J. Phys. Chem. 1988, 92, 2587-2591 EXAFS spectra gives a contradictory result; not all the species formed are MOO, crystallite, because of the Mo-Mo coordination number being much smaller in the catalyst, as is shown in Figure 5. The layered structure is often pointed out as a model of surface molybdenum species. This cannot, however, explain the small coordination number of Mo-Mo observed by EXAFS, because the coordination number in this model is 6 , which is the same as that of bulk. XRD results give a clue to explain the contradiction between the Raman and the EXAFS results, the existence of large amounts of X-ray amorphous species. These are in mixture with MOO, crystallite. In order to elucidate these X-ray amorphous species, we attempt to compare the scattering intensity of the Raman bands of crystallite MOO, with ammonium heptamolybdate, assuming that the X-ray amorphous species exist as heptapolyanion-type oxide. The physical mixture of the same Mo contents was used for Raman measurement. The peak intensity of the 822-cm-’ band of MOO, is about 7 times stronger than that of the 946-cm-I band of ammonium heptamolybdate. The result indicates that the polymolybdate species should be detectable with 20 wt % catalyst by Raman spectroscopy if the X-ray amorphous oxides, which account for about 80% of the supported Mo, exist as one kind of
2587
oxide such as heptapolyanion. However, Raman spectra do not show any bands except for the bands due to MOO, as is shown in Figure 4, suggesting that the X-ray amorphous species are not one kind of heptapolyanion. Most likely, they are a mixture of several kinds of Mo oxides. According to the EXAFS results, as is shown in Figure 5, Mo atoms are octahedrally coordinated by six oxygen atoms in X-ray amorphous species. Thus, we can conclude that, in addition to MOO, crystallites, there are many kinds of octahedrally coordinated Mo species on the 20 wt % catalyst. Finally, the above observation can be summarized in the following. At low loadings, the “interaction species” is formed already in the dried stage and it is stable and does not change by calcining at 773 K. At high loadings, free heptamolybdate is formed first and many kinds of Mo oxides exist after calcination at 773 K. One is MOO, crystallites and the others are the various kinds of octahedrally coordinated Mo oxide clusters.
Acknowledgment. We thank T. Kadowaki of the Research Institute for Catalysis, Hokkaido University, who kindly measured the surface area of silica gel. Registry No. Molybdenum oxide, 11098-99-0.
Photochemistry of Colloidal Semiconductors. 25. Quenching of CdS Fluorescence by Excess Positive Holes Ani1 Kumar, Eberhard Janata, and Arnim Henglein* Hahn-Meitner-Institut Berlin, Bereich Strahlenchemie, 1000 Berlin 39, Federal Republic of Germany (Received: September 22, 1987)
Colloidal CdS dissolves anodically (to yield CdZ+,SOPz-,and SO:- ions) when it is attacked by OH radicals produced by y-irradiation. The fluorescence intensity of CdS slightly decreases during the first stages of this dissolution, but drastically increases in the later stages as the particles become very small. This increase is attributed to the removal of surface sites at which the radiationless recombination of charge carriers takes place. Pulse radiolysis experiments were carried out in which the changes in fluorescence intensity upon the attack of a colloidal particle by one or a few OH radicals were studied. The OH radicals inject holes into the surface of the colloidal particles, and these holes decrease their ability to fluoresce. One excess hole introduced in this way has 50-100% quenching efficiency. The optical absorption of the deposited holes was also observed. Several holes cannot be stored on a particle as they rapidly react with each other. The longer the lifetime of fluorescence the more efficient is the quenching. The holes injected by OH radicals slowly leave the particles which then become fluorescent again. The consequences of this effect in photocatalysis are pointed out. A mechanism of quenching is discussed, in which the red fluorescence is attributed to the interaction of the electron generated by light absorption with various defect sites on the surface, trapped holes of the S- type acting as radiationless recombination centers.
