On the porosity of coldly condensed sers active Ag films

Surface Science 150 (1985) 367-385 North-Holland, Amsterdam

367

ON THE POROSITY OF COLDLY CONDENSED SERS ACTIVE FILMS 1. Characterization of the films by means of Xe adsorption

Ag

E.V. ALBANO

*, S. DAISER

Institut ftir Physikalische Rep. of Germany Received

15 August

**, R. MIRANDA

*** and K. WANDELT

Chemie, Universitiit Miinchen, Sophienstrasse

1984; accepted

for publication

9 October

I I, D - 8ooO Miinchen 2, Fed.

1984

In this paper we report on the geometric structure of Ag films, deposited under UHV conditions and annealed at temperatures (T,,) ranging from 58 to 430 K, as deduced from UPS, AES, TDS and work function measurements of adsorbed xenon. The macroscopic work function of the bare films increases continuously from 4.25 eV (T,, = 60 K) to 4.72 eV (T,, = 330 K). Evidence is provided that coldly deposited Ag films are highly porous and that the pores persist up to T,, = 170 K, but are irreversibly healed between 170 and 250 K. The minimum thickness of the evaporated Ag films needed to develop these pores is found to be 50 A. The width of the pores, which are most likely intercrystalline gaps, is estimated to be 5-15 A. Besides the “macroscopic” pores the films contain atomic scale defects, which, in contrast to the pores, are healed continuously with increasing T,,. Films annealed at 330 K are composed of (111) grains with still a few percent of defect sites. The implications of these structural features on the adsorption properties of pyridine as well as on the interpretation of SERS results from such Ag films are dealt with in part II of this work.

1. Introduction and message Raman cross sections of molecules adsorbed on metallic surfaces can be orders of magnitude larger than for the same molecule in the liquid or vapor phase [l]. The enhancement mechanisms producing this Surface Enhanced Raman Scattering (SERS) are under heavy debate. At least two contributing effects need to be considered in order to explain the overall enhancement [2,3]. The “classical enhancement” due to electromagnetic resonances of surface plasmons in a rough metal substrate accounts for the major part of the

* Permanent address: Instituto de Investigaciones Fisicoqulmicas Te&icas y Aplicadas (INIFTA), La Plats, Argentina. ** New address: Atomika T&n&he Physik, D-8000 Munchen 19, Fed. Rep. of Germany. l ** Permanent address: Departamento de Fisica Fundamental, Universidad Autonoma de Madrid, Madrid, Spain.

0039-6028/85/$03.30 0 Elsevier Science Publishers (North-Holland Physics Publishing Division)

B.V.

368

E. V. Albano et al. / Porosity of Ag films

I

observed effect. This classical enhancement depends only on the incident light and the structure of the metal surface under consideration, but is independent of the adsorbed species. Therefore a second contribution is required which accounts for the observed adsorbate specificity of SERS. This additional mechanism is based on the enhancement of the molecule-metal effective Raman polarizabilities through charge transfer excitations. Hence, it is a short-ranged mechanism which depends on the specific metal-molecule interactions. Both mechanisms and in particular their relative contributions to the overall enhancement have been extensively discussed in the literature [1,4,5]. Despite this controversy in the interpretation of the data, general agreement exists about the crucial role of surface roughness in SERS [5]. The relevant scale of roughness, however, remains unclear [7]. Many different kinds of surface protrusions ranging from micro-crystallites of few hundred A in diameter [5,6] to atomic scale roughness such as ad-atoms [5] have been taken responsible for the enhancement. Pyridine adsorbed on silver is the classic example of a system exhibiting SERS [1,5,8-lo]. In particular, strong enhancement of Raman signals has been reported from pyridine adsorbed on coldly condensed Ag films under UHV conditions. The SERS activity of these films, however, disappears irreversibly after warming the samples to room temperature [5,8-lo]. Surprisingly, no fruitful attempt to characterize experimentally the structure of the Ag films evaporated onto cooled substrates under UHV has been reported so far. The aim of this work is to characterize the structure of Ag films annealed in a wide range of temperature (58-430 IS) by means of Xe adsorption studied with surface techniques, such as Ultraviolet Photoelectron Spectroscopy (UPS), Auger Electron Spectroscopy (AES), Thermal Desorption (TDS) and Work Function (A+) measurements. The advantages of using Xe as probe atoms are the sensitivity of PAX (Photoemission of Adsorbed Xenon) towards the local work function, the high surface mobility of the Xe atoms even at temperatures below 50 K as well as the low and site distinctive desorption temperatures [ll], which allows desorption of the Xe probe atoms without irreversibly changing the respective Ag film structure. In order to help the reader to understand the wealth of the following results we want to adapt an uncommon procedure and present our conclusions first. In the light of these conclusions the great variety of data is easier to comprehend. Our message is the following: Ag films evaporated or annealed at temperatures below - 170 K contain pores which can host a certain amount of adsorbed xenon, depending on the thickness of the films and hence, the depth of the pores. Warming such low temperature condensed Ag films beyond - 250 K leads to irreversible annihilation of the pores due to silver self-diffusion. The data to support this structure model are given in section 3 of the present work and will be discussed in more detail in section 4. - The adsorption of pyridine on these well characterized Ag films and the relevance of our

E. V. Albano et al. / Porosity of Ag films. I

findings for the interpretation of existing SERS data [12] will be discussed detail in part II following this paper [13].

