac electrolysis in anodic aluminum oxide films, it is essential to know the sizes, shapes, and morphology of these particles. It is known, for example, that the metal deposits are concentrated at the bases of the pores in the form of zerovalent It has proven difficult todetermine confidently the particle morphology and geometry by means of transmission electron microscopy of film sections. However, electr0n3C+~.gand scanning tunneling microscopylo of particles removed from the oxide suggest that iron, nickel, and cobalt form columnar deposits in the pores while gold exists primarily as spheroidal particles. The degree of compactnessof the metallic columns is unknown. However, studies combining metal surface area measurements and atomic absorption measurements' suggest that iron, nickel, cobalt, and gold particles have equivalent areas per volume, implying particles with average radii in the range of 30-50 A, although these measurements a r e 0022-3654/93/2091-8495%04.00/0
somewhat uncertain since the adsorption stoichiometry of the probe gas on the particles is' not known with confidence. Transmission electron microscopy12of gold particles liberated from anodic aluminum oxide films indicate an average radius between 30 and 50 A; however, the extensive agglomeration present makes these numbers somewhat uncertain. Magnetic measurements on iron, cobalt, and nickel films reveal them to be highly anisotropic with far greater magnetization perpendicular to the surface of the film. This is consistent with columnar metal deposits whose long axes are perpendicular to the surface.7dAll of this implies that the nature of the metal deposits depends on the metal deposited. In this paper we present results obtained by reflectance spectroscopy,electron microscopy, and atomicabsorption analysis of metal content. By combining these techniques we are able to produce a far deeper understanding of the nature of the metal electrodeposited into the pores of anodic oxide films than what is possible from microscopy or spectroscopy performed separately. The optical properties of oxidelmetal composites has been the subject of many st~dies.~J3-'~ Noble metal (Cu, Ag, and Au) particles embedded in a dielectricmedium have an unusual optical absorptionin thevisible region of the spectrum due to the surface plasmon (SP) resonance. When the average dimension of particles falls below approximately 300 A, the conductivity of the noble metal becomes size-dependent due to the fact that electron collisions with the particle surface begin to outnumber those in the bulk.13J4J8Jg This is reflected, in turn, in an increased width of the SP absorption. Hence, measuring the SP spectrum can, in principle, provide an estimate of the particle size,1s1S at least in the size domain bounded above by the mean-free path of the 0 1993 American Chemical Society
8496 The Journal of Physical Chemistry, Vol. 97, No. 32, 1993
conduction electrons in the bulk and below by the onset of “quantum size effects”. This strategy wasapplied previously by Preston and Moskovitsl5 to estimate the average size of gold metal particles deposited in anodic aluminum oxide films, as well as the volume fraction of metal deposited in the oxide. However, the model used in ref 15 did not take into account the anisotropy of the metal-filled anodic film, the possible nonsphericity of the particles, nor the predominanceof the metal near the bases of the pores. These deficiencies are addressed in the present study.
The Optical Model The noble metal/oxide composite film is modeled in terms of a three-layer system (Figure 1) consistingof (1) an anodic alumina overlayer, film B, with a thickness dl; (2) a Maxwell Garnett20 layer composed of gold metal particles embedded in aluminum oxide, film A, with a thickness d2; and (3) an aluminum metal substrate. Although film B consists of aluminum oxide containing a large number of pores filled to an unknown extent with water, it is assumed to be optically homogeneoussince all of its structural components are small compared to the wavelength of light. Film A is assumedto consist of gold metal particles uniformly embedded in anodic aluminum oxide. Since the average radius of the metal particles and their average mutual separation are less than the wavelength, this layer also, can be assumed to be optically homogeneous. A simple method for calculating the optical constants of composite media is due to Maxwell Garnett.20 Its application to metal-filled anodicoxide films has been reported pre~iously.~J3--’~ The Maxwell Garnett model can be extended to allow for the possible nonsphericity of the metal particles. This problem was solved by Galeener21for aligned ellipsoidal particles. A dielectric film containing aligned metal ellipsoids will be optically anisotropic. Assuming one of the principal axes of the ellipsoid, a, lies along the normal to the surface of the film, while axes b and c are tangential to its surface, then the dielectric function components e,, eLr, and ety of the film for light polarized along these three directions are given by21
where L, is the depolarization factor associated with the ellipsoid principal axes, m = a, b or c; q is the metallic fraction of the film’s volume, assumed to be much less than 1 , ‘A and em are the dielectric constant of the host medium and the metal, respectively, and ef = e,, cW, or ety,according to the choice of L,,,. For spheroids the depolarization factors are given byZZ
1.
prolate spheroids ( a > b and b = c )
Preston and Moskovits
“A ,
.
.
d2“‘
’
/
Film 0 anodic alumina overlayer
”*’ A,
Film A Maxwell Garnett layer Aluminum substrate
Figure 1. Schematicrepresentationof a three layer film model consisting of an anodic aluminum oxide overlayer and a Maxwell Garnett layer above an aluminum substrate.
3.
spheres (a = b = c) L, = L, = L3 = ‘/,
In the first two cases the two tangential components of the dielectric function will be equal (Le., err = ety) and the film will be uniaxial. The Fresnel coefficient^^^ must also be modified in order to incorporate the film’s anisotropy as was done by Mosteller and W o ~ t e n .Finally, ~~ the specular reflectance of a uniaxial film oriented with its optic axis along the surface normal must be calculated according to the method of Dignam et aL25 The Fresnel coefficients of an anisotropic film depend upon two complex valued angles of propagation, and 4prand the tangential (&) and normal (&) componentsof the refractive index tensor. c $is ~ the complex angle between the propagation vector of the electric field in the medium and the optic axis of the medium (Le., the normal to the surface). 4pis defined through Snell’s law, in a manner similar to qjS. The Fresnel coefficients are
r3p
- A, cos 43 - A, cos 4p - 4 cos 4, + A, cos q5p
L, = L, = (1 - L1)/2
2.
