Fluorescence of Dye Molecules Adsorbed on Semiconductor Surfaces

403

FLUORESCENCE OF DYE MOLECULES ADSORBED ON SEMICONDUCTOR SURFACES A. M. PONTE GONCALVES

INTRODUCTION Adsorbed dye molecules may be used to sensitize wide bandgap n-type semiconductors to visible light, (1,2) and thus make them potentially useful as electrodes in photoelectrochemical cells for Sensitization occurs when the solar energy conversion ( 3 , 4 ) . lowest excited singlet state of the dye, D*, is above the conduction band of the semiconductor, in which case the molecule becomes oxidized by injecting an electron into the semiconductor. Bending of the conduction band at the semiconductor-electrolyte junction promotes escape of the injected electron into the bulk of the semiconductor. Operation of the cell is sustained by a reducing agent present in the electrolyte, which reduces D+ and regenerates the dye. The injection efficiency, @i, is governed by the competition between injection of an electron into the semiconductor (rate constant ki) and other processes which shorten the lifetime of D* (rate constant ka): 1.

The escape efficiency, @,, is determined by the competition between escape of the injected electron away from the surface (rate constant ke) and any other processes, such as trapping at surface states, which favor recombination of the electron with D+ (rate constant kr):

The overall scheme is

t

ka

kr

I

r31

It is assumed here that the light intensity is weak enough for the relative concentration of D+ to remain low. The conversion

404

efficiency, defined as 9 = [(number of electrons detected)/(number of photons absorbed)], is then simply (1,2,7)

Reported conversion efficiencies cover a wide range of values, from a few percent for single crystal electrodes ( 5 , 6 ) to near unity for sputtered thin film electrodes (7,8). One widely accepted explanation (1,2) for low 9 is that injection is made with near-unity efficiency, but most of the electrons are unable to escape the surface, i.e., ki >> ka but ke << kr. It is evident that a good understanding of the overall problem requires that $i and 9, be determined separately. Unlike the photocurrent, the dye fluorescence intensity (which is proportional to the population of D*) does not depend on kr or ke as long as the relative concentration of D+ remains small. Therefore, the fluorescence provides a fundamentally more direct look at injection than does the photocurrent. The fluorescence lifetime is rf

=

1/(ka

+

ki)

[51

and the fluorescence quantum yield is Cpf = kf Tf'

163

Both quantities are related to the injection efficiency by

In this chapter we focus our attention on the use of the fluorescence of adsorbed dyes to explore the competition between the processes which determine @i. Then, whenever comparison with 9 is possible, information can be obtained also about 9 ,. The basic strategy relies on measuring either rf or ef when injection is blocked (ki = 0, e.g., on an insulator surface) and when it is not, while assuming that ka is the same in both cases (7,9,10). Questions have been raised recently about the validity of this assumption, which attributes to injection any reduction in either rf or @if for dyes adsorbed on semiconductors. We postpone consideration of other processes which could play an important role until the experimental results are presented. Fluorescence quantum yields can, at least in principle, provide the same information as fluorescence lifetimes, as long as kf may also be assumed not to depend on the surface. However, if

405

the dye is adsorbed at more than one type of surface site, only a weighted average of Gf is obtained. By contrast, time-resolvea fluorescence measurements can provide both the lifetimes and the relative contributions of more than one adsorption site, and thus afford a much less encumbered look at the decay of D*. Unfortunately, while it is relatively easy to resolve contributions of widely different lifetimes, those with close lifetimes cannot be readily separated. The severity of this limitation depends on the quality of the data and on whether decay functions other than exponential have to be explored. We consider three decay channels for D* in addition to injection: Fluorescence (rate constant kf), intramolecular radiationless decay (rate constant ko), and energy transfer quenching within the adsorbed layer (rate constant kq): ka

=

kf

-+ ko + kq.