369

in

2. Experimental details The experiments were carried out in a stainless steel UHV apparatus (base pressure 5 x lo-” Torr) equipped with AES, UPS, A+ and TDS facilities as has been described previously [14]. The sample could be cooled down to - 40 K by means of a closed cycle helium refrigerator [15] and heated up to 1250 K by resistive heating. The temperature of the sample was measured with a chromel-alumel thermocouple spotwelded to the rear of the substrate. As reference point one of the junctions of the thermocouple was held at 0°C outside the chamber. A polycrystalline Pd foil (nominal purity of 99.999%) carefully cleaned by sucessive cycles of sputtering with Xe and treatments with oxygen to remove S and C has been used as substrate. Spectroscopically pure Ag (Demetron) was evaporated from a tungsten filament. The angle of incidence of the Ag vapour atoms with respect to the plane of the substrate was - 90”. Film thickness and evaporation rate were calibrated by measuring the attenuation of the Pd (substrate) Auger peaks caused by Ag adsorption using known mean free path lengths for the Auger electrons involved. After this calibration procedure a very thick layer of Ag (- 1 pm) was deposited onto the Pd foil at 400 K. All subsequent films employed for the adsorption experiments reported in this paper were freshly grown on top of this well annealed thick Ag film, and were free of contaminants (C, 0, S, etc) as judged by the Auger spectra recorded after deposition. Auger spectra were recorded with a single pass CMA (Varian) with coaxial electron gun. With UPS photoelectrons were detected between - 20” and - 65” off normal. In all experiments reported in this paper the temperature of deposition (Td) of the films was T, = 58 K. The films were subsequently annealed at selected annealing temperatures (T,,) for 10 min. For the annealing procedure the temperature of the substrate was raised from Td up to T,, at a constant rate of 4 K/s. As deduced from the adsorption and desorption properties the structure of films evaporated at Td= 58 K and subsequently annealed for 10 min at any T,,> Td was found to be the same as that of films evaporated directly at T,,. All the films were grown to a thickness of - 150 A with a condensation rate of - 25 A/mm, except those used to study the thickness dependence explicitely. Xenon (Messer Griesheim, 99.999 ~01%) was admitted into the chamber through a variable leak valve. The reported Xe exposures are not corrected for the gauge sensitivity factor 2.8 for Xe.

E. V. Albano et al. / Porosity of Ag films. I

370

TDS experiments were performed with a small-aperture differentially pumped quadrupole mass spectrometer in order to avoid detection of desorbing particles from the holder or the backside of the sample.

3. Results 3. I. Clean films The dependence of the photoelectric work function C#Iof clean Ag films (150 A thick) on the annealing temperature is shown in fig. 1 for two different films. The sample was held at the various temperatures for 10 min and then allowed to cool down again before C#Jwas measured. C#B continuously increases upon increasing Tan up to - 330 K as generally observed with evaporated metal films [16-l& and references therein]. The surface roughness of the films deposited at low temperature is usually taken responsible for their low work function due to the Smoluchowski effect (191. The irreversible elimination of defects (adatoms, small culsters, etc.) as well as the growth of close-packed faces taking place when raising T,, is then supposed to produce the work function increase [16,20]. For polycrystalline Ag samples +-values of 4.28 eV [21] and 4.26 eV [22] have been reported in excellent agreement with our results for films deposited at low temperatures. On the other hand, for films annealed at T,, = 330 K the work function was $J = 4.72 eV, which is virtually identical to the +-values reported for the Ag(ll1) single crystal surface [23], namely ~(111) = 4.74 & 0.02 eV. This result is not surprising since it is known that the crystallites of polycrystalline Ag films are oriented with the closepacked (111) plane nearly parallel to the surface [24]. A further increase of T,,

L 7‘L62” e&5_

-Agllll) -

Ag (1001 9/G

-AglllO)

t L L&3_

/.

0,’

, dd d.

4,

.

P .*

l’

‘0

,c

= 9.’ z &2150A Ag FILMS I? 4 lny a ~““‘,‘TTn,,,....,“‘.,“,‘,,,,,,,,, 3 50 150 350 250 ANNEALING TEMPERATURE -

Tan [K]

Fig. 1. The dependence of the macroscopic work function (A@) of clean Ag films on the annealing temperature. The lines show the work function of Ag single crystals taken from ref. [22] for comparison, Filled and open circles correspond to two different samples.

E. K Albano et al. / Porosity of Ag films.

I

371

from 330 to - 430 K causes the work function of the fiims to decrease again from 4.72 to 4.52 eV. This finding agrees with structural rearrangements in Ag films upon further increase of Ten as observed by in situ X-ray measurements [24]. In fact, Ag films grown onto a glass substrate and annealed at q, = 570 K show some [lOO] texture [24]. This explains the re-decrease of # (fig. 1) from #(ill) for temperatures T,, > 330 K. Recently Eickmans et al. [25] have reported that the work function of coldly deposited Ag films is 4.4 f 0.1 eV, independent on Ta,,. This result is in obvious disagreement not only with ours but also with practically all available experimental data on the T,, dependence of the work function of evaporated metal films [l&18, and references therein]. The angle integrated UPS spectra obtained using unpolarized light (hv = 21.2 eV) exhibit only minor changes in the d-band emission from the pure Ag films as a function of the annealing temperature between 58 and 430 K. Fig. 2 shows Ag 4d valence band spectra from films annealed at 60 and 330 K respectively together with the difference curve. 3.2. Adsorption of xenon

3.2.1. UPS spectra Fig. 3 shows 5p,,,, 5p,,, 150 A Ag films annealed

2.83

383

1.03

5.83

UPS difference spectra of Xe adsorbed at 58 I( on at 170 and 250 K, respectively. Because of the

6.83

783

Ek [eVl Fig. 2. Augular integrated UPS spectra of the 4d valence band of two Ag films annealed at (a) 58 K and (b) 430 K, respectively. The difference spectrum (c) shows only minor differences.