Introduction It was shown in the first papers of this series that colloidal cadmium sulfide in aqueous solution fluoresces, and that this fluorescence is quenched by certain The quenchers used were stable molecules or ions, such as methylviologen or T1’. Numerous reports on the fluorescence of colloidal CdS have appeared in the meantime4 in which similar quenchers were ap(1) Henglein, A. Ber. Bunsen-Ges. Phys. Chem. 1982, 86, 301-305. (2) Henglein, A. J . Phys. Chem. 1982,86, 2291-2293. (3) Henglein, A. Top. Cur?. Chem. 1987, 143, 113-180. (4) Rossetti, R.; Brus, L. J . Phys. Chem. 1982,86,4470-4472. Chestnoy, N.; Harris, T. D.; Hull, R.; Brus, L. E. J. Phys. Chem. 1986, 90, 3393-3399. Duonghong, D.; Ramsden, J. J.; Gratzel, M. J . Am. Chem. SOC.1982, 104, 2977-2985. Ramsden, J. J.; Gratzel, M. J . Chem. SOC., Faraday, Trans. 1 , 1984, 80, 919-933. Ramsden, J. J.; Webber, S. E.; Gratzel, M.J . Phys. Chem. 1985,89,2740-2743. Serpone, N.; Sharma, D. K.; Jamieson, M. A,; Gritzel, M.; Ramsden, J. J. Chem. Phys. Leu. 1985, 115, 473-476. Kuczynski, J. P.; Milosavljevic, B. H.; Thomas, J. K. J. Phys. Chem. 1984, 88, 980-984. Tricot, Y.-M.;Fendler, J. H. J . Phys. Chem. 1986, 90, 3369-3374.
0022-3654/88/2092-2587$01.50/0
plied. In the present paper, an unusual quenching effect is investigated, Le., the influence exerted by excess positive holes on the fluorescence of the colloidal particles. Such excess holes can be injected from OH radicals generated radiolytically.5 Experiments of this type are of interest with respect to our understanding of the reactions of the charge carriers produced by light absorption in small semiconductor particles. In many fluorescence studies, strong laser flashes have served as exciting light. Under these conditions, holes may be accumulated on the particles which in turn may influence the fluorescence. This often makes the results of such studies not comparable to those obtained at low light intensities. The main aim of the present work is to show how the damage produced in the attack of a small CdS particle by the OH radical influences its ability to fluoresce. Moreover, under conditions ~~
( 5 ) Baral, S.; Fojtik, A,; Weller, H.; Henglein, A. J . Am. Chem. SOC.1986, 108, 315-318.
0 1988 American Chemical Society
2588 The Journal of Physical Chemistry, Vol. 92, No. 9, 1988
(8
(8
Kumar et al.
EL
200
I
I
'r'
I,+AI
LOO 500 h[nml
600
700
Figure 2. Absorption and fluorescence spectra of sol B at various times of y-irradiation under an N 2 0 atmosphere. Dose rate: 2.5 X lo5 rad/h.
I H PM
300
t+ H AIxRL
BO
AMP
REC
ticles. Lamp EL in Figure 1 was flashed after time T and the fluorescence intensity I measured and compared with Z,,, the intensity observed without applying the electron pulse. The pulse apparatus was also used for normal optical absorption measurements to observe the absorption of the holes produced by O H attack which are known to have a broad absorption band at long wavelengths in the ~ i s i b l e . ~ Figure 1 shows schematically some details of the equipment. The same volume of solution was irradiated by the exciting light beam and the electron beam. The initial radical concentration could be varied between 0.1 and 10 pM, the concentration of colloidal particles amounting to several micromoles per liter. The light level of the exciting light source EL (XBO 450) was intensified for the measurement by a lamp pulser;' the level of the light pulse was flat during the measurement (maximum 2 ms). Only the UV region of the light, as selected by filter F2 (UGl), was used for excitation. The shutter (S2) prevented illumination between the pulse experiments. The analyzing light was generated by source AL. The emitted light was collected by lens L3 and converted into an electrical signal by photomultiplier PM. A five-stage circuit was used for the absorption measurements and an eleven-stage circuit for the emission measurements. Filter F3 prevented stray light from entering the monochromator MO. The background current Io was compensated for and measured by the back-off circuit BO.