Envlron. Sci. Technol. 1994, 28, 776-785
Photocatalytic Production of Semiconductor Collolds
H202
and Organic Peroxides on Quantum-Sized
Amy J. Hoffman, Elizabeth R. Carraway, and Mlchael R. Hoffmann'
Department of Environmental Engineering Science, W. M. Keck Laboratories, California Institute of Technology, Pasadena, California 91 125 Illuminated (320 S X S 370 nm), aqueous suspensions of transparent quantum-sized (Q-sized) ZnO semiconductor colloids in the presence of carboxylic acids and oxygen are shown to produce steady-state concentrations of HzOz as high as 2 mM. Maximum HzOz concentrations are observed only with added electron donors (i.e., hole scavengers). The order of efficiency of hole scavengers is as follows: formate > oxalate > acetate > citrate. Isotopic labeling experiments with l802 are consistent with the hypothesis that hydrogen peroxide is produced directly by the reduction of adsorbed oxygen by conduction band electrons. Quantum yields for H2Oz production are near 30% at low photon fluxes. However, the quantum yield is shown to vary with the inverse square root of absorbed light intensity a ( ( I a b ~ ) - ~with ) ~ / ~the ] , wavelength of excitation, and with the diameter of the Q-sized colloids. The initial rate of Hz02 production is 100-1000times faster with Q-sized ZnO particles (D,= 4-5 nm) than with bulksized ZnO particles (Dp= 0.1 pm).
Introduction Hydrogen peroxide plays a significant role as a primary oxidant in environmental systems, as indicated by a recent review of the atmosphericchemistry of peroxides ( I ) .Below pH 5, the oxidation of sulfur dioxide by Hz02 appears to be the major pathway for the formation of sulfuric acid in humid atmospheres (2-4). Field concentrations up to 60 mM, 250 mM, and 4.1 ppb have been measured in rainwater, cloudwater, and the gas phase, respectively (57). Hz02 can be generated in the gas phase by the disproportionation of two hydroperoxyl radicals (81,at the air-water interface by photoinduced redox processes (9),and in the aqueous phase via photocatalyzed reactions with humic/fulvic acid and green algae (10-12). In cloudwater droplets, scavenging of hydroperoxyl radicals from the gas phase leads to the in situ formation of HzOz. The combination of the in situ generation and scavenging of HzOz from the gas phase is thought to be the main source of HzOz in cloudwater droplets ( 1 3 , 1 4 ) ;however, photocatalytic reactions on naturally occurring metal oxides that are semiconductors may provide an additional source of HzOz. Some naturally occurring metal oxides, e.g., Ti02 and ZnO, have been shown to act as photocatalysts for a large variety of reactions (15), including the production of ammonia from dinitrogen (16,17). Earlier research by Kormann et al. (18) demonstrated the formation of H202 in illuminated aqueous suspensions of desert sand. H202 also plays an important role in pollution control technologies. Enhanced UY degradation rates of organic pollutants in systems containing H z 0 have ~ been reported (19). Semiconductor-assisted photooxidation of organic
* To whom correspondence should be addressed. 776 Environ. Sci. Technot., Vol. 28, No. 5. 1994
pollutants has been demonstrated for PCBs, simple aromatics, halogenated alkanes and alkenes, surfactants, and pesticides (20-27). Complete mineralization has been obtained in some cases (28,291. In most experiments with semiconductor photocatalysts, oxygen was present to act as the electron acceptor. As a consequence of the twoelectron reduction of oxygen, H202 was also formed. This reaction is of particular interest since Gerischer and Heller have suggested that electron transfer to oxygen may be the rate-limiting step in semiconductor photocatalysis (30). The detailed mechanisms of photooxidation on semiconductors are not fully understood; electron donors (i.e., pollutant) are oxidized either directly by valence band holes or indirectly by hydroxyl radicals. Hydroxyl radicals are formed by the reaction of holes with adsorbed HzO, hydroxide, or surface zincanol groups (>ZnOH). Hz02 may also contribute to the degradation by acting as an electron donor or as a direct source of hydroxyl radicals due to homolytic scission. A detailed study of the mechanism of photocatalytic production of peroxides on various metal oxide surfaces may provide more insight into the potential environmental implications and applications of peroxides. The present work examines the mechanism of Hz02 and ROOH formation using quantum-sized (&-sized)semiconductors as photocatalysts. Q-sized semiconductors exhibit a blue shift in absorption onset with decreasing particle size (3134). The advantage of using Q-sized semiconductors in this study is the efficacy of measuring quantum yields due to the negligible light scattering of the small diameter particles. In addition, the high surface area to volume ratios of Q-sized semiconductors enhances the rates of reactions that occur via surface states. We have examined the effect of the type and the concentration of hole scavengers, the concentration of the semiconductor, the pH, the concentration of oxygen, the semiconductor excitation wavelength, the semiconductor particle size, and the light intensity on the quantum yields of peroxide formation. Results of 180 isotopic labeling experiments are presented as evidence for hydrogen peroxide formation via oxygen reduction in the ZnO/formate system. In a companion paper (following paper in this issue), the oxidative pathways are examined in more detail through product analyses using acetate or formate as hole scavengers.