312

E. V. Albano et al. / Porosity of Ag films. I

somewhat uncertain shape of the broad and double-peaked 5p,,, signal [26,31] at lower binding energies, in this work we draw conclusions only from the total 5P,,D 5P,,, intensity as well as from the shape and energetic position of the 5P ,,* signal alone. The salient feature of the spectra from the Ag film annealed at T,, = 170 K (fig. 3a) is the fact that 5p intensity is clearly observed only beyond 7 L (1 L = 10v6 Torr s) Xe exposure (at 58K). Below 7 L the adsorbed Xe is obviously inaccessible to UPS. The full width at half maximum (FWHM) of the Xe 5p ,,2 peak is 0.70 eV, which is very broad compared to the value of 0.35 eV (0.24 eV) found for Xe adsorbed at the same temperature on homogeneous Ag(100) (Ag(ll1)) surfaces [27]. The Xe 5p,,, binding energy in the limit of zero intensity is E,,F = 7.41 eV, and this value shifts continuously towards higher binding energies by 470 meV upon increasing the coverage up to completion of the first monolayer. (Saturation of the first monolayer is clearly indicated by the sudden emergence of a new 5p,,, peak at even - 0.5 eV higher binding energy characteristic for emission from the second Xe layer [28].) Very similar results are obtained for Ag films annealed at 58, 100, 120 and 130 K prior to xenon adsorption. In particular, in all these cases again a minimum exposure of 6-8 L is required before the onset of Xe 5p emission can be detected with UPS. Fig. 3b, shows Xe 5p,,,, 5p,,, spectra for increasing exposures to a Ag film annealed at Tan= 250 K prior to Xe adsorption at 58 K. In contrast to the observations with all films annealed at T,, G 170 K, 5p emission is now readily observed after an exposure of only 0.5-1.0 L Xe. In the limit of zero intensity (= zero coverage) the 5p ,,* peak maximum emerges here at E& = 7.49 eV. This value increases by 310 meV up to completion of the first monolayer and

To,= 170 K

T,d=58K

Ih L.87

6.87

8.87 BINDING

L.87 ENERGY

-

I 6.87 [eV

I 8.87

.--I

I

difference spectra of xenon adsorbed on Ag Fig. 3. He I excited 5p,,,, 5p,,, UV photoemission films annealed at (a) 170 K and (b) 250 K. Note the different minimum Xe exposure to first detect Sp intensity.

E. V. Aibano et al. / Porosity of Ag films. I

373

the FWHM is 0.60 eV. Again very similar results were obtained with films annealed at temperatures higher than 250 K, namely at 300, 330 and 430 K. After annealing at 430 K the FWHM = 0.5 eV is still higher than on the Ag(l11) (0.24 eV) or the Ag(lO0) face (0.35 eV) [27]. Finally we just mention (without showing spectra) the results obtained with a Ag film 0 again annealed not higher than 58 K but which had only a thickness 5p,,, emission was immediately detectable after only of 10 A. Xe 5p,,,, 0.2-0.5 L Xe. The 5p,,, signal had a width of 0.7 eV and shifted in binding F - 6.95 eV (at the limit of zero coverage) to E& = 7.65 eV at energy from J?Z,,~ monolayer completion. As discussed in section 4 the described variations of the Xe 5p,,,, 5p,,, photoemission as a function of film thickness (10 versus 150 A) and annealing temperature (T,, G 170 versus T,, a 250 K) are indicative of strong variations in the porosity and surface roughness of the underlying films. Work function change rne~~r~rn~nts The work function changes A# induced by Xe adsorption

3.22.

at 58 K on 150 A Ag films annealed at five different temperatures are shown in fig. 4. The maximum work function decrease caused by the first monolayer of adsorbed Xe is observed on the film annealed at Tan= 300 K, namely A$ = - 530 meV. This value is slightly higher than the value of Acp = - 470 meV measured for Xe adsorption on a Ag(ll1) single crystal surface [28]. The difference may be explained in terms of the higher individual dipole moment of Xe atoms being adsorbed at surface defects such as steps and grain boundaries (e.g. refs. 127,291). Certainely 300 K are not high enough to

Xe /Ag

A$ ImeV

FILMS

, Tad

: 58 K

Ton 58K 0

1

:i++

,

\ 0

2

_

+1+

-500 4

6

c

8

10

12

lb

EXPOSURE

16

18

20

I I_1

Fig. 4. Xe induced work function changes (A+) as a function of Xe exposure to Ag films annealed at the indicated temperatures T,. Note the different initial slope of the curve for 7” < 130 K as compared to T, > 300 K.

eliminate all defects on this film (see section 3.2.4). In fact, the final Xe induced work function decrease on the Ag film annealed at 420 K is very similar to the value found on Ag(ll1) due to the more efficient healing of defects at this higher temperature. Also the steeper decrease of A+ for the film annealed at 300 K as compared to the one annealed at 420 K becomes clear from these arguments considering the higher adsorption energy of Xe atoms at defect sites (see below) and, hence, the selective popuIation of these sites at low exposures. Both A+ curves measured after annealing at 300 and 420 K deviate significantly from those recorded with films annealed at 58, 100 and 130 K in that they show a change in work function from the very beginning of exposure, whereas a minimum exposure of 5-8 L Xe is required to cause A+ to change on the latter three films. This distinct dependence on the temperature pretreatment parallels the observation that a minimum exposure of Xe was needed before any Xe Sp emission could be detected from the films annealed at Tan G 170 K (in contrast to those annealed at T.., > 250 K) as described in the previous section. The first adsorbed Xe atoms simply disappear into the pores. 3.2.3. Auger results The Xe(530)/Ag(351) .4uger intensity ratio as measured with three representative Ag films (150 .J$) annealed at 60, 170 and 300 K is piotted against the Xe exposure in fig. 5. These Xe uptake curves are in good agreement with the previous UPS and A# results obtained with the same films. For films annealed at T,, 6 170 K, exposures of 5-7 L Xe are needed in order to detect the onset of the Xe(530) Auger signal. Conversely, for films annealed at T,, G 250 K the presence of Xe on the surface was detected with AES right from the beginning (- 0.5 L) of exposure. The plateau reached on the 300 K film by - 15 L Xe

7

- XefAg

FILMS

Fig. 5. Variation of the Xe(530)/Ag(351) Auger intensity ratio as a function of Xe exposure at 60 K to Ag films annealed at the indicated temperatures. Saturation of the 300 K curve corresponds to monolayer completion on this film. Note the delayed onset of Xe(530) intensity from the films annealed at 60 and 170 K.