Since kq 0 at very low dye surface density (usually expressed in terms of surface coverage, B ) , the effect of energy transfer quenching on Gi is readily examined through the dependence of ef on 6. The inclusion of kq in [ 8 ] accounts for energy transfer only approximately, since energy transfer quenching leads to nonexponential fluorescence decays (11-13). Because early measurements giving qJ = were carried out at high B , Spitler and Calvin (5) suggested that the low values could be due to the effect of energy transfer quenching on qJi. One of the goals of work in this area has been to clarify the dependence of @i on B . The remainder of this chapter is organized as follows. Steady-state fluorescence measurements are reviewed briefly in Section 2 . Time-resolved fluorescence measurements performed in our laboratory are discussed in some detail and are related to other work in Section 3. A discussion of processes other than injection which might contribute to the decay of D* is given in Section 4 . In order to give the reader some feeling for the evolution of our views on this complex problem, the presentation is roughly chronological. .+

STEADY-STATE FLUORESCENCE MEASUREMENTS One of the earliest studies of quenching of the fluorescence of dyes adsorbed on wide bandgap semiconductors was reported by Arden and Fromherz (7). These authors also measured, in the same experiment, one of the highest conversion efficiencies reported to 2.

406

date, @ = 0 . 8 . In this elegant work, monolayers of a long-chain cyanine dye mixed with arachidic acid were first preassembled at air-water interfaces and then transferred to Sn-doped In203 electrode surfaces. The dye could be placed either in direct contact with the electrode or separated from it by two fatty acid monolayers. The fluorescence quantum yield of the dye in these preassembled monolayers was reduced five-fold by direct contact with the semiconductor. Arden and Fromherz used the approach ' . Although outlined in the Introduction to obtain ki = 7 x lo9 s B was not varied, the high @ value shows that energy transfer quenching cannot dominate @i in this system. Another early investigation was carried out by Tanimura et al. (lo), who vacuum deposited thin films of tetraphenylporphine on quartz and on Sb-doped Sn02. Contact with the semiconductor lowered ef by a factor of three, and a lower bound ki > lo9 s-l was estimated. The effect of energy transfer could not be assessed since B was not varied and @ was not determined. The dependence of @f on B was studied in our laboratory in an attempt to clarify the competition between energy transfer and injection for dyes adsorbed from solution (14). The fluorescence quantum yield of rhodamine B adsorbed on Sn-doped In203 and on glass was determined as a function of B (measured by the optical density of the adsorbed dye). A sharp difference was found between the behavior of qif on glass and on the semiconductor as B -* 0: ef increased by more than one order of magnitude on glass but only by a factor of two on the semiconductor. The increase in qif at low B reflects the decrease in the ability of energy transfer quenching to compete with the other D* decay channels. Assuming that neither kf nor ko depend on the surface, we estimated ki = 1.2 x lolo s-l by comparing @f on the semiconductor and on glass at the same B , i.e. , the same kq. This ki value differs by less than a factor of two from those obtained by Arden and Fromherz (7) as described above and by Nakashima et al. (11) for hole injection from rhodamine B into anthracene. Use of our ki value together with T~ = 2.7 ns, the lifetime in dilute ethanol solution, (15) gave @i = 1 at B = 0. At the highest optical These densities we estimated k = 1.9 x lolo s-' and @i = 0 . 4 . 4 results indicate that competition from energy transfer quenching is not the principal cause of low @ values, even at high B . Itoh et al. (16) conducted a detailed investigation of @f as a function of B for rhodamine B on glass, Sn02, and Ti02. From