Experimental Section Colloidal suspensions of ZnO with a mean diameter of 5 nm were synthesized via the controlled hydrolysis of Zn(acetate)z (Baker) in 2-propanol (EM Science) (34).All chemicals and solvents were analytical grade (199.0% purity) and were used without any further purification. For the hole scavenger experiments, acetate ions were removed by dialysis using Spectrapor membranes (molecular mass 12 000-14 000) in 2-propanol. Before illu0013-936)6/94/0928-0778$04.50/0
0 1994 American Chemical §oclety
mination, 2-propanol was removed by rotary evaporation. The solid ZnO was then immediately resuspended in water. Water was purified by a Milli-Q/RO system resulting in a resistivity >18 MQ cm. In a typical experiment, 25 mL of aqueous ZnO colloid in a quartz cell was saturated with oxygen (Po2= 1atm) for 30 min prior to illumination with either a Kratos 450W Xe arc lamp or a Spindler and Hoyer lOOOW Xe arc lamp. Continuous oxygen bubbling during the illumination provided a constant dissolved oxygen concentration and enhanced mixing. A CuSO4water filter, a glass IR filter, and a Corning GS7601 filter (bandpass 330-370 nm) were employed to limit heating of the sample and to prevent direct photolysis of hydrogen peroxide. Actinometry was performed using (E)-2-[1-(2,5-dimethyl2-furyl)ethylidene]-3-isopropylidenesuccinicanhydride in toluene (Aberchrome 540) according to the method of Heller and Langan (35).Light intensity was varied using neutral density filters. For Q-sized suspensions, the number of photons absorbed was calculated from the absorption spectrum of the colloid and the incident light intensity as measured by actinometry according to the equation: labs = I, - It,,, = Io(1-l0-fCI). Reaction temperatures were 24 f 1OC. Fluorescence spectra were measured on a Shimadzu RF-540, while absorption spectra were measured on a Hewlett-Packard 8451A diode array spectrophotometer. Hydrogen peroxide concentrations were determined by two different methods. The first method, sensitive to 1 pM HzO2, involved the oxidation of iodide to the triiodide anion (C352nm = 26 400 M-l cm-l) by peroxides catalyzed by ammonium molybdate (18). This method allows one to distinguish between Hz02 and most organic peroxides (ROOH) since the latter are expected to react more slowly with the molybdenum-iodide system (36, 37). Aliquots were diluted with water, then 0.1 M potassium biphthalate (Baker) was added, and the ZnO colloid was dissolved. At t = 0 min, the iodide reagent [0.4 M potassium iodide (MCB), 0.06 M NaOH (EM Science), 0.1 mM ammonium molybdate (Mallinckrodt)] was added and the absorbance vs time profile was recorded. The amount of HzOz formed is calculated from the absorbance at 2 min, while the total peroxide is calculated from the absorbance at 60 min. Method blanks were used to correct the absorbance at 60 min for the slow oxidation of iodide by oxygen. The difference between the total peroxide and inorganic peroxide was called the organic peroxide, which included monoalkyl or dialkyl hydroperoxides. The "inorganic" peroxide fraction could also include peroxyacetic acid, which reacts rapidly with iodide (38). The second method, sensitive to 10 nM HzOz, involved the dimerization of p-hydroxyphenylacetic acid (POPHA, Kodak) in the presence of H202 and horseradish peroxidase (Sigma) to yield a fluorescent product (Aex = 315 nm, A, = 406 nm) (18). A stock solution of HzOz was calibrated by titration with permanganate. Calibration curves from standards were used to calculate the response factors for each method. For the isotopic labeling experiments, 180-enriched water and oxygen were purchased from Isotec. All reactions were run in a closed cell using 100mL of solution and 10 mL of gas headspace. A total of 2 mM formate (EM Science) was added as the hole scavenger. Constant stirring was employed to ensure good liquid-gas mass transfer. The colloidal suspensions were degassed by three freeze-pump-thaw cyclesusing a vacuum line system prior to addition of the isotopically labeled oxygen or water.
The oxygen pressure was 1atm. The irradiation apparatus consisted of a lOOOW Spindler and Hoyer Xe arc lamp equipped with a 320-nm bandpass filter. Each colloidal suspension was irradiated to produce a 0.2 mM concentration of hydrogen peroxide, then acidified with formic acid to pH 3 to dissolve the ZnO colloid, and subsequently concentrated by two vacuum distillations. Blanks containing 2 mM formate and either labeled oxygen or water were stirred in the dark for the same time period as the ZnO/formate irradiated runs and were concentrated by a similar vacuum distillation procedure. The resulting solutions (typically 0.5 mL of 2-20 mM peroxide) were bubbled with argon and stored in septum-sealed vials at 4 "C prior to analysis by GC-MS. Samples of the colloidal suspension and headspace at the beginning and end of each irradiation were also saved for analysis. Peroxide isotopic analyses were performed on a Hewlett Packard Model 589015989A GC-MS operating in the single-ion mode. The separation of hydrogen peroxide from water, acetic acid, and formic acid was achieved using a Hewlett Packard HP-1 (25-m) capillary column. The GC-MS was operated under relatively low temperatures in order to minimize thermal degradation of hydrogen peroxide. The temperatures were 120, 100, 100, and 80 "C for the GC inlet, GC-MS interface, MS source, and quadropole chamber, respectively. The column temperature was held constant at 30 "C until the hydrogen peroxide eluted, and then it was ramped from 30 to 150 "C over a period of 10 min for the samples containing formic and acetic acids. The ionization mode was electron impact (70 eV). Ion mass-to-charge ratios of 18, 19, 20, 32,33, 34, 35,36,37,38,45, and 60 were monitored with a mass resolution of f0.3 amu. A series of hydrogen peroxide standards ranging in concentration from 0.05 to 100 mM was used to generate calibration curves. Typical injection volumes were 0.8 pL for liquid samples and 5-100 pL for gaseous samples. The retention times were 4.41, 5.18, and 5.85 rnin for formic acid, hydrogen peroxide, and acetic acid, respectively. Hydrogen peroxide was also separated from water using a Hewlett-Packard Model 1090(Series 11) HPLC equipped with a Hewlett Packard Hypersil ODS column (5-mm beads, 100 mm X 2.1 mm). The HPLC was connected to the mass spectrometer via a Hewlett Packard Model 59980Bparticle beam interface. The eluent was a mixture of 70% Hz0 and 30% MeOH, which also contained 1mM acetic acid. The temperatures of the particle beam desolvation chamber LC oven temperature were 50 and 45 "C, respectively. The injection volumes ranged from 6 to 16 pL, depending upon the concentration of HzOz in the sample. The mass spectrometer conditions were identical to those conditions used in the GC-MS analysis.
Results and Discussion Semiconductors can act as sensitizers for light-induced redox processes due to their electronic structure, which is characterized by a filled valence band an an empty conduction band (39). Absorption of a photon of energy greater than the bandgap energy leads to the formation of an electronlhole pair. In the absence of suitable scavengers, the stored energy is dissipated within a few nanoseconds by recombination (40). If a suitable scavenger or surface defect state is available to trap the electron or hole, recombination is prevented and subsequent redox Environ. Scl. Technol., Vol. 28, No. 5, 1994 777
0.20
-
A
f
I ImM ZnO.2 m M Acetate 1=2.4E-6 Molar Photow/Mln 0.Y.- 30.72 (HOOHI. 4.3% IROOHI
0
-
-
0.16-
Figure 2 shows an analogous profile for the formation of peroxides in an aqueous ZnO colloid containing 2 mM acetate using an lOOOW Xe arc lamp. In this case, the HzOz reaches a limiting concentration of 1.18 mM in 500 min. The ROOH is produced in much lower quantity relative to HzO2 than in the lower intensity case. The initial quantum yields are 3.4 % and 0.08 % for HzOz and ROOH, respectively. The solid line is a fit of the data to the equations suggested originally by Kormann et al. (18): excitation
-hy‘b + + hv
ZnO 0
100
200
300
Irradiation Time (Min)
Flgure 1. Peroxlde productionIn an aqueous 1 mM Q-slzedZnO colloid at low llght intensity. Detectlon via the iodometrlc method. Upper llne equals H202;lower llne equals ROOH. The open and closed circles represent two different batches of colloid.
