New method to determine the generation of hydroxyl radicals in illuminated TiO2 suspensions

Article abstract of DOI:10.1021/jp971315c

Hydroxyl (^·OH) radicals generated by the irradiation of aqueous TiO₂ suspensions were investigated using electron paramagnetic resonance (EPR) and product analysis employing a free nitroxide radical (3-carboxyproxyl) as a spin trap of ^·OH. The rate of formation of ^·OH was a linear function of light intensity at low intensities. Addition of selected anions to the photolysis mixtures resulted to diminished rate of disappearance of the spin trap resulting from the competition between the spin trap and the anion for ^·OH.

Full text of DOI:10.1021/jp971315c

J. Phys. Chem. B 1997, 101, 7127-7134
7127
A New Method To Determine the Generation of Hydroxyl Radicals in Illuminated TiO2
Suspensions
Peter F. Schwarz and Nicholas J. Turro*
Department of Chemistry, Columbia UniVersity, New York, New York 10027
Stefan H. Bossmann and Andre M. Braun
Lehrstuhl fu¨r Umweltmesstechnik am Engler-Bunte-Institut der UniVersita¨t Karlsruhe,
76128 Karlsruhe, Germany
Aboel-Magd A. Abdel Wahab
Faculty of Science, Assiut UniVersity, Assiut, A. R. E. Egypt 71516
Heinz Du1rr
Fb.:11.2 Organische Chemie, UniVersita¨t des Saarlandes, 66123 Saarbru¨cken, Germany
ReceiVed: April 16, 1997; In Final Form: June 24, 1997X
The generation of hydroxyl (•OH) radicals produced by irradiation of aqueous TiO2 suspensions was investigated
by electron paramagnetic resonance (EPR) and product analysis employing a stable free nitroxide radical
•
•
(3-carboxyproxyl) as a spin trap of OH. Product analysis demonstrated that the attack of OH on
3-carboxyproxyl produces a diamagnetic product (proxyl-NH) with a trapping efficiency of ca. 80%. It could
also be shown that a small amount of the nitroxides is reduced by conducting band electrons. The rate of
formation of •OH could be determined by monitoring the time dependence of the decrease in the concentration
of 3-carboxyproxyl monitored by EPR spectroscopy or the time dependence of the increase in the concentration
•
of proxyl-NH monitored by gas chromatography analysis. The rate of formation of OH serves as a
mechanistic parameter to investigate the mechanism of formation of this reactive species by the photoexcitation
of TiO2. The dependence of the rate of formation of •OH was shown to be a linear function of light intensity
at low intensities. It was observed that molecular oxygen, a good scavenger of conduction band electrons,
only slightly influences the rate of formation of •OH and supports the production of •OH from photogenerated
holes. Addition of selected anions to the photolysis mixtures results in a diminished rate of disappearance of
•
the spin trap as the result of a competition between the spin trap and the anion for OH. From a Stern-
Volmer analysis of the data, the rate constant for the reaction of hydroxyl radicals with the anions was
determined.
Introduction
SCHEME 1
During the past two decades following the discoveries of
Fujishima and Honda,1,2 considerable research has been devoted
to the investigation of the mechanism of the photoelectrochem-
istry of semiconductors such as TiO2, especially for air-saturated
aqueous solutions of organic molecules. This research has been
motivated in part because of the potential for the use of
irradiation of semiconductors as catalysts and as a means of
mineralization and purification of organic pollutants in envi-
ronmental aqueous media.3-6
From this research a useful working paradigm has evolved
(Scheme 1) which views the mineralization of organic pollutants
as originating from an initial process in which the absorption
of a photon, possessing an energy that is greater than that of
the band gap of the semiconductor particle, causes the promotion
of an electron from the valence band (vb) of the semiconductor
to the conduction band (cb) of the semiconductor. This process
generates a positive “hole”, hvb+, at the valence band edge of
the semiconductor particle and an electron, ecb-, in the conduc-
tion band of the particle (eq 1). A number of processes can
follow to initiate the mineralization of organic materials present
in the irradiated suspension containing the semiconductor, e.g.,
(1) hvb+ and ecb- can react directly with organic species adsorbed
+
-
on the surface of the particle, (2) hvb and ecb can be
transformed into high-energy chemical species which react with
+
-
the absorbed organic species, or (3) hvb and ecb can be
transformed into high energy chemical species which can diffuse
into the aqueous phase and then react with dissolved organic
X Abstract published in AdVance ACS Abstracts, August 15, 1997.
S1089-5647(97)01315-1 CCC: $14.00 © 1997 American Chemical Society

7128 J. Phys. Chem. B, Vol. 101, No. 36, 1997
Schwarz et al.
+
species. In the case of TiO2 it is commonly accepted that hvb
is transformed into hydroxyl radicals (•OH, eq 2) and that in
the presence of dissolved molecular oxygen e- is converted into
the superoxide anion (•O2-, eq 4). The quantum efficiency of
the production of reactive species depends on the efficiency of
eq 1 and the efficiency by which the reactive species produced
in eqs 2 and 4 can initiate mineralization. It is generally
accepted that hole/electron annihilation (eq 7) to produce
unproductive heat is a very fast process2 (occurring on the
nanosecond time scale) that determines the efficiency of
mineralization, especially under conditions for which the organic
material is not strongly adsorbed on the surface of the particle
(see Scheme 1).
