Surface-Enhanced Raman Spectroscopy of Colloidal Metal Systems

I thank Dr. H. R. Ihle for his support of this work ... Froschen for his technical assistance during the mea- ... of SERS studies using molecules and ...
0 downloads 3 Views 569KB Size
1540

J. Phys. Chem. 1983, 87, 1540-1544

ments with well-established values.l' This suggests that the measurements were carried out under equilibrium conditions. Acknowledgment. I thank Dr. H. R. Ihle for his support of this work and Dr. R. 0. Jones for his valuable discus-

sions. Many thanks are due to Prof. Paul von Rag& Schleyer for his interest. Acknowledgment is made to F. Froschen for his technical assistance during the measurements. Registry No. Li4, 12596-49-5;Li4+,12769-77-6.

Surface-Enhanced Raman Spectroscopy of Colloidal Metal Systems: A Two-Dimensional Phase Equllibrlum in p -Aminobenzoic Acid Adsorbed on Silver J. S. Suh, D. P. DILella, and M. Moskovits' Department of Chemistry and Erlndale College, University of Toronto, Toronto, Ontario M5S TAT, Canada (Received September 1, 1982; I n Final Form: December 1, 1982)

The surface-enhanced Raman spectrum of p-aminobenzoic acid (PABA) adsorbed on aqueous colloidal silver particles is reported. From the concentration and temperature behavior of the spectrum, one deduces that a two-dimensional "gas-olid" equilibrium exits, the spectrum being a composite of the spectra of the two phases. The PABA molecule is deduced to be adsorbed flat on the surface, r-bonding through the benzene ring. The degree of enhancement is estimated to be in excess of 4 X lo6.

Surface-enhanced Raman scattering (SERS) spectroscopy is now a well-established technique whereby unusually intense Raman signals are obtained from molecules adsorbed on or placed near rough metal surfaces, periodic metal structures produced lithographically or by coating gratings, metal island films, and colloidal metal partic1es.l Strong signals have been obtained from but a few metals including the group 1A and 1B metals' and indium.2 Reports of enhancement by mercury3 and by cadmium4 have encountered some d i ~ p u t e .Colloidal ~ metal particles have been used by several groups to produce S E W spectra. Creighton et ala6investigated the spectra of pyridine on silver and gold sols, as well as the physical properties of those sols. Wetzel and Gerischer' performed similar experiments but with sols produced in a different fashion. Kerker and co-workers* measured the spectrum of the citrate ion on silver sol. In addition two reportsBhave appeared describing SERS measurements of CO, ethylene, and acetylene obtained with matrix isolated colloidal silver particles. In this paper we describe the SERS spectroscopy and the surface phase equilibrium deduced from it for the molecule p-aminobenzoic acid. This is the first of a series of SERS studies using molecules and systems of some biological importance. (1) R. K. Chang and T. E. Furtak, Ed., "Surface Enhanced b a n Scattering", Plenum, New York, 1982. (2)D. P. DiLella and M. Moskovits, unpublished. (3)R. Naaman, S.J. Buelow, 0. Cheshnovsky, and D. Herschbach, J. Phys. Chem., 84, 2692 (1980);L. A. Sanchez, R. L. Birke, and J. R. Lombardi, Chem. Phys. Lett., 79,219 (1981). (4)B. H.Loo, J. Chem. Phys., 76,5956(1981). ( 5 ) Attempts by several groups including the present one to reproduce the mercury results have met with considerable difficulty. Van Duyne (personal communication) has moreover suggested that the Cd result wa8 due to the growth of a thick layer of pyridine complex of Cd. (6) J. A. Creighton, p 315 in ref 1, and references contained therein. (7) H. Wetzel and H. Gerischer, Chem. Phys. Lett., 76, 460 (1980). (8)M. Kerker, 0.Siiman, L. A. B u " , and D.4. Wang, Appl. Opt., 19, 3253 (1980). (9)H.Abe, K.Manzel, W. Schulze, M. Moskovits,and D. P. DiLella, J. C h m . Phys., 74,792(1981);K . Manzel, W. Schulze, and M. Moskovite, Chem. Phys. Lett., 86, 183 (1982). OO22-3654/83/2087-154O$01.50/0