reactions may occur. The valence band holes are powerful oxidants ($1.0 to +3.5 V vs NHE depending on the semiconductor and pH), while the conduction band electrons are good reductants (+0.5 to -1.5 V vs NHE) (41).Most organic photodegradation reactions utilize the oxidizing power of the holes; however, to prevent a buildup of charge, one must also provide a reducible species to react with the electrons. In bulk semiconductor electrodes, only one species, either the hole or electron, is available for reaction due to band bending (42). However, in very small semiconductor particle suspensions both species are present on the surface. Therefore, careful consideration of both the oxidative and the reductive paths is required. Due to the redox potentials of the electron/hole pair, HzOz can theoretically be formed via two different pathways in an aerated aqueous solution as follows:
0, + 2eib + 2H’
-
H202
overall stoichiometries for peroxide formation
0, + 2eib 2H’ 2H20 + 2h,b
-+
electron donor oxidation
H20, + 2H’
-
CH3CO; 4- h&
H,O,
CHi
co,
photochemical rate of peroxide production
overall stoichiometries for peroxide destruction H,O,
H,O,
+ 2eib + 2H’ 2qb
-
-
2H,O
0, 4- 2H’
photochemical rate of destruction
(1)
We will present evidence that in our system only the reductive pathway occurs to yield appreciable amounts of HzOz. Figure 1 shows the formation of peroxides in an aqueous ZnO colloid containing 2 mM acetate (pH = 7.8, buffered by the amphoteric mineral surface) using a 450W Xe arc lamp. Detection was by the iodometric method. At short irradiation times, both the H202 (top line) and ROOH (lower line) increase linearly with time. At long irradiation times (>200 min), the rate of peroxide production begins to slow down. The open and closed symbols represent two different batches of colloid, which gave similar results. The initial rate of peroxide production calculated from the slope of Figure 1 is 1.23 X 1o-S M s-l for HzO2 and 1.72 X M s-1 for ROOH. The initial quantum yield (@i) equals the ratio of the initial rate of peroxide production to the rate of photon absorption. The initial quantum yields are 30.7 % and 4.3 % for HzOz and ROOH, respectively. No peroxides are produced in the absence of light absorption by ZnO. It should be noted that in long illumination times, the ZnO colloid is unstable and begins to coagulate due to the intense mixing induced by oxygen bubbling. 778 Envlron. Scl. Technol., Vol. 28, No. 5, 1994
eib
net production peroxide rate
where @O is the quantum yield for peroxide production and @1 is the quantum yield for peroxide destruction at the illuminated interface. Solving eq 11 for the steadystate condition gives
From the steady-state concentration of hydrogen peroxide and the quantum yield of peroxide formation, the apparent quantum efficiency for peroxide degradation (@I) is calculated to be 28.8 M-1. The contribution of peroxide degradation via direct photolysis can be calculated from literature values of the hydrogen peroxide extinction coefficient and quantum efficiency as a function of wavelength (43,44). Using the values of E340 = 0.047 M-l cm-l and a340 = 0.96 gives a rate of degradation of 8.13 X 10-8 M min-l for a 1.18 mM HOOH solution. This represents only 1.2% of the actual degradation rate of 6.47 X lo4 M min-l.
1mM ZnO. 2mM Acetate l=Z.ZE-4 Molar PholonsAln lnltlal O.Y. = 3.4% (HOOH),0.08% (ROOHI
A
I
E
I
"
1.20-
P
0.12
B
0.80-
4
0
0.40-
0.04
U
0
0.001:,
I
ROOH
,
0 HOOH
,=
I
I
5.0
7.0
6.0
9.0
6.0
10.0
11.0
12.0
PH 11
I
10-
Table 1. Quantum Yields of Peroxide Production for Various Electron Donors. donor
HOOH (% )
ROOH (% )
none acetate oxalate formate
2.8 3.1 4.2 5.4
0.3 0.6 0.0
0.1
log Kz,,P+
1.1 3.9 0.7
9
I
I
b-
m
1.0 r M 2.0 mM NaAo
2sc
1
1'
-1
log
7.9 6.9 9.4
i
a 1 mM dialyzed ZnO colloids, pH 7.5-8.0,l mM electron donor, 1.b = 6.85 X 1W M photons min-l. -
1
-2
The effect of increasing the concentration of the electron donor was examined. As the acetate concentration increased from 0.2 to 2.0 mM at constant light intensity, the quantum yields increased from 2.8% to 3.1%. We were unable to increase the acetate concentrations to higher values due to the increased ionic strength causing coagulation of the ZnO colloid. In a previous study, using bulk size TiO2, increasing the acetate concentration from 1to 50 mM increased the peroxide concentration by a factor of 8 for similar light intensities and irradiation times (18). Irradiation of a dialyzed sample of ZnO colloid also produced low concentrations of H202, probably due to the presence of 0.2 mM acetate, which could not be removed by dialysis. A sample of bulk ZnO with no added hole scavengers gave an initial H202 production rate of 1.26 X 10-8 M s-l; however, the H202 concentration reached a maximum of only 10 pM versus the maximum H202 concentration of 1mM in the case of bulk ZnO with 2 mM acetate. This evidence suggests that the acetate is necessary for efficient H202 production, either by acting as a hole scavenger and preventing recombination or by its degradation to yield peroxide products. In order to further assess the importance of hole scavenging upon the rate of peroxide production, we examined the quantum yields of peroxide production with different electron donors. These donors included citrate, acetate, oxalate, and formate. The results are shown in Table 1along with the corresponding metal complexation constants for the 1:l complexes (45) and the hydroxyl radical reaction rates (46). All samples were dialyzed to remove 2 mM acetate, which remained from the synthesis of the ZnO colloids. A t the pH of the experiments (7.58.0), the surface is positively charged (pH,,, = 9.2), and the donors are all anionic. The quantum yields of H202 production increased in the order citrate < no hole
I
I
I
I
-1
0
1
2
Concentration of Acid/Base (mM) Figure 3. pH dependence of quantum yields: 1 mM Q-sired ZnO colloid, 2 mM acetate. (a) Effect of pH on quantum yield. (b) Titration of ZnO colloid.
