1977
J. Phys. Chem. 1989, 93, 1977-1984
Surface Potential Dependence of Enhanced Raman Bands of Aggregated Cyanine Dyes D. L. Akins,* C. K. Akpabli, and X. Li Department of Chemistry, The City College of The City University of New York, New York, New York 10031 (Received: October 20, 1987; In Final Form: June 21, 1988)
The surface potential dependence of Raman scattering by aggregated molecules is analyzed by using molecular vibroexcitonic states as excited states.in the scattering problem. When incident excitation is close to resonance with the J aggregate absorption, Raman bands attributable to the A term of Albrecht’s polarizability expression appear-in addition to bands attributed to the B term present also in the nonresonance case. These former bands are interpreted as due to Franck-Condon overlap of intramolecular modes of the molecules in the aggregate with intermolecular lattice modes of the aggregate. It is deduced that bands assigned to the A term can be used to normalize the intensity of bands due to the B term in terms of both excitation frequency and surface concentration, thus serving as internal Raman standards and thereby isolating the electric potential dependence of Raman bands assigned to the B term. We have found that a Stark-effect shifting the J aggregate band can explain the voltage dependence of excitation frequency and concentration normalized Raman band intensities for aggregated 2,2’-cyanine and 4,4’-cyanine.
I. Introduction Studies of the potential dependence of intensities of Raman bands of adsorbed molecules in electrochemical cells are quite prominent in the literature pertaining to surface-enhanced Raman scattering (SERS).ld In general, such studies have aimed at testing quantum mechanical theories and intuitive concepts of the interactions and processes that occur in the interfacial region. Of particular note is the theoretical formulat’ion, utilizing the Herzberg-Teller intensity borrowing scheme, from the Birke/ Lombardi group, involving charge transfer between adsorbed molecule and substrate which has gone a long way in providing a theoretical underpinning for SERS intensity and potential dependency.6 These latter postulates, as well as others, are insufficient for explaining enhanced Raman scattering and its potential dependency for adsorbed, aggregated molecules. One of us has shown,’ in fact, that the enhancement of Raman bands for adsorbed, aggregated molecules is mainly attributable to the existence of molecular exciton states and the concomitant enhanced polarizability of the aggregate structure, rather than to the widely accepted “electrodynamic” and “charge-transfer” SERS progenitors.*-1° The potential dependency of Raman bands of adsorbed, aggregated molecules in the vibroexciton picture has not been previously addressed. In this article, we apply our theoretical formulation for enhanced Raman scattering by aggregated molecules to the enhanced signal versus potential issue: specifically, the Raman scattering problem of aggregated molecules with the vibroexcitonic states used as excited states, and utilization of the Herzberg-Teller approach originally developed by Albrecht1’,12is applied. The resultant polarizability expression indicates that certain bands may be used as internal Raman scattering standards, thereby isolating the static ~
~
(1) Otto, A. In Topics in Applied Physics: Light Scattering in Solids; Cardona, M., Guntherodt, G., Eds.; Springer; Berlin, 1983; Vol. 4. (2) Weaver, M. J.; Barz, F.; Gordon, J. G., 11; Philpott, M. R.Surf. Sci. 1983, 125, 409. ( 3 ) Furtak, T. E.; Macomber, S . H. Chem. Phys. Lett. 1983, 95, 328. (4) Furtak, T. E.; Roy, D. Phys. Reu. Lett. 1983, 50, 1301. (5) Lombardi, J. R.;Birke, R.L.; Sanchez, L. A.; Bernard, I.; Sun, S. C. Chem. Phys. Lett. 1984, 104, 240. (6) Lombardi, J. R.;Birke, R.L.; Lu, T.; Xu,J. J . Chem. Phys. 1986,84, 4174. (7) Akins, D. L. J . Phys. Chem. 1986, 90, 1530.
electric field functionality from the dependence on excitation frequency and surface concentration-we refer to bands corrected in such a fashion as being normalized. We also find that a resonance is predicted for normalized bands when exciton-state band edges are tuned into energy coincidence with the exciting radiation as a result of changes in the static electric field at the electrode surface. Experimental testing of these conclusions (in section IV) is provided through investigation of the potential dependency of Raman bands of adsorbed, aggregated 2,2‘-cyanine (111’-diethyl-2,2’-quinocyanine chloride) and 4,4’-cyanine (1,l’diethyL4,4‘-quinocyanineiodide), where we find that experimental observations are consistent with the theory, strongly supporting the supposition that the phenomenon of enhanced Raman scattering of aggregated molecules is correctly described by using vibroexcitonic states in the Raman scattering problem. In section I1 of this article we discuss the development and interpretation of the polarizability expression. Key issues addressed are the relative importance of the Franck-Condon term ( A in Albrecht’s notation) and the Herzberg-Teller terms ( B and C in Albrecht’s notation) of the polarizability, how the Raman spectrum would be expected to change dependent on the excitation frequency, and the nature of bands that may be considered Raman internal (frequency and concentration) standards. In section 111 we discuss our experimental system and procedures that are used to acquire Raman spectra. In section IV we report experiments concerned with the voltage dependency of Raman scattering for 2,2’-cyanine and 4,4’-cyanine adsorbed onto a silver electrode (as indicated above), as well as analyses and interpretation of our findings. 11. Raman Scattering Involving Vibroexciton States
The fundamental equation for Raman scattering gives the intensity of scattered radiation in terms of the square of the polari~ability.’~J~ From this equation, Tang and Albrecht’* have shown that the polarizability can be expressed as ai,8”‘@’” =A
+B+C
(1)
With our notation,ls the individual terms become A = [ M ~‘g,r I [~jl’g,r [ M ~~ Ig , r [ ~Og.r j l r,v E,, + - ~ C K Erg+ + hCK
[
+
€rv,w,t
1.