* Only the change in current AZ(t) was amplified (AMP) and recorded (REC). The gain of preamplifier A M P was switchable via the data bus and an interface (not shown). The detection electronics were connected to the computer (CO) via the standard IEEE-488 bus. A high-precision circuit for monitoring the absorbed dose (to be published) and the timing circuit complete the experimental setup but are not shown in Figure 1. Signal averaging with base-line subtraction was applied. Steady-state luminescence spectra were recorded with a Shimadzu RF540 spectrofluorimeter. Colloids. Two different colloidal solutions were used: Sol A contained Q-CdS (colorless CdS showing strong size quantization effects), Le., very small particles. It was prepared as described p r e v i o ~ l y . The ~ solution contained 2 X lo4 M CdS, 2X M sodium hexaM excess Cd(ClO&, and 4 X metaphosphate (NaPO& (which mainly consisted of polyphosphates of different chain lengths), and had a pH of 10.2. Sol B consisted of larger particles (yellow CdS). It contained 5 X M CdS and 2 X M HMP, the pH being 10.7. Both sols had broad particle size distributions as could be shown by electron microscopy. Sol A contained mainly particles in the 18-24-A range, which corresponds to agglomeration numbers between 50 and 120, or a mean particle concentration between 4 and 1.6 KM. Sol B contained mainly particles in the 30-42-A range, which corresponds to agglomeration numbers between 300 and 600, or a mean particle concentration between 1.7 and 0.8 CLM.
u-zA DATA-BUS
Figure 1. Simplified diagram of the experimental setup for measuring the transient luminescence and optical absorption. AL, analyzing light source (XBO 450) used only in the absorption measurements; EL, exciting light source used in the luminescence studies; SI, S2, remote controlled shutter; F1,F2, F3, optical filter: L1, L2, L3, lens; CE, sampling cell; MI, mirror; MO, monochromator (Bausch and Lomb); PM, photodetector (photomultiplier R928); BO, base-line compensation cir-
cuit; AMP, dc-coupled broadband preamplifier, gain X l , X10, and X100; REC, digital recorder; A/D, analog-to-digital converter, 12-bit resolution: CO, PDP-1I computer: TE, computer terminal. where more than one radical can react with a colloidal particle, it is shown that the accumulated holes do not necessarily lead to a higher quenching efficiency as they react with each other. Finally, the reader is reminded that CdS colloids undergo photoanodic dissolution when they are illuminated under air.l In a previous study it was shown that the fluorescence intensity of a sol increased by a factor of about 2 in the beginning of dissolution.6 In the present work, dissolution experiments are described with sols y-irradiated under nitrous oxide, i.e., under conditions where OH radicals are generated, and a much more drastic increase in fluorescence intensity is reported.
Experimental Section Appnrutus. The principle of the measurements was not much different from that of ordinary pulse radiolysis. The colloidal solution was illuminated with the filtered light of a pulsed xenon lamp and the fluorescence light was observed at right angles as a function of time. OH radicals were generated in the solution by a 0.5-ps pulse of a 3.8-MeV electron Van de Graaff generator and the intensity of the fluorescence was recorded as the OH radicals reacted with the colloidal particles after the pulse. The pulse was applied 1 ms after the beginning of illumination. Quenching of the fluorescence resulted in an apparent absorption signal. The results are presented in the form of AZ(t)/Iovs time curves, where Io is the steady-state fluorescence intensity before the application of the pulse and AI(t) the change in intensity at time t . In another type of experiment, the solution was first irradiated with the electron pulse from the Van de Graaff generator to produce OH radicals which injected positive holes onto the par(6) Weller, H.; Koch, U.; GutiErrez, M.; Henglein, A. Ber. Bunsen-Ges. Phys. Chem. 1984, 88, 649-656.
(7) Janata, E. Notre Dame Radiation Laboratory Special Report, SR 59, 1980 (8) Janata, E. Rev. Sci. Instrum. 1986, 57, 213-215. (9) Fojtik, A.; Weller, H.; Koch, U.; Henglein, A. Ber. Bunsen-Ges. Phys. Chem. 1984, 88, 969-971.