scavenger < acetate < oxalate < formate. The observed order fails to match either the trends in the stability constants or the hydroxyl radical reaction rate constants. This result most likely indicates that both factors are significant in this system. Formate is unique in that it is known to exhibit a current doubling effect (47). Once oxidized, the C02' - radical can inject an electron into the conduction band of ZnO. Since COz' - radicals are also formed when oxalate is oxidized, this may explain the higher quantum yields for formate and oxalate. The rate of peroxide production was extremely slow when citrate was used as the electron donor; however, prolonged illumination produced an intermediate (detected by UVVis spectroscopy), and thereafter the rate of peroxide production increased rapidly. Although the intermediate was not characterized, it is likely to be acetone dicarboxylic acid, which is formed by the oxidative decarboxylation of citrate. The quantum yields of organic peroxide (ROOH) were essentially zero for oxalate and formate. This result is expected, since oxidation of these anions leads primarily to the formation of C02' - radicals, which are further oxidized to COP. In the case of acetate, the quantum yield of organic peroxides is 0.6%. Significant organic peroxides were also produced in the absence of added hole scavenger due to the 0.2 mM acetate remaining from the dialysis. Figure 3 illustrates the pH dependence of the quantum yield of hydrogen peroxide production using 0.2 mM acetate as the hole scavenger. The pH range of these experiments was limited by the stability of the colloid: below pH 6, the colloid dissolves, and above pH 9, the Environ. Sci. Technol., Voi. 28, No. 5, 1994 770
3
Irradiation Time (Min) Figure 4. Effect of 2-propanol concentration on hydrogen peroxide 10.0, (0)5.2, production rate: 1 mM Q-sized ZnO, 2 mM acetate. (0) (0) 0.10 M.
colloid rapidly coagulates. The behavior in Figure 3 cannot readily be explained in terms of surface charge since the pH,,, for Q-sized ZnO is 9.2, as determined from the titration curve in Figure 3b. If adsorption of the hole scavenger, acetate, to the surface was a controlling factor, an increase in the pH from 6 to 11 would decrease the positive surface charge and should lower the quantum yield. Instead, the opposite result was observed. As the pH increased, the quantum yield of hydrogen peroxide formation increased. This behavior can be explained in terms of changing redox potential of the conduction band electron with pH. The redox potential shifts negative by 59 mV per each unit increase in pH (41). Thus at high pH, the driving force for oxygen reduction to form hydrogen peroxide should increase, and the quantum yields would be expected to be higher, as is experimentally observed. The pH effect is actually more complex, due to the fact that the OH- added to increase the pH is also a good hole scavenger and could form additional H202 by OH' reactions. In the case of bulk ZnO, we observed that the initial rate of peroxide production in the presence of 0.2 mM hydroxide ions was 4.7 X lo4 M min-l versus 6.09 X lo4 M min-' in the presence of 2.0 mM acetate ions. The effect of 2-propanol on peroxide production is presented in Figure 4. As the concentration of 2-propanol increased from 0.1 to 10.0 M, the initial rate of peroxide production increased from 1.1X lo4 M s-l to 2.49 X lo4 M s-1. Thus, we utilized an additional vacuum step to remove any traces of 2-propanol prior to resuspension. Increasing the 2-propanol concentration resulted in higher concentrations of organic peroxides relative to the inorganic peroxide. If the water oxidation pathway was primarily responsible for the production of HzO2, then increasing the concentration of 2-propanol could decrease the rate of peroxide production since 2-propanol is a very efficient hole scavenger. This effect would be offset by the fact that hydrogen peroxide is also formed efficiently by a secondary reaction of the 2-propanol ketyl radicals with oxygen (48). The observation that the rate of peroxide production only increased by a factor of 2.5 when the %propanol concentration increased by a factor of 100 indicates a weak dependence on the hole scavengerlsolvent in this system and suggests that the reduction of oxygen is the more important pathway for the formation of hydrogen peroxide. Figure 5 demonstrates the oxygen dependence of peroxide production. As the partial pressure of oxygen 780 Envlron. Sci. Technol., Vol. 28,
No. 5, 1994
01 0
I
I
I
1
20
40
I
I
I
60
80
100
Oxygen (%I
Figure 5. Effect of oxygen on hydrogen peroxide production. Solid line is fit of data to Langmuirian model: 1 mM Q-sired ZnO; 2 mM acetate.
increased in the bubbling gas, the quantum yield of HzOz reached a limiting value of 2.8%. The solid line is a fit of the data to a Langmuirian adsorption model (49). The quantum yield is not zero at 0% oxygen due to the presence of a small amount of oxygen that could not be removed by sparging. The data suggest that the oxygen, which is reduced by conduction band electrons to produce Hz02, is adsorbed on the surface of the Q-sized semiconductor. At 20 % 0 2 where the quantum yield of hydrogen peroxide formation begins to saturate, the ratio of dissolved oxygen molecules to ZnO particle is calculated to be 300. The Langmuirian dependence on oxygen concentration has also been observed by Okamoto et al. (50) and by Turchi and Ollis (51) for the photodegradation of organic pollutants using Ti02 powder as the photocatalyst. In order to elucidate the mechanism of peroxide production, isotopic labeling experiments were performed. Formate (2 mM) was chosen as the hole scavenger because ZnO/formate systems demonstrated the fastest rate of peroxide production and produced no organic peroxides. In 1953, isotope studies on bulk ZnO suspensions were performed by Calvert et al. (52). The experimental procedure utilized by Calvert et al. was cumbersome; it involved a thorough degassing of the reaction solution and subsequent oxidation of HzOz by ceric sulfate to form oxygen, which was then analyzed at mass 34 and mass 32. In addition, the isotopic enrichments were very low, so the sensitivity was low. The analysis was based upon the premise that the oxidation of HzOz by ceric sulfate did not scramble the isotopic label; however, prior research by Cahill and Taube (53) suggested that a small amount of fractionation did occur. By increasing the isotopic enrichment and developing a method for hydrogen peroxide analysis via GC-MS, we were able to analyze all the expected product mass-to-chargeratios without converting the hydrogen peroxide to oxygen. Although this method requires an increased investment in isotopes, it has the advantage of simplicity of sample preparation and ease of analysis. The hydrogen peroxide calibration curve for a 25-m HP-1 column is shown in Figure 6. The 25-m column exhibited better sensitivity than a 10-m column due to increased separation of peroxide from the solvent peak. Long-term exposure to hydrogen peroxide irreversibly decreases column sensitivity by 70% or more. Under optimal
8.0
,
I
I
I
0.0006
l
m
Slopel.1997 (0.0281
0.V.
lnleroepl-7.988 10,078)
l
l
l
l
~
1b26.4 Mclu P h O l O n r N n
= 2.00,2.06.2.?6, B.U %
0.00061
5.0
4
2.0
I
I
I
Flgure 6, GC-MS hydrogen peroxide calibration curve uslng 25-m HP-1 column. ~~
~~~
Table 2. Mass Ratios from Experiments.