( 8 ) Surface Enhanced Raman Scattering Chang, R.K., Furtak, T. E., Eds.; Plenum Press: New York, 1982. (9) Adrian, F. J. J . Chem. Phys. 1982, 77, 5302. (10) Lippitsch, M. E. Phys. Reu. 1984, B29, 3101. (11) Albrecht, A. C. J. Chem. Phys. 1961, 34, 1476. (12) Tang, J.; Albrecht, A. C. In Raman Spectroscopy, Theory and Practice; Szymanski, H. A,, Ed.; Plenum Press: New York, 1970; Vol. 2; p
(13) Craig, D. P.; Thirunamachandran, R.Molecular Quantum Electradynamics; Academic Press: New York, 1984. (14) Eyring, H.; Walter, J.; Kimball, G. E. Quantum Chemistry; Wiley:
33.
New York, 1944.
0022-3654/89/2093-1977$01.50/0
(xrvlx,,,)
(xgv4xrv)
0 1989 American Chemical Society
(2)
1978 The Journal of Physical Chemistry, Vol. 93, No. 5, 1989
SO
-
s’q
Akins et al.
SO
M IS’
(3)
Figure 1. Excitation scheme and importance of terms in Albrecht’s polarizability expression: (a) incident radiation’s energy leads to raising electron from the ground state of the aggregate to the exciton band (this scheme leads to the B term being important);(b) incident excitation leads to raising electron from exciton band to higher excited states (this scheme leads to the C term being important);(c) situation in which the frequency of the incident radiation is in resonance with the single-moleculevibronic absorption.
(4) where the [MI’Sare the electric dipole moment vectors integrated over electronic coordinates for the ground-state equilibrium configuration, v is the excited molecular-state vibrational quantum number; v’ and v” refer to the upper and lower vibrational states in the scattering problem; the x’s are the vibrational wave functions; t, with its respective subscripts, is the vibrational energy difference referenced to the ground-state g; t represents states that can mix with the ground state of the scatterer; r represents an excited vibronic state of the scatterer (in the case of an aggregate, the excited vibronic state is made up of substates of the lowest vibroexciton state created by the incident radiation and, due to conservation of energy, has an energy close to that of the incident excitation); s is taken to represent, in general, other excited vibronic states of the scatterer as well as metal substrate states that can be in near resonance with a particular excited state r. In addition, haKtis the coupling term between the states g and t (for the molecule with equilibrium ground-state configuration-as indicated by the superscript zero on the [MI’S),and Q, is the displacement of mode a. Specialization of the above expressions to a system in which analytical forms of aggregate excited states (Le., vibroexciton states) are provided7 has been discussed elsewhere.7J5 However, here, in addition, the resultant expressions for A, B, and C are expanded by appending to them appropriate summations over single molecule states to allow for the excitation condition where laser radiation is tuned to a specific single-moleculeelectronic state. In this latter situation, the excitation will create (initially) a spatially localized, perturbed vibronic distribution in the aggregate (since the single molecule’s excitation is intercalated into that of the aggregate). The expressions for A , B , and C, specialized to aggregates as well as the single-molecule excited states, are obvious extensions of those provided in ref 15. For brevity and since their explicit forms are not important in what follows, we do not produce the full-blown expressions here. A . Importance of A, B, and C Terms of the Polarizability. A pictorial view, somewhat akin to that used elsewhere,6is helpful in assessing the meaning and relative importance of the various contributions ( A , B, and C) to the polarizability. Of necessity, any graphical representation of the scattering process must take into account the energy of the incident radiation since absorption resonances are important factors in determining the magnitude (15) Akins, D. L.; Lombardi, J. R. Chem. Phys. Lett. 1987, 136, 495.