407

the results in the limit 9 + 0, they calculated ki = 7.4 x lo8 s-' and rpi = 0.72 for Sn02. These ki values are over one order of magnitude lower than those obtained by us and by Arden and Fromherz (7) for In203. The authors also reported that very little fluorescence quenching was observed on Ti02, leading to ki < 3 x lo7 s-' and ei < 0.1. The wide range of ki values obtained from steady-state fluorescence experiments suggested that injection may vary significantly from semiconductor to semiconductor. TIME-RESOLVED FLUORESCENCE MEASUREMENTS In order to examine more closely the processes which determine @i, time-resolved fluorescence measurements on dyes adsorbed on semiconductor and glass surfaces were carried out in our laboratory. The first set of experiments used a low repetition rate, mode-locked Nd:glass laser and streak camera detection (17). For rhodamine B adsorbed from 4 x M aqueous solutions, we obtained rf = 55 ps on Sn-doped In203 and rf = 46 ps on glass. Because these experiments were carried out at high 8 , we concluded that the short rf on both surfaces was determined mostly by efficient energy transfer quenching. The low sensitivity of the experimental system did not permit experiments at low e . In order to overcome this obstacle, we used a synchronously pumped, mode-locked dye laser, cavity-dumped at 4 MHz and timecorrelated single-photon counting detection (18). Because of the higher sensitivity of this experimental system we were able to work at low 9 , using aqueous rhodamine B solutions with concentrations down to M. To examine the dependence of the fluorescence decays on 9 we chose to work with surface-solution interfaces, so as to minimize the problems associated with inhomogeneous surface coverage, which arise with dry surfaces (14). The semiconductor surfaces were those of thin films of Sndoped In203 or Sb-doped Sn02, vacuum deposited on glass substrates. Each I1wetl1sample was a sandwich of two surfaces of the same type with a thin layer (a few microns thick) of dye solution trapped between them. lrDrytl samples were prepared by separating wet samples and letting the solution film evaporate. The fluorescence intensity was analyzed by an iterative nonlinear least-squares reconvolution of the instrument response function and biexponential decays of the form 3.

408

F(t)

=

A1 exp(-t/rl) + (1

- A1)

exp(-t/r2).

[91

The functional form [9] was chosen because two fluorescent species (on the surface and in solution) were expected. No decay component approaching the instrument response function was found in any experiment. The goodness of each fit was judged by the x2, the distribution of weighted residuals, and the autocorrelation function. The fits used not only the fluorescence decay but also all of the rise, which gave higher x2 than when only the decay was included. Because all fits were quite good (i.e. , had low x2 and apparently random distributions of weighted residuals) no attempt was made to explore functional forms more complex than [9]. The results of typical individual fits are shown in Table 1. TABLE 1 Parameters of best fits of [9] to rhodamine B fluorescence decays for surfaces in contact with loe7 M aqueous solutions of the dye. Surface

r1 (ns)

72

(ns)

A1

x2

glass Sn02

0.68

2.6

0.36

1.31

0.41

0.71

1.35

InZ03

0.40

1.4 1.3

0.75

1.21

It is clear from these results that the fluorescence decays are considerably different at glass-solution and at semiconductorsolution interfaces. We note, in particular, an important point: The dominant component is the slow one on glass and the fast one on both semiconductors. (A detection polarizer set at the magic angle ruled out rotational depolarization as the source of the fast component.) No effect of solution concentration on r1 for semiconductor surfaces was found in the M range, which shows that interference from energy transfer quenching may be neglected at loe7 M. Experiments were also performed on dry samples in order to avoid interference from solution dye. A dry glass sample, prepared from the same wet sample which gave the results at the top of Table 1, had rl = 0.60 ns, r 2 = 3.1 ns, and A1 = 0.22. This is typical of results from dry glass samples prepared with M solutions: The dry sample always had r1 slightly shorter than the original wet sample, r 2 = 3.1 ns, and A1 5 0.25. The

409

shorter r 2 obtained with wet glass samples reflects the unresolved contributions from dye molecules on the surface (rf = 3.1 ns) and in solution (rf = 1.5 ns) (11). The fact that as many as 25% of the molecules adsorbed on glass have a lifetime considerably shorter than in solution is less readily understood. We attributed r1 on glass to molecules which inject into surface states or other localized states near the surface. On the other hand, Kemnitz et al. (12) observed a similar two-component fluorescence decay for rhodamine B adsorbed on silica ( r l = 0.9 ns, r 2 = 3.8 ns), and attributed the fast component to molecules loosely attached to the surface. We return to this point in the Discussion. M solutions Dry Sn02 and In203 samples prepared from were always found to have r 1 slightly shorter than the original wet sample, r 2 = 3.1 ns, and A1 z 0.97. We take particular note of the fact that the same r 2 was determined for dry glass as for both dry semiconductors. Also, the smallness of the slow component on dry semiconductor surfaces explains why r 2 obtained with wet semiconductor samples was close to the solution lifetime. The similarity of the decays on the two semiconductors does not support the earlier suggestion that the disparity in the steadystate results was due to the semiconductor. We attributed also to injection the dominant fast component of the fluorescence on the semiconductors. The minor slow component was attributed to molecules adsorbed at a few semiconductor surface sites where injection cannot take place. The rate constants ka and ki were estimated from the data obtained at M ) , so as to minimize energy the lowest concentration transfer effects. The values r f = r2(dry) = 3.1 ns and ki = 0 were used in [5] to derive ka = 3.2 x 108 s-’. This ka value and