Photochemistry of Colloidal Semiconductors
The Journal of Physical Chemistry, Vol. 92, No. 9, 1988 2589
-0,001
A =460nm
-0.002 -0.003 -0,004
1 7
m
:-0.005
m
n L
0.003
::
n L
x
n
0.002
m
a
0.001
-0,004 -0.005
a
I/
0'0°' 0 ..
& 20 a 10 300 0
600nm
0.002
20
40 t [PSI
60
80
I/
u 650nm
0
Figure 3. Pulse radiolysis of sol A. Decrease in 310-nm absorption (upper), increase in 500-nm absorption (middle), and decrease in fluorescence intensity (lower) as a function of time after the pulse.
1
2 3 4 [OH1 [ p M l
5
6
Figure 5. Final values of the 460-nm absorption decrease (upper), the 600-nm absorption increase (middle), and the decrease in luminescence (lower) as functions of the concentration of radicals. Sol B.
30
t 0 '
'
450
1
500
I
550 h Inml
I
600
1
650
Figure 4. Quenching of fluorescence (expressed a AIf/&,) as a function of the wavelength of fluorescence. Sol B.
The sols had an orange-red fluorescence, the quantum yield being about 1%. Before irradiation, the sols were purged with a stream of nitrous oxide. It is known from radiation chemistry that OH radicals, which are generated with a yield of 5.4 radicals/ 100 eV absorbed radiation energy, are practically the only reactive products of radiolysis under these conditions.
Results Figure 2 shows how the absorption spectrum of a CdS solution (sol B) changed upon y-irradiation. As the colloid dissolved ions,1° anodically, i.e., in the form of Cd2+ and SO3*-or Sod2the intensity of the absorption spectrum faded away. At the same time, the shape of the spectrum changed as the particles became smaller. Note the blue shift in the absorption edge which is a typical size quantization effect?," The fluorescence band, which had maxima at 590 and 660 nm, became slightly less intense during the first 5 min of irradiation. After 45 min, it was drastically increased. After 75 min, the fluorescence band was further increased and strongly blue-shifted, an effect, which is typical for very small CdS particle^.^ Kinetic traces from the pulse radiolysis studies are shown by Figure 3 for solution A. The upper and middle parts of the figure show ordinary absorption measurements. At 310 nm (upper), where the Q particles have an absorption maximum, a decrease in absorption was observed which is explained by the consumption of CdS. At 500 nm, where the injected positive holes absorb, an increase in absorption was recorded. The lower part of Figure 3 shows how the luminescence at 650 nm changed. The increase in signal height (apparent absorption) corresponds to a decrease (10) Pfanner, K. Thesis, Naturwk-Techn. Akademie Isny and HahnMeitner-Institut Berlin, 1986. (11) Brus, L. E. J . Chem. Phys. 1983, 79, 5566-5571; 1984, 80, 4403-4409.
0
1 [OH I IpM1
2
Figure 6. Quenching of fluorescence as a function of the OH radical concentration for freshly prepared and aged sol B.
in luminescence intensity. All three changes occurred in the same time range. The efficiency of quenching by O H attack of CdS particles, however, depended on the wavelength at which the luminescence was recorded. This is seen from Figure 4, where AZf/Zo is plotted as a function of wavelength, AZf being the change in luminescence intensity at 80 ks, Le., after the reaction of the O H radicals is finished (Figure 3). Below about 550 nm, the efficiency of quenching decreased. In the experiments of Figure 3, the concentration of O H radicals produced was smaller than the concentration of colloidal particles. A particle therefore was not attacked by more than one free radical. As the rate constant of reaction of the O H radicals with the colloidal particles is greater by a factor of at least 10 than ,~ no O H radicals were that of the O H + O H r e a ~ t i o npractically lost. In the experiments of Figure 5, the final changes in absorbance and luminescence were measured as functions of the concentration of free radicals produced. At the higher radical concentrations, the colloidal particles were attacked by several radicals. However, a certain amount of radicals were lost because of the increased rate of the O H + OH reaction. Knowing the specific rate of this reaction (5 X lo9 M-I & ) I 2 and that of the radical-colloid reaction (5 X 10" M-' s-' for sol B), which is diffusion controlled, one could calculate the loss of OH radicals for each pulse strength. The loss of radicals was 30% at the highest dose applied. In the diagrams, the concentration of radicals that reacted with the colloidal particles is given on the abscissa. It is seen from Figure 5 that all three signals, Le., that of consumption of CdS (upper), the formation of holes (middle), and luminescence (lower) behaved in a similar way. Moreover, taking into consideration the absorption coefficient of the hole at 600 nm5 it can be concluded from the observed plateau values of Figure 5 that more than one hole produced by OH attack cannot be stored on a particle. The signals changed linearly with dose at small doses
Kumar et al.