of each contribution to the polarizability. The density of states for the exciton, arising from exciting a molecule in the adsorbed aggregate, is anticipated as having a distribution similar to that suggested in Figure 1. The effective energy spread of the exciton’s density of states is diminished beyond the intrinsic width associated with the dipoledipole interaction energy16 by the smaller spread of states associated with allowed dipole transitions (which leads to the narrowness of aggregate bands (J bands, H bands, etc.) in fluorescence and absorption s p e c t r o s c ~ p i e s ~ and ~ J ~ by ) conservation of energy vis-&vis the incident photons. The understanding of such observations was one of the incentives for much of the original theoretical interest in the molecular exciton model for molecular aggregate^.'^-^' Also, we speculate that the excitonic states can experience a “Stark effect” due to the coupling between the permanent dipole moment of the aggregate and the applied potential in our electrochemical system; as a result, vibroexcitonic states are expected to be tunable by applied potential. This tunability is incorporated into the scattering problem through the dependence of psin eq 3 on potential. This latter issue will be developed in more detail later. With the view of the energy states of the aggregate and the substrate as given above, the relative importance of A , B, and C in the polarizability expression can be assessed from the frequency-dependent scattering schemes of Figure 1 . Figure l a represents the case where the incident radiation’s energy is insufficient to reach the center and higher states of the exciton band. For J aggregate formation, the exciton band is of lower energy than the isolated molecule energy indicated by the symbol so. This situation, nevertheless, leads to electronic excitation from the ground-state aggregate to the exciton, which is analogous to the molecule-to-metal excitation scheme of ref 6 and corresponds to the B term of the polarizability being important. In addition, when the incident radiation’s energy can raise electrons from the aggregate ground state to exciton states greater than represented by the band center, yet still lower than the single molecule state so, one again has the energy picture of Figure l a except that [MIog,,corresponds to a transition dipole moment for excitation into a more highly excited vibroexciton band. Again, this picture corresponds to the B term but would be expected to (16) McRae, E. G.; Kasha, M. In Physical Processes in Radiation Biologv; Academic Press: New York, 1963; p 23. (17) Cooper, W. Photogr. Sci. Eng. 1973, 17, 217. (18) Her.?, A. H. Ado. Colloid. Interfacial Sci. 1977, 8, 237. (19) McRae, E. G.; Kasha, M. J . Chem. Phys. 1958, 28, 721. (20) Kasha, M. Rev. Mod. Phys. 1959, 31, 162. (21) Kasha, M. Radiat. Res. 1963, 20, 5 5 .
Raman Bands of Aggregated Cyanine Dyes give a different vibronic excitation, with resultant different matrix elements for B. This latter point is made in recognition of the fact that a particular germinal electronic distribution upon formation of the exciton will be reflected in the vibrational wave functions which are the eigensolutions of the nuclear motion in the field of the electrons. Figure 1b corresponds to electron transfer from the exciton to other excited states and is the schematic description of the C term.6 This term is considered to be less important than B, in the present scattering problem, mainly for two reasons. First, the energy spacing Pg- Ptin the denominator of eq 4 (with t s), in general, will be greater than the energy difference E“r in the denominator of eq 3. Second, the excitonic state only exists once a single molecule in the aggregate has been excited-thus, a twephoton process is necessitated by this scheme and is anticipated to have a small cross section. For the above reasons, of the Herzberg-Teller contributions, the B term but not the C term is expected to be important in the present scattering system. The A term can be expected to become enhanced when the exciting frequency is close to a “resonance”. As pointed out earlier,15 nonzero overlap integrals in eq 2 lead to fundamental, overtone, and combination bands, usually for totally symmetric vibrational modes.22 However, even though the. excitation conditions may correspond to a resonance situation, both A and B, in general, can lead to the presence of vibrational bands in the Raman spectrum. For certain bands, contributions from the A term may vanish. This latter situation is expected for the “strong-coupling case” for exciton formation, as detailed by Kasha,21which corresponds to little impediment for the excitation leaving the isolated single molecule and “roaming” through the aggregate structure. In such a case, the Born-Oppenheimer separability of intramolecular electronic and vibrational motions is necessitated. Further, since the excitation would be spread over many molecules, each individual molecule would have essentially the same electronic structure as a ground-state molecule. From the Franck-Condon principle it would follow that the A term would only be nonzero for upper state modes identical with ground-state modes. In other words, the A term would not contribute to the appearance of Raman bands resulting from the overlap of intramolecular ground and vibroexcitonic modes. On the other hand, the excited-state lattice modes of the aggregate, resulting from the intermolecular potential function, can be expected to give rise to nonzero overlap integral products with ground-state intramolecular modes. The B term, in addition to the situation discussed above, can become dominant under other conditions. For example, B alone contributes to Raman bands in the nonresonant Raman case where vibrational closure over the excited-state vibrational modes relegates the A term to make a contribution only to Rayleigh scattering and for non-totally symmetric modes (in both the resonance and nonresonance cases) for which the A term vanishes through symmetry. It is to be noted that intensity borrowing through the h“, factor allows B to contribute to Raman scattering, and, because of its explicit dependence on the vibrational mode (a),the B term contributes only to intramolecular Raman modes. Figure I C depicts the situation in which the frequency of the incident radiation is strongly resonant with the single-molecule vibronic absorption. A highly perturbed electronic distribution is expected to be created in the aggregate in this case and the full complement of excitonic subbands allowed by conservation of energy. As a result, the A term might be expected to lead to overtone and combination bands in the Raman spectrum. Indeed, we have found this to be the case, with an extensive overtone/ combination pattern.23 It is to be noted that the role of the substrate, for the molecular vibroexciton scheme as here discussed, is simply to be a support (22) Hamaguchi, H. In Advances in Infrared and Raman Spectroscopy; Clark, R. J. H., Hester, R. E., Eds.; Wiley: New York, 1985; Vol. 12. (23) Experimental results for 2,2’-cyanine and 4,4’-cyanine presently being prepared for publication. In the former, excitation is at 514 nm, while in the latter, 580-nm excitation is used.