Experimental Section Silver sols were made according to the recipe reported by Creighton.'j A 60-mL solution of 2 X M sodium borohydride was mixed with a 22 f 2 mL solution of 1 X M silver nitrate; both solutions were chilled to ice temperatures. D20 colloids were made in exactly the same fashion, substituting D20 (98%) for water. The adsorbate was introduced into the colloid by adding one drop of a solution of alkaline (NaOH, pH 10) p-aminobenzoic acid (PABA) of varying concentration to 1.5 mL of (PH -7) colloid. The drop size was estimated by measuring the volume of 50 drops. In this way samples of low overall concentration in PABA could be made. The colloids were pale, opalescent yellow before PABA was added. Upon adding the adsorbate the color of the colloid changed to blue. Spectra were recorded with a SPEX 1401 monochromator equipped with photon counting and interfaced to a Tektronix 5041 computer. The sample temperature could be varied by placing the sample cuvet in a glass jacket through which dry air heated to the desired temperature was circulated. Results A series of surface-enhanced Raman spectra of PABA solution adsorbed on silver colloid are shown in Figure 1. Each of the spectra in Figure 1corresponds to a different overall concentration of PABA in solution. As the concentration increases several changes are observed in the spectra including the growth of bands of which the most easily observable is the one at 1452 cm-', and an increase in the slight fluorescent background. The change brought about by increasing concentration is best presented as a difference spectrum between spectra of two different concentrations. When such a difference is taken between two spectra of nearly the same concentration one notes (Figure 2) that the spectrum consists of two types of features: one type increases almost linearly (Figure 3) with increasing PABA concentration, the other remains almost unchanged or decreases very slightly. When the difference between two spectra is taken, one corresponding to a 0 1983 American Chemical Society

The Journal of Physical Chemistty, Vol. 87, No. 9, 1983 1541

SERS of Colloidal Metal Systems

Y W

-

50

9-

B>

t-

z

k L

z

-

0

300

700

I

0.22

1 0.26

Flgure 3. The intensity of the 1452-cm-' peak plotted against the concentration of PABA (arbitrary units).

1500

I100

1

1

0.14 0.18 CONCENTRATION

0.10

0.06

I

i

I

I

ih~icm-')

Flgure 1. A series of SERS spectra of PABA adsorbed on silver sol particles, excited with 514.5-nm argon ion laser radiation: A-F correspond respectively to 0.292, 0.590, 1.17, 4.67, 9.17, and 18.2 mM overall PABA concentration.

I

100

I

503

I

900

no (cm-')

I 1300

I

I I 700

Flgure 4. As in Figure 2 but with the indicated concentrations of PABA. loo

500

900

1300

1700

Oo(cm-'i Figure 2. Two SERS spectra of PABA adsorbed on silver sol particles, excited with 514.5-nm argon ion laser radiation. Bottom and middle curves correspond to 9.35 and 2.34 mM overall PABA concentration. The top curve is the dtfference between the bottom and mlddle spectra, respectively.

colloid sample dilute in PABA while the other is concentrated in PABA (Figure 4),one finds that the spectral features which before (Figure 2) were almost unchanged are now slightly negative. In both cases one should keep in mind that the fluorescent background also changes; hence, when one speaks of negative or positive changes it is with respect to the fluorescence background as baseline. Plotting difference spectra also revealed spectral features which were difficult to discern in the original spectrum. A particularly clear case in point is the band at 1139 cm-', which the difference spectra clearly shows to be composed of two closely spaced bands. Raman spectra were also taken as a function of temperature. Two such spectra are shown in Figure 5 together with their difference. In it we see that those bands which increased with increasing concentration also increased upon cooling the sample while those bands which did not change in intensity or decreased

slightly upon increasing the concentration decreased markedly upon cooling the sample. A Raman spectrum obtained with a silver colloid sample containing a very low PABA concentration is shown in Figure 6; Figures 7 and8 show the Raman spectra of PABA solution without silver colloid and of solid PABA, respectively. In addition, in Figure 9 the frequencies and assignments of the bands are included in order to facilitate the following discussion. Figure 10 shows the SERS spectrum of PABA added to deuterated silver sol. Ita main utility was to identify modes involving NH, motion since that group will be totally deuterated by hydrogen-deuterium exchange in a fully deuterated aqueous medium. A series of UV-visible spectrum of our sols, before and after addition of PABA, is shown in Figure 11. Before the adsorbate was added the spectrum consisted of a broad band which peaked at 385 nm with a weak shoulder at approximately 425 nm. After PABA was added the 385-nm peak was diminished in intensity and a broad absorption developed, centered at approximately 580 nm. With some colloids a color change from orange to blue (together with the concomitant growth of the 580-nm band) was not observed upon addition of PABA. No SEW spectrum was detectable under those circumstances. (This observation was generally the case with most adsorbates tried to date.)