2590 The Journal of Physical Chemistry, Vol. 92, No. 9, 1988 100
I
1
0
2
4
6
8
1
E
0
\ o - ~ io-’
I O H I [pMI
Figure 7. Quenching of fluorescence and hole absorption as functions of the OH radical concentration. Sol A. 80
_----
0
1
2
3
I
10
T[sl
4
5
6
IOHIIpMI
Figure 8. Quenching of fluorescence in sol A as a function of the concentration of OH radicals. Generation of the radicals before measuring the luminescence intensity. T, time interval between radical production and luminescence measurement. The dashed curve was obtained in the normal mode of operation, Le., switching on of the exciting light first and applying the electron pulse 1 ms later.
and finally reached a limiting value at large doses. Sol B was used in these experiments; note that the limiting value of luminescence quenching was 50%. The efficiency of quenching depended on the preparation of the sol. Figure 6 shows a typical example. AZf/Iois plotted here as a function of the radical concentration for sol B, the relationship being measured for a freshly prepared and an aged sol. It is seen that fluorescence quenching at the higher radical concentrations was more efficient in the case of the aged sol. Similarly, it was found that the fluorescence of a sol to which 2 X IO” M AgC104 had been added was more efficiently quenched. Aging and doping with Ag+ ions affect the colloid in a similar way. In both cases, the fluorescence intensity is increased’~’~ and the lifetime of fluorescence is prolonged. In Figure 7 one finds the dependence of AIf/Zo on the radical concentration for sol A. The figure also contains the absorbance due to positive holes. The absorbance and luminescence quenching have the same dependence on the radical concentration. The final value of AZr/Zo is close to 80%, which indicates that the fluorescence of the Q particles can more readily be quenched by positive holes than that of large particles. Experiments were also carried out to follow the 600-nm signal (positive hole absorption) and the 650-nm signal (luminescence quenching) in sol B at much longer times than shown in the figures. As the lamp of the analyzing light could not be flashed, the signals were rather weak and also stability problems of the lamp arose. However, it could be seen that both the absorption at 600 nm and the luminescence quenching faded away in the millisecondsecond time range. This indicates that the holes produced by OH attack are slowly detached from the colloidal particles. The experiments of Figures 8 and 9 were carried out to decide on whether fluorescence quenching was due to the attack of excited CdS particles by OH radicals or of CdS particals in their ground state. Sol A was exposed first to the electron pulse and then to
Figure 9. Efficiency of fluorescence quenching as a function of the time interval T.
the increased light intensity of lamp EL (Figure I ) , and the luminescence intensity was determined at time T after the pulse. In case O H attack on ground-state CdS particles was responsible for fluorescence quenching, the same AI/Zo vs dose curves would result as in the experiments where lamp EL was flashed first and the electron pulse applied after 1 ms. On the other hand if fluorescence quenching was due to reaction of OH with long-lived excited CdS particles, no quenching should be observed in the reverse mode of pulsing. The curves in Figure 8 are similar to those obtained in Figures 6 and 7, Le., in the normal mode of operation. This proves that OH attack on ground-state particles is the reason for quenching. Note that the quenching efficiency in Figure 8 depends on the time interval T. This indicates that the positive holes created by the electron pulse do not live forever but are detached from the particles as was already concluded above from the observation of the hole absorption. In Figure 9 the quenching efficiency is plotted vs T o n a semilogarithmic scale. The damage induced by OH attack fades away in a multiexponential manner, the colloidal particles recovering with respect to their ability to fluorescence.