The Journal of Physical Chemistry, Vol. 93, No. 5, 1989 1979 onto which aggregates can form. B. Association of Albrecht’s A and B Terms with Specific Bands. Here we focus on attributing the appearance of specific bands in the Raman spectrum to the A and B terms of Albrecht’s polarizability expression. In general, association of bands-with the A and B terms depends on the scatterer. However, the excitation frequency dependence of Raman scattering by aggregated cyanines exhibits a behavior typified by that of 2,2’-cyanine. Thus, here we confine our discussion to 2,2’-cyanine but realize that other cyanines studied (4.4’-cyanine, 2,2’-carbocyanine, etc.) are subject to the same type of analysis. Figure 2 shows typical Raman spectra for aggregated 2,2’cyanine excited by nonresonant (top) and resonant (bottom) radiation. Figure 2b, further, presents the changes in the spectra as a function of potential. The experimental system and p r d u r e s are addressed in section 111. The specific exciting radiation frequency (in these cases 488 and 583 nm, respectively) is not the important issue. Rather, it is observed that off-resonance excitation, either higher frequency than the single-molecule state (ca. 520 nm) and adsorbed aggregate state (ca. 575 nm), as is the 488-nm excitation,’* or of lower excitation frequency (e.g., 647-nm krypton excitation; spectrum not shown here) leads to spectra with the same bands in the region between 300 and 1700 cm-’. These latter bands, though they have slightly different intensities dependent on the frequencies of the off-resonance excitation, are attributed to the B term of the polarizability and, as a result, are associated with intramolecular modes. Upon excitation with 583-nm radiation (Figure 2, bottom), which overlaps the adsorbed J aggregate’s absorption, two bands at 232 and 274 cm-’ for 2,2’-cyanine (while 4,4’-cyanine exhibits only one band at 210 cm-’ upon excitation with 647-nm radiation as shown in Figure 3), broader than those attributed to the B term, undergo significant intensity enhancements; also, all the B term bands present with nonresonant excitation appear as well. In line with the discusssion in section HA, we attribute the two bands to the A term of the polarizability and their origins to overlaps of single-molecule modes and aggregate (Le., lattice) modes. In addition, we conclude that all B term modes are most likely fundamentals, since with nonresonant excitation, these same modes, as a result of closure over excited-state modes, must be fundamentals (in the harmonic oscillator approximation).’ Since A (eq 2) has the same excitation frequency dependence as B (eq 3), but does not depend on the factor E“r - E“$, which B has in its denominator, while in addition all Raman intensities are proportional to the concentration of the scattering species, it follows that bands that depend only on A can be used as internal standards to isolate the E“, - EOs dependence of Raman bands that depend only on B. However, an additional frequency dependent factor v0v3, for photon-counting measurements,22 where vo is the incident laser frequency and v is the frequency of the Raman band, must also be taken into account. Identification of an internal standard for Raman intensity versus potential measurements, as suggested here, makes possible a study that is impossible otherwise and differs from the use normally made of internal standards (which are hopefully innocuous added scattering species) to normalize just the frequency dependence of Raman scatterer^.^^^^^ 111. Experimental System and Procedures
Raman scattering was excited in an electrochemical cell consisting of a 45’ angle Ag working electrode, Pt counter electrode, and saturated calomel electrode (SCE). The cell was designed with a 45’ Pyrex window in near contact with the working electrode to minimize aperture effects associated with absorption of scattered radiation.26 Incident radiation impinged from below (24) Albrecht, A. C.; Hutley, M. C. J. Chem. Phys. 1971, 55, 4438. (25) Dudik, J. M.; Johnson, C. R.; Asher, S. A. J . Chem. Phys. 1985.82, 1732. (26) Li, X.;Gu, B.; Akins, D. L. Chem. Phys. Lert. 1984, 105, 263.
1980 The Journal of Physical Chemistry, Vol. 93, No. 5, 1989
Akins et al. I
100
300
500
700
1100
900
av
1500
1300
1700
("1)
r
1
I
I
t 100
300
500
700
900
aY
1100
1300
1500
1700
("1)
Figure 2. Raman spectra of aggregated 2,2'-cyanine excited by nonresonant (top) and resonant radiation (bottom). Electrode pretreatment was employed in both cases and is detailed in the text. At top is the spectrum of 5 X M 2,2'-cyanine chloride, with a Ag working electrode, excited by 488-nm argon ion laser radiation. The spectrum below 1100 cm-' has a full-scale setting of 4 X 10' counts/s, while higher wavenumber bands have a scale setting of I X lo4 counts/s. Laser power was ca. 30 mW at the sample. At bottom is the spectrum of 5 X M 2,2'-cyanine chloride, excited by 583-nm laser radiation from an argon ion pumped, rhodamine 6G dye laser. The voltage-dependent spectra, for potentials -0.8, -0.95,-1.10, and -1.2 V all vs SCE,are provided from bottom to top, respectively. The scale setting in all cases is 3 X lo4 counts/s and the laser power at the sample was
ca. 30 mW. and was collected by an optical system with its axis perpendicular to the propagation direction and polarization of the incident laser excitation. The actual angle of collection of scattered radiation (as measured with respect to the incident direction) was selected by orienting the electrode to obtain maximum Raman signal while sitting on a particular band: typically, the resultant angle was significantly less than 90°, as is to be expected from reflection calculation^.^^ The laser radiation, when exciting at the frequency of the aggregate band's absorption, was supplied by a Coherent (27) Mullins, D. R.; Campion, A. J . Phys. Chem. 1984, 88, 8.
tunable dye laser (Model CR-599) using Rhodamine 6G pumped by a continuous-wave 4-W Spectra Physics Model 2000 argon ion laser. A three-plate birefringent filter was used with the dye laser to give radiation of ca. '/.,-A width. Other features of the experimental apparatus were essentially the same as reported elsewhere.26 Solutions and chemicals were prepared and purchased as indicated earlier,28while the electrode pretreatment (when used) was accomplished by a potential step sequence in which the initial (28) Gu,B.; Akins, D. L. Chem. Phys. Lett. 1985, 113, 558.