E. K Ahmo

et al. /

Porosity

ofAgfih.

I

375

corresponds to monolayer saturation on this non-porous film. The films annealed at T,, 6 170 K are still porous and, hence, reach this value only after an exposure of - 22 L. The behavior of films which are thinner than 150 A and which were held at 58 K is shown in fig. 6. A Xe uptake corresponding to 0.5 L is only detectable if the film thickness d is less than 50 A. In thicker films the pores may extend to a depth which is larger than the escape depth of the Xe(530/542) Auger electrons. For comparison fig. 6 includes a spectrum from 0.5 L Xe adsorbed on a non-porous (jr,, = 330 K) film of 150 A thickness. (Xe TDS spectra from Ag films with thicknesses 10 G d G 150 A are described in section 3.2.5.)

The most striking finding so far was that neither UPS and AES nor A+ measurements were able to detect the presence of < 7 L Xe on coldly deposited Ag films which are thicker than - 100 A and which were annealed not higher than 170 K. Evidence, however, that under these conditions xenon is, indeed, adsorbed is obtained from TDS. Fig. 7 shows selected series of thermal desorption spectra from 150 A Ag films annealed at 58 and 330 K, respectively. In particular fig. 7b demonstrates that Xe desorption from a porous film (T,, = 58 K) is detected throu~out the whole exposure range from 0.2 L (see inset) to 15 L. The same is true for all porous films with T,, Q 170 K and also, of course, the non-porous films with T,, z 250 K (see fig. 1 in ref. jl21). Beyond the qualitative finding that Xe in fact is adsorbed on the porous

0.5L

Xe

, Tad= 58K

e

d

Ton

loh

58K

rsoh

330K

&h2. 500

520

560

560

Ek,”

id]

Fig. 6. Xe(530/542) Auger spectra from xenon adsorbed on Ag films of different thickness d deposited at 60 K. For comparison the result from a 150 A film annealed at 330 K is also shown.

316

E. K Albano et al. / Porosity of Ag films. I

films even though it is undetectable with UPS, AES and A+ up to exposures of - 7 L the TDS spectra contain much more information about the detailed topography of the differently pretreated films. We first consider the Xe desorption traces from a non-porous film which had been annealed at 330 K prior to xenon adsorption at - 60 K. Fig. 7a clearly shows two desorption maxima, (rn and CY,.The au-state desorbs around Tp= 100K (FWHM = 18 K) but saturates already after an exposure of 2 L Xe. The cy,-state peaks at Tp= 83 K (FWHM = 5 K) independent on Xe coverage and continues to grow up to 10 L Xe (fig. 7a, spectrum g) corresponding to monolayer saturation. Shape and constant peak posirion of the cy,-state suggest first order desorption kinetics, in which case the desorption temperature Tp= 83 K can be converted into an activation energy of desorption Ed(a,)= 5.6 kcal/mol using the simple Redhead analysis [30] with a pre-exponential of v = lOI SC’ [28,31] and a heating rate p= 4 K .s-‘. With a perfect Ag(ll1) surface Behm et al. [28] found (beyond a certain minimum coverage) zero order Xe desorption with &(lll) = 5.2 kcal/mol in good agreement with equilibrium measurements by Unguris et al. [32] who derived 5.13 kcal/mol. The close agreement of Ed(aI) with E,(lll) suggests to assign (Y, to the desorption of Xe atoms from the flat, (Ill)-like (see section 3.1.) surface facets of the annealed (T,, = 330 K) Ag film. Zero order kinetics, which may be explained with desorption from the edges of large two-dimensional Xe islands at higher coverage, however, is not expected to develop on the film due to remnant defects which prevent island formation [27,28]. In fact, we associate the an-state with desorption of Xe

J I T -[Kl Fig. 7. Thermal desorption spectra of Xe from Ag films annealed at (a) 330 K and (b) 58 K and exposed to increasing amounts of Xe. The insets show the enlarged TDS spectra for low coverages in either case. The heating rate was 4K/s. The various desorption states, a, and aI, from the annealed film and B,, /3,,. Blr, from the unarmealed, porous film, are discussed in the text.

E. V. Albano et al. / Porosiiy of Ag films. I

311

atoms from defect sites. This assignment is supported by the higher desorption temperature corresponding to Ed (at,) = 7.0 kcal/mol and the fact that an saturates after 2 L Xe. Strongest support for this assignment, however, stems from Xe desorption traces from a low temperature Ag film of only 10 A thickness (see fig. 8). This film contains mainly atomic surface roughness (no pores yet) and consequently gives only one desorption peak at - 100 K (at,). A clear differentiation of defect sites by means of Xe adsorption energies has already been demonstrated with defect and vicinal surfaces of Ru [29] and Pd [14,33]. Quantitative comparison of the areas under the an-peak (fig. 7a, spectrum d) and the a,-peak (spectrum g) suggests - 2% of the Xe atoms of a complete monolayer to occupy defect sites (steps, grain boundaries, etc.), which have not been healed at T,,= 330 K. Xe desorption traces from 150 A films annealed at Tan= 58 K are very rich in structure as shown in fig. 7b. With increasing coverage, however, various desorption states can easily be disentangled. After low exposures (5 1 L) two (see inset of fig. 7b). At the states, /I, and p,r, can be clearly distinguished lowest coverage (0.2 L) the maximum of the p,-state occurs at Tp= 139 K. This /?, peak temperature shifts to lower values as the coverage is increased, and reaches T,(p,) = 110 K after an exposure of 5 L Xe. The /In maximum shifts from 105 K (0.2 L) to 95 K (1OL). This is the same temperature interval as found for the an-state above, and is therefore assigned to the same origin.