(2)
oblate spheroids (a C b and b = c )
L, = ( ( 1 + f > / f ~ 1 - ( l / ntan-’A
L, = L3 = ( 1 - L,)/2
(3)
(8)
where cos 4s= (At2 - nA2 sin2 41)1/2/&, cos 4p= (hn2- nA2 sin2 41)lf2/h,,, and 2 and 3 denote the second and third interfaces, respectively, as reckoned from the vacuum toward the aluminum substrate. The Fresnel coefficientsfor the air/alumina interface, rlsand rlPhave their usualdefintions.23 A polarization-dependent complex phase change must also be defined: 6, = (2~d/A& cos 4” where v = s or p
e’ = 1 - (bz/a2)
(4)
(9)
Since s-polarized light has only tangential field components, it is not sensitive to the anisotropy. With p-polarized light, the components normal and tangential to the surface will change in relative magnitude with the angle of incidence, providing a better probe of the anisotropy. The average diameter of the pores in anodic aluminum oxide formed in phosphoric acid is approximately 500 A. This restricts the average radius of the metal particles deposited in them to less than 100 A.zB Because this diameter is smaller than the electronic mean free path in bulk gold, the refractive index of the gold must be corrected for the size of the particle. An approach used previously13-*7is to assume that the electronic relaxation rate, y,
Gold in Anodic Aluminum Oxide Films is increased over that of the bulk, Yb, through collisions at the surface of the particle; so that for a particle of average radius, X, 7 = Yb 2uF/x. This results in the following expression:15
+
where e, is the dielectric function of bulk gold at w, i.e., e, = C2 = (n, - iQ2, emEorris the dielectric function at w corrected for gold particle size, wp is the plasma frequency and UF is the Fermi velocity for bulk gold metal. The size-correctedform of the metal dielectric function, ,e-, is then used in place of e, in eq 1. The optical reflectivity of the film system is given by23
+
R, = [rlv rZvexp(-2i6,) rlJ2J3v rlJ3,
+ r3vexp(-2i(6, + ),6 +
exp(-2i62v)l/ [ l + r172v exp(-2i61) + exp(-N61 + 62,)) + rZJ3, exP(-2i6,)1 (1 l a )
wherev denotes s- or p-polarization;rlv,rzV,and r3vare the Fresnel reflection coefficients for the air/film B, film B/film A and film A/alnminum interfaces, respectively, and 61 = (27r/X)n~dlcos41 where 41 is the angle of propagation in film B. The reflectance is equal to the square modulus of R,. However, in order to apply eq 11a to simulate measured reflectance spectra onemust takeaccount of light scattering due to surface roughness and film inhomogeneity. Accordingly another parameter S,,the specular fraction: is introduced so that the effective reflectance, Re, becomes
Re = S,RI2 (1 1b) The specular fraction will differ, in general, for s- and ppolarized light. The optical model described above is simplistic in a number of ways. No attempt is made to account for the localization of the metal in pores27 nor for the scalloped structure of the aluminum/film interface. Moreover, the Maxwell Garnett model assumes dipoldipole interactions between metal particles. This is not adequate for closely packed particles. Finally the “corrected bulk” approach to the modification of the optical properties of gold due to the small particle size will become progressively less applicable as the particle size decreases. The bulk gold optical constants used in the calculations were28,291.39 X lo8 cm/s for the Fermi velocity, 1.385 X 1016 Hz (or 9.1 16 eV) for the plasma frequency, and 1.26 X 1014s-l (or 0.08326 eV) for the bulk gold relaxation rate. Experimental Section Sample Preparation. Samples prepared for reflectance and some TEM measurements were produced on highly polished, 127pm thick, superpurity aluminummetal foil. All other samples were prepared on commercial, 25 pm thick aluminum foil. The thinner substrate allowed for a smaller error in thedetermination of gold content of the samples by AAS. Samples of equal area were used throughout in order to ensure consistent electrochemical deposition conditions. The aluminum foil substrates were degreased ultrasonically at room temperature for 20 min in trichloroethylene, etched in 25 g/L sodium carbonate at 80 OC for 1 min to remove the airformed oxide, rinsed in distilled water for 30 s, dipped into 1:l nitric acid/water for 15 s to neutralize the sodium carbonate, and finally rinsed in distilled water for 1 min. The cleaned foils were anodized in 10%(w/w) aqueous H3P04 solution at room temperature (20-25 “C) for 35 min at 20 V dc using a high-purity lead foil cathode. Following anodization gold depositionwas carried out using ac electrodepositionin an aqueous solution of gold trichloride dihydrate, buffered with boric acid and lowered to a pH of 1.1 with sulfuric acid (30 g/L boric acid and 0.93 g/L AuCly2H20).