r 1 were then used in [5] and in [7] to derive ki = 2.1 x lo9 s-’ and @i = 0.88 both for Sn02 and for In203. This ki value is rf

=

six times lower than that estimated from our steady-state measurements for In20g and three times higher than calculated by Itoh et al. (16) for Sn02. Because the time-resolved measurements are much more direct, we consider this ki value to be more reliable than those derived from steady-state measurements. As expected, r l became progressively shorter as the solution M, a consequence of energy concentration increased above A dry Sn02 sample prepared from a M transfer quenching. solution gave r 1 = 0.07 ns and no detectable slow component, in

410

good agreement with our streak camera results (obtained with 4 x M solutions). With this rl value we calculated @i = 0.13, which shows that energy transfer lowers @i to a greater degree than previously estimated. Nevertheless, the lower @i is still well above the @ values measured for single crystal electrodes. Only three other time-resolved measurements of the fluorescence of dyes adsorbed, on wide bandgap semiconductor surfaces seem to have been reported to date. In the earliest of these, Kamat and Fox (19) investigated erythrosin B solutions containing Ti02 colloidal particles. The decays followed [9], with r1 in the 0.20 0.36 ns range attributed to adsorbed dye, quite similar to our results for rhodamine B on Sn02 and In203. On the other hand, the results of Kamat and Fox are in sharp contrast to those of Itoh et al. (16), who found that sputtered TiOZ thin film surfaces had almost no effect on the fluorescence of rhodamine B. This may mean that some physical or chemical characteristic of these poorly defined surfaces is more important than the identity of the semiconductor. More recently, Crackel and Struve (20) investigated the fluorescence decay of cresyl violet adsorbed at high -9 on Ti02 single crystal surfaces. The decay could be fitted by a sum of three exponentials with lifetimes 36.8 ps (78%), 209 ps (19%), and 1555 ps (3%). Similar results were reported later by Anfinrud et al. (21) for rhodamine 3B adsorbed at low .9 on ZnO single crystal surfaces: 79 ps (55%), 337 ps (32%), and 1221 ps (13%). The extremely short lifetime of the dominant component in both decays (not observed in our experiments at low R ) was attributed by the authors to energy transfer from the dye to the semiconductor. Both decays also have an intermediate component in the same range as the r1 values obtained by Kamat and Fox (19) and by us. Finally, the smallness of the slow component is in general agreement with our results for dry semiconductor surfaces.

-

DISCUSSION We have outlined experimental results which show that adsorbing dye molecules on semiconductor surfaces (with D* above the conduction band) will lead to lower fluorescence quantum yields and shorter fluorescence lifetimes than on insulator surfaces. While it has been generally assumed that this is due to injection from D*, other processes have been suggested which are discussed in this Section. 4.

411

4.1 Dimers Energy transfer among identical molecules does not, by itself, quench the fluorescence: quenching occurs when the excitation is transferred to a trap, a species of lower excited state energy. It has been suggested that non-fluorescent dimers (or larger aggregates) serve as excitation traps for adsorbed rhodamine B (11). For two-dimensional Forster energy transfer quenching by isotropically distributed traps, the fluorescence decay is expected to be non-exponential and of the form F(t) = A exp[-t/rf