80

100 120 140 T IKI

Fig. 8. Dependence of the low coverage TDS spectra of Xe on Ag film thickness. The spectra were recorded after an exposure of 0.5 L Xe at Tsd = 6Oi2 K to films of the indicated thicknesses and annealed at T, = 60 K. For comparison a spectrum from a 150 A film annealed at 330 K is also shown. Heating rate: 4 K/s.

378

E. V. AIbano et al. / Porosity of Ag films. I

Beyond saturation of the &,-state at - 10 L a further peak /?,,, grows on the low temperature side (spectrum h, fig. 7b) with the peak maximum at 83 K similar to the at-state from fig. 7a. We like to make special note of the strictly successive population of states &, &i and &it in fig. 7b; & grows basically between 0 and 5 L, pit between 5 and 10 L and PIrl between 10 L and 15 L (up to - 25 L). A very similar sequence of TDS sectra was recorded with films annealed at T,, = 170 K, except for the fact that the low coverage onset of &-desorption was shifted to low temperatures by - 20 K, while T,(p,) at p,-saturation remained unaffected. It is the &-state which is inaccessible to UPS, AES and Alp measurements on 150 A Ag films which had been held below 7!,, = 170 K. p, will therefore be associated with desorption out of the pores (see section 4.2). This as well as the assignment of /3,, corresponding to desorption from surface defects on the surface is strongly supported by the dependence of the TDS behavior on the thickness of the low temperature deposited Ag film described in the following section.

3.2.5. Dependence of Xe TDS on Ag film thickness The dependence of both the TDS and AES (see section 3.2.2) spectra of Xe on the thickness d of Ag films annealed at T,, = 58 K has been investigated in the range 10 I d 5 150 A. These Ag films were condensated on top of a very thick well-annealed and, hence, flat Ag substrate film. A low Xe exposure (0.5 L) was chosen for this experiment in order to avoid the shifts and overlaps of the TDS peaks observed at higher coverages and therefore to facilitate the assignment of the different desorption states. The TDS spectra shown in fig. 8 are taken immediately after the corresponding AES spectrum from fig. 6. From the 10 A/58 K Ag film only a single desorption peak is observed at Tp = 97 K. Desorption from the 30 A/58 K Ag film occurs in two peaks, with the dominant peak maximum at 130 K stretching out to 160 K and a small peak at 110 K. From the 50 A/.58 K Ag film the major portion of Xe desorbs at Tr = 140 K again with a tail up to 160 K, while a small amount desorbs already at 110 K. Basically identical desorption spectra as this latter one were detected for all 58 K Ag films thicker than 50 A, independent of thickness. Note also from fig. 6 that for d 2 50 A the Xe uptake of 0.5 L is no longer visible with AES. In fig. 8 we have also included for comparison the 0.5 L TDS spectrum from fig. 7a from a 150 A Ag film annealed at 330 K. Note again that the an-state desorbs within the same temperature interval (considering coverage dependence) as the /3,,-state from the 30 and 50 A film as well as the sole desorption peak from the 10 A/58 K Ag film.

E. V. Albano et al. / Porosity of Ag films. I

379

4. Discussion 4. I, General remarks The structure and the growth mechanisms of evaporated metal films have been extensively studied in the past [24,34] and have been correlated with other physical properties such as conductivity and optical properties. It has been established that the morphology and topography of such films depend on (a) the nature of the substrate; (b) the substrate temperature during deposition (TY); (c) the rate of deposition; (d) the angle of incidence of the metal vapor atoms with respect to the substrate surface; (e) the pressure and composition of the gas phase; (f) the temperature of annealing (T,,) following deposition; (g) the rate of heating from T, to T,,; (h) the annealing time (sintering) and (i) the film thickness. In the present study most of the above mentioned parameters were held constant and have been specified in the experimental section. Only the annealing temperature Tan and in a few special cases the film thickness d were varied. It is generally accepted that for low values of T,, the discrete nuclei originated during the earliest stage of film condensation grow sideways incorporating atoms adsorbed on the substrate and outwards by direct capture from the vapor. Because surface self-diffusion is negligible, spaces between crystallites formed when they touch or overlap cannot be filled in. Therefore, the columnar crystallites are separated by gaps and the films are said to be porous [24]. Upon increasing Tan, when the surface self diffusion becomes appreciable the narrow gaps between crystallites are filled. Hence, normal grain boundaries are formed and the continuous film consists of adjacent columnar crystals. The filling of the intercrystallite gaps is generally accompanied by a sudden change in electrical resistivity, optical reflectivity etc. Our combined UPS, AES, A+ and TDS results from adsorbed xenon will allow us to design a quite detailed structure model of the coldly deposited Ag films, in particular, as a function of annealing temperature. 4.2. Assignment

of Xe adsorption states

The Xe desorption spectra suggest the distinction of basically three adsorption states which can be assigned to three different classes of adsorption sites present on the various Ag films. To indicate the origin of the TDS spectra we have labeled TDS maxima observed from films annealed at Tan z 250 K with (Y and those from films annealed at T,, < 170 K with fi. The dominant desorption maximum (or from a film annealed at T,, = 330 K occurs independent of coverage at 83 K (fig. 7a). This value and the calculated activation energy of desorption of 5.6 kcal/mol are very close to the corresponding values found for Xe desorption from a perfect Ag(ll1) single crystal surface [28,32].

380

E. V. Albano et al. / Parody

of Ag films.