The Journal of Physical Chemistry, Vol. 97,No. 32, 1993 8497
TABLE I: Gold/Phosphoric Acid Anodic Alumina Sample Preparation Conditions sample time (s) frequency (Hz) sample time (s) frequency (Hz) 15200si 15 200 305Osi 30 50 200 30200si 30 30100si 30 100 200 30200si 60200si 60 30 200 90200si 90 200 30500si 30 500 ac electrodeposition,k n ~ w n ~togenerate . ~ . ~ metal particles inside the pores of the anodic oxide, was carried out at 50 V peak-topeakusing twographitesheets, oneon either sideof the aluminum sheet, as counter electrodes. The graphite was cleaned prior to use to remove undesired metal ions by applying an ac voltage across the two carbon electrodes in a weak solution of aqueous gold trichloride for approximately 30 min. Metal is deposited during the cathodic half cycles of the ac voltage. The anodic oxide is rectifying so that the current is suppressed during the anodic half cycle. The effects of varying the deposition time and the frequency of the ac voltage during gold electrodeposition were investigated. The former was investigated in order to determine the connection between the quantity of gold deposited and the deposition time and the latter because it had been supposed3 that the frequency of the ac electrolysiscould affect the average particle size of the deposited metal since the duration of the deposition (cathodic) half cycles decreases as the frequency increases. The deposition times and frequencies used are given in Table I. Transmission electron micrographs were recorded from microtome sections of samples 30200si and 60200si. Detailed discussion of the morphology of the films derived from electron microscopy and other analytic techniques will be presented elsewhere.27 Reflectance Measurements. Light from a tungsten lamp was collimated before being reflected from the surface of the sample film. Reflectance measurements were usually carried out at 45O incidence; however, measurements were performed at various angles of incidence (34S0, 45.0°, 52S0, 61S0, and 70.0°)on sample 3050si. The reflected beam was passed through a polarizing filter so that only s- or p-polarized light was detected by the spectrometer. Spectra were measured by a SPEX 14018 double monochromator in the range 25 000-14 000 cm-l. Unanodized, highly polished aluminum foil was used as the reference. Computations. The measured reflectance spectra were fit to the expressions (eqs 1-1 1) outlined above using a nonlinear leastsquares routine, Harwell VA05, which applies a combination of Newton-Raphson, steepest descent, and Marquardt algorithms to determine the required partial derivatives numerically. The s- and p-polarized reflectance spectra were fit separately by adjusting the following parameters: S, (s-polarized specular fraction), Sp(p-polarized specular fraction), q (Maxwell Garnett volume fraction of metal particles), dl (thickness of alumina overlayer), dz (thicknessof Maxwell Garnett layer), A (semimajor axis of the gold ellipsoids perpendicular to the surface of the film), B (two equal semimajoraxes of the gold ellipsoidstangential to the surface of the film), and nA (refractive index of alumina). For ease of discussion it is convenient to express the average spheroid dimensions derived from the fit in terms of the radius of a sphere with an equivalent volume. The optical constants reported by Schulz30were used for aluminum and those of Johnson and Christy28 and Kitte129for gold. Atomic Absorption Spectroscopy. Atomic absorption measurements of gold/alumina films were performed on a Varian Techtron spectrometerusing the 242.8-nmline of a hollow cathode gold lamp and 300-pm slits for samples containing 0-40 pg of gold per mL. The fuel was acetylene in an oxidizing atmosphere. Background was determined with deionized water. A linear Beer law calibration curve of absorbance vs ppm gold was obtained using a series of standard gold concentrations between 2 and 100
8498 The Journal of Physical Chemistry, Vol. 97, No. 32, 1993
Preston and Moskovits
TABLE 11: Gold AAS Results ~
sample
wt 96 gold
gold content (n/cm2)
gold vol fraction
15200siO 30200sia 60200sia 90200siu 15200sib 30200sib 60200sib 90200sib
0.10 0.20 0.25 0.3 1 0.61 0.96 1.69 1.97
3.38 X 6.40 X 10-5 8.06 X 10-5 10.3 X le5 3.08 X 10-5
0.06 0.1 1 0.14 0.17
4.85 X
8.64 X 10-5 10.1 x 10-5
0.05 0.08
0.14 0.17
a Samples prepared from 127-pm aluminum foil. Samples prepared from 25-pm aluminum foil.
ppm. Samples were prepared by dissolving ca. 0.05 g of gold/ alumina/aluminum sample (25 pm aluminum) or ca. 0.10 g of gold/alumina/aluminum sample (127 pm aluminum) in 10 mL of aqua regia and making up the volume to 50 mL with distilled water. Weight percent gold measurements obtained using AAS were compared with the values determined from the reflectance studies. In addition, when the AAS results were combined with film thickness values determi1d2~ by transmission and scanning electron microscopy, the volume fraction of gold metal in the gold/alumina layer could be determined. The effective mass thickness of the gold (i.e., the thickness of the metal layer that would result if all of the gold in the film were compressed to a compact layer of equal area and with bulk density) was also determined from the AAS measurements. Transmission Electron Microscopy of Isolated Gold Particles. Samples for transmission electron microscopy were prepared from films previously used in the reflectance measurements by soaking the films in a mixture of 0.5 M phosphoric acid and 0.2 M chromic acid at 80 OC for 5 min to dissolve the aluminum oxide. The gold particles remaining on the surface of the aluminum substrate, were wiped off with a cotton swab, transferred and dispersed in acetone, a drop of which was then placed on a TEM carbon substrate.
Results and Discussion Atomic AbsorptionMeasurementsof Gold/Ahunina Films. Two series of samples were analyzed by atomic absorption. The first was prepared on 127-pm aluminum with gold deposition times varying from 15 to 90 s (samples 15200si, 30200si, 60200si, and 90200si). The second was prepared on 25-pm foil with similar gold deposition times and conditions. The gold content of the films as determined by AAS analysis is given in Table 11. The second column expresses the gold content as wt 96 of the total sample, including both the aluminum substrate and the alumina film. Scanning and transmission electron micrograph^^^ of sections of the samples indicate the thickness of the alumina to be approximately 1.2 pm for all samples anodized in phosphoric acid. TEMs of samples containing gold (e.g., samples 30200si and 60200si) indicate that for the range of deposition times used, the metal is concentrated in a region of average thickness approximately equal to 300 nm near the aluminum/alumina interface. Values of the gold volume fraction, q, were calculated by assuming thevalue of d2 to be 300 nm, independentof deposition time. These are given in the fourth column of Table 11. The weight percent of deposited gold and the volume fractions calculated from them increase with deposition time. Moreover, the quantity of deposited gold is found to be independent of the substrate thickness. Transmission Electron Micrographs of Isolated Gold Particles. TEM images of gold particles liberated from the oxide films of a series of samples (152OOsi, 30200si, 60200si, and 90200si) prepared with increasing metal deposition time were obtained with magnifications ranging from 200 OOOX to 1 600 OOOX. Images of other samples were also recorded. Two examples are
Figure 2. Transmission electron micrograph of gold particles dissolved out of sample 60200si.