180
Isotopic Labeling
oxygen analysis
331326
34/32
35/32
36/32
ZnO/18Oz start ZnOl18Oz end HOOH/180z start HOOH/l80z end ZnO/H21S0start ZnO/H2180end HOOH/Hz'W start HOOH/Hzls0 end
6.0598 6.0294 6.0122 6.0231 0.0040 0.0041 0.0040 0.0040
10.4532 10.3919 10.3571 10.3791
3.8567 3.9297 3.9696 3.9439 0 0 0
0.4050 0.4248 0.4359 0.4288 0 0 0
0
0
37/34
38/34
HOOH analysis
35/34
O.OOO8 O.OOO8 O.OOO8 O.OOO8
36/34
3.7485 0.3770 6.1046 10.5447 ZnO/18Oz O.OOO9 0 0 0.0042 HOOH/1802 O.OOO8 0 0 0.0040 ZnO/Hs180 0 0 0.0040 O.OOO8 HOOH/Hz*80 a 1 mM $-sized ZnO, 2 mM formate. Peak area ratio of mass 33 to mass 32 ion signals. HzOz peak area ratios are corrected for fractionation of HzOz to HOz+.
concentrations, we were able to detect hydrogen peroxide concentrations as low as 50 pM. We believe this to be the lowest hydrogen peroxide concentration detected by mass spectrometry to date. Most peroxide research utilizing the MS technique reports using concentrations of 1-99 % HOOH,which corresponds to 0.3-33 M (54-56). This detection limit is not as low as that reported for GC and IC systems (10-9 M HOOH),which utilize either a redox chemiluminescence detector or postcolumn derivatization techniques to detect the hydrogen peroxide (57-59). Fragmentation of the hydrogen peroxide in the mass spectrometer ionization chamber is responsible for the higher detection limits in our case. We were also able to detect hydrogen peroxide in a LC/MS thermospray system using a watedmethanol mixture as the solvent; however, the sensitivity and signal to noise ratios were lower probably due to the lack of good mass separation in the desolvation chamber. fn addition, the LC/MS technique had the disadvantage of requiring larger sample sizes. The results of the labeling experiments are presented in Table 2. Control experiments demonstrated that the l80z and H P O isotopic labels do not scramble with Hz02, HzO, or 02.Thus, labeled H202 resulting from the labeled oxygen or water identifies the precursor. At the short irradiation times employed in these experiments, the rate of peroxide production in the closed system was equal to
Irradiation Tkne (Mh) Figure 7. Effect of semiconductor (ZnO) concentration on hydrogen peroxide production. (W) 0.8, (0)0.2, (0)0.5, (0)1.0 mM ZnO.
the initial rates obtained in the open, continuously oxygen bubbled system. Within an experimental error of f6%, the mass ratios of the oxygen analysis were identical at the beginning and end of each run. Furthermore, the observed mass ratios were equal to the calculated ratios for isotopically enriched oxygen gas containing 21 % l60, 65% ITO, and 14% 180. The similarity of oxygen mass ratios between the runs containing ZnO and those without ZnO suggests that the rate of oxygen isotopic exchange between oxygen and water is slow compared to the time scale of the irradiation. Comparison of the mass ratios from hydrogen peroxide and oxygen analysis of the runs without ZnO indicates no detectable oxygen isotope exchange between water and hydrogen peroxide in 120 min and a slight exchange between hydrogen peroxide and oxygen in 120 min. No isotopically enriched hydrogen peroxide product was detected in the runs containing isotopically enriched water. However, the enrichment factors were low, and it is possible that the higher masses were below detection limits. More importantly, the isotopically enriched hydrogen peroxide product was detected only in the run containing ZnO and isotopically enriched oxygen gas, and eachmass ratio from the peroxide analysis matched within 6%the corresponding mass ratio from the oxygen analysis. Thus, based on the above evidence, it appears that the oxygen in the photoproduced hydrogen peroxide originates entirely from the oxygen gas according to the reactions: e;b
+ 0,
-
0,'-
Note that these isotopic labeling experiments cannot distinguish between hydrogen peroxide formed directly via reduction of oxygen by conduction band electrons and hydrogen peroxide formed in a secondary reaction via electron injection of the initially formed formyl radicals into the semiconductor followed by oxygen reduction. Figure 7 presents the semiconductor concentration dependence of peroxide production. As the concentration of ZnO increased, the rate of HZOZproduction also increased, as might be expected since increasing the semiconductor concentration also increases the number of surface sites available for reaction. An additional effect Envlron. Sci. Technol., Vol. 28, No. 5,
1994
781
40
I
'
I
'
I
'
I
'
I
'
band electron transfer to a surface trap, valence band hole transfer to a surface trap, trapped-state recombination back to the ground state, and oxygen reduction:
a
hv
ZnO c b
i
+ eib
-
,
I
c b
0.0012
Intensity (Molar photons/Min) I
ZnO + hv (or heat)
(17)
-
+ ecb
ZnOHl:O,
(18)
>Zn'OH,'
(19)
>Zn"OH'+
(20)
-
P,
0 , , , , , , , , , , , 0.0000 0.0002 0.0004 0.0006 0.0008 0.0010
40
(16)
>ZnOHi + 0, >Zn"OHzf
0 0
+ eib
c b
>Zn'OH,'
I
>Zn"OH
+ >Zn"OH'+
---*
kr -+
>Zn"OHi
+ >Zn"OH
(21)
b Zn(II) j - O >, - 0e" 0-2
__c
Z j - O - H4 2zzn(Ir)
(22)
*
If the rate-determining step is the reduction of the adsorbed oxygen with surface trapped electrons, the rate of reduction is given by
0
600
1200
I800
2400
Inverse Square Root of Intensity
Using a steady-state analysis on ZnI-HzO' gives the expression:
Flgure8. (a) Light intensity dependenceof quantum yields of hydrogen peroxlde production. Open and closed circles represent different batches of colloid: 1 mM Q-sized ZnO, 2 mM acetate. (b) Solid line is fit of data to inverse square root of intensity.