1542

The Journal of Physical Chemistry, Vol. 87, No. 9, 1983

Suh et al.

' ' sC-0- SOLUTION

H2N-Q-

1

/

/

j

l

~

900

503

IO0

~lsOpcm_,,2100

I

~

I303

l

~

/

l

,

2500

,

2900

l

~

I

3300

Figure 7. Raman spectrum of PABA in aqueous solution in the absence of silver sol.

500

900 A 3 (cmll

I300

I 700

Flgure 5. SERS spectra of PABA (approximately 1 mM overall concentration) adsorbed on silver collold partlcles recorded at 12 and 43 O C and the difference of the two spectra.

I1 li 2.5 x

M

1

1

lo0

1

500

I

,

900

,

,

1300

,

,

17

.

/

2100

a 4TC m- 11

,

,

2500

,

,

2900

,

,

3300

.

Figure 8. Raman spectrum of solid, crystalllne PABA.

t t

100

500

900

I300

1x0

A V (cm-')

c I

Fbure 6. SERS spectrum of a low concentration of PABA adsorbed on sliver sol particles.

Discussion The Raman spectrum of PABA adsorbed on the surface of colloidal silver at higher concentrations is dominated by benzene ring vibrations; the most prominent features are vlga (1515 cm-'), vlgb (1452 cm-l) (Wilson notation will be used throughoutlo), vsa (1598 cm-'), and vg, (1172 cm-l), all in-plane benzene ring stretching vibrations. Next in order of intensity are a band a t 1139 cm-l probably due to NH, bending, a band at 1397 cm-' due to COzstretching, followed by lesser vibrations in the 300-900cm-' region, almost all of which are assignable to benzene ring modes. The SERS spectrum of PABA is considerably richer than its counterpart in either solution or in the (10) G. Varsanyi, 'Vibrational Spectra of Benzene Derivatives", Academic Press, New York, 1969.

IO0

300

500

700

900

AO ! c d )

HOO

1300

I500

4%

Flgure 9. SERS spectrum of PABA adsorbed on aqueous colloidal silver, indicating vibrational assignments.

crystalline solid. This is due to the fact that vibrations which are normally rather weak have been rendered strong upon adsorption. The carbon-hydrogen and nitrogenhydrogen stretching regions showed only very very weak bands and could not be used to determine the ionic form of the molecule, although vibrations assignable to 7(NH,) at 1139 cm-' and 6(NH2)at 1630 cm-' (together with the

,

/

The Journal of Physical Chemistry, Vol. 87, No. 9, 1983 1543

SERS of Colloidal Metal Systems

spectra shown here are for solutions containing NaOH used primarily to solubilize PABA. The predominent species at the pH used (around 7) is the anion. Attempts to record spectra without base or with added acid produced only a very weak spectrum and an enormous increase in the fluorescence background.) Presumably there are cations coadsorbed (or preadsorbed) onto the silver in order to counteract the increase in negative charge attendant upon adsorption of the anionic PABA. In contrast to the high concentration case the very low adsorbate concentration SERS spectrum of PABA is dominated by two bands: one at 1372 cm-I most likely corresponding to a COO- stretching vibration and one at 1600 cm-l, a benzene ring stretching vibration. These two bands are close to, but not precisely at, the same frequencies as their counterparts in the SERS spectra of the more concentrated samples. Before attempting to explain these observations let us consider briefly the structure of the adsorbate. Molecule I is almost planar. The carboxylate group is conjugated to the benzene ring only when its plane lies in the same plane as the ring. Consequently, it adopts that geometry1’ with a substantial energy barrier to rotation about the C-C bond. Likewise the NH2group which is almost tetrahedral about the nitrogen in alkyl amines (with the lone pair forming the fourth “substituent”)is in this instance almost