Figure 3. Transmission electron micrograph of gold particles dissolved out of an anodic film prepared in oxalic acid.
shown in Figures 2 and 3. Clustering was observed in all of the TEM images recorded. Since large aggregates are also observed in the TEM images of film sections,27it is not unreasonable to conclude that at least some of the clustering already exists within the anodic alumina film. The cross sections of most of the gold particles imaged by TEM are found to be elliptical rather than circular. However, several of the smaller particles appear to have circular cross sections. TEM. images of silver particles embedded in gelatin obtained by Skillman and Berry” displayed topographies almost identical with those observed in the present study with gold. Based on the various projections observed in their micrographs, they concluded that their silver particles were prolate spheroids randomly disposed rotationally. BecauseTEM micrographs show only two-dimensional projections it is difficult to determine the shapes of the particles unequivocally. However, since we do not observe very large circles,as would be expected for oblate particles,
Gold in Anodic Aluminum Oxide Films
The Journal of Physical Chemistry, Vol. 97,No. 32, 1993 8499
TABLE IIk Dimensions of Gold Particles Liberated from Aluminum Oxide Determined by TEM sample avg rad (A) smallest rad (A) largest rad (A) 152OOsi 15.0 f 2.2 10.6 19.8 302OOsi 14.0 f 2.7 11.1 19.3 60200si 25.3 k 5.7 19.5 42.3 90200si 23.4 f 4.9 16.8 33.7
I
I
1-
I
I
I
Tf
1
1
WAVELENGTH (n m )
Figure5. Visible reflectance spectra (measured-; calculated - - -) for WAVELENGTH ( n m )
Figure 4. Visible reflectance spectra (measured -; calculated - - -) for the following samples: (a) 30100si, s-polarized;(b) 30100si, p-polarized;
the followingsamples: (a) 3050si, s-polarized; (b) 305Osi, gpolarizcd; (c) 3020Osi,s-polarized,(d) 302OOsi,p-polarized; (e) 152OOsi,s-polarized, (f) 152OOsi,p-polarid, (g) 602OOsi,s-polarized;(h) 60200si, ppolarizcd.
(c) 30500.4, s-polarized; (d) 30500si, p-polarized.
I
I
while several small circles and large ellipses are observed, it is very likely that the particles are prolate spheroids. The average dimensions of the gold particles were determined from the TEM images measuring the diameters of between 5 and 15particles. Clusteringis anobvioussourceof error here, resulting in a larger apparent value of the average size. The results, given in Table 111, are converted to the average equivalent spherical radius for each sample for ease of comparisonwith values extracted from the reflectance spectra. The average particle radius is found to increase with increasing deposition time (Table 111). The particle size distribution is also altered with deposition time, the dimensions of the largest particles increasing rather dramatically as deposition times increase. Additionally, the particles become more eccentric as the deposition time was increased. None of the particle diameters exceed the 500-A average pore diameter as measured from TEM images of film sections.27 Reflectance Spectroscopy. The reflectance spectra recorded for the samples listed in Table I are shown in Figures 4-6. An additional series of spectra were recorded with sample 3050si at various angles of incidence. Only one of these (45') is reproduced here (Figure 5a). However, the results of the analysis of this series will be presented below. In Figures 4-6, the lower-most spectrum is reproduced as recorded. All others are progressively shifted by 0.2 units for clarity. Fits of the measured reflectance spectra to eq 11 were carried out by first obtaining the best fit by adjusting the specular fraction, S, or S,, the gold volume fraction, q, and the thickness of the two oxide layers, dl and dz, while the average particle dimensions, B and A, and the refracture index of the alumina, nA, were held constant at plausible values based on previous work.l5 The three latter parameters were, then, successively allowed to vary, and the parameters that produced the best fit at each stage were used as startingvalues for the next. Thisstepwise procedurewas found to reduce the tendency for the fit to become trapped in local minima.
I
I
I
I
I
(0.20
i
0.50
0.40 W
t 0.20 I
-
c
. ' 7
WAVELENGTH ( n m )
Figure 6. Visible reflectance spectra (measured -; calculated - - -) for
the following samples: (a) 90200si,s-polarized;(b) 902OOsi, ppolarized.
Although nA, dl, q, and d2 were treated as separate adjustable parameters it had previously been shown4that the product qdz, the effective mass thickness of the gold, is a strong parameter while q and d2 independently are weak parameters, Le., almost equally good fits are possible over a range of values of these parameters so long as their product is fixed. A similar situation is true for the product nAd1, which determines the optical path length responsible for the interference. Hence no confidence should be placed in the values of these parameters separately but only in the appropriate pairwise products. Moreover, the values of S , and S , were not restricted to be less than unity in order to allow them greater lattitude to correct for the defficienciesof the model. As a result, S,and S, can no longer be interpreted strictly as the fraction of the light that is specularly reflected. The s- and p-polarized spectra were fitted separately because they appear to carry different information. The p-polarized spectra are more sensitive to the particle dimensions, due to the greater contribution of the size-sensitive SP absorption to spectra
Preston and Moskovits
8500 The Journal of Physical Chemistry, Vol. 97, No. 32, 1993
TABLE IV: Parameters Obtained from pPolarized Reflectance Spectra for Samples with 30-s Metal Deposition at Variable Sine Wave Freauency ellipsoid equivalent F q& nAd1 vol(A3) sphererad (A) sample frequency (Hz) S, 4 d~ (nm) dz(nm) B(A) nA p3050si p301OOsi p30200si p30500si
50 100 200
500
1.0835 2.1906 1.0239 1.8369
0.025755 0.03518 0.034846 0.032158
805.57 1087.14 1379.98 791.54
48.0
537.06 493.19 554.37 482.70
38.8
30.7 48.1
21.9 17.8 13.3 21.1
1.093 1.114 1.087 1.137
0.001035" 0.021437 0.002885 0.03074
14.0 878. 21oooO. 17.4 1210. 112000. 19.3 1500. 52400. 15.5 900. 20400.
36.9 30.0 23.2 36.6
112 points in this calculation, 147 points in the other calculations.