The Journal of Physical Chemistry, Vol. 93, No. 5, 1989 1981
Raman Bands of Aggregated Cyanine Dyes
> I-
VI
z w I-
f,
100
300
500
700
1100
900
a3
1300
1500
1700
(cm-l)
Figure 3. Raman spectrum of 1 X lod M 4,4’-cyanine at pH 12, with a Ag working electrode at a controlled potential of -0.4 V vs SCE, excited by 647-nm krypton ion laser radiation. The spectrum below 900 cm-’ has a full-scale setting of 1.2 X lo4 counts/s, while higher wavenumber bands have a scale setting of 5 x io3 counts/s. .E6
.E3
ABS
.?9
-76
.72
0 WAVELENGTH (nm)
Figure 4. Potential dependence of diffuse reflectance of 2,2’-cyanine absorbed onto a silver-coated glass slide. The silver-coated slide was made by M solution that also contained 0.1 M KCI; 30 min was allowed for the procedure detailed in ref 29. 2,2’-Cyanine was adsorbed from a 2 X equilibrium adsorption to occur. The resultant coated slide was then used in a three-electrode reflectance cell designed for the sample port of a Perkin-Elmer Lambda 3A spectrophotometer; the concentration of 2,2’-cyanine in the sample cell solution was lowered to ca. lo-’ M to diminish the monomer absorption, which did not affect the aggregate peak absorbancy in the time it took to acquire spectra. The supporting electrolyte was maintained at 0.1 M in both the reference and sample ports, and nitrogen gas was bubbled for 15 min before the spectra were acquired. Diffuse reflectance taken at three potentials are shown: solid line a t -0.1 V, broken line a t -0.8 V, and dotted line at -1.2 V, all versus SCE.
potential was -0.1 V, stepped to +0.3 V, and then reversed to -0.1 V, all versus SCE. Pretreatment of the electrode, however, was not essential for these studies. The supporting electrolyte was potassium chloride at a concentration of 0.1 M.
IV. Experimental Results and Discussion for t,Z’-Cyanine and 4,4’-Cyanine To be sure that observed potential dependency of normalized Raman bands is indicative of potential effects on the excitonic state and not a response to shifting or complex change of the aggregate absorption spectrum, we have conducted UV-vis, diffuse reflectance studies, under controlled potential. In these studies, glass slides have been coated with silverz9and used as electrodes
in a diffuse reflectance configuration of a Perkin-Elmer, Lambda-3A spectrophotometer. We have found that the J aggregate band for 2,2’-cyanine does not change its basic shape or shift in frequency with applied potential. However, the dye does exhibit a potential-dependent peak absorbancy that goes through a maximum a s the potential varies from -0.1 V vs SCE to -1.4 V vs SCE. Such an observation simply indicates the effect of surface potential on surface concentration, which is explicitly corrected for in our normalization procedure. Figure 4 provides UV-vis diffuse reflection spectra for 2,2’-cyanine. (29) Ni, F.;Cotton, T.M. Anal. Chem. 1986, 58, 3159.
1982 The Journal of Physical Chemistry, Vol. 93, No. 5, 1989
.
I
. . 0.4
Akins et al.
06
'
~
+ + + + . a . L
12
1.0
*
X
/
I
.
.
0.8
*
00
02
-Volts vs SCE
I
* *
04 06 -VoltageiSCE)
08
Figure 6. Intensity versus potential plots for several bands of 4,4'-cyanine excited by 647-nm radiation. Data plotted have been corrected for background signal-defined by drawing a smooth curve beneath spectra such as shown in Figure 3 and subtracting. Spectral bands plotted are X (210 cm", which is used in the analyses procedure to normalize the other bands), + (456 cm-I), (478 cm-'), (532 cm-I) W (626 cm-I), 0 (650 cm-I), and A (1364 cm-I).