planar, the N atom being almost sp2-hybridizedand the lone pair occupying a p orbital almost perpendicular to the plane of the molecule. The molecule is, therefore, very well suited to a-bonding in a flat orientation and less likely to “stand up”, binding through the lone pairs on either the carboxylate or amine group. This might explain the dominance of benzene vibrations in the SERS spectrum. We have argued12that those vibrations which result in the most charge transfer in and out of the metal in synchronism with the nuclear motion accompanying them will be the most enhanced. When r-bonded through the benzene ring of PABA, benzene ring vibrations become good candidates for such charge transfer and should therefore be strongly represented in the SERS spectrum. The observed concentration and temperature behavior of the SERS spectrum of PABA suggest the existence of PABA in two forms on the surface. The following two possibilities were considered and rejected because they could not be made to explain our observations. (1) An equilibrium between a monomeric and dimeric form of PABA. (2) The existence of two types of surface sites, a more active site which saturates with increasing coverage and a less-active site which is populated thereafter. Up to now we used the overall PABA concentration in solution and the surface coverage of PABA almost interchangeably. This is because there is presumably an absorption isotherm which relates the two. At the low concentrations which we have used one should, moreover, be on the portion of the isotherm for which the surface coverage would be more or less proportional to the concentration in solution. For this reason we will continue to expect the overall concentration to be a measure of the coverage. (We could not, in fact, attain PABA concentrations sufficiently high to achieve complete monolayer coverage because the colloid became unstable and began to flocculate before that point was reached.) The model which accounts reasonably well for our observations is one in which we assume an equilibrium to exist between a two-dimensional (2-D) gaslike form and a two-dimensional solid- (or perhaps liquid-) like form of adsorbed PABA. The latter is visualized to be islands of the molecules linked together by hydrogen bonds between NH2 and COO- groups on neighboring molecules. Such linkages are in fact known to exist in the ordinary solid form of PABA,12and the bonding to the carboxylate group may account for the decrease in intensity and shift in frequency encountered for the symmetric stretching vibration of that group on going from the 2-D gas to the 2-D solid form of the adsorbed molecule, i.e., from the low concentration to the high concentration spectrum. The 2-D gas molecules are seen as single molecules skating across the silver surface, in dynamic equilibrium, of course, with PABA islands some of which “condensed” onto the islands while others “vaporized” from them. Such an assignment is strengthened by the fact that the SERS spectrum of the surface solidlike phase of PABA resembles that of crystalline PABA in many respects, while the surface gaslike phase is much more similar to (presumably single) PABA molecules in solution. For example, the prominent band at 1452 cm-’, in the SERS spectruum of the 2-D solid, has its counterpart (at 1432 cm-’) in crystalline PABA. No such band is seen in the solution spectrum. The solution spectrum, on the other hand, is dominated by the two bands at 1383 and 1607 cm-l close to their counterparts (at 1372 and 1600 cm-I) in the SERS spectrum of the adsorbed molecule in the “gaslike” surface

(11) R. K. Mackenzie and D. D. MacNicol, J. Chem. SOC.,Chem. Commun., 1299 (1970).

(12) D. P.DiLella and M. Moskovita, J.Phys. Chem., 85,2042 (1981), and ref 9.

Figure 10. SERS spectrum of PABA adsorbed on aqueous silver colloidal particles prepared in 99.8% D20.

i WAVELENGTH

(n m)

Flgwe 11. UV-visible spectrum of aqueous Ag colldd prepared in (A) H20, (B) D20 before the addition of PABA, and (C) as In B but after the addition of PABA to approximately 1 mM overall concentration.

COO- vibration) suggest that the molecule is adsorbed in its anionic form (I) rather than the zwitterion. (All the

1

1544

The Journal of Physical Chemistry, Vol. 87,No. 9, 1983

phase. Furthermore, the COO- vibration (at 1383 cm-l) is very weak both in the b a n spectrum of the crystalline solid and in the SERS spectrum of the adsorbate in the “solidlike” phase. Finally, both of these spectra have a large number of vibrations below 900 cm-l, many of which at almost coincident frequencies while the Raman spectra of PABA in solution and of PABA in the “gaslike”surface phase are almost devoid of bands in that region. Let us consider what behavior one expects for such a two-surface-phase system. At any given temperature one will have a characteristic surface vapor pressure which will be a function of temperature only. At low surface coverage all PABA molecules will be ”gaslike”,and they will remain so until the surface pressure reaches the equilibrium surface vapor pressure a t which point a 2-D condensed phase appears. Increasing the surface coverage further will cause only the condensed phase to grow in size. The number of “gaslike” molecules, on the other hand, decreases slowly with increasing surface covering, reflecting only the decrease in the total surface area available to it by the increasing size of the condensed phase. Eventually a monolayer is achieved, i.e., the condensed phase occupies the entire surface. Since the colloidal particles are small compared to the laser beam cross section the SERS signal reflects a constant fraction of all the “gaslike” and ”solidlike”adsorbed PABA molecules corresponding to the given concentration and temperature. Considering for the moment the case in which two phases coexist while yet far from the complete monolayer stage one expects to see only the solid phase spectrum increase in intensity upon increasing the surface coverage at constant temperature while the gaslike spectrum remains almost unchanged. Decreasing the temperature a t constant coverage, on the other hand, causes condensation, thereby increasing the solidlike phase at the expense of the gaslike phase. Now let us consider this process more quantitatively. Consider a colloidal metal particle whose surface area is A cm2. Let ng and n, be the number of molecules of PABA in the gaslike and solidlike phases, respectively, and let n = ng + n, (1) be the total numer of molecules of PABA on the surface of the colloidal particle. If the equilibrium two-dimensional pressure corresponding to the ambient temperature is J3 then for values of IZ such that n C IIA/kT where k is Boltzmann’s constant and “ideal two-dimensional gas” behavior is assumed, all the PABA molecules are gaslike i.e., n, = n and n, = 0. When n > IIA/kT then we must partition A into an area A, covered by solid and an area A, covered by gas, and A, + A, = A (2) If the surface density in the 2-D solid phase is p (molecules cm-2) then n, = PA, (3) IIA, = n,kT (4) combining 1, 2, 3 and 4, one obtains