TABLE V Parameters Obtained from s-Polarized Reflectance Spectra for Samples with 30-s Metal Deposition at Variable S i e Wave Freauencv ellipsoid sample frequency (Hz) 50 s3050si s30100si 100 s30200si 200 500 s30500si
S, 1.4473 1.581 0.9129 1.5977
d~(nm) 0.03555 989.94 1555.27 0.0351 0.03978 1836.49 0.02007 1592.83
4
dz(nm) 410.15 319.29 343.79 473.79
B(A) A(A) 21.6 88.9 45.3 96.6
9.9 28.6 18.7 31.3
equivalent
vol (A3) sphererad (A)
nA
F
qd2
JIA&
1.058 1.019 1.029 0.993
0.00114' 0.0506 0.0216 0.08575
14.6 11.2 13.7 9.5
1047. 19300. 1584. 94700. 1889. 16100. 1582. 1220000.
16.7 60.9 33.7 66.4
112 points in this calculation, 147 points in the other calculations.
with this polarization (Figures 4-6). The s-polarized spectra were found to be more sensitive to the optical path in the anodic alumina overlayer (film B) due to the dominance of interference effects in these spectra (Figures 4-6). A. Reflectance Spectra Measured at Varying Angles of Incidence. The reflectance spectra of sample 3050si, measured at varying angles of incidence, illustrate the necessity of including film anisotropy in the analysis. The variance, F, which is a measure of the goodness of fit, decreased by 2-3 orders of magnitude when these spectra were fit using the present model as opposed to one which assumed the particles to be spherical and the film optically isotropic.15 Fitting the reflectance spectra measured for this sample at five angles of incidence, concurrently but separately for the s- and p-polarized samples, yielded the following parameters:19 nAd1 = 1873 nm from the s-polarized spectra; qd2 = 8.2 rim, A = 25.8 A, and B = 40.1 A from the p-polarized set. The average equivalent spherical radius obtained from the s-polarizedset (1 26 A) is much too large, illustrating the aforementioned point regarding the polarization sensitivity associated with the various parameters. The corresponding parameter obtained from the p-polarized spectra produced a value of 35 A, close to the value obtained in an earlier study,l5 which yielded a maximum spherical particle radius of 33 A, and in substantial agreement with the value determined by TEM and from surface area measurements' 1912 of electrochemicallydeposited transition metal particles in anodic alumina films. The average refractive index, 1.10, determined by the fit is a little low. Estimates of the refractice index of the porous anodic oxide may be made using the Lorentz-Lorenz equation which assumes that the quantity (n2- l)/(n2 + 2) is proportional to the film density. Assuming the anodic film to be composed primarily of y-alumina2s26with an average density 0.6 that of the bulk and taking the refractive index32 of solid y-alumina to be 1.63 one expectsthe refractive index of the anodicoxidetobeapproximately 1.3.
It is instructive to consider why the value of the refractive index, nA, is a weak parameter. This quantity affects the overall reflectance weakly through the Fresnel coefficients (eqs 5 and 6) and very weakly through 61,where it occurs as the product nAd1. It can have a potentially strong effect on Re through eq 1. It was shown4that for spherical particles the wavelength at which the SP absorption is maximum depends on the value of q and on nA. The volume fraction, q, is also independently affected by the total absorbance. Hence for spherical particles n A is a strong parameter as was shown in ref 15 where a value of 1.42 was reported for n A on the basis of reflectance measurements and assuming spherical particles. On allowing the particle geometry to change from spherical to spheroidal one introduces another
degree of freedom which strongly affects the position of the SP absorbance maximum. (In the present study it occurs at approximately 520 nm; the precise value is given by the maxima in the imaginary part of et in eq 1). Hence nA becomes a considerably weaker parameter and theaspect ratioof the particle takes over as the major determiner of the position of the SP, while the breadth of the SP absorption is primarily determined by the average dimension of the particles. Thequantity nAd1 is found to bea much more robust parameter. The overall anodic film thickness, dl d2 was measured by TEM to be approximately 1.2 pm, while d2 was found to be approximately 0.3 pm suggesting that dl = 0.9 pm, Combining this value with the value of n A estimated above, 1.3, suggests that nAd1 2 1.2 pm, close to the value obtained from the s-polarized spectra. In summary, the anisotropic optical model is better suited to the gold/anodic alumina films prepared for this study than the isotropic model used previou~ly.~~ Analysis of the p-polarized spectra yields more reliable values for the size and shape of the gold metal particles embedded in the compositefilm, while analysis of the s-polarizedspectra produces a better measure of the optical thickness of the alumina overlayer, nAdl. B. Varying Deposition Frequency. The parameters obtained by fitting the reflectance spectra of the samples prepared with a fixed metal deposition time of 30 s but varying frequency of deposition voltage (Figures 4 and 5) are summarized in Tables IV and V. The parameters that are considered strong are shown in bold letters as a mnemonic. The only robust parameters obtainable from fitting the s-polarized reflectance spectra (Table V) are the four values of nAd1, which fall between 1 and 2 pm. These are somewhat larger than the value previously estimated, except for the sample prepared at 50 Hz. However, they lie within the range previously reported for phosphoricacid anodic alumina films.2326.27The larger values observed with the higher deposition frequencies may suggest that there is some deposition of aluminum oxide during the ac electrolysis,as was previously suggested to o c ~ u r . TEM ~ . ~ ~images of sections of two of the samplesZ7indicate no significant increase in the film thickness during the metal deposition process, hence the increase in the value of nAd1 implies that there is deposition of optically dense material within the pores, probably in the form of amorphous alumina which could form during the anodic half cycles of the depositionprocess. The valuesof all of the parameters returned by the fit are included in the tables for completeness. Fitting the p-polarized spectra yields a number of robust parameters (Table IV);chief among them are the average particle dimensions and the metal content qd2, expressed as the equivalent mass thickness. Even this last parameter must be processed further in order to yield a value of the metal content approaching
+
The Journal of Physical Chemistry, Vol. 97, No. 32, 1993 8501
Gold in Anodic Aluminum Oxide Films
TABLE VI: Parameters Obtained from p-Polarized Reflectance Spectra for Samples with Variable Metal Deposition Time at 200
Hz
p152OOsi p30200si p60200si p902OOsi
15 30 60 90
1.7123 1.0239 0.3754 0.2392
0.06315 0.03485 0.01842 0.07214
1496.73 1379.98 1407.33 1690.39
283.17 554.37 802.42 236.53
22.0 30.7 64.6 73.4
10.3 13.3 17.9 16.0
1.148 1.087 1.078 1.183
0.01548 0.002885 0.003068 0.001335
17.9 19.3 14.8 17.1
1718. 1500. 1518. 1999.