arises from differences in the absorbed light intensity with the variation in semiconductor concentration. The extinction coefficient of Q-sized ZnO is approximately 320 M-l cm-l at 350 nm. When the quantum yields, which were corrected for differences in absorbed light intensity, were calculated, CP was found to decrease from 6.1% to 2.0% with increasing concentrations of the ZnO. This result led us to examine the effect of light intensity on the quantum yields of peroxide production. By using a combination of neutral density filters and different wattage Xe arc lamps, we were able to vary the light intensity over 3 order of magnitude. As exhibited in Figure 8, the quantum yield decreased rapidly as the light intensity increased. The solid line is a fit of the quantum yield to the inverse square root of light intensity (vide infra). When the absorbed light intensity is varied by changing to ZnO concentration while keeping the incident light intensity constant, the values of the quantum yields fit to the same line. Thus, the controlling factor in this case is the difference in absorbed light intensity, not the change in concentration of surface sites. The quantum yield is expected to vary as the inverse square root of the absorbed light intensity under conditions in which secondorder recombination reactions compete with first-order peroxide production. In order to explain these observations, we propose the following model for photoreactivity, which includes photophysical excitation, direct band gap recombination, surface adsorption of dioxygen, conduction 782
Environ. Scl. Technol., Vol. 28, No. 5, I994
Likewise the steady-state analysis on [ecb-l gives die,] -dt
- I&- k,,[eibbl
[h&1 - k,,[>Zn"OH,+I [eibl -t-
= 0 (25)
k-,,[>Zn'OH,'l
Since the photogeneration rates of h,b+ and ecb- are equal and the intrinsic carrier density is comparatively low, [eeb-l = [h,b+], and the equation becomes labs
= k,,[>Zn"OH,+] [eib] - k-,,[>Zn'OH,'l
+ kl,[eib12 (26)
At high carrier concentrations (Le., high light intensity), the third term in eq 26 is much larger than the other two terms and the equation simplifies to (27)
For very low light intensities, the third term in eq 26 is much less than the other two terms and the equation simplifies to
When the back-reaction rate of eq 19 is slow compared to the rate of light absorption, then the equation for the
concentration of electrons simplifies further to give [e,]
=
labs
klg[>Zn'IOH,I
o"21
The rate of reduction of oxygen at high light intensity is then given by
- kobfl(Iabs)1'2
[>Zn"Ozl ads 1+ K'[>Zn"Ozlads
390
o,18i 0.12
(32)
1
1
0.06
where kob= klgk17-1/2 [>ZnlIOHz+]andK' = kZdk-19, while for the conditions of low light intensity the following equation is obtained:
The quantum yield is defined as the ratio of the rate of the reaction to the rate of photon absorption. Therefore, the final expressions for the quantum yields at high light intensity and at low light intensity are, respectively, as follows:
(35)
This dependence has also been observed in this laboratory for the case of photodegradation of chloroform (22) and for the case of photoinduced polymerization of methyl methacrylate (60). The semiconductor excitation wavelength dependence of the rate of peroxide production was also examined using acetate as the hole scavenger. Excitation wavelengths between 320 and 380 nm were selected using a Jarrell-Ash Monospec 27 triple-grating monochromator with a resolution of 1.0 nm using a 0.5-mm slit width. The results are shown in Figure 9. The production rate decreases as the wavelength increases, in a manner which loosely follows the absorption spectrum of the colloidal solution as shown in Figure 9b. The additional feature at 358 nm may be due to the overlap of an 02- Zn2+ charge transfer band with the colloid's absorption spectrum. In a study of the photoinduced reductive dissolution of &-Fez03by bisulfite, Faust and Hoffmann (61)observed a sharp increase in
-
370
360
Wavelength tnm)
Rearranging eqs 30 and 31 and collecting constants gives the followingsimplified equation for the conditions of high light intensity: d[Oz' dt
330
310
and the corresponding equation at low intensity is given by
0.00, 310
I
1
I
,
I
I
330
,
1
360
,
'
I
I
I
v
I
370
390
Wavelength (nrn) Flgure 9. (a) Effect of semiconductor excitation wavelength on rate of hydrogen peroxlde prductlon: 1 mM Q-sized ZnO, 2 mM acetate. (b) Absorption spectrum of the ZnO colloid.
Table 3. Wavelength Dependence of Quantum Yields of Hydrogen Peroxide Production. wavelength (nm)
quantum yield (% )
wavelength (nm)
quantum yield (% )
320 9.15 354 4.23 330 16.32 360 6.86 340 3.99 364 4.35 344 4.08 370 1.45 350 3.86 a Corrected for differences in absorbed light intensity. 1 mM &-sizedZnO, 1 mM acetate.
quantum yields in the spectral region dominated by the lowest energy charge-transfer band (Ama = 375 nm for 02Fe3+) of hematite. The quantum yields of peroxide production as a function of wavelength are listed in Table 3. The effect of the size of the semiconductor particles and the type of semiconductor were also examined using 2.0 mM acetate as the hole scavenger. The initial rates of HzOz production were 100-1000 times slower for bulk ZnO (diameter 0.1 pm, 1 mM) than for 1 mM Q-sized ZnO. (We were unable to compare quantum yields due to the light scattering by the bulk ZnO.) When Q-sized ZnO and Q-sized Ti02 [synthesized according to the method of Kormann et al. (34),diameter 2 nml were compared, the quantum yields of both H202 and ROOH production were significantly higher for ZnO than for Ti02 (Le., 2.1% and 0.5 ?6 for ZnO vs. 0.1 % and 0.1 % for TiOz). These results are complicated by the increased coagulation rate of Ti02
-
Envlron. Sci. Technol., Vol. 28, No. 5, 1994 783
1 i i
I
I
i
1
-
0.1 -
0.0 I
22
, 24
I
26
'
I
28
'
I
30
'
I
32
'
I
34
'
I
36
'
38
Conclusions
Size (Angstroms)
/A c
O n
20-
2
.