It is generally assumed that the •OH radicals produced in eq
2 are the major species responsible for the mineralization of
organic pollutants.7,8 This premise is rendered plausible because
of the lower reducing potential of a conduction band electron
compared to the oxidizing power of a valence band hole and
because most reducible substrates may not compete kinetically
with molecular oxygen in trapping photogenerated conduction
band electrons. Convincing evidence for the generation of
•OH upon irradiation of aqueous solutions of TiO2 has been
obtained from the use of spin traps9-13 capable of scavenging
the •OH radicals and producing a characteristic nitroxide which
can be detected by electron paramagnetic resonance (EPR).
Jaeger and Bard9 first demonstrated that •OH could be scavenged
by spin traps such as 5,5-dimethyl-1-pyrroline N-oxide (DMPO),
Figure 1. EPR signal and structure of 3-carboxyproxyl in an aqueous
solution (pH ) 3.8). EPR spectrometer settings: center field, 348 mT;
sweep width; 6 mT; number of scans, 10. Microwave frequency: 9.78
GHz. The arrow indicates the peak used for height measuring.
purification system) and sonicated for 10 min. The pH values
for the flash photolysis experiments were adjusted with HClO4
and NaOH.
EPR Experiments. The EPR spectra were recorded using
a Bruker ESP 380 ESR spectrometer, operating at the X band.
Data acquisition and instrument control were achieved by using
Bruker’s ESP 380 E software. The samples were examined in
a fused silica cell inside a rectangular TE102 (Bruker ER 4102ST)
cavity of the EPR instrument. In order to record the consump-
tion of 3-carboxyproxyl in the irradiated solutions, the height
of the low-field EPR line (Figure 1) was monitored as a function
of time.
O2 Concentration. To control the concentration of molecular
oxygen, the samples were bubbled with air for 15 min, prior to
the EPR experiments. To examine the influence of the oxygen
concentration on the 3-carboxyproxyl consumption, the samples
were bubbled with a mixture of prepurified O2 and N2 for 40
min before the experiments. During the equilibration process,
the samples were isolated from surrounding air by rubber
stoppers.
•
•
for which a OH-trapping efficiency (chemical yield of OH
adduct based on DMPO) of ca. 33% has been estimated.13 In
addition to serving as a means of demonstrating the formation
of •OH, the spin-trapping technique has been usefully employed
to investigate the mechanism of the processes implied in eqs
1-7, including the absolute quantum efficiency of the formation
•
of OH.
In the present report we evaluate the technique employing
an EPR active nitroxide, 3-carboxyproxyl, as the spin trap for
Photon Flux. The samples inside the optical cavity of the
EPR instrument were illuminated by a Xe lamp (LX300 UV)
whose output passed through a H2O infrared filter of 10 cm
path length with a UV filter WG 280 (CVI Laser Cooperation)
inside and then focused onto the cell with a quartz lens. To
examine the influence of the light intensity on the 3-carboxy-
proxyl consumption, neutral density filters were placed between
the lamp and the IR cell. For the flash photolysis experiments,
a frequency-tripled Nd-YAG laser (λex ) 355 nm with flash
duration, ca. 15 ns) was used. The TiO2 loading for these
•
which reaction with OH produces diamagnetic EPR inactive
products. The premise behind the selection of the method is
the validity of the paradigm described above for the role of
•OH as the dominant oxidizing species produced upon irradiation
of aqueous solutions of TiO2 and the expectation that the
radical-radical coupling reaction between a nitroxide and •OH
would be extremely rapid (eq 3). Under these assumptions,
the time dependence of the disappearance of 3-carboxyproxyl
provides a parameter directly related to the concentration of
•OH. It is also expected that the validity of the technique could
be tested by comparing results to those achieved by the
traditional spin-trapping methods recently reported.12 To pro-
vide further mechanistic information a quantitative analysis of
the products produced by photolysis of aqueous solutions of
3-carboxyproxyl in the presence of TiO2 was undertaken. The
experiments was 0.002 g L-1
.
Photolysis. The photolysis experiments were carried out
using a TQ 150 (Heraeus) lamp illuminating an aqueous,
unbuffered, air-saturated 400 mL solution of 0.20 g L-1 Degussa
P-25 TiO2 and 8.0 × 10-4 mol L-1 3-carboxyproxyl. The
spectral emittance of the lamp between 200 and 400 nm was
determined by uranyl oxalate actinometry to be 11 W. The
solution samples (5 mL) were taken at approximately 0, 60,
120, 180, 240, 300, 360, 420, 480, 540, and 600 s and filtered
through 0.22 µm nylon-syringe-filter (Roth). The TiO2 residue
was then washed with 5 mL CHCl3. The solution of 3-car-
boxyproxyl and the reaction products were injected into a gas
chromatograph (GC) (HP 5971A MSD (mass selective detector),
coupled with a HP 5965B ID (infrared detector)). The injection
volume was 0.10 mL. An HP-INNOWAX capillary column
(cross-linked polyethylene glycol) was employed. All reaction
products were identified by IR and mass spectroscopy (MS).
For the structural determination of the reaction products formed,
authentic samples of 3-carboxyproxyl and proxyl-NH were
employed as standards. The same GC/MS response factor of
•
quantum yield for formation of OH and the photon flux
dependence of the rate of formation of •OH are all found to be
consistent with recently published results confirming the
quantitative validity of the 3-carboxyproxyl spin trap method.