20800. 52400. 312000. 360000.
17.1 23.2 42.1 44.2
TABLE VII: Parameters Obtained from p-Polarized Reflectance Spectra for Samples with Variable Metal Deposition Time at 2M H z
ellipsoid sample
time (s)
S,
4
dl (nm)
dz(nm)
B(A) A @ )
s152OOsi s30200si s60200si s90200si
15 30 60 90
1.4806 0.9129 0.3495 0.3793
0.3437 0.03978 0.01276 0.0564
1681.13 1836.49 1296.92 1319.28
168.53 343.79 621.76 237.72
15.1 45.3 138.7 202.5
quantitative significance. This will be dealt with below. The values of the parameters in Table IV imply that the gold particles are oblate with very uniform aspect ratios averaging 0.44 and an equivalent spherical radius ranging from 23 to 37 A. The average equivalent spherical radius of 31.7 A agrees well with those reported previously19as well as withvalues determined by applying other techniques. The oblate shape is surprising in view of the TEM images discussed above which suggest prolate particles. The average particle dimensions appear to be more or less independent of frequency, providing a definite answer to the question regarding the influence of the duration of the cathodic half cycle on the average particle size. Clearly particles that are nucleated in one half cycle continue to grow in subsequent half cycles. Likewise the frequency does not seem to influence greatly the quantity of gold deposited as measured by the parameter qd2, at least in the range 50-500 Hz. This is consistent with visual inspection of the samples which showed no obvious differences in color among the samples. C. Varying Deposition Times. The parameters obtained by fitting the measured reflectance spectra (Figures 5 and 6) of samples (1 5200si, 30200si, 60200si, and 90200si) prepared at varying metal deposition times (15-90 s), keeping the frequency of the deposition voltage constant at 200 Hz are given in Tables VI and VII. The nAd1 values obtained from the fits to the s-polarized spectra (Table VII) are in substantial agreement with the values derived for the previous samples (Table V). The slight decrease in the optical thickness of the alumina overlayer with increasing deposition time, if it is, in fact, significant, is unclear. Analysis of the p-polarized spectra provided information regarding the dependence of the total quantity of gold deposited and the variation of the aspect ratio and the mean particle size on the metal deposition time. ( i ) Spheroid Size. As before, the spheroids are found to be oblate. Their aspect ratio is found to decrease from 0.47 to 0.22 while their mean spherical radii increase from 17 to 44 A on increasing the metal deposition time from 15 to 90 s. That is, the parameters derived from fitting the p-polarized reflectance spectra imply that the particles become flatter and larger as metal deposition time increases. This increase in size is consistent with what was observed in the TEM images (Table 111). But the particle geometry, once again, is at odds with the implication of the TEM micrographs that the particles are prolate. The averagesize of the particlesdetermined from the reflectance spectra agree remarkably well with those obtained by TEM (Table 111); well within the uncertainities associated with the model. Even assuming the correctness of the overall form of eq 10, the precise definition of parameter X in terms of the average dimensions of the particles is equivocal. X has been variously defined4-15-34as the mean radius or 4 f the mean radius of the particle. In our analysis we choose X to be the number average
16.6 18.7 30.4 42.5
nA
F
4d2
DAdf
vol (A3)
equivalent sphererad (A)
1.08 1.029 1.003 1.114
0.0289 0.0216 0.000789 0.01169
57.9 13.7 7.9 13.4
1816. 1889. 1302. 1470.
15900. 161000. 244oooO. 730000.
15.6 33.7 83.6 120.4
of the semimajor axes of the spheroids. However, the volume average spheroid radius would be preferable. A more troubling problem is the apparent lack of accord between the shape of the spheroid derived from the reflectance spectra (oblate) and from the TEM images (prolate). While the reasons for this discrepancyare uncertain, a plausible explanation is that the gold particles are actually prolate but that their orientation within the pores is such that, on average, their (long) axes of revolution are perpendicular to the walls of the pores, i.e., perpendicular to the film normal. This peculiar orientation may arise from the increased deposition rate in the direction perpendicular to the pore wall in response to the greater electrochemical field component in that direction.2.26 In a cluster of particles a significant fraction may, therefore, be disposed in this manner causing the cluster to appear oblate. The model that we use does not allow for such an arrangement of particles. Modifying it to do so would introduce serious difficulties. The model assumes that the axis of revolution coincides with the normal to the film surface. Adopting an oblate shape for the “particle” may have been the model’s way of coping with the above particle disposition. Assuming that the gold particles are, in fact, prolate but arranged in the above manner one can calculate a modified value for the equivalent spherical radius by interchanging the meaning of semiaxesA and B. Using this approach, the average equivalent spherical radii of the metal particles in samples 15200si,30200si, 60200si, and 90200si are determined to be 13.3, 17.5,27.5, and 26.5 A in even better agreement with the values determined by TEM (Table 111). Interpreted in this way one concludes that the particles become larger and