I

Therefore the cut-state is assigned to Xe atoms adsorbed on flat, (Ill)-like surface regions. The au-state occurs between 105 K and 95 K and is assigned to Xe atoms adsorbed at defects on the surface. This assignment is supported by the sole desorption peak observed from a 10 A/58 K Ag film at 95-100 K (fig. 8). This thin unannealed film is expected to be only atomically rough without pores yet. Also the desorption peak from a slightly sputter-roughened (below 100 K) Ag(ll1) surface was shifted by - 10 K to higher temperature as compared to the perfect Ag(lll) face 1281. Finally, (r,t saturates at quite low dosages, which also decrease with increasing T,, (2 250 K), that is the dominant (Y, state still grows at the expense of ar,, as T,, increasingly exceeds 250 K. The prevalence of flat, (Ill)-like facets at the surface of films annealed at 330 K was also con.cluded from the macroscopic work function of these films. The desorption states & and &tt from the films held beiow Tan= 170 K emerge at the same temperatures as (it, and (Y, from the films annealed at Ta, > 250 K, and therefore are ascribed to the same adsorption sites as those, namely /3,i, corresponds to desorption from flat portions of films never heated higher than 170 K, while /3t, is due to desorption from atomic-scale defects on the surface of these films, such as steps, kinks, ad-atoms, vacancies etc. In contrast to the OLDstate #3t, = (rt, shows a slight shift to lower desorption temperature (from 105 to 95 K) as the population of this state increases (see fig. 7 and 8). This indicates a certain ~j~t~i~utio~of these defects sites as regards their adsorption energy. This distribution of surface defect sites manifests itself also in the very broad Xe 5p,,, p eak width of the UPS spectra. As shown in a‘series of earlier papers the 5p,,, UPS (He I) electron binding energy of adsorbed Xe atoms reflects the local work function of the respective adsorption site [11,29,35]. As demonstrated with a variety of regularly stepped surfaces [29,33,36] as well as with sputter-roughened metal surfaces [29,33,28J the electrostatic surface potential is IocuRy lowered at defect sites in agreement with the Smoluchowski electron smoothing effect at surfaces of different roughness [19]. Hence, the very broad 5p ,,2 peaks from Xe adsorbed on the rough Ag surfaces can be regarded to encompass emission from a distribution of defect sites of differing local work function. With increasing annealing temperature this distribution narrows somewhat as suggested by the decrease of FWHM (5p,,,,) from 0.7 eV (T,, = 58 K) to 0.5 eV (T,, = 430 K). The latter value, however, indicates that even after annealing at such high temperatures there is still some remnant atomic scale roughness on the surface (in agreement with the corresponding TDS spectra from figs. 7a and 8). For comparison FWHM (5~,,~) from Xe adsorbed on Ag(ll1) and Ag(lOO) is as small as 0.24 and 0.35 eV, respectively [27]. Even this difference in FWHM (5~t,~) on the two perfect Ag surfaces seems to be significant in view of the different mis-registry of the hcp Xe overlayer with the (100) and (111) structure, respectively. A very similar change

E. Y. A&am

et al. f Porosity of Ag jiims. i

381

in FWHM (Sp,,,) has been observed with Xe adsorbed on Pd(100) to different coverages [31], where up to a critical coverage corresponding to a Xe c(2 X 2) overlayer all Xe atoms may fit into equivalent fourfold hollow sites, from which they are necessarily displaced at higher coverages. (Note that at low coverages Xe tends to form hcp islands on silver but not on palladium.) The of interpretation of the broad Xe 5p,,,, signal as being due to a distribution defect sites is also supported by corresponding observations for Xe adsorption on a sputter-roughened Ag(ll1) surface as compared to the perfect (111) face 1281. The remaining Xe adsorption state on the coldly deposited Ag films is the p, state (figs. 7b and 8). This state is only observed with films which have never been heated higher than r,, = 170 K It is also exactly this state which is inacessible to AES (and UPS and A+) as long as these films are thicker than - 50 A and the Xe exposures are low. This becomes obvious from a comparison of figs. 6 and 8. Consequently & is assigned to Xe atoms being trapped in the pores of films annealed at T,, I 170 K owing to the high mobility of the rare gas atoms even at 50 K. Beyond the non-detectability of this state with AES, UPS and A+ below a critical exposure (5-7 L) and, hence, below a minimum filling of the pores, the assignment of the /I,-state is further supported by its rather high initial adsorption energy (- 9.5 kcal/mol), which is compatible with a very high coordination of the Xe atoms to the silver substrate as offered by sites inside the pores. In addition to the “macroscopic” pores there is plenty of atomic scale roughness on these films (see assignment of LY,,= pi, above) as well as in the pores, which produces a further strong adsorption site heterogeneity also within the pores and which explains the enormous width of the TDS spectra. The shift of the p, peak to lower temperature upon increasing the Xe exposure again indicates the successive population of adsorption sites within the pores according to the variation of Ead_For low exposures Xe adsorption begins at sites offering the highest degree of coordination, i.e. more likely at the bottom of the intercrystallite gaps. On raising the exposure, sites of increasingly lower coordination become occupied. Consequently, the pores are filled from the bottom to the mouth. This manifests itself in the strong coverage dependent decrease of Ead(/3,) from - 9.5 kcal/mol at 0.2 L to - 7.5 kcal/mol at 5 L Xe exposure. We like to mention that a contribution arising from diffusion limitation of the Xe desorption out of the pores can in principle also not be excluded. This should also contribute to a strong decrease in Tp (Pi) from 140 K to 110 K as the pores become more and more filled. A quantitative evaluation of the TDS intensities of the various P-desorption states yields the following picture: For Xe exposures I 10 L at which the completion of the first layer of the non-porous films annealed at T,,= 330 K takes place, all the TDS peaks from differently annealed samples (58 I T,, s 330 K) enclose the same area proving that the total amount of adsorbed Xe is