Experimental Section
3-Carboxy-2,2,5,5-tetramethyl-1-pyrroldinyloxy (3-carboxy-
proxyl) and its sodium salts were purchased from Aldrich and
used without further purification. 2,2,5,5-Tetramethyl-3-pyr-
rolidine-3-carboxamide was also purchased from Aldrich and
hydrolyzed to 3-carboxy-2,2,5,5-tetramethylpyrrolidine (proxyl-
NH). The samples containing 0.20 g L-1 Degussa P-25 TiO2
and 8.0 × 10-4 mol L-1 3-carboxyproxyl, unless otherwise
noted, were prepared in deionized water (Millipore water

Generation of Hydroxyl Radicals in TiO2 Suspensions
J. Phys. Chem. B, Vol. 101, No. 36, 1997 7129
3-carboxyproxyl and 2,2,5,5-tetramethyl-3-carboxypyrollidine
oxime (proxyl-OH) was assumed for quantitative determination
of products. For the dimer, it was estimated that the GC/MS
response is 2 times that of 3-carboxyproxyl.
teristic of the fast motion conditions (τc < 2-3 × 10-9 s), the
three nitrogen hyperfine lines of the spin trap 3-carboxyproxyl
are modified in their relative widths for variations of the
environmental microviscosity.21 In this case it is possible to
calculate the widths of each hyperfine component by the
simplified formula22,23
Results and Discussion
The Spin-Trapping Method for Determining the Genera-
tion of •OH. The EPR spin-trapping technique, developed by
Janzen and Blackburn,14 was adapted by Jaegar and Bard to
2
∆H(mN) ) A + BmN + CmN
(9)
In this formula A represents both contributions from the
anisotropies of the g tensor (Zeeman coupling) and the AN tensor
(hyperfine coupling) and terms which are not motionally based.
The remaining B and C terms also contain the anisotropic
components of the g and A tensors and the spin densities in
terms of two correlation times for motion, τB and τC, respec-
tively. With the known line widths of the three mN manifolds
and assuming an anisotropic Brownian motion where τB * τC,
the values of τB and τC can be calculated by the simplified
formulas
•
trap OH produced in the irradiation of aqueous solutions of
TiO2. The basic concept of an EPR spin trap is that a
diamagnetic radical scavenger (e.g., DMPO)10 serves as an
efficient scavenger of reactive free radicals to produce an EPR
signal of the nitroxide adduct that results from addition. In
favorable cases the structure of the trapped radical can be
inferred from the hyperfine coupling observed in the nitroxide
•
adduct. For example, the addition of OH to DMPO (eq 8)
τB ) B*∆H0[(h0/h1)1/2 - (h0/h-1)1/2]
(10a)
(10b)
τC ) C*∆H0[(h0/h1)1/2 + (h0/h-1)1/2 - 2]
In the above equations, hi represents the peak heights. Using
the value B* ≈ C* ≈ 6.5 × 10-10 for nitroxide radicals in water
solution24-26 and τc ) (τBτC)1/2, we have calculated the
correlation time, τc, for motion of 3-carboxyproxyl in water at
pH ) 3.7 to be 38 ps. Addition of 2.3 gL-1 to the sample
causes an increase of the value of τc to 45 ps. Though the effect
is not large, the increased values of τc are indicative of an
interaction of the spin probe with the TiO2 surface.
Irradiation of Aqueous Solutions of 3-Carboxyproxyl
Containing TiO2. Irradiation (280 nm < λ < 400 nm) of an
aqueous solution of 3-carboxyproxyl (0.8 mM) in the presence
of TiO2 (0.2 g L-1) results in a decrease of the EPR signal but
not in a measurable change in the shape of the signal. A
straightforward adaptation of the conventional scheme for the
mechanism of steady-state photomineralization of organic
molecules by TiO2 is given in Scheme 1 in which an elementary
reaction step and a corresponding rate expression are given.
Asmus, employing the pulse radiolysis technique, measured the
rate of addition of •OH radicals to 3-carboxylproxyl derivatives
to be close to that of diffusion-controlled reactions (ca. 3.7 ×
109 M-1 s-1), an expected value from literature values of
reactive radicals to nitroxides.27-29a The structure of the primary
product of addition was postulated to be the adduct proxyl-
NOOH (eq 11). However, no product analyses were reported.
occurs with a rate constant close to that of a diffusion-controlled
reaction (ca. 3-4 × 109 M-1 s-1) to form a nitroxide (DMPO-
OH) with a characteristic EPR splitting pattern. Sun has
determined the fraction of DMPO-OH among all the products
to be ca. 33%, a quantity termed the “trapping efficiency”, η.19
The trapping efficiency is a chemical yield based on the spin
trap converted to product. The efficiency η should not be
confused with the quantum efficiency Φ, which is based on
the number of photons absorbed, a difficult quantity to measure
in a highly scattering medium. Colussi et al.13 employed DMPO
to determine the quantum yield for photogenerated hydroxyl
radicals in irradiated TiO2 solutions to be Φ ) 0.04. This value
was in good agreement to that determined by Sun and Bolton,19
employing a quantitative analysis of the products derived from
•
the attack of OH on methanol.
In many respects, the spin-trapping method employing
3-carboxyproxyl (eq 3) in this report to track the production of
•OH is directly analogous to the conventional method employing
DMPO (eq 8). The main difference is a technical one in that
the disappearance of an EPR signal is monitored rather than
the appearance of an EPR signal. The technique reported here
has an advantage in simplicity, but suffers from the lack of direct
information on the chemical fate of the spin trap. It was
therefore of prime importance to quantitatively determine the
products derived from reaction of 3-carboxyproxyl and to
correlate the results with those reported for DMPO.