Suh et al.

constant while n, is linear in n. When n l p = A , full coverage is achieved, hence ng = 0 and n, = n. As the temperature is raised n will increase in more or less an Arrhenius fashion, increasing ng and decreasing n,. The increase in II greatly exceeds the decrease due to the l l k T term. The observed behavior is consistent with this analysis. Let us now estimate the degree of enhancement obtained with our colloid. We assume the colloidal particles to be about 200 A in diameter.6 Using the value 1.59 A for the radius of silver one calculates a volume of 16.8 A3 and a cross-sectional area of 8 A2 for a silver atom. The volume of a colloidal particle is therefore 4.2 X lo6A3, which corresponds to 2.5 x lo5 atoms. The surface area of the colloidal particle is 1.3 X lo5A2. This corresponds to 1.6 X lo4 surface sites, assuming one surface adsorbate site per surface silver atom. Since one began with roughly 1 X M AgN03 in making the colloid one has a concentration of colloidal silver particles of 4 X M. The focal volume, i.e., the volume of the focussed laser beam from which Raman signal is collected is roughly 1 cm in length and 0.001 cm in radius, corresponding to a volume of 3 X lo4 cm3. Such a volume contains 7.3 x lo6 colloidal particles which, a t full coverage, would carry approximately 1.2 X 10” molecules of PABA. These yield a signal of about 105 counts s-l on the strongest (1600 cm-l) band. The same focal volume contains 1.8 X l O I 7 molecules of PABA in a 0.1 M solution of the compound. Such a solution produces a signal of about 40 000 counts s-l on the most intense band. Thus, the enhancement is at least 4 X lo6. We claim this a lower limit since we have ignored several factors which would increase the enhancement. These include (i) the shadowing of the laser beam by the colloidal particles preventing half of the molecules from being illuminated by the laser and half of those illuminated from scattering in the direction of the spectrometer; (ii) the fact that full monolayer coverage was never achieved; and (iii) the severe attenuation of the laser beam intensity by the highly scattering colloid reducing the effective focal volume for that sample. We finally turn our attention to the UV-visible spectrum. The 385-nm feature is characteristic of the surface plasma resonance absorption for a silver sphere in water. The presence of the shoulder at 425 nm indicates the presence of particles with nonspherical ge~metries.’~We interpret the change in color upon the addition of the adsorbate, also reported by Creighton,6 as being due to adsorbate-induced partial coagulation of spherical colloidal particles to particles more prolate in shape. The stability of colloidal metal particles results partly from surface charges upon them. The nature of these charges are changed by the presence of adsorbed molecules (presumably even more so by adsorbed ions), causing the colloid particles to coalesce. Prolate ellipsoids are known to have their plasma resonance shifted toward the red. This has the advantageous result of placing the most favorable excitation conditions within the range of visible lasers. Acknowledgment. We thank NSERC and the donors of the Petroleum Research Fund, administered by the American Chemical Society, for financial support. We are also indebted to Professor A. J. Kresge for explaining to us the intricacies of the bonding in PABA. Registry No. Ag, 7440-22-4; PABA, 150-13-0.

n, = n - ng

(6)

It is clear from eq 5 that, while n / p C A , ng is almost

(13) A. Thgorgt, Spectrochim. Acta, Part A , 27, 11 (1971). (14)M. Kerber, D.-S. Wang, H. Chew, 0. Siiman, and L. A. Bumm, p 109 of ref 1, and references therein.