+
d 20.1
06
08
10
12
-Volts s SCE
Figure 5. Intensity versus potential plots for several bands of 2,2'-cyanine chloride excited by 583-nm radiation. Data plotted have been corrected for background signal-defined by drawing a smooth curve beneath the spectra of Figure 2, bottom, and subtracting. Figure 5A displays data for some of the weaker bands (A (844 cm-I), (1514 cm-'), and W (1622
+
cm-I)), while Figure 5B provides data for some of the stronger bands (0 (1348 cm-I), + (1368 cm-I), and X (1384 cm-I)). We have also acquired Raman excitation spectra of 2,2'-cyanine with numerous exciting frequencies in the vicinity of the J aggregate band at 575 nm. In these studies, we have observed intensity variations of bands but no new bands as the excitation frequency is changed. The most dramatic effect in this study is a resonance exhibited by the low-frequency bands attributed to the A term of the polarizability. Specifically, the bands at 232 and 274 cm-l exhibited an enhanced signal as the dye laser frequency (supplied by the coherent tunable dye-laser mentioned earlier) was tuned through the J aggregate band. For our potential dependency studies, Raman spectra for 2,2'-cyanine and 4,4'-cyanine were acquired with the apparatus and procedures indicated earlier. The potential dependence of the intensities of Raman bands of 2,2'-cyanine was measured for potentials ranging from -0.4 to -1.3 V vs S C E (see Figure 2, bottom) for spectra at 4 potentials out of the 13 actually used), while the potential range for 4,4'-cyanine was -0.2 to -0.9 V vs SCE. The general trend for the spectral bands (with background subtraction applied) was a gradual rise in intensity to some maximum intensity (-1.1 V vs SCE for 2,2'-cyanine and -0.8 V vs SCE for 4,4'-cyanine) followed by a diminution of intensity at more negative potentials. Figures 5 and 6 show data for several bands of 2,2'-cyanine and 4,4'-cyanine, respectively. Such plots of the data do not allow discrimination between models of the potential dependence of the Raman signal. For example, such disparate concepts as the adatom model',* and that of Lombardi et a1.,6 which involves a potential-induced resonance with substrate states, predict the same type of behavior. The former model, however, essentially relates surface concentration (which depends on applied potential) to scattering intensity. The latter incorporates a quantum mechanical resonance mechanism but does not exclude the former from playing a role in the potential dependence. Our present model is a quantum mechanical one and, in addition, since our analysis procedure involves a ratioing of intensities associated with B and A terms of the polarizability, thereby factoring the potential-dependent surface concentration and excitation frequency dependency, agreement with our model can be taken as strong
1-
t
... t
r
". 0.6
0.4
0.8
1.2
1.0
-Volts vs. SCE
'"1
iB)
6.0
x
x
21 0 1 04
06
; x
i i',
10
08
-Volts
bs
,
, 12
SCE
Figure 7. Intensity versus potential plots for frequency and concentration normalized bands of 2,2'-cyanine chloride excited by 583-nm radiation. Respective data for Figure 5 have been frequency corrected (Le., divided by v0v3), and divided by the intensity of the 274-cm-' band (which is also frequency corrected and defined by subtracting the background signal). Figure 7A displays data for the same bands as Figure 5A, however,
divided by the frequency corrected intensity of the 274-cm-l band. Figure 7B displays data for the same bands as Figure 5B but treated in the same fashion as data in Figure 7A. affirmation of its theoretical concepts. In Figure 7, for the same Raman bands of 2,2'-cyanine as used in Figure 5 , we show the residual potential dependencies of the relative intensities that result upon using the 274-cm-' band (attributed to the A term of the polarizability) as an internal standard and accounting for the frequency dependency of Raman intensities as detailed in section IIB, while Figure 8 provides similarly treated data for 4,4'-cyanine using the 210-cm-' band (see Figure 3) as an internal standard. We interpret the potential dependencies shown in Figures 5 and 6 as principally reflecting the increased aggregation on the surface induced at potentials more positive than the point of zero charge (PZC), followed by a subsequent dissolution of adsorption sites as the substrate assumes a negative potentialz6 The potential dependency behaviors for the ratioed bands shown in Figures 7
The Journal of Physical Chemistry, Vol. 93, No. 5, 1989 1983
Raman Bands of Aggregated Cyanine Dyes
t
04 0.2
.
.
m
m
t
m
I
'
0.4
0.6
0.8
-Voltage(SCE)
Figure 8. Intensity versus potential plots for frequency and concentration normalized bands of 4,4'-cyanine iodide excited by 647-nm radiation.
Respective data for Figure 6 have been frequency corrected (Le., divided by voy3) and divided by the intensity of the 210-cm-' band (which is also frequency corrected and defined by subtracting the background signal). and 8, in the present theoretical model, must represent the residual tuning of the exciton state in relationship to the incident radiation. This latter assertion is based on the supposition that narrow-band incident radiation will excite only a subset of the total number of exciton subbands that would be available if a single molecule in an N-molecule aggregate were fully excited to its first excited singlet level. An applied potential can then be expected to Stark shift the exciton bands, because of the expected large dipole moment of the aggregate, creating a different subset of exciton subbands in resonance with the incident radiation. Since exciton subbands are equally spaced (at zero applied electric field) with "allowed" transition dipole moments existing only for some, the subset excited by absorption of incident radiation would be part of the "allowed" constituency of exciton subbands. Tuning by varying the applied potential, in the above picture, would correspond to Stark-effect tuning of the subset of incident radiation created exciton bands through the range of allowed exciton substates. The above concepts can be applied to the B term to gain information concerning exciton parameters, such as permanent dipole moments and effective exciton band-edge energies. The analysis begins with the assumption that the product of transition moments and Herzberg-Teller constants for the B term of eq 3 can be written as a constant times the density of states function of the exciton state. This assumption is the same as used by Lombardi et aL6 in their treatment of metal conduction band states. As a result, the summation over exciton substates can be replaced by the integral