EPR of 3-Carboxyproxyl in Aqueous Solutions Containing
TiO2. The EPR spectrum of aqueous solutions of 3-carboxy-
proxyl show the well-known three-line spectrum of a nitroxide
(Figure 1) resulting from 14N hyperfine coupling (aN ) 1.6045
mT). Two standard parameters are readily derived from the
EPR spectra of nitroxides: the average polarity of the micro-
scopic environment of the spin probe derived from the analysis
of the coupling constants, and the rotational mobility of the spin
probe derived from the simulation of the spectrum and
computation of the rotational correlation time, τc, of the spin
probe. Both parameters have the potential to report interactions
of the spin trap with the TiO2 surface. The correlation times,
τc, for motion of 3-carboxyproxyl were evaluated from the
observed EPR spectra by means of the simulation program of
Schneider and Freed.20 Assuming that the spectra are charac-
Ollis et al. found that in the case of hydroxyl radicals the
average distance of molecular diffusion depends on the con-
centration of the substrate [S] and the rate constant for the
•
reaction between OH radicals and the substrate.29b Using [S]
) 10-3 M and a second-order rate constant on the order of 109
M-1 s-1, the characteristic diffusion length is around 10-8 m.29b
On the basis of these findings, we would expect hydroxyl
radicals not to diffuse very far into the solution, but to react
directly or at least considerably close to the surface. In the

7130 J. Phys. Chem. B, Vol. 101, No. 36, 1997
Schwarz et al.
Figure 2. Lifetime of the conducting band electrons with (4) and
without (3) 3-carboxyproxyl (0.8 M). [TiO2] ) 0.015 g L-1. Plot also
indicates the correlation time τc (b) for the spin trap at different pH
values. The lifetimes were determined from the kinetic absorption traces
recorded at λabs ) 620 nm (λexc ) 355 nm, 22 °C). EPR spectrometer
settings: center field, 348 mT; sweep width, 6 mT; number of scans,
5. Microwave frequency: 9.78 GHz.
Figure 3. Photolysis of 3-carboxyproxyl (0.8 mM) in the presence of
TiO2 (0.2 g L-1). For experimental details see Photolysis part in
Experimental Section.
semiconductor surface. In the pH range 4-6 the probe
molecules are strongly attached to the surface due to the
deprotonated carboxyl group (pKA ≈ 3.9) and the positively
charged TiO2 surface (isoelectric point of Degussa P25 TiO2 )
6.129c). For pH ) 3.8 we found a binding constant of 1.4 ×
104 M-1. When the pH is increased above 6, the surface charge
becomes negative and the negatively charged spin probe is
repelled from the surface, which is displayed by a decrease of
the correlation time τc. Lowering the pH below 4 results in
the protonation of the carboxyl group and therefore in a weaker
interaction between adsorbent and adsorbate which also leads
to lower values for τc. The reduction of nitroxides with
electrons is expected to produce reduction products such as
proxyl-OH (eq 12).
case of surface-bound substrates the characteristic diffusion
length should even be smaller due to the closer distance to the
reaction sites. We have determined the binding constant for
3-carboxylproxyl (8 × 10-5 M up to 8 × 10-3 M) on TiO2 (0.2
g L-1) in an aqueous solution at pH ) 3.8 to be 1.4 × 104
M-1, using the Langmuir model of adsorption. At a 3-carboxy-
proxyl concentration of 0.8 mM, 95% of all spin trap molecules
are adsorbed on the TiO2 surface, and therefore we expect eq
11 to take place with hydroxyl radicals either still bound to the
particle surface or very close to it. Further work detailing the
influence of the pH on the trapping process will be reported
elsewhere.
Although the disappearance of the signal of 3-carboxyproxyl
can be reasonably attributed to eq 11, it is also plausible that
the disappearance could be due to reaction with conduction band
•
-
electrons or species derived from these electrons, e.g., O2
.
Indeed, the reaction of nitroxyl derivatives related to 3-carboxy-
proxyl are reported28 to react with hydrated electrons with rate
constants of the order of 1010 M-1 s-1
. As a result, we
investigated the reaction of our spin trap with conduction band
electrons of TiO2 produced by photoexcitation.
Product Analysis. Product analyses are of considerable
importance in any mechanistic analysis. In the case of products
formed from strongly oxidizing and strongly reducing conditions
as those present when TiO2 is photoexcited, primary products
may rapidly be converted to secondary products, so there is no
guarantee that isolated products are primary unless conversions
are kept low and proper controls are conducted.
Aqueous solutions of 3-carboxyproxyl kept in the dark are
stable for weeks. With added TiO2 (0.2 g L-1) no observable
change in the 3-carboxyproxyl could be observed in the dark
within 2 weeks. The illumination (200 nm < λ < 400 nm) of
a TiO2-free solution of 3-carboxyproxyl produces a decrease
of the spin trap concentration from the initial 8 × 10-4 M to
7.2 × 10-4 M within 10 min due to the reactions of photoexcited
nitroxides with themselves.30a
A quantitative analysis was made of the products produced
by the steady-state photolysis (200 nm < λ < 400 nm) of
aqueous solutions of 3-carboxylproxyl (0.8 mM) in the presence
of TiO2 (0.2 g L-1) at pH ) 3.8. Aliquots of samples were
taken from the photolysis medium at minute intervals and
subjected to GS, MS, and IR analysis. The results are
summarized in Figure 3. Three experiments were run per point
and the maximum deviation was 7% of the reported value. Two
products, proxyl-NH and proxyl-OH, were identified unam-
biguously by comparison of the GC retention time, MS cracking
pattern, and IR spectrum to authentic materials.
Conducting band electrons give a pH-sensitive absorption
spectrum revealing a broad absorbance with a maximum around
800 nm under alkaline conditions.2 Lowering the pH shifts the
spectrum to the blue, showing a new maximum around 600 nm.
Taking advantage of the characteristic absorption band, we have
recorded the temporal evolution of the transient spectrum (λabs
) 620 nm) at different pH values after excitation of the
semiconductor particles with a frequency-tripled (355 nm) ND
laser pulse.