more elongated as the metal deposition time increases. (ii) Total Gold Content in the Maxwell Garnett Layer. The values of the parameter qdz determined by fitting the p-polarized spectra did not show the expected increase with increasing metal deposition times (Table VI) in contradiction both to the AAS results and to the visual evidence that samples darkened in color with increasing deposition time. Moreover, inspection of the spectra in Figures 5 and 6 indicates that the absorbance associated with the SP, which should depend on the value of qd2, increases with increasing metal deposition time. The problem seems to be associated with the specular fraction, S,, which variesdramatically with metal deposition time (Table VI). Attempts to fit the series of spectra with a fixed value of S , (and the corresponding s-polarized set with a constant value of S,) produced unsatisfactory fits. Clearly another approach was required. Disentangling the value of S, from the value of qd2 is a difficult problem since the reflectivity of the film system depends upon the value of qd2 through 62, and on q through the Fresnel coefficients. An approximate approach is possible which makes use of the fact that overall reflectance is proportional to thevalue of S,. The
Preston and Moskovits
8502 The Journal of Physical Chemistry, Vol. 97, No. 32, 1993 TABLE VIII: Comparison of Gold Content in Oxide Films Determined by AAS and by Reflectance Spectroscopy Corrected for Light Scattering sample
qdzdC
p152OOsi p30200si p60200si p90200si s152OOsi s30200si s60200si s90200si
17.88 19.32 14.78 17.06 57.92 13.68 7.93 13.41
S, 1.7123 1.0239 0.3754 0.2392 1.4806 0.9129 0.3495 0.3593
q(AAS) 0.06 0.11 0.14 0.17 0.06 0.11 0.14 0.17
qdzM 17.3 32.4 40.6 51.6 17.3 32.4 40.6 51.6
qdzcoK 6.3 18.8 35.8 47.8 46.4 16.4 38.7 43.4
light transmission, I , through a film of thickness d2 will depend upon the product of d2 and the absorption coefficient kr. The dependence of the transmission, I , on kfd2 will be given approximately by the Beer-Lambert absorption law:
Z = doexp{4rk+f2cos $/A]
(12) where 01 is a constant that depends on the film reflectivity, X and Io are the wavelength and the intensity of the incident radiation, respectively, and 4 is the angle of propagation in the film. To first order, 4 may be assumed to be equal to the angle of incidence, 45O, since n A is close to unity. Equation 12 may be rewritten to include a specular fraction,
s
Z = SZoaexp(\k)
(13)
where
0 = 4rk+f2cos $/A - In S For a dilute compositefilm composed of a dielectric with metal particles distributed uniformly within it, the absorptioncoefficient will be approximately proportional to the metal volume fraction q. Therefore kr = 84, where 8 will be only slightly dependent on the wavelength of the incident radiation, the particle dimensions, and the volume fraction, q. For the sake of simplicity, j3 will be taken to be constant. Therefore \k can be written as
0 = 4rj3qd2 cos # / A - In S
= (474 cos $/X)(qdzWn- X In S I ( 4 r p cos 4))
Hence qdZw" = qd?lC
+ x In S,where x = h/(4a@cos 6).
(14)
Using eq 14 a value of qd2corrected for the influence of S, may be determined from the values for qd2 given in Tables VI and VII. In applying this correction, values of kf were calculated from the absorption coefficients determined at 520 nm, the SP absorbance maximum, for the Maxwell Garnett film associated with sample 30200si. Separate kr values were determined from the s- and p-polarized spectra. Although only qd2 values derived from the p-polarized spectra are robust, the analysis was carried out for the s-polarized spectra for completeness. The two values of kr,assumed to be constant for all the samples, are found to be -0.3267 and -0.2397, respectively for p- and s-polarization at 520 nm. The negative signs arise from the manner in which kf is defined in eq 12. The values of 8 were calculated by combining these numbers with the value of q determined by AAS for sample 30200si. The values of qd2 corrected for the effects of S, and q d p K are given in Table VIII. The corrected values based on the p-polarized spectra indicate that the quantity of gold deposited increases with increasing deposition time. They are in fair agreement with
the corresponding values determined by AAS, as well as with previously reported results.15 Corrected values of qd2 obtained from the s-polarized spectra are included for completeness. While these are of the right order of magnitude they do not show the expected trend very markedly.
Conclusions Gold-filledanodic aluminumoxide filmsprepared in phosphoric acid were studied by a combination of visible reflectance spectroscopy, transmission electron microscopy both of film sections and of metal particles chemically extracted from the film, and by atomic absorption spectroscopy to determine total metal content. The TEM images imply that the gold metal particles are concentrated near the bottom of the pores as clusters of prolate spheroids. Analysis of the reflectance data in terms of a model that relates the width of the surface plasmon absorption to the particle dimensions suggests that the particles are oblate spheroids of approximately the same size as reported by TEM. This discrepancy was explained in terms of a model in which prolate particles grow with their major axes perpendicular to the pore walls producing oblate clusters of particles. The equivalent spherical radii of the gold particles increase with metal deposition time, from 14 to 25 A, for 15-90 s of deposition. These