-

B(t/rf)1/3],

[lo1

in which rf is the lifetime in the absence of energy transfer and Other functional forms may be appropriate to other types of energy transfer (13,22). Even though the dye fluorescence is appreciably quenched at high 8 , Spitler and Calvin (5) have shown that 9 is independent of B (at least at high 8 ) . This led Itoh et al. (16) to suggest that dimers (which trap the excitation energy) must inject as efficiently as monomers, Nakashima et al. (11) reached a similar conclusion for hole injection from rhodamine B into anthracene. Spectral evidence of dimers offered by Kemnitz et al. (12,23) for rhodamine B adsorbed at high 8 on glass and on organic molecular crystals. These authors found strong evidence of room temperature dimer emission (exponential decay), in addition to the quenched monomer emission (non-exponential decay following [lo]). The different spectrum and temperature dependence of the two emissions helped distinguish between them. The dimer was found to have rf = 120 ps at room temperature and rf = 3.8 ns at 77K, which suggests that the room temperature lifetime is determined by dissociation (23). If we use the monomer ki value derived from our time-resolved experiments together with rf = 120 ps, we obtain @i = 0.20 for dimers, which is significantly lower than @i = 0 . 8 8 we calculated for monomers. Thus, while dimers may harvest most of the energy from monomers, as proposed by Itoh et al. (16), the injection efficiencies of the two species will be similar only if ki is much greater for dimers than for monomers. Our time-resolved experiments focused on the low 8 limit, in which case the fluorescence decays were found to be independent of wavelength. For the few experiments performed at high 0 we did not examine the wavelength dependence of the fluorescence. B is proportional to the trap concentration (11,12).

WAS

412

Therefore, it is quite possible that the value r1 = 0.07 ns we measured at high 8 on dry Sn02 results from an unresolved combination of dimer and quenched monomer contributions. An investigation similar to that of Kemnitz et al. (23) for semiconductor surfaces is needed to help clarify the role dimers play in the sensitization at high 8 . 4 . 2 . Multiele Sites Indications that molecules adsorbed at different sites on semiconductor surfaces may contribute very differently to I$ have been in the literature for a long time (6). However, recent work by Spitler (24) offers valuable new insight. The optical density of the adsorbed dye and the photocurrent were measured simultaneously in a cell using a ZnO single crystal electrode. Operation at low B eliminated problems associated with energy transfer quenching and dimer formation. A rhodamine B solution was first placed in contact with the electrode until equilibrium adsorption was reached. The dye was then desorbed by flushing the surface with electrolyte solution. With this simple but ingenious approach Spitler found that, although less than one-tenth of the adsorbed dye could be removed, cp decreased in the process by more than one order of magnitude. The results showed that 9% of the surface dye was weakly adsorbed and had cp = 0.15, while the rest was strongly adsorbed and had cp = 0. Spitler's work suggests time-resolved fluorescence experiments before and after weakly adsorbed dye is removed from single crystal surfaces, as a means to clarify the relationship between cp and I$i at each type of site. Although the non-exponentiality of a fluorescence decay may be clear, the distinction between [ 9 ] (or even more complex multiexponential forms) and [lo] can be rather difficult. This problem is alleviated when the experiments are performed at low 8 , so that components of the form [lo] arising from energy transfer within the adsorbed dye layer need not be considered. However, the fits are not unique even then: For example, a reasonably good fit with [ 9 ] will often be improved if a sum of three exponentials is used instead (25). Whatever the details of the functional form used in the fit, there is clear evidence for more than one adsorption site in the fluorescence decays of dye molecules adsorbed on a variety of surfaces (12,18,25). At low 8 the emitting species are monomers with lifetimes which vary from site to site. The existence of multiple adsorption sites on the same surface raises the concern that ka may not be surface independent.