The lifetime τ of the conduction band electrons in the
presence and absence of 3-carboxylproxyl as a function of pH
is shown in Figure 2 (left ordinate). There is no significant
change in lifetime in the pH range 7-11 and a maximum change
in lifetime (from ca. 120 ns to ca. 90 ns) around pH 4.8. The
value of τc for 3-carboxylproxyl was determined (Figure 2, right
ordinate) in the same pH range, and a maximum value of τc
was found at pH ) 4.8, i.e., the same pH for which the
conduction band electrons undergo maximum quenching by
3-carboxyproxyl. These results suggest that there is a certain
amount of reaction of 3-carboxylproxyl with conduction band
electrons. The increased value of τc indicates an increased
interaction of the spin trap at pH ) 4.8, consistent with a more
efficient reaction with the conduction band electrons that have
made their way to the TiO2 surface. These findings also reflect
the electrostatic interactions between the spin trap and the

Generation of Hydroxyl Radicals in TiO2 Suspensions
J. Phys. Chem. B, Vol. 101, No. 36, 1997 7131
Under the conditions employed for photolysis, the initial
concentration of 3-carboxylproxyl decreased from 0.8 to 0.11
mM within 10 min. At the lowest conversions the disappearance
of 3-carboxylproxyl is accompanied by the appearance of a
single major product, proxyl-NH (eq 13), and a minor product,
proton (eq 9). In this way proxyl-OH represents the reduced
nitroxide species produced in the reaction leading to the different
lifetimes of the conducting band electrons with and without the
spin trap (Figure 2). The reductive pathway to proxyl-OH was
also confirmed by running the same photolysis experiment under
a nitrogen atmosphere. In the absence of any traces of oxygen,
the quantity of generated proxyl-OH is remarkably higher as
the result of the removal of an important path (eq 4) that
competes with 3-carboxyproxyl for conducting band electrons.
We also consider the possibility that the illumination with
wavelengths λ < 200 nm excites both the π f π* and the n f
π* transitions of the N-O bond.30a Nitroxides in the photo-
excited state are much better hydrogen atom abstractors than
in the ground state. Therefore, it is possible, that a certain part
of proxyl-OH molecules have been formed via hydrogen
abstraction from 3-carboxyproxyl or other species with ab-
stractable hydrogens. A reaction involving hydrogen abstraction
also explains the appearance of the third isolated product, with
a mass peak ) 370.44 mu corresponding to a “dimer” of
3-carboxyproxyl. From the mass spectrum and the IR absorp-
tion bands we propose that this product is a dimer based on
two 3-carboxyproxyl molecules and coupled via the C4 position
of the first molecule and the N-O bond of the second molecule.
proxyl-OH. A second minor product corresponding to a
“dimer” of 3-carboxylproxyl (vide infra for proposed structure)
was detected by MS (m/e ) 370.44 mu).
To examine the possibility of a direct oxidation of the spin
trap by photogenerated holes, we carried out the photolysis in
an acetonitrile/methanol (1:1) mixture, where hydroxyl radicals
are not expected to have a major impact on the reaction. An
air-saturated suspension containing 0.8 mM 3-carboxylproxyl
and TiO2 (0.2 g L-1) was illuminated (200 nm < λ < 400 nm)
for 10 min. Under these conditions the initial concentration of
the spin trap decreases from 0.8 to 0.68 mM. Besides traces of
unidentified products only one major product appears, which
could be identified as proxyl-OH. While the generation of
proxyl-OH can be attributed to either a reduction process with
conducting band electrons (eq 12) or the reaction of photoexcited
spin traps with themselves,30a the absence of any major oxidation
product clearly indicates that hole trapping by the spin probe
does not effectively compete with direct oxidation of the solvent.
Carrying out the same experiment under a nitrogen atmosphere,
we found almost the same results with slightly higher amounts
of proxyl-OH produced.
•
Up to this point we have assumed that OH is the reactive
species involved in reaction with 3-carboxylproxyl, yet the
structure of the major product, proxyl-NH, appears to indicate
that a reduction has occurred. This apparent discrepancy may
be rationalized by consideration of the most likely first step in
the reaction of •OH with 3-carboxylproxyl to produce proxyl-
NOOH as a primary product (eq 11) and proxyl-NH as a
plausible secondary reaction of this primary product (eq 14).
The unit R2NOOH which is present in proxyl-NOOH is
reminiscent of the ROOOH unit present in trioxides.30b The
later species undergo facile cleavage (eq 15), even at low
temperatures. The formal disproportionation shown in eq 16
is a plausible pathway for reaction of the species produced in
eq 15.
This observed dimer could arise from the reaction of a
biradical formed via hydrogen abstraction with a second
3-carboxyproxyl molecule, similar to the addition of carbon-
centered radicals to nitroxides. Corresponding reaction products,
with the exception of proxyl-NH, were found by Keana from
the photolysis of 4-hydroxy-2,2,6,6-tetramethylpiperidine-1-oxyl
in toluene in the absence of a semiconductor.30a The traces of
unidentified products could be attributed to a photodecompo-
sition of the spin trap, eliminating nitrogen oxide and leading
to an olefin derivative.31
From the data of Figure 3 the trapping efficiency or chemical
yield of proxyl-NH at low conversion can be computed.
During the first 5 min of photolysis, the mass balance of the
three identified products, proxyl-NH, proxyl-OH, and the
dimer, almost exactly corresponds to the amount of 3-carboxy-
proxyl consumed, indicating that the amount of unidentified
products is negligible. In this way we have calculated the
trapping efficiency η for each time from eq 17.