results were confirmed by transmission electron microscopy of gold particles which had been removed from the film by dissolution of the oxide. When the frequency of the ac metal-deposition voltage was increased from 50 to 200 Hz, there was no apparent change in the metal particle size. The incorporation of gold into the pores of the anodic oxide as clusters of rather separate particles contrasts with the situation found for other metals. Transition metals such as Ni and Fe are found to form compact wire-like deposits which can almost fill the oxide pore.277'J Calculations in this study also determined that the equivalent mass thickness of deposited metal, qd2, was more or less independent of depositionvoltage frequency, but the gold content was found to increase with increasing deposition time. Atomic absorption measurements showed that the metal contents determined from the reflectance spectra were substantially correct. Finally, the effective optical thickness of the metal-free oxide overlayer, that constitutes the major portion of the film, as determined from the reflectance spectra is found to be in acceptable agreement with the values determined by TEM.27 Acknowledgment. The authors are grateful to NSERC and to The Centres of Excellencefor Molecular and Interfacial Dynamics for financial support. C.K.P. would like to thank theGovernment of the Province of Ontario for a scholarship. Thanks are also due to Dr. Diyaa Al-Mawlawi for valuable discussions and assistance and to Dr. Neil Coombs for his assistance in obtaining electron micrographs. References and Notes (1) Diggle, J. W.; Downie, T. C.; Goulding,C. W. Chem.Rev. 1%9,69, 365. (2) Thompson, G. E.; Wood, G. C. Anodic Films on Aluminum. In Treatise on Materials Science and Technology;Scully, J. C., Ed.;Academic Pres: New York, 1983; Vol. 23, Chapter 5 and references therein. (3) (a) Brace, A. W.;Sheasby,P.G. TechnologyofAnodizingAluminum; Technicopy: Stonehouse, 1979, p 197. (b) Sandera, L. Aluminum 1973,49, 533. (c)Sautter, W.;Ibe,G.;Meier,J.Aluminum1974,50,143. (d)Doughty, A. S.; Thompson, G. E.; Richardson, J. A.; Wood, G. C. Trans. Insf. Metal Finish. 1975. 53, 33, 161. (4) Goad, D. G. W.; Moskovits, M. J. Appl. Phys. 1978.49, 2929. (5) Moskovits, M. US.Patent 5,202,290. (6) Haruta, M.; Kobayashi, T.; Sano, H.; Yamada, N. Chem.Soc. Jpn., Chem. Lett. 1987,405. Miller, D.; Moskovits, M. J. Am. Chem. Soc. 1989, 1 1 1 , 9250. (7) (a) Kawai, S.; Ueda, R. J. Electrochem. Soc. 1975, 122, 32. (b) Kawai, S.In Proceedings of the Symposium on Electrochemical Technology in Elecrronics; Ramankiw, L. T., Osaka, T., Eds.; Electrochem. Soc.: Pennington, NJ, 1987; Vol. 88-23, p 389. (c) Tsuya, T.; Tokushima, T.; Wakui, Y.; Saito, Y.; Nakamure, H.; Hayano, S.; Furugori, A.; Tanaka, M. IEEE Trans. Magn. 1986, MAG-22, 1140. (d) AlMawlawi, D.; Moskovits, M. J. Appl. Phys. 1991, 70, 4421.
Gold in Anodic Aluminum Oxide Films (8) Wefers, K.; Evans, W. T. Plat. Surf. Finish. 1975, 62, 951. (9) Furneaux, R. C. Trans.Insr. Metal Finish. 1983,61,35. Fumeaux, R. C.; Thompson, G. E.; Wood, G. C. Corros. Sci. 1978, 18,853. (10) Pontifex, G. H.; Zhang, P.; Wang, Z.; Haslett, T. L.; AIMawlawi, D.; Moskovits, M. J. Phys. Chem. 1991, 95, 9989. (11) Moskovits, M.; Miller, D.; Darsillo, M.; Preston, C. Unpublished results. (12) Moskovits, M.; Al-Mawlawi, D.; Pontifex, G.; Preston, C. Unpublished results. (13) Granqvist, C. G.; Hunderi, 0. Solid Stare Commun. 1976, 19,939 and references therein. (14) Granqvist, C. G.; Hunderi, 0. Phys. Rev. B 1977, 16, 3513 and references therein. (15) Preston, C. K.; Moskovits, M. J. Phys. Chem. 1988, 92,2957. (16) Cohen, R. W.; Cody, G. D.; Coutts, M. D.; Abelts, B. Phys. Rev. B 1973,8,3689. Cohen, R. W.; Priestly, E. B.; Abelts, B. Phys. Rev. B 1975, 12, 2121. (17) Kreibig, U.; Althoff, A.; Pressmann, H. Surf. Sci. 1981, 106, 308. Kreibig, U.; Genzel, L. Surf. Sci. 1985,156,678. Kreibig, U.; Quinten, M.; Schonauer, D.; Genzel, L. Surf. Sci. 1985, 156, 741. (18) Preston, C. K., MSc. Thesis, University of Toronto, 1986. (19) Preston, C. K., Ph.D. Thesis, University of Toronto, 1990,University Microfilms, Ann Arbor, MI. (20) Maxwell Garnett, J. C. Philos. Trans. R. SOC.London A 1904,203, 385. Maxwell Garnett, J. C. Philos. Trans.R. SOC.London A 1906,205,247.
The Journal of Physical Chemistry, Vol. 97, No. 32, 1993 8503 (21) Galeener, F. L. Phys. Rev. Lerr. 1971, 27, 421. (22) van de Hulst, H. C. tight Scattering by Small Particles; Dover: New York, 1981. (23) Heavens, 0.S . Optical Properties of Thin Solid Films; Dover: New York, 1965. (24) Mosteller, L. P.; Wooten, F. J. Opt. Soc. Am. 1968, 58, 511. (25) Dignam, M. J.; Moskovits, M.; Stobie, R.W. J. Chem.Soc.,Faraday Trans. 2 1971, 67, 3306. (26) @Sullivan, J. P.; Wood,G. C. Proc. R. Soc. London A 1970,317, 5 11 and references therein. (27) Preston, C. K.; Moskovits, M. To be published. (28) Johnson,P. B.; Christy, R. W. Phys. Rev. B 1972, 6, 4370. (29) Kittel, C. Introduction to Solid State Physics; Wiley: New York, 1976. (30) Schulz, L. G. J. Opr. Soc. Am. 1954, &A, 357. Schulz, L. G. J. Opt. Soc. Am. 1954,44B, 362. (31) Skillman, D. C.; Berry, C. R. J . Chem. Phys. 1968,48, 3297. (32) CRC Handbook of Physics and Chemistry, 60th ed.; CRC Pras: New York, 1979. (33) Tachihara, K.;Itoi, Y.Spectrochim. Acta 1981, 26, 1299. (34) Kleeman, W. Z . Physik 1968,215,113. Kreibig, U. J.Phys. F 1974, 4,999. Moskovits,M.; McBreen,P. J. Chem.Phys. 1!?77,68,4992. Moskovits, M.; Dignam, M. J. J. Chem. SOC.,Faraday Trans. 2 1973, 69,65.