413

Internal Conversion A simple view of the effect of placing a dye molecule on a surface is that adsorption inhibits some of the low frequency skeletal motions which contribute to the radiationless decay of D* to the ground st?te (internal conversion). Thus, it is easy to understand how adsorption of rhodamine B can lead to a lifetime which is longer than in solution, just as the solution lifetime increases significantly with solvent viscosity (15,26). We have taken r 2 to be l/ka and recall, somewhat reassuringly, that r 2 was the same on glass and on both semiconductors. The origin of the short lifetime r1 on glass is not as clear. We attributed r1 to injection and suggested that surface states, or some other localized states near the surface, serve as electron acceptors on insulators as well as on semiconductors (18,27). Kemnitz et al. (12,25) proposed instead that I1irregular1lsurface sites enhance internal conversion and lead to rl, while llidealll sites hinder internal conversion and lead to r 2 . These authors supported their view with two extremely interesting observations: (a) The introduction of imperfections into molecular crystals increases the contribution of fast fluorescence decays, and (b) dyes with greater skeletal rigidity have smaller contributions from fast decays. The view that adsorption may enhance as well as suppress internal conversion was also expressed by Nakashima and Phillips (28). The work of Kemnitz et al. (25) suggests a similar investigation for semiconductor surfaces. Dyes of various degrees of skeletal rigidity should be used in time-resolved fluorescence experiments in order to clarify the processes responsible for r 1 on semiconductors. The effects of mechanical polishing and of etching of the semiconductor surface (which introduce and remove surface traps, respectively) on the fluorescence decay might help identify the surface sites. 4 . 4 Intersvstem Crossinq Skeletal motions can also induce transitions from the lowest excited singlet state to the lowest triplet state (intersystem crossing). Thus, changes in ka may be due not only to internal conversion but also to intersystem crossing. It should be noted that injection can also take place from the triplet state, as discussed by Spitler et al. (29). Although the question of injection from the triplet state remains largely unexplored, the possibility of a heavy atom effect 4.3

414

(which enhances intersystem crossing) on Sn-doped Inj02 surfaces was considered by Arden and Fromherz (7). These authors placed a dye monolayer on a LaF3 film (expected to produce a heavy atom effect comparable to that of In and Sn) but found no evidence of enhanced intersystem crossing. Another exploration of possible heavy atom effects was conducted by Tanimura et al. ( l o ) , also with negative results. 4.5 Enercrv Transfer to the Semiconductor Quenching of D* by energy transfer to the semiconductor has not received much attention as a possible cause of low @. This is because interest in dyes adsorbed on semiconductors has focused on the sensitization of wide bandgap semiconductors. In this case the molecular excitation energy is smaller than the bandgap (although D* is above the conduction band) and cannot generate electron-hole pairs. The early work of Tanimura et al. ( 1 0 ) examined tetraphenylporphine thin films separated from Sn02 surfaces by KC1 spacers of varying thickness. These authors found that @f decreased as the distance to the semiconductor surface was reduced, and suggested the possibility of energy transfer to surface trap states below the conduction band. The strategy of using spacers of varying thickness to probe energy transfer to the semiconductor has been used since by other authors. (Unlike energy transfer, electron transfer is a very short range process.) Hayashi et al. ( 3 0 ) measured cpf and r f f o r 50 A thick tetracene films separated from narrow bandgap semiconductors by LiF spacers. Because the tetracene excitation energy was greater than the bandgap, energy transfer to the semiconductors could readily produce electron-hole pairs ( 3 1 ) . Surprisingly, although cpf for the tetracene films was found to decrease as the distance to the semiconductor decreased, rf did not become shorter. Deri ( 3 2 ) interpreted these results according to the Chance-Prock-Silbey model ( 3 3 ) for molecule-metal interactions, and suggested interference effects (radiative) rather than energy transfer (radiationless). Alivisatos et al. ( 2 2 ) measured r f for pyrene layers separated from Si single crystal surfaces by Xe spacers. In contrast to the results of Hayashi et al. ( 3 0 ) , rf was found to be considerably shorter at smaller separations: 2 3 ns at 1 9 6 A and 3 . 3 ns at 17 A. Detailed analysis of the results was hindered by energy transfer within the pyrene layer and by the formation of excimers, both due to high 8 . Nevertheless, these experiments