R2NOOH f R2NH + O2
(14)
ROOOH f RO• + •OOH
RO• + •OOH f ROH + O2
(15)
(16)
η ) [proxyl-NH]/([proxyl-NH] +
[proxyl-OH] + [dimer]) (17)
We propose that for proxyl-NOOH, the reactions analogous
to eqs 15 and 16 could be possible. Notice that proxyl-NH is
the product expected from disproportionation of a R2N• •OOH
radical pair (eq 14). Recent calculations by Sumathi et al.
have supported the possibility of the cleavage of unsubstituted
H2NOOH into NH3 and O2.30c However, at this point the
mechanism of product formation must be considered as specula-
tive.
The values derived from eq 17 for the trapping efficiency η
of 3-carboxyproxyl with hydroxyl radicals vary between 0.82
and 0.78 giving an average trapping efficiency η ) 0.79. This
value may be compared to η ) 33% reported for DMPO.12 In
this way it is possible to calculate the concentration of proxyl-
NH at any given time by measuring the signal intensity of the
spin trap 3-carboxyproxyl at that time. The difference between
the corresponding concentration and the initial concentration
of 3-carboxyproxyl multiplied by the trapping efficiency (η )
The appearance of proxyl-OH in the reaction process can
be attributed to the reduction of 3-carboxyproxyl molecules by
conducting band electrons and the subsequent addition of a

7132 J. Phys. Chem. B, Vol. 101, No. 36, 1997
Schwarz et al.
Figure 6. Effect of oxygen concentration on the rate constant k (10-7
M s-1) of hydroxyl radical formation. Continuous irradiation (280 nm
< λ < 400 nm) of a TiO2 (0.2 g L-1) suspension containing
3-carboxyproxyl (0.8 mM).
Figure 4. Hydroxyl radical concentration vs time at different light
intensities, employing the EPR signals obtained by steady-state
illumination (280 nm < λ < 400 nm) of a suspension containing
3-carboxyproxyl (0.8 mM) and TiO2 (0.2 g L-1).
yield for the generation of DMPO-OH with an increasing light
intensity are confirmed.13
Effect of Oxygen. Figure 6 shows the effect of the oxygen
concentration on the rate constant of the hydroxyl generation.
From degradation experiments with organic pollutants it is
known that the mineralization efficiency with TiO2 is dependent
on the concentration of dissolved oxygen in the suspension.32
Oxygen acts as an electron acceptor, scavenging electrons from
the conducting band (eq 4) and thus reducing the recombination
rate (eq 7). The result in Figure 6 shows that an increasing
amount of dissolved oxygen increases the rate constant of the
•OH radical formation. The result is in good agreement with
results derived by the DMPO method19 and by the mineralization
of phenol in aqueous TiO2 suspensions.32
Figure 5. Dependence of the rate constant k (10-7 M s-1) of the
formation of hydroxyl radicals on the photon flux. The data are from
Figure 4.
Effect of Added Sodium Salts. We have also examined the
effect of dissolved sodium salts on the decrease of the spin trap
signal and thus on the generation of proxyl-NH. An air-
saturated suspension containing 3-carboxyproxyl (0.8 mM), TiO2
(0.2 g L-1), and different sodium salts ([NaX] ) 0.005-0.1
M) at pH ) 3.8 was illuminated (280 nm < λ < 400 nm) for
7 min and the decrease in the signal intensity measured by EPR.
From electrochemical potentials, sodium ions are not expected
to interfere with the reduction process or the oxidation reactions
by holes or hydroxyl radicals. Therefore upon addition of
sodium salts, changes in the rate constant k for the production
of proxyl-NH should be due to the influence of the corre-
sponding anions. For the generation of DMPO-OH from
DMPO, it was found that oxidizable compounds have a strong
impact leading to lower rate constants.37 Kochany has shown
that the competition of various chlorophenols with DMPO
0.79) is the concentration of proxyl-NH. In the absence of
other oxidizable compounds, the concentration of hydroxyl
radicals directly corresponds to the proxyl-NH concentration
(eq 3) and therefore has the same value.
•
Effect of Photon Flux on the Rate of Formation of OH.
During recent years, several contradicting results dealing with
the dependence of photocatalytical rates on the photon flux I0
were reported.33-35 Okamato found a square root dependence,36
but recent results favor, at least at lower light intensities, a linear
correlation between the photon flux and the quantum yields for
the production of hydroxyl radicals in illuminated TiO2 suspen-
sions.13 Applying 3-carboxyproxyl (0.8 mM) as the spin trap
and using the derived trapping efficiency η ) 0.79, we have
•
calculated the production of OH radicals at different light
•
intensities (Figure 4) for an air-saturated suspension with TiO2
(0.2 g L-1) at pH ) 3.8. The photon flux was in the range
1.64 < I0 (× 10-5 einstein s-1) < 16.4. The illumination time
chosen was 6 min, wherein the conversion of 3-carboxyproxyl
to proxyl-NH, and thus the detectable generation of hydroxyl
radicals, is almost linear. Figure 4 indicates that the criteria of
a linear slope is reasonably fulfilled. With higher light
intensities, the generation of hydroxyl radicals is increased. From
the slopes in Figure 4, we have calculated the rate constant k
(M s-1) for the production of •OH radicals at different
illumination strengths. Despite the fact that at longer time scales
the production of proxyl-NH reaches a maximum and can be
described with an exponential function, a linear fit to evaluate
the rate constant k seems valid for the first few minutes. The
result displayed in Figure 5 clearly shows that for low light
intensities, the rate constants k for the hydroxyl production
increase linearly with an increasing photon flux. Therefore,
recent results supporting a linear dependence of the quantum
toward OH radicals in H2O2 solution obeys a linear Stern-
Volmer relationship.38 However, recent results derived by
Colussi with DMPO in aqueous TiO2 suspensions display a
nonlinear effect of dichlorobenzene toward the generation of
DMPO-OH.13
To investigate the effect of different sodium salts, 3-carboxy-
proxyl was used as a spin trap for hydroxyl radicals. Within
the first 5 min of each illumination, the EPR signal of
3-carboxyproxyl decreased linearly, allowing linear fits from
which the rate constant R for the generation of proxyl-NH can
be derived. According to Kochany38 and modified to our
system, the rate R of the formation of proxyl-NH can be
expressed as follows by eq 18.