Discussion Photoanodic Dissolution of CdS. When OH radicals attack CdS particles, a dissolution takes place. The first step of radical attack leads to a positive hole (possibly through an intermediate OH a d d ~ c t )the , ~ absorption of which can be seen in a pulse radiolysis experiment. However, the final products are sulfite and sulfate which has been shown by ion chromatography.” At some point after O H radical attack the oxidation product must leave the colloidal particle. This could already happen at a very early stage. A trapped hole on the surface is an S’- radical that may remain on the colloidal particle only for a limited time period. Our observations about the fading away of the hole absorption and of the rebvery of the fluorescence ability of the particles after OH attack indicate that most of the holes produced by OH attack do not reside longer on the colloidal particles than one-tenth of a second. This finding is important with respect to the use of CdS particles in p h o t ~ a t a l y s i s .A~ hole which is detached in the form of S- may diffuse in the solution and interfere with a cathodic prqcess on another catalyst particle, especially when this cathodic process requires more than one electron, such as the reduction of water. During the later stages of the photoanodic dissolution the fluorescence intensity increases tremendously (Figure 2). The radiationless recombination of the charge carriers obviously becomes less important as the particles become smaller. It seems that in the dissolution process those centers on the surface are destroyed from which the recombination of charge carriers without emission of light takes place. Part of the increase in luminesence may be due to the fact that excess Cd2+ions were formed during the photoanodic dissolution itself. It is known that Cd2+ions in alkaline solution form a protecting layer of cadmium hydroxide around the colloidal particles, which results in a stronger luminescence. 3314
(12) Farhataziz; Ross, A. B. Natl. Stand. Ref: Data Ser., Nafl.Bur. Stand. 1977, 59. (13) Spanhel, L.; Weller, H.; Fojtik, A,; Henglein, A. Ber. Bunsen-Ges. Phys. Chem. 1987. 91, 88-94.
(14) Spanhel, L.; Haase, M.; Weller, H.; Henglein, A. J . Am. Chem. SOC. 1987, 109, 5649-5655.
The Journal of Physical Chemistry, Vol. 92, No. 9, 1988 2591
Photochemistry of Colloidal Semiconductors
I
60 1
50 40
5 30 L
20
10
c
0
1
Z
R
3
I
I
I
I
1
2
3
4
R
L
I 5
Figure 10. Statistical distribution of the radicals on the particles as a function of the ratio of radical to particle concentration. Sum of the fractions of particles carrying an odd and even number of radicals. Inset: fraction of particles having n radicals.
Fluorescence Quenching by OH Radical Attack. In usual fluorescence quenching experiments, the quencher interacts with the excited state of the emitting molecule. Quenching by OH radical attack is brought aboutbby another mechanism. The OH radical does not interact with an excited CdS particle but produces a quenching site on a ground-state particle. A subsequent light absorption by this particle then produces less fluorescence. Under certain pulse radiolysis conditions many more OH radicals were produced than colloidal particles are present. In these experiments, one particle can multiply be attacked and the question may be asked whether the various S- spots produced behave independently or rapid reactions of the type 2s- S S2-(or S22-)take place. S (or S22-)spots may be different from S- spots with respect to their ability to quench the luminescence. If the hole-hole reaction is fast, one would expect colloidal particles carrying an even number of damages to behave differently from those carrying an odd number. This is taken into account in the following discussion of luminescence quenching. The occupancy of a colloidal particle by one, two, etc. radicals is ruled by statistics. The fraction of particles carrying n radicals is derived from Poisson’s equation
-+
fin, = e-RR”/n!