415

showed the importance of energy transfer to the semiconductor. Because the pyrene excitation energy is not much lower than the Si bandgap, some energy transfer leading to electron-hole pair formation was expected. Two recent papers have suggested quenching of D* by energy transfer to the semiconductor even though the excitation energy was much lower than the bandgap. In the first of these, Crackel and Struve (20) recorded the fluorescence decay of cresyl violet separated from Ti02 single crystal surfaces by arachidic acid The multilayer spacers of thickness between 80 A and 509 A. authors assigned the two-fold reduction in r f at the smallest separation to energy transfer to the semiconductor according to the Chance-Prock-Silbey model (33). Analysis was again difficult For dye molecules adsorbed directly on the because of high 8 . semiconductor the fluorescence decays were fitted, as we saw in Section 3, by tri-exponential functions for cresyl violet on TiOZ However, (20) and for rhodamine 3B on both ZnO and Ti02 (21). the fluorescence decays could also be fitted with [lo], an example of the lack of uniqueness of the fits to which we alluded earlier. Anfinrud et al. (21) chose the fit with [lo] to assign the fast decay to energy transfer to discrete trap states on the surface of the semiconductor, as suggested earlier by Tanimura et al. (10). Kavassalis and Spitler (34) have discussed a model in which injection takes place into acceptor traps on the surface, rather than directly into the conduction band; escape would then be from the traps to the conduction band. Fluorescence experiments cannot alone determine whether quenching of D* for adsorbed dye takes place through electron injection or through energy transfer. Anfinrud et al. (21) argued that the ultrashort lifetimes (under 100 ps), and an even faster ground state recovery (13 ps, measured by a pump-and-probe technique for rhodamine 640 adsorbed on a ZnO single crystal), ruled out injection. The argument was that the decay of photogenerated charge carriers is known to be several orders of magnitude slower than this. However, electron injection into surface traps may be followed by extremely fast recombination, so that the observed ground-state recovery could be due to fast injection followed by even faster recombination. If the decay of D* (by either energy or electron transfer) is trap-mediated, then polishing and etching of the surface (which change the trap It has been shown density) should have opposite effects on rf.

416

that such surface treatment can change @ significantly (35,36). Different evidence for energy transfer to the semiconductor was provided by Nakao et al. ( 8 ) , who noted that ef decreased by a factor of two when the doping level of the semiconductor was increased. This effect was attributed to quenching of D" by energy transfer to conduction electrons, which reach closer to the surface at higher doping levels (thinner space charge layer). This is of particular interest because the authors also found that @ increased by more than one order of magnitude as the doping level increased. For highly doped Sn02, sensitization with rose bengal and rhodamine B yielded @ = 0.9 and 0.3, respectively. Since the doping level affected C#I much more than @fl the effect on @ was attributed to enhanced escape arising from the thinner space charge layer. An examination of the effect of doping level on dye fluorescence decays could establish whether adsorption in the vicinity of dopant atoms affects the lifetime of D*. 4 . 6 Summarv A great deal of knowledge has been gained from work in this field, although it has served largely to better define the problem and to unveil its complexity, rather than to provide final answers. The results surveyed in this chapter raise several important questions and suggest future experiments which may help answer them. It will be important to extend the approach of Arden and Fromherz (7) and tie more closely together time-resolved fluorescence and @ measurements in the experimental design, especially where the distribution of adsorption sites can be manipulated. It would also be extremely useful if semiconductor surfaces used in time-resolved fluorescence experiments were better characterized physically as well as chemically. Thus, some future experiments should take advantage of ultra-high vacuum (22) and surface analysis techniques to prepare clean and well characterized surfaces prior to deposition of the dye. REFERENCES 1

H. Gerischer and F. Willig, Top. Current Chem., 61 (1976),

2

H. Gerischer, M. T. Spitler, and F. Willig, in: S. Bruckenstein (Ed.), Proceedings of the Third Symposium on Electrode Processes 1979, The Electrochemical Society, Princeton, NJ, 1980, p. 115. A. Heller, Acc. Chem. Res. , 14 (1981) 154. J.H. Fendler, J. Phys. Chem. 89 (1985) 2730.

3 4

31..

417

5 6

M. T. Spitler and M. Calvin, J. Chem. Phys., 67 (1977) 5193. M. Spitler, M. Lubke, and H. Gerischer, Ber. Bunsenges. Phys.