'
s,, Ps
dE4
(5)
with PIcorresponding to the laser excitation energy (Le., E', = hvo),the prime specifying the absence of s = 1 from the sum, and ps the density of states populated by the incident excitation. In a w r d with our discussion immediately above, the density of states can be assumed to be a constant in the ranges E'e IE', IEol - 6 and Eo1 6 5 Eo, I Eo,,, with EOe and E",, representing the lower and upper effective band-edge limits, respectively, for the "allowed" exciton substates, and 6 a positive differential energy. The voltage-dependent behavior of the B term is determined by the logarithmic expression that results from the integration shown in eq 5:
+
KE0e- Eoi>/(@i-
(6)
Thus, resonances in the applied potential dependency occur when the "pumped" exciton substrates-of energy E'1= hu,, with uo the laser frequency-become the substrates that define the effective band-edge energies p cand E',,. Hence, in general, we can expect two resonances in the plot of normalized band intensity versus potential. However, our studies to date have involved a metallic silver electrode and excitation frequencies that allow us to observe the resonance that occurs at negative potential but not
the resonance at more positive potential because of substrate oxidation. We propose (vide supra) that the voltage tuning is a Stark-effect tuning of P,, into coincidence with the energy of the incident radiation. E?, is taken as the active limiting energy because we envisage the exciton electric dipole moment as forming with positive charge density closest to the negative electrode, leading to the dipole pointing in the direction of the positive field. As a result, one gets a lowering of states with increasingly negative electrode potential as indicated by the Stark-effect energy relationship AE = -p[. The Stark-effect explanation appears resonable as a simple calculation can support. This calculation starts from the realization that the potential gradient a t the working electrode in our electrochemical cell likely ranges from to lo+' V/cm, while the electric dipole moment of the excitonic state, because of its large spatial extent and delocalized electrons, should be substantially larger than the ground-state electric dipole moment, probably by at least l D. With these parameters, the shift effected by the potential gradient can be calculated from the Stark conversion factor 5.954 x 104-
D V/cm cm-
=1
(7)
We find an energy shift of 168 cm-I for a potential gradient of 10'' V/cm and a conservative electric dipole moment increase of 1 D (for the excited aggregate relative to that of the ground state). This energy shift corresponds to a wavelength shift, centered on the incident excitation wavelengths (583 nm for 2,2'-cyanine and 647 nm for 4,4'-cyanine), of ca. *60 A. This shift is of a magnitude that can tune the excitonic state of 2,2'cyanine (with zero applied potential state at 575 nm) through a resonance with the exciting radiation, but of insufficient magnitude for 4,4'-cyanine (with zero applied potential state a t 710 nm). Moreover, since E', is lowered as 2: becomes more negative, the original mismatch between the frequencies of the exciton band edge and the incident radiation (at 583 nm) for 2,2'-cyanine is diminished and then increased, leading to a resonance maximum as shown in Figure 7, while, the absence of residual tuning for 4,4'-cyanine, as shown in Figure 8, is attributable to Stark shifting of its I?, band edge to even longer wavelength with negative potential, thus not allowing for a resonance with the fixed incident excitation at 647 nmS3O A possible use for the voltage resonance measurement is to determine relative dipole moment increases of excitonic states (above those of the ground states) for different cyanine dyes. This application necessitates that only a narrow excitonic state subband be identifiable, region be excited, that the upper band edge P,, and that dipole moments be aligned in the field direction. With these conditions, the equation
applies for the different aggregates, where h"',,(O) is assumed constant and is the band-edge energy in the absence of an applied field,,,[ is proportional to the voltage at which resonance occurs, and PIis the energy of the incident radiation. Relative p's (for different aggregated molecules) can be determined by fitting respective data to eq 8 and taking a ratio of slopes.
V. Conclusions In conclusion, we interpret our experiments to suggest that the potential dependence of enhanced Raman scattering by aggregated cyanine dye molecules can be explained by a Stark-effect shifting of molecular exciton states. We have also ascertained that certain bands in the Raman spectrum are due to Franck-Condon overlap of intramolecular modes of the molecules in the aggregate with intermolecular lattice modes of the aggregate and that such bands can be used as internal Raman standards to normalize surface concentration. In addition, we have determined that it is not necessary to implicate the substrate states in a potential-induced (30) Li, X.; Akpabli, C. K.; Akins, D. L., manuscript in preparation.
1984
J . Phys. Chem. 1989, 93, 1984-1987
resonance scheme in which, for example, substrate states are tuned into resonance with exciton or single-molecule states. The substantive agreement between theory and experiments for enhanced Raman scattering by aggregated molecules validates the utilization of molecular exciton states as excited states in the scattering problem. Our studies suggest to us that Raman scattering by aggregated molecules on surfaces can provide a spectroscopic means of acquiring structural information for single molecules forming aggregates and electron injection dynamical information for aggregated structures; both types of determinations would not be
complicated by strong substrate/molecule couplings.
Acknowledgment. Support for this research by the National Science Foundation (NSF) under Grants RII-8305241 and RII-8504995 is gratefully acknowledged. Also, support at the inception of this effect by the N S F under grant PRM-8211023 and by the City University of New York under a PSC-CUNY grant are acknowledged. Last, heartfelt thanks are extended to a colleague, J. R. Lombardi, for significant conversations that have provided insight into the Stark-effect arguments used here. Registry No. 2,2’-Cyanine,2402-42-8; 4,4’-cyanine, 4727-49-5.