(1)
where R is the ratio of radical to particle concentrations. The inset of Figure 10 shows the fraction of particles carrying n radicals as a function of R . In the main part of the figure, one finds the sums of the fractions with odd and even numbers n. At high radical concentrations, the number of particles carrying odd and even number of radicals is the same, Le., 50%. At lower concentrations of radicals the fraction of odd numbers dominates. The curves in Figure 10 may now be compared to the experimental AZf/Iocurves (Figures 6 and 7 ) . In all these experiments, a linear relation between AZf/Ioand the radical concentration was observed for low radical concentrations, where R < 1. If several radicals were required to produce noticeable quenching, the curves should be bent upwards as in the case for n > 1 in the inset of Figure 10. In the initially linear part of the curves, where R C 1, the relation A I f / I o = Rq
(2)
should be valid, q ( < f ) being the efficiency of quenching by one-radical attack. By measuring AIf/Io as a function of the radical concentration, and knowing the concentration of colloidal particles, one can calculate 9. The particle concentration was known with an uncertainty of 50%. Within this range, 9 values between 0.5 and 1.0 were obtained.
If only an odd number of damages were capable of quenching, AIf/Ioat higher radical concentrations should not exceed 50%. In the case of the small particles a much higher value of 80% was attained, while it was 50% in the case of freshly prepared larger particles. However, when the fluorescence of the latter was increased by aging or doping with Ag’ ions, higher AIf/Iovalues were also observed. These results can perhaps be interpreted by saying that an even number of damages also produce fluorescence quenching, although to a lesser degree than an odd number. The curves in Figures 6 and 7 which fit the experimental data were calculated. In the case of freshly prepared particles in Figure 6, the fit was obtained with a particle size of 40.6 A (or particle concentration of 0.85 gM) assuming that only an odd number of damages led to quenching with q = 1. In the case of Figure 7 , the fit was obtained with a particle size of 18.7 A (or particle concentration of 3.5 pM) assuming q = 1 and 0.8 for odd- and even-numbered particles, respectively. The fact that the absorption of the positive hole produced by OH attack had the same dependence on the concentration of radicals (Figures 5 and 7 ) as the fluorescence quenching shows that only a small number of such holes can exist independently on a colloidal particle. When holes are accumulated, they destroy themselves, possibly by the process mentioned above: 2s- S S2- (or sl2-). After having described the phenomenological aspects, we have to discuss a conceivable molecular mechanism of fluorescence quenching by OH attack. The fluorescence of CdS is attributed to the recombination of charge carriers generated by light absorption. If these carriers have not relaxed into traps before their recombination, the emitted light will be blue-green, Le., correspond to a photon energy equal to the band gap energy. The red fluorescence has to be attributed to the recombination of trapped charge carriers. As the nature of these traps is not exactly known it is difficult to present a definite mechanism of quenching. The trapping sites are very probably on the surface as surface modification procedures strongly change the fluorescence behavior of the colloidal particle^.'^*'^ It is also believed that anion vacancies play an important role as the red fluorescence is stronger in the presence of excess Cd2’ ions.’ The interaction of the electron generated by light absorption with the preexisting hole generated by OH attack is an event in which no light is emitted. This preexisting hole must be different from the hole, whose reaction with the electron produces the red fluorescence. There could be all sorts of empty orbitals on the surface, and so there could be many different “holes” for the electron to decay into before it is finally captured by the trapped hole produced by light absorption. If light was emitted in these decay processes, but no light was emitted in the reccmbination of the electron with a trapped hole S-,one could understand why the hole produced by OH attack acts as a quencher. The band of the red fluorescence is very broad. In the case of such broad bands the fluorescence lifetime depends on the wavelength, shorter times being observed at shorter wa~elengths.6~~~ In the present experiments it was found that the quenching efficiency of the holes produced by OH attack was smaller at shorter wavelengths (Figure 4). Similarly, more efficient quenching occurred in aged sols or sols doped with Ag’ ions where the lifetimes also were longer. Thus one arrives at the conclusion that quenching does not consist of the interaction of the electron with the excess hole produced by OH attack before the electron is trapped but of interacting with it after trapping. Registry No. CdS, 1306-23-6; OH, 3352-57-6.
+
-
(15) Chestnoy, N.; Harris, T. D.; Hull, R.; Brus, E. J . Phys. Chem. 1986, 90. 3393-3399.