R ) d[proxyl-NH]/dt ) k3[3-carboxyproxyl] [•OH] (18)
According to Scheme 1, the concentration of [•OH] is

Generation of Hydroxyl Radicals in TiO2 Suspensions
J. Phys. Chem. B, Vol. 101, No. 36, 1997 7133
[•OH] ) (I + k4[h+])/(k3[3-carboxyproxyl] +
k6a[substrates]) (19)
Neglecting eq 7, it follows that
R ) k3[3-carboxyproxyl](I + k4[h+])/
(k3[3-carboxyproxyl] + k6a[substrates]) (20)
R0 is the rate of proxyl-NH formation when the concentration
of the salts ([substrates]) ) 0 and can be expressed as
R0 ) (I + k4[h+])
(21)
0
Figure 7. Inverse initial rate ratio R0 /R0 (see text) vs salt concentration
[S] for different salts.
By solving eqs 20 and 21 with the initial concentrations [S]0
and [3-carboxyproxyl]0, eq 22 is obtained.
TABLE 1: Rate Constants k6a (108 M-1 s-1) for the
•
Reaction of OH Radicals with Sodium Salts. Data Is
Derived from Eq 22 and Figure 7
0
R0 /R0 ) 1 + k6a[substrates]0/k3[3-carboxyproxyl]0 (22)
NaCl
1.49
NaBr
3.83
NaI
NaNO2
14.4
0
10.9
The slopes of the plot R0 /R0 vs [S0] give a value of k6a/k3[3-
carboxyproxyl]0 with an intercept of 1 (see Figure 7). The
competition of the anions with 3-carboxyproxyl for hydroxyl
radicals obeys a linear Stern-Volmer kinetic relationship. From
the slopes in Figure 7 we have also estimated the rate constants
k8a for the reaction of •OH radicals with the anions. Using the
rate constant k3 ) 3.7 × 109 M-1 s-1 for the reaction of
3-carboxyproxyl with hydroxyl radicals (eq 3)28 and an initial
concentration [3-carboxyproxyl]0 ) 0.8 mM, we obtained the
rate constants k8a shown in Table 1.
The results in Table 1 show that the rate constants k6a for the
reaction between •OH radicals and the anions are close to those
that are diffusion-controlled. Among the rate constants for the
three halides, the rate constants k8a for the reaction
Furthermore, 3-carboxyproxyl as a spin trap was employed to
examine the effect of oxygen as a scavenger for conducting
band electrons. With an increasing concentration of dissolved
O2, the rate for the recombination between electrons and holes
is diminished, and in this way, the generation of hydroxyl
radicals slightly improved. In the presence of different sodium
salts, the conversion of 3-carboxyproxyl to proxyl-NH is
strongly affected by the anions, which compete with the spin
•
trap for OH radicals. The rate constants for the reaction
between hydroxyl radicals and the anions, derived from the
Stern-Volmer plots, are diffusion-controlled and dependent on
the electrochemical potential of the redox couple.
-
•OH + X- f OH- + •X (+ X- f •X2
)
(23)
Acknowledgment. The authors thank the National Science
Foundation, NATO, Fond der Chemischen Industrie (foundation
of the German Chemical Industry (VCI)), and BMBF for
financial support of this work. P.F.S. thanks the Deutsche
Forschungsgesellschaft (DFG) for a postdoctoral fellowship.
correspond with the decreasing potential of the corresponding
-
•
-
•
-
• -
redox couple X-/•X2 in the sequence Cl2 > Br2 > I2
.
The results derived with our method are therefore in good
agreement with results obtained by Henglein using pulsed laser
techniques.39 However it should be noted that the author has
attributed the reaction to a direct oxidation of adsorbed halides
by valence band holes at pH ) 1. We can not completely
exclude a direct oxidation of surface-adsorbed halides, but
according to results derived by Grabner and Quint the quantum
yield for halide oxidation by holes and the concentration of
adsorbed halides decrease with increasing pH of the solution.40
We have employed a pH ) 3.8, where the reaction of holes
with surface OH groups is thermodynamically favored over
halide oxidation. Therefore, we assume that halide oxidation
in our case can be attributed primarily to the reaction with
hydroxyl radicals.
References and Notes
(1) Fujishima, A.; Honda, K. Bull. Chem. Soc. Jpn. 1971, 44, 1148.
(2) Graetzel, M. Heterogeneous Photochemical Electron Transfer; CRC
Press, Inc.: Boca Raton, FL, 1989.
(3) Fox, M. A.; Dulay, M. T. Chem. ReV. 1993, 93, 341.
(4) Serpone, N. Res. Chem. Intermed. 1994, 20, 953.
(5) Photocatalytic Purification of Water and Air; Ollis, D. F., Al-Ekabi,
H., Eds.; Elsevier: Amsterdam, 1993.
(6) Photocatalysis: Fundamentals and Applications; Serpone, N.,
Pelizetti, E., Eds.; Wiley: New York, 1989.
(7) Kraeutler, B.; Bard, A. J. J. Am. Chem. Soc. 1978, 100, 2239.