.cu
bl
4
1
'
i
W d
0 22
1
__
I
26
30
34
30
Size (Angstroms) Flgure 10. (a) Particle size dependence (colloid diameter) of quantum yields: 1 mM Q-slzed ZnO, 2 mM acetate, in 2-propanol. (b) Fit of inverse quantum yield to particle diameter.
in the presence of 2.0 mM acetate and the tendency of peroxides to adsorb to the Ti02 surface (62). Peroxide adsorption is not a problem with ZnO because addition of the biphthalate buffer dissolves the colloid prior to iodometric analysis for peroxides. Under these conditions Ti02 does not dissolve, thus the peroxide quantum yields may be artificially lower for TiO2. This may be partially compensated by the lower absorbed light intensity for Ti02 versus ZnO, which results in higher quantum yields as demonstrated in Figure 8. Kormann et al. have also observed lower quantum yields of peroxide formation for Ti02 vs ZnO (18) and have shown that the lower steadystate concentration of peroxide for Ti02 arises from a decreased quantum yield of peroxide formation combined with an enhanced degradative quantum yield. Figure 10 illustrates in greater detail the dependence of the quantum yield of peroxide formation upon the semiconductorparticle size. Different diameters of Q-sized ZnO in 2-propanol were obtained by synthesis at 0 "C followed by slower aging at low temperatures. The absorption spectrum of each colloid was recorded before and after each irradiation, and the absorption onset was used to calculate the average particle diameter according the method of Brus (31-34). The absorption spectra were also used to integrate the colloid absorption as a function of wavelength over the light output of the lamp/filter system as a function of wavelength. This was necessary to obtain the best value of the absorbed light intensity for each colloid since the absorption spectra change in intensity and red shift as a function of aging time. The 784
Envlron. Scl. Technol., VoI. 28, No. 5, 1994
quantum yields are much higher than before (approaching 1at the lowest particle diameter) because 2-propanol was used as the solvent. The quantum yields decrease rapidly as the Q-sized particle diameter increases. The data fits an inverse quantum yield versus particle diameter relationship, as shown in the Figure lob. Decreasing quantum yields with increasing particle size has also been observed by Anpo et al. (63) for the photocatalytic hydrogenolysis of CH3CCH with water using Ti02 as a photocatalyst; however, these quantum yields were calculated using incident light intensity instead of absorbed light intensity. The effect can be due to the increased surface area of the smaller diameter colloids or due to the enhanced redox potential of the conduction band electron.
We have demonstrated the production of HzOz using oxygenated aqueous suspensions of ZnO semiconductors. Quantum yields as high as 30% have been obtained in the case of low light intensities with aqueous suspensions of ultrasmall particles, while quantum yields as high as 100% have been obtained using Q-sized ZnO in 2-propanol. At high light intensities and long irradiation times, limiting steady-state concentrations as high as 2 mM have been observed. With bulk-sized ZnO, H202 was produced even in the absence of added hole scavengers; however, the concentrations were very low presumably due to backreactions of the peroxide with electrons or holes. Quantum yields increased as the concentration of the hole scavenger (e.g., electron donors such as acetate or 2-propanol) increased. Under these conditions, the rate of hole transfer to the electron donor increases relative to the rate of electronlhole recombination in a system with no added scavenger, thus increasing the rate of peroxide production. The quantum yields depend on the type of hole scavenger and increase in the order: citrate C no added scavenger C acetate < oxalate C formate. The quantum yields for peroxide production depend strongly upon the absorbed light intensity and fit an inverse square root model. The dependence of the quantum yield on oxygen pressure was Langmuirian, suggesting that the primary formation of H202 occurs via the reduction of adsorbed oxygen by conduction band electrons. isotopic labeling experiments demonstrate that all the oxygen in the hydrogen peroxide product arises from dioxygen. The lack of scrambling of the higher mass ratios indicates that the 0-0 bond stays intact. The initial rates of peroxide production on bulk ZnO are 100-1000 times slower than those on Q-sized ZnO. Furthermore, the quantum yields decrease as the diameter of the Q-sized ZnO increases. Results of the isotopic labeling experiments clearly show that hydrogen peroxide is formed as the result of the reduction of dioxygen by conduction band electrons. Acknowledgments We are grateful to the EPA (815041-01-0;815170-01-0) and ARPA (NAV5 HFMN N0001492J1901) for financial support, and we appreciate the assistance of Drs. Detlef Bahnemann, Claudius Kormann, and Peter Green in various aspects of this research. Literature Cited (1) Gunz, D. W.; Hoffmann, M. R. Atmos. Enuiron. 1990,24A (7), 1601.