382

E. V. Albano et al. / Porosity of Ag films. I

independent of T,,. In contrast to the flat and well annealed films, however, the area under the desorption peak from porous films continues to grow with exposures > 10 L. For a film annealed at T,, = 58 K, the completion of the first monolayer is reached after - 25 L Xe (see figs. 3a, 5 and 7b). Taking the monolayer (ML) capacity of the flat film annealed at T,, = 330 K as unity, the total surface area of a 58 K Ag film is evaluated to be 2.8 k 0.4 times that of the geometric surface area. This “roughness factor” was found to decrease approximately linearly to unity with increasing T,,. The area under the /?, peak from a 58 K Ag film obtained after the critical exposure of - 7 L represents the amount of Xe adsorbed below the surface in deeper layers than the escape depth of the detected Xe photoelectrons or Auger electrons (- 20 A), since these Xe atoms cannot be detected either by UPS and AES or by A$ measurements. This value was estimated to be 0.7 f. 0.1 ML. Taking into account, that the total surface area of a 58 K film was “titrated” to be 2.8 ML, the remaining - 2 ML are distributed in the upper part of the pores as well as on the surface of the crystallites. 4.3. Conclusions: the structure of differently

annealed Ag-films

Our combined AES, TDS, A+, UPS and PAX data lead us to suggest the following structural properties of the coldly deposited and differently annealed Ag films. These films are characterized by two major types of surface imperfection: Firstly the large scale roughness (> 50-100 A), namely the Ag grains separated by the pores or gaps, and secondly the atomic scale roughness like steps, kinks, ad-atoms, vacancies etc. on the overall surface of individual Ag grains. The existence of pores is concluded from the luck of Xe induced UPS and AES signals as well as work function changes on these porous films, although TDS proves the presence of a strongly bound Xe species on (in) the films. The formation of pores in cold film deposition is also predicted from molecular dynamic calculations of this process [34]. The existence of atomic scale roughness follows from the adsorption site heterogeneity on the surface as suggested by the broad Xe 5p,,, UPS peak width indicating a corresponding distribution of local work functions across the respective surface as well as by the broad and coverage dependent /3, and /3,, = (Y,, desorption maxima. Both types of roughness exhibit quite a different dependence on the temperature T,, at which the films are annealed. Pores are present only in films which have been annealed not higher than 170 K; films annealed at T,, > 250 K no longer contain pores, that is, the pores are irreversibly annihilated between 170 and 250 K. This finding is in full agreement with earlier results on the electrical and optical properties of such films. For Ag films condensated at - 110 K the sudden drop of the electrical resistivity as they are warmed through - 210 K [37] indicates the onset of appreciable surface diffusion and the filling of the intercrystalline gaps. In spite of the quite different experimen-

E. V. Albano et al. / Porosity of Ag films, I

383

tal conditions employed by Allpress et al. [37] (background pressure of - lo-’ Torr, d = 1000 A., condensation rate - 30 A/s, glass substrate, etc.), their results are in remarkably good agreement with ours obtained under UHV conditions. A sudden change in optical properties of clean Ag films freshly condensed at 120 K, namely an irreversible loss of Raman scattered intensity intensity, was observed by Pockrand at 160 cm-’ as well as of the background and Otto [39,40], when warming the films beyond - 220 K. Conversely, the atomic scale roughness is continuously healed over the whole temperature range (58 c T,, G 430 K) and shows nothing like the abrupt disappearance of the pores between 170 and 250 K (that is around 220 K). This is concluded from the persistance of strong adsorption site heterogenity (as suggested by the Xe 5P ,,z UPS FWHW value as well as the TDS data) beyond room temperature as well as the approximately linear decrease of the total surface area (as measured in terms of Xe up take) up to 330 K. Typical average dimensions of the crystallites (grains) and the intercrystalline gaps of coldly deposited Ag films can also be inferred from the UPS, AES and TDS measurements with adsorbed Xe. Coadsorption experiments of Xe and pyridine [13] indicate that from a top view of the sample at most 10% of the surface correspond to the open intercrystalline gaps. The width of the gaps must be at least 5 A in order to host Xe atoms and pyridine molecules [13] of this diameter. The upper limit for the gap width is roughly estimated to be - 15 A. This estimate is based on the Xe uptake of the pores, the dependence of the porosity on the film thickness, i.e. the depth of the pores, and the fact that Xe photoelectrons and Auger electrons are hindered to leave the pores. From these arguments we arrive at the following average dimensions for the 150 A thick films: depth of the pores loo-120 A, size of the crystallites 50-150 A. This mean crystallite size is in very good agreement with the relevant scale of roughness deduced from SERS experiments on coldly deposited Ag films, i.e. - 50 A [8,38]. Our data demonstrate unambiguously that the pores of coldly deposited Ag films are heated out in the range 170 G T,, < 250 K. Precisely within this interval the SERS activity of the films is irreversibly lost [5,8-lo]. Diffusion experiments of pyridine into the pores, theoretical results on the classical electromagnetic enhancement of the Raman cross section within the pores as well as a survey of published SERS data lead us to conclude that the SERS active sites of coldly deposited Ag film are within the pores [12]. These topics will be discussed in more detail in part II following this paper [13].

5. Summary Combined UPS, AES, TDS and work function change data of adsorbed xenon yield the following structural informations about coldly condensated Ag

384

E. V. Albano et ai. / Por&ty

of Ag jih.s.