(8) Frank, S. N.; Bard, A. J. J. Phys. Chem. 1977, 81, 1484.
(9) Jaeger, C. D.; Bard, A. J. J. Phys. Chem. 1979, 83, 3146.
(10) Harbour, J. R.; Chow, V. S. F.; Bolton, J. R. Can. J. Chem. 1974,
52, 3549.
Conclusions
(11) Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. J.
Phys. Chem. Ref. Data 1988, 17, 717.
3-Carboxyproxyl may serve as an alternate spin-trapping
compound to detect hydroxyl radicals in illuminated TiO2
solutions by electron paramagnetic resonance (EPR). The
products generated during a photolysis of a 3-carboxyproxyl-
containing TiO2 suspension were analyzed, and the main species
found was proxyl-NH. For the given 3-carboxyproxyl con-
centration (0.8 mM), the trapping efficiency η ) 0.82 remains
almost constant during the first 5 min of the illumination and
allows the estimation of hydroxyl radical concentration released
by the semiconductor particles. We could also confirm recent
results that at lower light intensities, the rate for the •OH radical
production rises linearly with an increasing photon flux.
(12) Riegel, G.; Bolton, J. R. J. Phys. Chem. 1995, 99, 4215.
(13) Grela, M. A.; Coronel, M. E. J.; Colussi, A. J. J. Phys. Chem. 1996,
100, 16940.
(14) Janzen, E. G. Acc. Chem. Res. 1971, 4, 31.
(15) Moza, P. N.; Fytianos, K.; Samanidou, V.; Korte, F. Bull. EnViron.
Contam. Toxicol. 1988, 41, 678.
(16) Neta, P.; Steenken, S.; Janzen, E. G.; Shetty, R. V. J. Phys. Chem.
1980, 84, 532.
(17) Sun, L. Quantum Yield and Mechanism in TiO2 Mediated Photo-
catalysis. Ph.D. Thesis, The University of Western Ontario, London, Ontario,
Canada, 1994.
(18) Carmichael, A. J.; Makino, K.; Riesz, P. Radiat. Res. 1984, 100,
222.
(19) Sun, L.; Bolton, J. R. J. Phys. Chem. 1996, 100 4127.

7134 J. Phys. Chem. B, Vol. 101, No. 36, 1997
Schwarz et al.
(20) Schneider, D. J.; Freed, J. H. In Biological Magnetic Resonance.
Spin Labeling. Theory and Applications; Berliner, L. J., Reuben, J., Eds.;
Plenum Press: New York, 1989; Vol. 8, p 1.
(21) Ottaviani, M. F.; Cossu, E.; Turro, N. J.; Tomalia, D. A. J. Am.
Chem. Soc. 1995, 117, 4387.
(22) Kivelson, D. J. Chem. Phys. 1960, 33, 1094.
(23) Jolicoeur, C.; Friedman, H. L. Ber. Bunsen-Ges. Phys. Chem. 1971,
75, 248.
(24) Martini, G.; Ottaviani, M. F.; Ristori, S.; Lenti, D.; Sanguineti, A.
Colloids Surf. 1990, 45, 177.
(25) Ristori, S.; Ottaviani, M. F.; Lenti, D.; Martini, G. Langmuir 1991,
7, 1958.
(26) Ottaviani, M. F.; Ghatlia, N. D.; Turro, N. J. J. Phys. Chem. 1992,
96, 6075.
(27) Gerischer, H. Electrochim. Acta 1993, 38, 3.
(28) Nigam, S.; Asmus, K. D.; Willson, R. L. Trans. Faraday Soc. 1976,
43, 2324.
(29) (a) Bowry, V. W.; Ingold, K. U. J. Am. Chem. Soc. 1992, 114,
4992. (b) Turchi, C. S.; Ollis D. F. J. Catal. 1990, 122, 178. (c) Furlong,
D. N.; Wells, D.; Sasse, W. H. F. J. Phys. Chem. 1986, 90, 1107.
(30) (a) Keana, J. F. W.; Dinerstein, R. J., Baitis, F. J. Org. Chem. 1971,
36, 209. (b) Nangia, P. S.; Benson, S. W. J. Am. Chem. Soc. 1980, 102,
3105. (c) Sumathi, R.; Engels, B., Peyerimhoff, S. D. J. Chem. Phys. 1996,
105, 8117.
(31) Keana, J. F. W.; Baitis, F. Tetrahedron Lett. 1968, 365.
(32) Okamoto, K.; Yamamoto, Y.; Tanaka, M.; Itaya, A. Bull. Chem.
Soc. Jpn. 1985, 58, 2015.
(33) Hoffmann, A. J.; Carraway, E. R.; Hoffmann, M. R. EnViron. Sci.
Technol. 1994, 28, 776.
(34) Lepore, G. P.; Pant, B. C.; Langford, C. H. Can. J. Chem. 1993,
71, 2051.
(35) Hong, A. P.; Bahnemann, D. W.; Hoffmann, M. R. J. Phys. Chem.
1987, 91, 2109.
(36) Okamato, K. Bull. Chem. Soc. Jpn. 1985, 58, 2023.
(37) Savel’eva, O. S.; Shevchuk, L. G.; Vysotskaya, N. A. J. Org. Chem.
USSR (Engl. Transl.) 1972, 8, 293.
(38) Kochany, J.; Bolton, J. R. J. Phys. Chem. 1991, 95, 5116.
(39) Henglein, A. Ber. Bunsen-Ges. Phys. Chem. 1982, 86, 241.
(40) Grabner, G.; Quint, R. M. Langmuir 1991, 7, 1091.