(2) Hoffmann, M. R.; Boyce, S. D. In Advances in Experimental Science and Technology;Schwartz, S. E., Ed.; Wiley: New York, 1983; Chapter 12, pp 147-189. (3) Hoffmann, M. R.; Jacob, D. J. in SO,, NO, and NO2 Oxidation Mechanisms: Atmospheric Considerations; Calvert, J. G., Ed.; Acid Precipitation Series; Butterworth: Boston, MA, 1984; Vol. 111, pp 101-172. (4) Jacob, D. J.; Hoffmann, M. R. J . Geophys. Res. C: Oceans Atmos. 1983, 88, 6611. (5) Kelly, T. J.;Daum, P. H.; Schwartz, S. E. J . Geophys. Res. D: Atmos. 1985, 90, 7861. Zika, R. G.; Saltzman, E.; Chameides, W. L.; Davis, D. D. J . Geophys. Res. C: Oceans Atmos. 1982,87, 5015. Heikes, B. G.; Kok, G. L.; Walega, J. G.; Lazrus, A. L. J . Geophys. Res. D: Atmos. 1987, 92, 915. (8) Chameides, W. L.; Davis, D. D. J . Geophys. Res. C: Oceans Atmos. 1982, 87, 4863. (9) Zika, R. G.; Saltzman, E.; Chameides, W. L.; Davis, D. D. J . Geophys. Res. C: Oceans Atmos. 1982,87, 5015. (10) Zika. R. G. Elsevier Oceanogr. Ser. (Amsterdam)1981,31, 299. (11) Cooper, W. J.;Zika, R. G. Science (Washington,D.C.)1983, 220, 711. (12) Moffett, J. W.; Zika, R. G. Mar. Chem. 1983, 13, 39. (13) Graedel, T. E.; Goldberg, K. I. J . Geophys. Res. C: Oceans Atmos. 1983,88, 10865. (14) Chameides, W. L. J . Geophys. Res. D: Atmos. 1984, 89, 4739. (15) Fendler, J. H. J. Phys. Chem. 1985, 89, 2730. (16) Schrauzer, G. N.; Strampach, N.; Hui, L. N.; Palmer, M. R.; Saleshi, J. Proc. Natl. Acad. Sci. U.S.A. 1983, 80, 38733876. (17) Soria, J.; Conesa, J. C.; Augugliaro, V.; Palmisano, L.; Schiavello, M.; Sclafani, A. J . Phys. Chem. 1991, 95, 274. (18) Kormann, C.; Bahnemann, D. W.; Hoffmann, M. R. Environ. Sci. Technol. 1988, 22, 798. (19) Ollis, D. F.;Pelizzetti, E.; Serpone, N. Enuiron. Sci. Technol. 1991, 25 (9), 1523. (20) Ollis, D. F.; Pelizzetti, E.; Serpone, N. In Photocatalysis: Fundamentals and Applications; Serpone, N.; Pelizzetti, E., Eds.; Wiley: New York, 1989; pp 604-637. (21) Chemseddine, A.; Boehm, H. P. J. Mol. Catal. 1990, 60, 295. (22) Kormann, C.;Bahnemann,D. W.;Hoffmann,M.R. Environ. Sci. Technol. 1991, 25, 494. (23) Hidaka, H.; Zhao, J.; Pelizzetti, E.; Serpone, N. J . Phys. Chem. 1992,96, 2226. (24) D’Oliveira, J.-C.; Al-Sayyed, G.; Pichat, P. Environ. Sci. Technol. 1990, 24, 990. (25) Kochany, J.; Bolton, J. R. Environ. Sci. Technol. 1992,26, 262. (26) Pelizzetti, E., Minero, C.; Carlin, V.; Borgarello, E. Chemosphere 1992, 26, 343. (27) Mills, G.; Hoffmann, M. R. Enuiron. Sci. Technol. 1993,27, 1681. (28) Pruden, A. L.; Ollis, D. F. J . Catal. 1983, 82, 404. (29) Serpone, N., Pelizzetti, E., Eds. Photocatalysis: Fundamentals and Applications; Wiley: New York, 1989;p 604. (30) Gerischer, H.; Heller, A. J . Phys. Chem. 1991, 95, 5261. (31) Henglein, A. Chem. Rev. 1989, 89, 1861. (32) Brus, L. E. J . Chem. Phys. 1984,80, 4403. (33) Brus, L. E. J . Chem. Phys. 1983, 79,5566. ~I
(34) Bahnemann, D. W.; Kormann, C.; Hoffmann, M. R. J . Phys. Chem. 1987,91,3789. (35) Heller, H. G.; Langan, J. R. J . Chem. Soc., Perkin Trans. 1981, 2, 341. (36) Lazrus, A. L.; Kok, G. L.; Gitlin, S. N.; Lind, J. A.; McLaren, S. E. Anal. Chem. 1985,57 (4), 917. (37) Kok, G. L.; Thompson, K.; Lazrus, A. L.; McLaren, S. E. Anal. Chem. 1985, 58 (6), 1192. (38) Frew, J. E.; Jones, P.; Scholes, G. Anal. Chim. Acta 1983, 155, 139. (39) Boer, K. W. Survey of Semiconductor Physics; Van Nostrand Reinhold: New York, 1990; p 249. (40) Rothenberger, G.; Moser, J.; Gratzel, M.; Serpone, N.; Sharma, D. K. J . Am. Chem. SOC.1985,107,8054. (41) Gratzel, M. Heterogeneous Photochemical Electron Transfer; CRC Press: Boca Raton, FL, 1989; p 101. (42) Memming, R. In Topics in Current Chemistry; Steckhan, E., Ed.; Springer-Verlag: Berlin, 1988; Vol. 143, pp 79-113. (43) Schumb, W. C.; Satterfield, C. N.; Wentworth, R. L. Hydrogen Peroxide;Reinhold Publishing Corp.: New York, 1955; p 291. (44) Zellner, R.; Exner, M.; Hermann, H. J . Atmos. Chem. 1990, 10, 411. (45) Smith, R. M.; Martell, A. E. Critical Stability Constants; Plenum Press: New York, 1977; Vol. 111, p 1. (46) Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. J . Chem. Phys. Ref. Data 1988, 17 (2), 513. (47) Gomes, W. P.; Freund, T.; Morrison, S. R. J . Electrochem. SOC.:Solid State Sci. 1968, 115 (8), 818. (48) Bothe, E.; Schuchmann, M. N.; Schulte-Frohlinde,D.; Von Sonntag, C. Photochem. Photobiol. 1978,28, 639. (49) Adamson, A. W. Physical Chemistry of Surfaces; Wiley and Sons: New York, 1982; p 533. (50) Okamoto, K.-I.; Yamamoto, Y.; Tanaka, H.; Itaya, A. Bull. Chem. SOC.Jpn. 1986,58, 2023. (51) Turchi, C. S.; Ollis, D. F. J . Catal. 1990, 122, 178. (52) Calvert, J. G.;Theurer, K.; Rankin, G. T.; MacNevin, W. M. J . Am. Chem. SOC.1954, 76,2575. (53) Cahill, A. E.; Taube, H. J . Am. Chem. SOC.1954, 74, 2312. (54) Foner, S. N.; Hudson, R. L. J . Chem. Phys. 1962,36 (lo), 2676. (55) Roberson, A. J. B. Trans. Faraday SOC.1952,48, 228. (56) Lindemann, L. P.; Guffy, J. C. J . Chem. Phys. 1958,29,246. (57) Lazrus, A. L.; Kok, G. L.; Gitlin, S. N.; Lind, J. A.; McLaren, S. E. Anal. Chem. 1985,57,917. (58) Hewitt, C. N.; Kok, G. L. J . Atmos. Chem. 1991,12,181. (59) Nyarady, S. T.; Barkley, R. M.; Sievers, R. E. Anal. Chem. 1985,57, 2074. (60) Hoffman, A. J.; Mills, G.; Yee, H.; Hoffmann, M. R. J . Phys. Chem. 1992,96,5540. (61) Faust, B. C.; Hoffmann, M. R. Environ. Sci. Technol. 1986, 20 (9), 943. (62) Kiwi, J.; Gratzel, M. J. Mol. Catal. 1987, 39, 63. (63) Anpo, M.; Shima, T.; Kodama, S.; Kubokawa, Y. J . Phys. Chem. 1987,91,4305.
Received for review May 12,1993. Revised manuscript received November 29,1993. Accepted January 5, 1994.’ @Abstractpublished in Advance ACS Abstracts, February 15, 1994.
Environ. Scl. Technol., Vol. 28, No. 5, 1994
785