Photoreactions at the Semiconductor/Eiectrolyte Interface under Diminished Field Conditions. Transient Currents and Charge Collection in Pulsed Laser Irradlated TiO, Electrodes Michael Neumann-Spallart, Albin Schwarz, and Gottfried Grabner* Institut fuer Theoretische Chemie und Strahlenchemie der Universitaet Wien, Waehringerstrasse 38, A-1090 Wien, Austria (Received: February 18, 1988; In Final Form: July 18, 1988)
Photocurrent transients in n-TiOz electrodes under pulsed laser illumination (355 nm, 10 ns) were measured as a function of radiant exposure (dose), which was varied over 5 orders of magnitude (up to >0.1 J/cmz), and as a function of applied potential. The electrode potential was stepped to up to 40 V vs SCE during less than 1 ms. Within this time (after RC relaxation) the laser pulse was applied. Only at sufficiently low doses and high potentials were the transients determined by the RC constant of the circuit. At medium and high doses they assumed space-charge-limited features, Le., saturation of current with increasing dose and lengthening of lifetime. A limiting value of the quantum yield of charge collection was reached at doses below 10” J/cmZ ( 5 V). Above this dose, the quantum yield fell off steeply due to higher order (bulk) recombination. Upon increase of potential the lifetime of the transients decreased and a considerable increase of quantum yield was observed. At 1.3 mJ/pulse (0.007 85 cm2 electrode surface) a total charge of 0.014 mC could be collected at 20 V.
Introduction The dependence of the photoelectric response of illuminated semiconductor devices on light intensity has so far received only little attention with respect to the very high photon fluxes encountered in pulsed laser irradiation. As to the semiconductor/electrolyte junction, the significance of pulsed laser measurements lies in the possibility of direct observation of kinetics and the detection and identification of intermediary species arising from electrochemical reactions of photogenerated charge carriers. From this, new insight in processes occurring under continuous illumination can be gained. Investigations of transient photopotentials at semiconductor/ electrolyte junctions yielded information on the time course of charge transport within the semiconductor and interfacial transfer On the other hand, the shape of photocurrent transients has been shown to reflect in most cases the RC constant of the photoelectrochemical provided the dose (radiant exposure) is sufficiently low. No investigations on the quantitative aspects of the fate of charge carriers created by light pulses have ,been reported. This paper deals with the quantum efficiency of charge collection in laser flash irradiated TiOz electrodes at average light ( I ) Perone, S . P.; Richardson, J. H.; Deutscher, S. B.; Rosenthal, J.; Ziemer, J. N. J. Electrochem. SOC.1980, 127, 2580. (2) Kamat, P. V.;Fox, M. A. J . Phys. Chem. 1983, 87, 59. (3) Hartig, K.J.; Grabner, G.; Getoff, N.; Popkirov, G.; Kanev, St. Ber. Bunsen-Ges. Phys. Chem. 1985, 89, 8 3 1 . (4) Harzion, Z . ; Croitoru, N.; Gottesfeld, S. J. Electrochem. SOC.1981, 128. 551. ( 5 ) Wilson, R. H.; Sakata, T.; Kawai, T.; Hashimoto, K. J . Electrochem. SOC.1985, 132, 1082.
( 6 ) Sakata, T.; Janata, E.; Jaegermann, W.; Tributsch, H. J. Electrochem. SOC.1986, 133, 339.
0022-3654/89/2093-1984$01.50/0
fluxes ranging from lozoto 3 x quanta/(cm2.s). Preliminary experiments with semiconductor electrodes irradiated with light pulses of a frequency-tripled Nd:YAG laser (half-width 10 ns) had shown us that quantum yields of collected charge (integrated photocurrent) around 1 V vs SCE were surprisingly low and no photoproducts could be detected by transient absorption measurements. This is in contrast to semiconductor materials suspended in colloidal form in a liquid, where high quantum yields of photoproducts can be obtained upon irradiation with pulsed laser light.’,* (This situation essentially corresponds to a photoelectrochemical experiment carried out in a cell at open circuit .9J0) In a recent paper,” we studied the effects of pulsed laser light on a Si photodiode in order to see if the above-mentioned discrepancy originates mainly from differences in topology between colloidal and bulk semiconductor systems. The main finding was that quantum yields were as low as with a semiconductor/electrolyte junction. At the light fluxes encountered in these experiments the condition for space-charge-limited currents (SCL) as discussed by Weisz et a1.12 for insulators and by Tove and Andersson for Schottky junction^'^ was fulfilled, thus favoring higher order recombination reactions in competition with charge collection. However, we found that the quantum yields can be considerably increased by applying bias voltages up to 60 V. In this work we wanted to see how the results from Si photodiodes (7) Duonghong, D.; Ramsden, J.; Graetzel, M. J. A m . Chem. SOC.1982, 104, 2977.
(8) Ramsden, J. J. Proc. R. Soc. London, A 1987, 410, 89. (9) Neumann-Spallart, M.; Enea,0.J . Electrochem. Soc. 1984,131, 2767. (1 0 ) Desilvestro, J.; Neumann-Spallart, M. J . Phys. Chem. 1985,89, 3684. (1 1) Neumann-Spallart, M.; Schwarz, A,; Grabner, G. Appl. Phys. A 1988, 46, 9.
(12) Weisz, S. 2.;Trester, S.; Many, A. J. Appl. Phys. 1968, 39, 2296. (13) Tove, P. A.; Andersson, L. G . J. Appl. Phys. 1973, 44, 2690.
0 1989 American Chemical Society