1

films. Ag films deposited at low temperatures onto a flat Ag substrate to a thickness of 150 A and subsequently annealed at temperatures 58 G Ta,,G 430 K are characterized by two major types of roughness. Films evaporated or annealed at T,, G 170 K are highly porous. The pores (intercrystalline gaps) are already fully developed after deposition of - 50 A. The width of the gaps is estimated to be 5-15 A producing an open surface of - 10%. These pores are irreversibly annihilated within the narrow interval 170 _< T,, s 250 K. Besides these “macroscopic” pores the Ag films contain atomic scale surface roughness on the crystallites as well as within the pores, the density of which decreases approximately linearly with T,,. A film annealed at Tan = 330 K exposes mainly (111) crystal facets with still a few percent of atomic scale defects. The implications of these structural properties on the interpretation of “Surface Enhanced Raman Scattering (SERS)” results from such Ag films are discussed in detail in the accompanying paper.

Acknowledgements Financial support of this work by the Deutsche Forschungsgemeinschaft (SFB 128) is gratefully acknowledged, Three of us (R.M. and E.V.A. and S.D.) thank the A.v. Humboldt-Stiftung, the Consejo National de Investigaciones Cientificas y Tkcnicas de la Repitblica Argentina (CONICET) and the Fond der Chemischen Industrie for respective fellowships. On of us (K.W.) also gratefully acknowledges fruitful discussions with T.J. Chuang and W. Seki from IBM, San Jose. The high purity Ag wire used for evaporation was a gift from DEMETRON.

References fl] R.K. Chang and T.E. Furtak, Eds., Surface Enhanced Raman Scattering (Plenum, New York, 1982). [2] T.E. Furtak and D. Roy, Phys. Rev. Letters 50 (1983) 1301. [3] K. Billmann and A. Otto, Solid State Commun. 44 (1982) 105. [4] T.E. Furtak and J. Reyes, Surface Sci. 93 (1980) 351. [S] A. Otto, in: Light Scattering in Solids, Vol. IV, Eds. M. Cardona and G. Giintherodt (Springer, Berlin, in press). [6J J.E. Rowe, C.V. Shank, D.A. Zwemer and A. Murray, Phys. Rev. Letters 44 (1980) 1170. f7] A. Otto, Phys. Rev. B27 (1983) 5231. T.H. Wood, Phys. Rev. B27 (1983) 5137. [8] T.H. Wood, Phys. Rev. B24 (1981) 2289. [9] I. Pockrand and A. Otto, Solid State Commun. 38 (1981) 1259. [lo] H. Seki, Solid State Commun. 42 (1982) 695. [ll] K. Wandelt, J. Vacuum Sci. Technol. A2 (1984) 802. [12] E.V. Albano, S. Daiser, G. Ertl, R. Miranda, K. Wandelt and N. Garcia, Phys. Rev. Letters 51 (1983) 2314.

E. V. Albano et al. / Porosity

[13] [14] [15] [16] [17] (181 [19] 1201 [21] [22] [23] 1241 1251 [26] (271 [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40]

of Ag films. I

385

E.V. Albano, S. Daiser, R. Miranda and K. Wandelt, Surface Sci. 150 (1985) 386. R. Miranda, S. Daiser, K. Wandelt and G. Ertl, Surface Sci. 131 (1983) 61. K. Wandelt, S. Daiser, R. Miranda and H.J. Forth, J. Phys. E (Sci. Instr.) 17 (1984) 22. J.M. Heras and E.V. Albano, Thin Solid Films 106 (1983) 275. P.E.C. Franken and V. Ponec, Surface Sci. 53 (1975) 341. B.G. Baker, B.B. Johnson and G.L.C. Maire, Surface Sci. 24 (1972) 572. R. Smoluchowski, Phys. Rev. 60 (1941) 661. J. H&l and F.K. Schulte, in: Springer Tracts of Modem Physics, Vol. 85, Ed. G. Hohler (Springer, Berlin, 1979). D. Dayal, P. Rudolf and P. Wissmann, Thin Solid Films 79 (1981) 193. Handbook of Chemistry and Physics, 63rd ed. (CRC Press, Cleveland, OH, 1982-83) p. E-78. A.W. Deweydari and C.H.B. Mee, Phys. Status Solidi (a) 17 (1973) 247. J.V. Sanders, in: Chemisorption and Reactions on Metallic Films, Ed. J.R. Anderson (Academic Press, London, 1971) pp. l-38. J. Eickmans, A. Goldmann and A. Otto, Surface Sci. 127 (1983) 153. J. Kuppers, F. Nits&k.+ K. Wandelt and G. Ertl, Surface Sci. 88 (1979) 1. A. Jablonski, S. Eder and K. Wandelt, to be published. J. Behm, C.R. Brundle and K. Wandelt, in: Proc. IVC-9 and ICSS-5, Madrid, p. 13; J. Behm, C.R. Brundle and K. Wandelt, in preparations. K. Wandelt, J. Hulse and J. Ktppers, Surface Sci. 104 (1981) 212. P.A. Redhead, Vacuum 12 (1982) 203. K. Wandelt and J. Hulse, J. Chem. Phys. 80 (1984) 1340. J. Unguris, L.W. Bruch, E.R. Moog and M.B. Webb, Surface Sci. 87 (1979) 415. K. Wandelt, J. Hulse, J. Ktppers and G. Ertl, in: Proc. ICSS-4 and ECOSS-3, Cannes, 1980, p. 104. H.J. Leamy, G.H. Gilmer and A.G. Dirks, in: Current Topics in Materials Science, Vol. 6, Ed. E. KaIdis (North-Holland, Amsterdam, 1980) ch. 4, pp. 311-344. J. Hulse, J. Kippers, K. Wandelt and G. Ertl. Appl. Surface Sci. 6 (1980) 453. S. Daiser and K. Wandelt, Surface Sci. 128 (1983) L213. J.G. Allpress and J.V. Sanders, J. Catalysis 3 (1964) 528. T.H. Wood, D.A. Zwemer, C.V. Shank and J.E. Rowe, Chem. Phys. Letters 82 (1981) 5. I. Pockrand and A. Otto, Solid State Commun. 38 (1981) 1159. A. Otto, Appl. Surface Sci. 6 (1980) 309.