10

Chem., 83 (1979) 663. W. Arden and P. Fromherz, Ber. Bunsenges. Phys. Chem. 82 (1978) 868; J. Electrochem. SOC., 127 (1980) 370. M. Nakao, K. Itoh, T. Watanabe, and K. Honda, Ber. Bunsenges. Phys. Chem., 89 (1985) 134. T. Iwasaki, T. Sawada, H. Kamada, A. Fujishima, and K. Honda, J. Phys. Chem., 83 (1979) 2142. K. Tanimura, T. Kawai, and T. Sakata, J. Phys. Chem. 83

11

N. Nakashima, K. Yoshihara, and F. Willig, J. Chem. Phys., 73

12

K. Kemnitz, T. Murao, I. Yamazaki, N. Nakashima, and K. Yoshihara, Chem. Phys. Lett., 101 (1983) 337. F. Willig, A. Blumen, and G. Zumofen, Chem. Phys. Lett., 108

7 8 9

13

(1979) 2639.

(1980) 3553.

(1984) 222.

18

Y. Liang, P. F. Moy, J.A. Poole, and A.M. Ponte Goncalves, J. Phys. Chem., 88 (1984) 2451. M. J. Snare, F. E. Treloar, K. P. Ghiggino, and P.J. Thistlethwaite, J. Photochem., 18 (1982) 335. K. Itoh, Y. Chiyokawa, M. Nakao, and K. Honda, J. Am. Chem. SOC., 106 (1984) 1620. Y. Liang, A. M. Ponte Goncalves and D. K. Negus, J. Phys. Chem., 87 (1983) 1. Y. Liang and A. M. Ponte Goncalves, J. Phys. Chem., 89 (1985)

19 20

P. V. Kamat and M. A. FOX, Chem. Phys. Lett., 102 (1983) 379. R. L. Crackel and W. S. Struve, Chem. Phys. Lett., 120 (1985)

21

28

P. A. Anfinrud, T. P. Causgrove, and W. S . Struve, J. Phys. Chem. , 90 (1986) 5887. A. P. Alivisatos, M.F. Arndt, S. Efrima, D.H. Waldeck, and C. B. Harris, J. Chem. Phys., 86 (1987) 6540. K. Kemnitz, N. Tamai, I. Yamazaki, N. Nakashima, and K. Yoshihara, J. Phys. Chem., 90 (1986) 5094. M. T. Spitler, J. Phys. Chem., 90 (1986) 2156. K. Kemnitz, N. Tamai, I. Yamazaki, N. Nakashima, and K. Yoshihara, J. Phys. Chem., 91 (1987) 1423. T. Karstens and K. Kobs, J. Phys. Chem., 84 (1980) 1871. T. Kajiwara, K. Hasimoto, T. Kawai, and K. Yoshihara, J. Phys. Chem. 86 (1982) 4516. N. Nakashima and D. Phillips, Chem. Phys. Lett. 97 (1983)

29

M. Spitler, M. Lubke,

14 15 16 17

22 23 24 25 26 27

3290.

473.

337.

56 (1978) 577.

and H. Gerischer, Chem. Phys. Lett.,

30

T. Hayashi, T. G. Castner, and R. W. Boyd, Chem. Phys. Lett.,

31 32 33

D. L. Dexter, J. Lumin., 18/19 (1979) 779. R. J. Deri, Chem. Phys. Lett., 98 (1983) 485. R. R. Chance, A. Prock, and R. Silbey, in: S. A. Rice and I. Progogine (Eds.) , Advances in Chemical Physics, Vol 37 , Wiley-Interscience, New York, 1978, p. 1. C. Kavassalis and M.T. Spitler, J. Phys. Chem., 87 (1983)

34 35 36

94 (1983) 461.

.

3166.

H. Gerischer, F. Hein, M. Lubke, E. Meyer, B. Pettinger, and H.R. Schoppel, Ber. Bunsenges. Phys. Chem., 77 (1973) 284. M. Matsumura, Y. Nomura, and K. Honda, Bull. Chem. SOC. Jpn., 52 (1979) 1559.